[Objective] Named entity recognition (NER), a central task in the information extraction realm, aims to precisely identify various named entity types in textual content, including personal names, locations, and organizational names. In Chinese NER domain, deep learning techniques are crucial for character and vocabulary representations and feature extractions, yielding remarkable research achievements. Common deep learning models for NER include sequence labeling, span-based approaches, generative methods, and table-based strategies. Nevertheless, this task suffers from the scarcity of lexical information. Hence, this challenge is perceived as a primary hindrance limiting the development of high-performance Chinese NER systems. Despite developing extensive lexical dictionaries encompassing rich vocabulary boundaries and semantic insights, effective incorporation of this lexical knowledge into Chinese NER task remains a considerable challenge. Particularly, the seamless integration of semantic information from matching vocabulary and its contextual cues into Chinese character sequence remains intricate. Moreover, ensuring the accurate delimitation of named entity boundaries is still a remarkable concern. In the realm of intelligent judicial systems, the NER task within legal documents has garnered significant attention. Nonetheless, prevailing sequence labeling models predominantly rely on character information, constraining their capacity to capture semantic and lexical contextual nuances and inadequately addressing entity boundary constraints. To resolve these challenges, this paper introduces an innovative model called semantic and boundary enhanced named entity recognition (SBENER). To enhance the semantic features of legal documents within the SBENER model, external information containing vocabulary pertinent to theft crimes is smartly integrated. Initially, word vectors for theft crime terms are acquired through pretraining. Subsequently, a vocabulary dictionary tree is constructed, enabling the potential vocabulary candidate identification for each character. Further, these candidates are amalgamated into a final external information vector via a bilinear attention mechanism. Additionally, a linear gating structure is introduced to mitigate interference from external information in the original text. To overcome the limitations of sequence labeling models for managing entity boundary constraints, this study designs a boundary pointer network within the model to confine entity boundaries. This involves embedding the character sequence into hidden layer representations via bidirectional long short-term memory followed by decoding to introduce probability constraints for each entity span. Ultimately, contextual and boundary information is inputted into a conditional random field for obtaining the ultimate entity classification outcomes. This design adroitly tackles the issues of vocabulary loss and boundary constraint scarcity within sequence labeling models. Experimental results on the CAILIE 1.0 and LegalCorpus datasets corroborated the effectiveness of the proposed method, yielding F₁ scores of 88.70 % and 87.67 %, respectively, surpassing other baseline models. Additionally, the study conducted ablation experiments to validate the effectiveness of each component. The experimental results showed that integrating external semantic information related to theft, enhancing entity boundary constraints through pointer networks, and incorporating gating mechanisms to restrict irrelevant information fusion were all effective approaches for achieving higher F₁ scores for the model. Furthermore, this paper applied dimensionality reduction to external semantic word vector information and conducted experimental analysis on different fusion layers. Single-layer fusion outperformed multilayer fusion, while fusion at intermediate levels yielded better results. This underscored the marked enhancement in judicial NER facilitated by the proposed approach. The SBENER model effectively enhances the proficiency of recognizing named entities in legal documents through the fusion of external information and reinforcement of boundary constraints. This pioneering method substantially contributes to advancements within the intelligent judicial systems.

[Objective] In the contemporary “We Media” era, the simplification of news production and dissemination has elevated every individual to the status of news producer and disseminator, and a large amount of false information also follows. Despite the increasing and abundant information on the Internet, the regulation of false information is relatively weak. Consequently, fact-checking is becoming more and more important work, while traditional related work tends to simply label predictions without explaining the reason for the label. The generated explanation in a few studies is also relatively primitive which is hard to comprehend. Because Fact-checking demands a substantial amount of common sense, reasoning, and background knowledge about claims. Prompt learning may further utilize common sense and reasoning ability in pre-trained language models. It may also incorporate the relevant information or additional details within the explanation for claims. In all, it is essential to generate high-quality smooth explanations and further leverage generated explanations for improving classification performance through prompt learning. [Methods] To address this multifaceted challenge, we propose the PromptWE model (Prompt With Evidence) that uses the prompt learning paradigm to integrate auto-generated explanations with claims. We not only provide natural language explanations that enhance the explainability of the classification result but also further improve the model performance by combining explanation into prompt learning. The model performs hierarchical evidence distillation on many related new reports for every claim to obtain relevant evidence, then uses the BART-CNN model to summarize these incoherent pieces of evidence into one smooth explanation. Consequently, it integrates the claim and explanation into six self-designed templates for prompt learning. Finally, we ensemble the result from different templates to predict the authenticity of the news. Moreover, we replace the generated explanation with the professional explanation from the dataset to investigate the impact of expert evidence on the prompt learning models. [Results] Our method achieves good results on two fact-checking datasets: Liar-RAW and RAWFC. Its F1 score is 5 % higher than the state-of-the-art model on both datasets at least. We also find that ensemble learning with multiple templates can effectively improve the F1 score of the model. For explanation generation, the model has a higher ROUGE-2 score than the former model. After integrating professional evidence into the prompt templates, the model achieves significant improvement in the classification results on the two datasets, with a maximum improvement of 15 % when compared to the results of the PromptWE model. Also, we find that for multi-class classification task, the model with integrated professional evidence exhibited exhibits significant performance improvement on more challenging categories, such as half-true and mostly true. [Conclusions] Related experiments indicate that incorporating extracted explanations as supplementary background knowledge about claims, along with the common sense and reasoning abilities learned from pre-trained models, into prompt learning templates can further enhance classification performance for claim veracity. Moreover, sequentially employing the methods of hierarchical evidence extraction and text summarization makes explanations more concise, coherent, and comprehensible. Also, the explanation extracted from unrelated evidence is better suited for integration into prompt learning methods. The further improvement in classification performance after incorporating professional evidence underscores that this approach could swiftly identify accurate and informative prompt templates, facilitating subsequent more efficient utilization of general large models like ChatGPT.

[Objective] In the digital age, brief titles are critical for efficient reading. However, headline generation technology is mostly used in news rather than in other domains. Generating key points in classroom scenarios can enhance comprehension and improve learning efficiency. Traditional extractive algorithms such as Lead-3 and the original TextRank algorithm fail to effectively capture the critical information of an article. They merely rank sentences based on factors such as position or text similarity, overlooking keywords. To address this issue, herein, an improved TextRank algorithm—text ranking combining keywords and sentence positions (TKSP)—is proposed. Extractive models extract information without expanding on the original text, while generative models generate brief and coherent headlines, they sometimes misunderstand the source text, resulting in inaccurate and repetitive headings. To address this issue, TKSP is combined with the UniLM generative model (UniLM-TK model) to incorporate text topic information. [Methods] Courses are collected from a MOOC platform, and audio are extracted from teaching videos. Speech-to-text conversion are performed using an audio transcription tool. The classroom teaching text are organized, segmented based on knowledge points, and manually titled to generate a dataset. Thereafter, an improved TextRank algorithm—TKSP—proposed here is used to automatically generate knowledge points. First, the algorithm applies the Word2Vec word vector model to textrank. TKSP considers four types of sentence critical influences: (1) Sentence position factor: The first paragraph serves as a general introduction to the knowledge point, leading to higher weight. Succeeding sentences have decreasing weights based on their position. (2) Keyword number factor: Sentences with keywords contain valuable information, and their importance increases with the number of keywords present. The TextRank algorithm generates a keyword list from the knowledge content. Sentence weights are adjusted based on the number of keywords, assigning higher weights to sentences with more keywords. (3) Keyword importance factor: Keyword weight reflects keyword importance arranged in descending order. Accordingly, sentence weights are adjusted; the sentence with the first keyword has the highest weight, while sentences with the second and third keywords have lower weights. (4) Sentence importance factor: The first sentence with a keyword serves as a general introduction, more relevant to the knowledge point. The sentence weight is the highest for this sentence and decreases with subsequent occurrences of the keyword. These four influencing factors of sentence weight are integrated to establish the sentence weight calculation formula. Based on the weight value of the sentence, the top-ranked sentence is chosen to create the text title. Herein, the combined TKSP algorithm and UniLM model, called the UniLM-TK model, is proposed. The TKSP algorithm is employed to extract critical sentences, and the textrank algorithm is employed to extract a topic word from the knowledge text. These are separately embedded into the model input sequence, which undergoes transformer block processing. The critical sentence captures text context using self-attention, while the topic word incorporates topic information through cross-attention. The final attention formula is established by weighting and summing these representations. The attention mechanism output is further processed by a feedforward network to extract high-level features. The focused sentences extracted by TKSP can effectively reduce the extent of model computation and data processing difficulty, allowing the model to focus more on extracting and generating focused information. [Results] The TKSP algorithm outperformed classical extractive algorithms (namely maximal marginal relevance, latent Dirichlet allocation, Lead-3, and textrank) in ROUGE-1, ROUGE-2, and ROUGE-L metrics, achieving optimal performances of 51.20 %, 33.42 %, and 50.48 %, respectively. In the ablation experiments of the UniLM-TK model, the optimal performance was achieved by extracting seven key sentences, with specific indicator performances of 73.29 %, 58.12 %, and 72.87 %, respectively. Comparing the headings generated by the UniLM-TK model and GPT3.5 API, the headings generated by UniLM-TK were brief, clear, accurate, and more readable in summarizing the text topic. Experiments were performed for real headings using a large-scale Chinese scientific literature dataset to compare the UniLM-TK and ALBERT models; the UniLM-TK model improved the ROUGE-1, ROUGE-2, and ROUGE-L metrics by 6.45 %, 3.96 %, and 9.34 %, respectively. [Conclusions] The effectiveness of the TKSP algorithm is demonstrated by comparing it with other extractive methods and proving that the headings generated by UniLM-TK exhibit better accuracy and readability.

[Objective] Several studies highlight that negative sentiment dialogues within the family remarkably impact individuals' mental and physical well-being. Conversely, positive sentiment dialogues offer individuals constructive feedback, motivating learning and personal growth. Such dialogues aid in building self-confidence and positive attitude, enabling better coping with life's challenges. Text style transfer is an effective tool to shift negative sentimental dialogues to positive sentimental dialogues. The goal of text style transfer is to retain the content of the text while imbuing the generated text with specific attributes. Sentiment style transfer is an important research direction in natural language processing, and sentiment style transfer in the context of family dialogues holds practical value. However, the current literature on sentiment style transfer has mainly focused on English datasets with relatively limited research within the Chinese domain. We constructed a dialogue-based Chinese sentimental text dataset in this study. The initial data was extracted from dialogues in the TV series “Home with Kids”, where considerable sentiment differences were observed between dialogues involving characters Liu Mei and Liu Xing as well as Liu Mei and Xia Xue. While interactions between Liu Mei and Liu Xing were primarily critical, interactions between Liu Mei and Xia Xue were characterized by encouragement and respect. Preprocessing was applied to this dataset in the following steps: (1) Data cleaning, filtering, and format conversion were performed to ensure data quality and consistency. (2) A recurrent modeling annotation method was employed using suitable algorithms and models to annotate the data, identifying key information and features. Six iterations were performed, with the classifier being fine-tuned using the data updated from the previous iteration each time. (3) Manual annotation was also conducted, meticulously reviewing and labeling the data manually to further enhance accuracy and reliability. Furthermore, the final dataset comprises 30 836 sentences, including 11 562 sentences with positive sentiment content and 19 274 sentences with negative sentiment content. In this dialogue dataset, most texts explicitly contain sentiment-related words. Based on the characteristics of this dialogue dataset, research involving dialogue-based Chinese positive sentiment style transfer was started using editing-based delete-retrieve-generate (DRG), tagger and generator (TAG), conditional Bert (CondBert), and tagging without rewriting (TWR) models. In addition, the improved TWR (TWR^*) Transformer model was introduced. The original TWR model used a multilayer perceptron to train a style classifier. To improve the ability to accurately identify specific styles, a style classifier was trained based on RoBERTa-Large-Chinese model for distinguishing different text styles. These experiments demonstrated that using the pretrained language model RoBERTa-Large-Chinese produced enhanced classification results, which was attributed to the close relationship between the attention weights of the penultimate layer in the Transformer model and words commonly associated with positive and negative sentiments. RoBERTa-Large-Chinese model presented a higher accuracy in recognizing textual sentiment style attribute words. Experimental results confirm that the style classifier trained on our dataset can effectively identify negative content within text. Through both automated and manual evaluations, this TWR^* model outperforms baseline models in identifying textual sentiment attributes, achieving positive sentiment style transfer, thus verifying the effectiveness of model enhancements and the validity of the dataset.

[Objective] Existing emotion distribution label enhancement (EDLE) methods construct the emotion distribution based on a discrete spatial emotion model; hence, expressing the correlation between emotions in a granular manner with continuous values is challenging. Therefore, herein, a valence-arousal-dominance(VAD) emotion knowledge-based text emotion distribution label enhancement (VADLE) method is proposed based on the VAD continuous-dimensional psychology emotion model. Unlike existing EDLE methods, VADLE uses VAD emotion knowledge in a three-dimensional continuous space to model emotion correlations and generate a more nuanced emotion distribution. The VADLE method comprises several steps: (1) Extraction of emotion word information via referencing lexicon and extracting emotion words from a given sentence. (2) Generation of priori emotion distributions for emotion labels using a local linear-weighting algorithm. The algorithm measures the effect of secondary emotion on the primary emotion based on the VAD emotional spatial distance and assigns weights to nearby emotions using a Gaussian kernel. (3) Construction of sentence-level emotion distribution by combining the prior emotion distributions of sentence and textual emotion words. Furthermore, this study uses joint loss to train a multitask emotion distribution learning model based on the robustly optimized bidirectional encoder representations from transformers pretraining approach (RoBERTa) pretrained language model. This approach simultaneously optimizes the prediction of emotion distribution and classification. The sentence text features extracted using the RoBERTa pretrained model are then passed through a fully connected layer to generate a probability distribution over all emotion labels. Based on this probability distribution, the model utilizes the Kullback-Leibler (KL) loss for measuring the distance between the predicted and actual distributions, optimizing the emotion distribution prediction task. Simultaneously, cross-entropy loss is employed for optimizing the emotion recognition task. To evaluate the performance of the proposed VADLE method, extensive comparative experiments is performed on several single-label English datasets using four baseline EDLE methods: emotion wheel and lexicon-based emotion distribution label enhancement (EWLLE), lexicon-based emotion distribution label enhancement (LLE), Mikels emotion wheel-based emotion distribution label enhancement (MWLE), and One-Hot. Moreover, this study explores the effect of the bandwidth parameter (τ) in the local linear-weighting algorithm on the balance between the primary and secondary emotions in the generated emotion distribution. The performance of the model's emotion prediction was assessed using four classification evaluation metrics (Precision, Recall,F₁-score, and Accuracy) and four emotion distribution prediction metrics (Canberra, Chebyshev, Cosine, and Intersection). The experimental results demonstrated that the VADLE method was superior to the baseline methods. Specifically, the VADLE method achieved superior performance on the emotion classification task over the EWLLE, LLE, and MWLE methods across all four indicators. The VADLE method also exhibited excellent performance for the emotion distribution prediction task. For instance, on the Cosine metric, the VADLE method outperformed the suboptimal EWLLE method by 2.6 % and exhibited considerable improvements over the LLE, MWLE, and One-Hot methods. The results showed that the optimal balance could be achieved by setting τ to 0.6, enabling the highest level of performance in the emotion distribution generation. Unlike existing EDLE methods, the VADLE method employs a fine-grained approach to studying emotions. It combines the prior emotion knowledge in the VAD continuous space with the linguistic information inherent in the sentiment words for generating more reasonable emotion distributions. Experimental results reveal that the VADLE method outperforms existing methods in terms of enhancing the emotion distribution labels in emotion prediction tasks.

[Objective] In the context of China's promotion of the development and utilization of renewable energy, ground source heat pump (GSHP) systems based on shallow geothermal energy has been extensively used. With increasing service life of GSHP systems, the risk increases if the operation and maintenance are not properly performed. The operation analysis and operation strategy optimization of existing large-scale GSHP systems are of great value in standardizing the design and operation management of GSHP systems. In this work, the GSHP system of a residential community in Beijing is used as a case study. The heating, ventilation and air conditioning (HVAC) system in the community is a compound system, which is composed of a GSHP system, a chiller, a cooling tower and a gas-fired hot water boiler. The GSHP system has been used for heating and cooling buildings since 2014 with no other equipment. According to its operational records from 2014 to 2020, the current operating characteristics and rules of the system are summarized. A 3-D finite element model (FEM) was developed to perform back analysis of the soil thermophysical parameters. A 2-D FEM was designed to analyze the heat transfer of the buried heat exchanger group. The characteristic parameters of the heating and cooling loads of the GSHP system were determined by comparing the 2-D numerical results with the operating data. Furthermore, using 2-D FEM, the changes of ground temperature in the borehole area, shutdown situation, and water supply/return temperature of the ground source end were predicted for the next 5 years for optimal operation strategies. Finally, the long-term stability of the system was assessed using these strategies. As an introduction, there are two GSHP subsystems in the residential community, one (#53) with 108 boreholes and the other (#54) with 304 boreholes. Each borehole is 120 m deep, 15—18 cm in diameter, and 3.6 m in spacing. The operational data involved the supply and return water temperature and flow velocity in the ground source. From 2014 to 2020, every year, heat extraction was higher than heat rejection. For #54, the average heat imbalance rate reached 16.2 % from 2017 to 2019. The ground temperature in the borehole area decreased from 14.78 ℃ to 13.00 ℃. For the #54 system, the results of the 2-D FEM analysis revealed the following. (1) Using current operational measures to continue for 5 years, it was observed that the temperature of the water supply would still be at a low level in winter, there was a risk of shutdown, and the heat extraction and output level would be further reduced. (2) Three possible operation strategies were predicted: (a) stop the operation of the GSHP system in one winter season; (b) increase the use of the GSHP system in summer (considering the two unbalance rates of 5 % and 10 %); and (c) guarantee a certain source of heat and use gas-fired boiler peak regulation. All three measures alleviate further decreases in ground temperature and enhance downtime. The heat imbalance rate is maintained at 5.0 %—8.3 % for (b) and (c). In this case, because of imbalanced heating and cooling loads, after the GSHP system had been running for several years, the ground temperature had decreased and energy efficiency of the system had decreased in winter. The operational strategy requires adjustments. In view of the change in ground temperature, shutdown situation, water supply temperature at the ground source, and heat extraction/rejection in the next 5 years, using the gas-fired boiler set in the composite system is the most appropriate operation strategy.

[Objective] Energy utility tunnel is a new type of energy underground structure based on a soil-source heat pump system buried-pipe heat exchanger and underground utility tunnel. Currently, the thermal response mechanism of an energy utility tunnel is unclear, specifically the heat transfer efficiency and thermal stress of the energy pipe corridor floor. Based on the comprehensive pipe corridor project of Longyuan Road located in Jiaozuo City, a thermal response test was conducted under various test conditions to discuss heat transfer and mechanical properties of the energy pipe corridor. Heat exchange tubes were laid at the bottom of the pipe corridor to form an energy pipe corridor. The inlet and outlet water temperatures and the temperature and strain of the pipe corridor floor were measured. Subsequently, heat exchange performance and mechanical properties of the pipe corridor floor were discussed. The average initial temperature of the energy pipe gallery floor at a depth of 7 m was approximately 21.4 ℃ in summer and 12.4 ℃ in winter. Moreover, the initial average temperature of the soil layer under the bottom floor was approximately 20.2 ℃ in summer and 13 ℃ in winter. The maximum thermal compressive stress under the heat removal condition in summer is 1.35MPa, and the maximum thermal tensile stress under the heat extraction condition in winter is 0.89 MPa, both did not exceed the strength value of concrete in the pipe corridor bottom plate. Inlet water temperature increased from 30 ℃ to 35 ℃, and heat transfer power increased from 22 W/m to 28.7 W/m, resulting in a heat transfer power increase of approximately 30 %.Moreover, when the flow rates were 300 L/h, 600 L/h, and 900 L/h, heat transfer power were 14.6 W/m, 29.3 W/m, and 28.7 W/m, respectively. Compared with a continuous operation, an intermittent operation increased the heat transfer power from 30.9 W/m to 36.9 W/m on the second day and from 30.6 W/m to 35.4 W/m on the third day. When the initial average temperatures were 21.3 ℃ and 12.5 ℃, the heat transfer power were 23.9 W/m and 51.6 W/m, respectively. The heat transfer power of winter heating conditions was 14.3 W/m, and that of summer heat removal conditions was 22 W/m. Field test results show that the temperature at various locations of the base plate of the corridor is the same in the process of heat transfer; however, temperature stress is different. The transverse temperature stress is greater than the longitudinal temperature stress, and the transverse temperature stress gradually decreases on moving from north to south. Furthermore, the longitudinal temperature stress is greater in the center and lesser on both sides. The heat transfer power decreases with the test time extension and gradually stabilizes, and the heat transfer power fluctuates greatly in the first two days; therefore, the test duration should be more than 48 h. The heat transfer power increases with the increase in water inlet temperature. Increasing the flow rate can improve the heat transfer power; however, a large flow rate can make the heat transfer insufficient, resulting in a decrease in the heat transfer power. Thus, an intermittent operation can ensure higher heat transfer power compared with that during a continuous operation. However, even when the operation time was extended, heat transfer power continued to decline compared with that on previous day. Therefore, the energy pipe gallery floor is more suitable for summer cooling.

[Objective] Pile foundations are one of the most commonly used foundation types in permafrost regions and are characterized by low thermal disturbance and high bearing capacity. Reducing engineering thermal disturbances and improving the long-term stability of pile foundations are key concerns in permafrost engineering. For friction pile foundations, bearing capacity mainly depends on the freezing strength at the interface between the pile and permafrost. Recently, global warming has increased, leading to an acceleration of the degradation of permafrost. It appears that the traditional design method relying solely on increasing pile diameter and length to improve pile bearing capacity is too conservative. Additionally, these methods do not have the bearing capacity reserves and may not be able to address the challenges of climate warming. Therefore, pile foundation settlement frequently occurs in permafrost regions. This study introduces solar cooling technology into permafrost engineering by proposing a solar cooling pile foundation. It comprises a solar power generation system, a vapor compression refrigeration system, and a concrete pile. This new structure actively cools the permafrost around the pile using a solar cooling system to protect the permafrost from climate warming effects. In this study, onsite experiments were conducted using a model pile in the Qinghai-Tibet Plateau permafrost region. The model pile had a diameter and length of 0.16 and 4.5 m, respectively. We analyzed the actual cooling effects of the solar cooling pile, including the cooling temperature, cooling radius, and cooling power. Furthermore, we established a numerical model of the temperature field of solar cooling piles using finite element software (COMSOL Multiphysics). We conducted long-term cooling performance simulations under different cooling durations (6, 9, and 12 h/d). The field test results demonstrated that the cooling temperature of the solar cooling pile sidewall could be reduced to a negative temperature, and the cooling radius reached 0.65, 1.24, and 1.5 m after operating for 3, 10, and 30 days, respectively. The adequate cooling power of the solar cooling pile was estimated to be approximately 180 W through theoretical analysis and numerical simulation. The coefficient of performance was approximately 0.9. The simulation results revealed that the longer the cooling duration is, the greater the amplitude of the pile-side temperature and the lower the stable temperature is. The pile temperature corresponding to cooling durations of 6, 9, and 12 h/d were reduced to -2.39 ℃, -3.48 ℃, and -4.45 ℃, respectively; moreover, after ten years, the influence radius increased to 6.68, 8.34, and 9.46 m, respectively. Even if the solar cooling pile stopped operating, the permafrost around the pile could remain in a stable low-temperature state for 2—4 years, providing ample time to maintain the solar cooling system. The solar cooling pile can significantly reduce the permafrost temperature around the pile, effectively preventing permafrost degradation. In the future, it can be combined with remote control of the cooling temperature and duration, offering the prospect of achieving precise supplementary cooling for permafrost. This study provides a new method for designing and constructing piles in permafrost regions.

[Objective] The global energy crisis is an urgent issue, highlighting the importance of shallow geothermal energy as a pivotal renewable energy source to support sustainable social development. Energy piles, a key method for harnessing shallow geothermal energy, hold enormous potential for growth and widespread application in global energy conservation and emission reduction projects. Research into the thermal performance of precast high-strength concrete (PHC) energy piles has advanced rapidly. Notably, phase change materials (PCMs) can absorb or release latent heat at a constant temperature during phase transitions, which makes PCMs an ideal backfill material to improve the heat transfer performance of PHC energy piles. However, a key drawback of PCMs is their low thermal conductivity, which could compromise the heat transfer performance of the PHC energy piles. To optimize the use of PCM in backfill materials, researchers have been exploring the addition of substances with high thermal conductivity. Graphite, in particular, has emerged as one of the most preferred fillers to enhance PCM thermal conductivity. Notwithstanding these advancements, there is limited research on how varying graphite volume fractions in PCM backfill materials affect the heat transfer performance of precast PHC energy piles. We established an indoor model of PHC energy piles and conducted tests using different backfill materials. To calculate the thermal conductivity of the backfill materials with different graphite volume fractions, we proposed and validated a theoretical calculation model based on the Maxwell model. A finite numerical model was developed using COMSOL Multiphysics, underpinned by empirical validation. Leveraging these models and experimental investigations, we analyzed the effects of different graphite volume fractions on several parameters, including the temperature difference between the inlet and outlet of the PHC energy pile, pile temperature, and the phase change state of the backfill material. The temperature difference increases with the graphite volume fraction. During the heat transfer process, the temperature difference gradually decreases until it stabilizes. A higher graphite volume fraction accelerates the heat exchange process within the PHC energy pile system, thus contributing to improved heat transfer performance. We observed a strong linear correlation between the thermal conductivity of the backfill materials and the temperature difference. As the graphite volume fraction increases, the pile temperature rises rapidly during the heat transfer process. When the operation time of the PHC energy pile remains constant, the pile temperature increases with increasing graphite volume fraction, whereas the axial temperature increment of the pile shows no significant change and exhibits a relatively uniform distribution. The use of the backfill materials designed in the numerical models ensures the stable operation of the PHC energy piles. The rate of phase change can be accelerated, and the recovery time can be shortened by increasing the graphite volume fraction of the backfill materials. Phase change backfill materials with a high graphite volume fraction can improve the heat transfer performance of PHC energy piles. This improvement is pivotal in meeting the high endurance energy requirements of buildings, thereby facilitating the efficient extraction of geothermal energy.

[Objective] With the rapid increase in aircraft thermal loads and the flight Mach number, combustion fuel and ram air can no longer effectively cool the fuel thermal management system (FTMS). Recently, hot fuel return has become an important way to enhance the cooling capacity of the FTMS, and the regulation of the hot fuel return flow has garnered significant attention. However, the impact of the hot fuel return flow on the limited temperatures still requires systematic research, and the thermal load limits of the FTMS have not been explored. Therefore, in this study, the hot fuel return characteristics of the FTMS with multiple temperature limit points were investigated to improve system performance. Considering the scalability of the models, a steady-state simulation model for the FTMS was developed using Python based on the thermal fluid network method and solved using the damped Newton method to guarantee the convergence of the flow and heat transfer iterations. In addition, the program was verified via experiments to guarantee its accuracy. A complete FTMS flow path was designed, and a standard condition was set for subsequent calculations. First, the temperature variations of the temperature limit points with the hot fuel return flow were calculated under the standard condition. Subsequently, by increasing the airborne thermal load under the standard condition, the maximum airborne thermal load that the system can withstand under the action of the hot fuel return (airborne thermal load limit) was investigated, and the occurrence condition of the limit state was analyzed. Lastly, the maximum total thermal load for normal operations (total thermal load limit) was explored under the standard condition, and the condition for reaching the total thermal load limit was discussed by changing the aeroengine thermal load. The increase of the hot fuel return flow may not decrease all the limited temperatures, even inducing the outlet temperature rises of the fuel tank and the hot fuel return valve. Excessive or insufficient hot fuel return flow may result in the overtemperature of the FTMS, and there exists a change interval for it to meet the multiple temperature limitations. For the outlet temperatures of the airborne thermal load heater and the fuel nozzle, there exists a critical flow of the hot fuel return, which indicates that the two outlet temperatures will not change once the hot fuel return flow reaches the critical flow. Combined with the outlet temperature rises of the fuel tank and the hot fuel return valve, when the hot fuel return flow surpasses its critical value, further increasing the hot fuel return flow will only increase the risk of system overtemperature. Moreover, the FTMS exhibits similar hot fuel return characteristics under different airborne thermal loads, and the critical flow of the hot fuel return rises with increasing airborne thermal load. However, as the critical flow of the hot fuel return rises slower than the lower boundary of the limited interval for the hot fuel return flow with the increased airborne thermal load, the airborne thermal load limit corresponds to the critical state when the size of the limited interval for the hot fuel return flow mutates into zero, and the lower boundary of the limited interval of the hot fuel return flow is just the system critical flow of the hot fuel return in this condition. Furthermore, the calculation results reveal that when the outlet temperatures of the airborne thermal load heater and the fuel nozzle reach their respective limit values in the limit state, the total thermal load limit can be achieved. In addition, to fully utilize the total thermal load limit of the FTMS under the action of an unreasonable aeroengine thermal load, the intermediate loop in this paper can be used to achieve the mutual transfer of the system thermal loads by the heat exchangers and refrigerating devices. As long as the total thermal load does not exceed the total thermal load limit, the FTMS can ensure that the system works normally through the intermediate loop to adjust the new aeroengine thermal load transferred into the fuel. This study explains the temperature variation regularity of multiple temperature limit points and the thermal load limits under the effect of the hot fuel return, providing a reference for the design of the thermal load distribution and the regulation strategy of the hot fuel return flow.

[Objective] The Yarlung Zangbo River contains numerous hydropower resources, with high head and large flow rates in the downstream region, which is conducive to power generation by Pelton turbines. Pelton turbines convert the kinetic energy generated from the water potential energy into mechanical energy for rotating a runner. The runner is the core component for the flow and work of the Pelton turbine, and the shape of its bucket is crucial for the runner's performance, which is uniformly arranged along the hub of the runner. As the surface shape of the bucket is complex, several parameters are required to determine its geometry model, undoubtedly posing a huge obstacle to work. In this paper, a design method is proposed to address the problem of designing and improving the bucket based on the Bézier curve. The design space is simplified as much as possible based on geometry, and the Bézier curves are utilized for designing the bucket shape. An orthogonal analysis is applied for the optimization of bucket parameters, while the computational fluid dynamics method is employed for analyzing the energy characteristics and three-dimensional flow field of the Pelton turbine. In the bucket design method, the three-dimensional geometry of the bucket can be divided into contour, flow profile, and guidelines, and several characteristic parameters can be determined for those lines. Each type of line includes several biquadratic Bézier curve connections. The number of characteristic line parameters is decreased by establishing a connection between five control points of the Bézier curves. Thus, a three-dimensional design method for the bucket of the Pelton turbine is proposed based on the five controlled characteristic parameters. The main optimization parameters are chosen by the geometry. Subsequently, bucket depth, width increment, outflow angle, splitter angle, and cutout diameter are chosen to conduct orthogonal optimization for the Pelton turbine bucket. For further analysis of the flow characteristics of the optimized bucket, the runner is modeled based on the optimum parameters. In the computational fluid dynamics method, grids are meshed by ICEM, and computational fluid dynamics is performed with ANSYS FLUENT. The results of the polar analysis and three-dimensional unsteady flow field revealed that width had the maximum influence on runner efficiency; outflow angle, cutout diameter, and bucket depth had a smaller influence; and splitting angle had the minimum influence. After optimization, the hydraulic efficiency of the Pelton turbine was increased by 6.71 %. The optimized bucket demonstrated a larger torque peak than the prototype bucket. The bucket always showed large torque when its torque decreased to zero and exhibited smoother curve transition and longer work time. Thus, the optimized bucket demonstrated greater total torque than the prototype bucket; furthermore, the former's high-pressure area was larger, making the energy conversion of water from the nozzle to the bucket more effective. This paper proposes a three-dimensional design method for the Pelton turbine bucket based on the controlled characteristic parameters. The energy performance of the Pelton turbine was enhanced by the orthogonal optimization and three-dimensional flow simulation.

[Objective] Shell-and-tube steam condensers are widely used in power stations and ships; their performance, which requires an in-depth understanding of the internal flow and heat transfer process, ensures device efficiency and safety. The distribution of tube-side cooling water in its flow (z -direction) and transverse (x-y plane) directions influences the distributions of various physical quantities inside a condenser and its design optimization, which cannot be calculated by purely two-dimensional (2D) computational fluid dynamics (CFD) simulation. Moreover, the complexity of the two-phase flow, turbulence, phase change, and heat transfer mechanisms inside the condenser makes full 3D CFD simulations computationally expensive. Herein, a quasi-3D algorithm considering the transverse distribution of cooling water was proposed and applied to simulate the condensation process in a shell-and-tube steam condenser. The tube bundle region was simplified through porous media assumption, introducing extra resistance source terms to momentum transport equations. The condensation source term in the continuity equation was computed based on equivalent thermal resistance, referring to the summation of cooling-water convection, tube wall, condensate water, and noncondensable gas (air) thermal resistance. Considering the temperature rise along the z -direction, the quasi-3D algorithm split the condenser into several sections along the z -direction, performed 2D simulation in the midplane of each section, and used the secant iteration method to balance the steam pressure drop. The simulation was conducted based on ConDesign-2D, a self-developed 2D CFD code for the condensation process of shell-and-tube condensers, which adopted unstructured meshes and the collocated-grid-based SIMPLE algorithm. The temperature rise of cooling water along the flow direction directly reduced the heat transfer temperature difference by 10 % and affected the distribution of important physical quantities such as the condensation rate. Compared with the reference condensation rates of different sections, the calculated values considering the transverse distribution of cooling water are more accurate than those that do not. Additionally, the transverse distribution of the cooling water led to a more uniform condensation rate distribution due to the negative feedback between cooling-water temperature and condensation rate. However, this variable had less influence on the distribution of noncondensable gas (air). The comparison with 2D results revealed that 2D and quasi-3D simulations gave similar results on the midplane of the condenser, which illustrates the linearity of field distributions in the z -direction. According to the computation complexity analysis, the complexity of the full 3D simulation can be 2-3 orders of magnitude higher than that of 2D simulation, whereas quasi-3D simulation can be only 1 order of magnitude higher. Compared with 2D simulation, the proposed quasi-3D algorithm can compute the 3D distribution information of the flow field in the cooling-water flow direction and avoid the high computational cost brought by full 3D simulation. Therefore, for the rapid and accurate prediction of 3D flow and the heat transfer process inside a practical shell-and-tube condenser, the proposed quasi-3D algorithm is preferred over 2D and full 3D simulations.

[Objective] The stage-stacking method, which is based on stage characteristic curves and uses a sequential calculation scheme, is a valuable tool for predicting the performance of multistage axial-flow compressors. A slight variation in mass flow can cause considerable changes in the pressure ratio because the compressor operates at high speed, and constant speed lines exhibit near-vertical orientation. To avoid this issue, a traditional approach is to convert the mass flow rate boundary into pressure ratio boundary conditions. This undoubtedly increases the complexity of calculations. Furthermore, practical monitoring parameters and boundary conditions linked with other components in an entire gas turbine system are state variables, namely, pressure and temperature, rather than the mass flow rate. Consequently, if we adopt the mass flow rate boundary condition, an initial value must be assumed, and the result may be obtained through an intricate iterative process. Thus, the calculation frequently becomes highly inefficient. Targeting the complexities and inefficiencies inherent in the traditional approach, a modified stage-stacking method is developed. The modified stage-stacking method, similar to the traditional approach, is based on two generalized stage performance curves, namely, pressure coefficient and efficiency curves. Each stage is considered as an independent control volume, delineated by its physical boundary. This method uses thermodynamic parameters—static temperature, static pressure, and axial velocity—along meridional streamline at a mean radius of all stage inlets and outlets as unknown variables. When conservation of mass, momentum, and energy is applied to each stage, a nonlinear system with 3n governing equations is obtained for a compressor of n stages. These equations involve 3n variables with the inlet total pressure, the total temperature, and the outlet total pressure as boundary conditions. Thus, the group of equations can be simultaneously solved. The Newton-Raphson method is used as the iterative numerical solver for the nonlinear algebraic equation set. The thermodynamic properties are determined by functions from the Multiflash library. Furthermore, while assuming a linear mathematical correlation between the variable stator vane and the inlet guide vane (IGV), the impact of the variable geometry of a modern heavy-duty gas turbine compressor on the performance is investigated. In addition, to analyze the effect of air bleeding on compressor performance, the bleeding quantity is deducted from the pertinent continuity equation. According to this approach, a model is developed to predict the performance of a multistage axial-flow compressor featuring variable geometry. To validate the accuracy of the model, four representative compressors for fixed geometry, variable geometry, and interstage bleeding with distinct parameters are selected as research subjects. Compared with field data, the results are in good agreement with an average relative error of only 1.593 % for fixed geometry compressors. For variable geometry compressors, excellent agreement is observed between the predicted results and field data, with a maximum relative error of 3.856 % at high constant speed lines. For low constant speed lines, despite the largest relative error of 10.834 %, the absolute error remains small and within an acceptable range. Of great importance is the strong conformity between the trends of compressor performance with speed variation and IGV adjustments obtained from this model and field data, providing a substantial indication of the result accuracy. If a suitable IGV schedule is chosen, the relative error can be as low as 0.450 %. In addition, the model can accurately estimate the geometric and thermodynamic parameters with limited design parameters, with root mean square errors of 0.022 and 0.918, respectively. These results show that the modified stage-stacking method can not only precisely calculate the overall performance of axial-flow compressors and assess the impact of the variable geometry on compressor performance but also obtain the geometric and thermodynamic parameters of each stage of such compressors. This method serves as a valuable framework for developing steady-state and dynamic compressor models.

[Objective] Perovskite solar cells have drawn considerable attention in recent years. The addition of additives to perovskite precursor solutions is an important method to improve the quality of perovskite films for enhancing the performance of perovskite solar cells. In the past, alkali metal ions were extensively used as additives. Rubidium ions (Rb⁺) were generally added into perovskite films alongside other kinds of cations, following which the photovoltaic performance of the solar cells was clearly improved. However, few researchers studied the effects of only adding various proportions of Rb⁺ on perovskite films. In this study, rubidium chloride (RbCl) was used as an additive in perovskite precursor solutions and the morphology and structure of perovskite films were analyzed. Perovskite films were fabricated using a two-step method. RbCl was used as an additive into lead (II) iodide (PbI₂) precursor solutions with the RbCl proportions 2 %, 4 %, 6 %, 8 %, 10 %, and 12 % versus PbI₂, and a PbI₂ precursor solution with no RbCl added was used as the control. Scanning electron microscopy (SEM), energy-dispersive spectroscopy (EDS), and X-ray diffraction (XRD) analysis were employed to analyze the difference in surface morphology and structure of the perovskite films. Steady-state photoluminescence (SSPL) and time-resolved photoluminescence (TRPL) spectra were recorded using devices of fluorine-doped tin oxide (FTO)/SnO₂/perovskite films to study the carrier-transporting properties. The photovoltaic performances of the perovskite solar cells were studied through a solar simulator and external quantum efficiency testing. UV-visible (UV-vis) absorption spectra were recorded to explore the change in light absorption. The crystalline grain size is clearly enhanced upon adding 4 % RbCl. The grain size is 1.61 μm in the control and 2.14 μm upon adding 4 % RbCl. However, a high addition proportion (>8 %) damages and distorts the crystal structure, decreasing the film quality. Adding RbCl at a low proportion is beneficial for guiding the growth of perovskite grains, increasing grain size, and forming a dense film with fewer holes. The XRD patterns reveal that the peak at 12.6° corresponding to PbI₂ is suppressed upon adding RbCl, whereas a new peak appears at 11.3°. The suppression of the PbI₂ peak and the appearance of the new peak can be attributed to the formation of the RbCl complex and excessive PbI₂, and the complex can be observed in the SEM images, which is confirmed by EDS results. The TRPL results reveal that adding RbCl at a low proportion enhances the transport and extraction of charge carriers, which is consistent with the SSPL results. Furthermore, the photovoltaic performance results reveal that with RbCl as an additive, the photoelectric conversion efficiency of the perovskite solar cells increases from 18.88 % to 20.06 %, and photoelectric properties such as open-circuit voltage, short-circuit current density, and filling factor are considerably improved. However, the UV-vis absorption spectra show that the absorption is not improved upon adding RbCl and even decreases with a high addition proportion, which is due to the increasing roughness of the perovskite films with increasing RbCl proportion. The enhancement of the photoelectric properties is due to the increase in transport and extraction of charge carriers caused by the improvement in film quality. This research demonstrates that adding RbCl at low proportions can enhance the grain size and transport of the carriers, improving the photovoltaic performance. The optimal RbCl addition proportion is ~4 %. This study has considerable potential for improving the performance of perovskite solar cells.

[Objective] Owing to the airplane requirements for lightweight, simple system structures and easy maintenance, air cooling is the primary thermal management for airplane power batteries. Previous studies have explored the impact of air-cooled thermal management on battery performance, but the research combining the effect on performance with the underlying mechanism is limited. To investigate the impact of air cooling on the performance of ternary lithium batteries in airplanes, a lithium battery thermal management test platform was developed, with a wind speed-adjustable power. The platform is independently designed and consists of a check valve, inlet connecting section, battery placing middle section, outlet connecting section, cooling fan, and pulse width modulator (PWM) speed control module, among other components. The system includes four main functions: rectification, insulation, stability, and high-precision wind speed control. The rectification function regulates the airflow direction for uniformity. The insulation function is achieved by installing adiabatic check valves at the entrance and exit. The airflow stability is ensured by utilizing the Venturi effect to maintain low inlet pressure and pulsation. Finally, the high-precision wind speed is controlled by a PWM speed control module. The platform has the additional benefit of a stable and uniform wind speed, which strongly correlates with the current and enables precise wind speed control by adjusting to the current size. Based on this platform, experiments were conducted to study the impact of air cooling on the performance of lithium batteries, including thermal, electrical, and material performance, and explore the relationship and impact mechanisms among them. The experimental results indicate the following: (1) applying air cooling can effectively reduce the surface temperature of the cell body, maintaining it within a suitable working temperature of 45 ℃ and a temperature difference of 5 ℃. Concurrently, air cooling significantly minimizes temperature fluctuations, resulting in lower temperature stresses throughout the battery and greater stability of its internal structure; (2) the thermal performance of the battery is improved to weaken its impact on material performance, preventing the fragmentation of cathode particles in ternary lithium batteries, maintains a stable and orderly layered structure, and suppresses the loss of cathode active material and active lithium; (3) under suitable wind speed conditions, such as 6 m/s, the battery material maintains a stable and significant inhibition of resistance growth, impeding the decline in battery capacity and effectively extending the service life of the battery. The capacity degradation rate of air cooling is significantly lower than that of the no-air-cooling condition under the same number of cycles. Herein, we address the research gap on the effect of air cooling on the performance of ternary lithium batteries in airborne power batteries. Experimentally, we investigate the impact of air cooling on the thermal, electrical, and material performance of ternary lithium batteries, providing insights into the intrinsic mechanism of air cooling on battery performance. The results can guide the design of power battery systems for airplane operations. Additionally, it offers data support and a theoretical basis for developing next-generation power battery thermal management systems.

[Objective] Emergency sensemaking can control a situation by identifying the warning signs early and before the situation deteriorates further. However, the different benefit paths of multiagents in emergency sensemaking can easily lead to sensemaking failure and additional escalation of risk event losses. Therefore, it is vital to study how to influence the strategic choice of the relevant agents in sensemaking in the real world. [Methods] Based on the prospect theory-mental account, this report applied the evolutionary game model to study the following complex behavior mechanism among the multiagents in emergency sensemaking: the public (truthfully providing information or falsely providing information), frontline personnel (efficiently reporting or inefficiently reporting), and emergency command (group wisdom or group myth). According to different strategic choices, the payoff matrix is constructed, and the evolutionary replication dynamic equation is obtained using the matrix; however, the three evolutionary replication dynamic equations cannot directly determine the equilibrium point of the tripartite strategy. In this study, the Jacobian matrix is obtained by partial derivation of the forward decision probability of three differential equations, and the stability of the strategy choice of each game agent is analyzed by calculating the eigenvalues of the matrix. In order to further analyze the influence of key elements on the evolution process and results of the game. The computer simulation software MATLAB was used to establish a game model and assign initial values to each parameter in the model. The key parameters affecting the evolution path of decision-making behavior are extracted and analyzed by adjusting the initial values of the parameters. [Results] Solving the game model provides the following findings: 1) When the labor cost of frontline personnel is high, or the punishment is weak, the cost of the emergency command choosing a speculative strategy is substantially reduced. In this case, emergency sensemaking completely depends on the spontaneous behavior of the public, resulting in a lack of coordination efficiency and accurate judgment from frontline personnel and the emergency command. 2) By increasing the reward subsidies for positive decision-making by frontline personnel and increasing the penalties for negative decision-making, the signs of emergencies can be better understood and constructed by multiagents, and the emergency command can respond more rapidly. 3) When the perceived benefit of the emergency command to fully respond to the emergency is increased, the penalty of being tracked for choosing the speculative strategy is increased, and the system will stabilize at the optimal equilibrium point (1, 1, 1). 4) Simulation studies of the game model reveal that by adjusting the initial probabilities of the agents' strategies, perceptual reference points, and the intensity of punishment, the agents' strategic choices gradually tend toward positive choices. The convergence speed improves substantially with the increase of these parameters. [Conclusions] The results of this study showed that the perception of benefits influences the initial probability of the agents' strategy choice, and the initial probability of the choice is the factor that affects the agents' positive strategy choice. While expanding the incentive policy to encourage the agents' positive sensemaking behavior, the supervision and reporting measures should be increased to neglect the public interest behavior, and the corresponding agent should be severely punished. In addition to the established reward and punishment measures, subjective factors significantly influence decision-making. In practice, excellent individuals should be commended, and public welfare campaigns should be organized to reduce the perceptual reference points of income, allowing the agents to understand the necessity of emergency sensemaking work and make positive decisions spontaneously.

[Objective] With the establishment of high-voltage transmission lines across forested areas, their inspection becomes crucial to reduce the fire risk of transmission lines and forest areas. At present, few studies have studied the path planning for unmanned aircraft inspection of transmission lines based on the fire risk in forest areas, but they do not address the security of the operation and maintenance of the power grid system or consider the interactions between different influencing factors. Therefore, an unmanned aerial vehicle path planning framework for forest power grid inspection is proposed based on the analytic network process method and genetic algorithm. Moreover, a path optimization method based on the maximum deflection angle constraint is developed. [Methods] After determining the assessment routes, the framework integrates field research and historical data to determine the objective data of these routes and identifies six classes of factors affecting the risk of forest fires: combustible factors, terrain factors, meteorological factors, human factors, surface wet conditions, and rescue conditions. These factors are subdivided into 18 typical factors by researching the historical accident cases and related literature. The forest fire risk indicator system is developed using typical factors, and to guarantee that this indicator system can effectively reflect the actual risk level, herein, the typical factors selected are those that are commonly used and recognized by previous researchers. Subsequently, weights for these typical factors are computed based on the analytic network process. Compared with the hierarchical analysis method, which is traditionally applied in earlier works, the network analysis method has the advantage of considering the interactions between the factors. The weights with objective data are combined to calculate the fire risk value for each grid. The high fire risk grids are employed as inspection nodes, and the shortest inspection path is acquired using path planning via the unmanned aerial vehicle inspection based on the genetic algorithm to reach the objective of obtaining real-time data in a short time, at low cost, and with high coverage. For the nodes in the path that do not meet the maximum deflection angle constraint, path optimization is conducted by adding new optimization nodes and the shortest path is ensured under the condition that the roadbed meets the maximum deflection angle constraint. [Results] Sections #3542—#3547 of the line of an important transmission channel in Anhui are taken as an application object. Ten high fire risk areas around the line are determined, and path planning is performed on them. The proposed framework yields an optimal path length of 5 391.72 m, and the path length optimized based on the maximum deflection angle is 5 401.36 m. Here, the path length is only increased by 0.179 % compared with the original one. This indicates that the path optimization method not only makes the original path satisfy the constraint of maximum deflection angle, but also increases the path length to be shorter, which has good optimization effect. [Conclusions] This work presents a path planning framework for the unmanned aerial vehicle inspection based on the results of fire risk assessment considering the interactions between various forest fire risk factors. In addition, the proposed path optimization method can make the path satisfy all constraints with a small increase in the path length. The proposed framework and optimization method offer reference and future ideas for realizing the unmanned aerial vehicle inspection of transmission lines in forest areas.

[Objective] An accurate measurement of the heat release rate (HRR) of a fire source is crucial for thoroughly understanding the fire evolution process. However, the commonly used oxygen consumption method requires expensive equipment, leading to high operational costs. According to the energy conservation principle, heat released per unit of time during material combustion is closely related to the increase of the room temperature. Machine learning methods have demonstrated considerable potential for exploring relationships between several independent and dependent variables, and attempts have been made to predict fire parameters based on temperature. Therefore, based on previous studies, this paper proposes a comprehensive machine learning framework to predict the HRR using temperature data as input. Furthermore, feature selection techniques are innovatively introduced to obtain key location temperatures to maximize HRR prediction accuracy. First, fire scenarios with different parameters are simulated in an ISO 9705 room using the fire dynamics simulator (FDS) software. Temperature data at different locations are obtained by gridding thermocouples, and a fire database is constructed. Subsequently, feature selection using recursive feature elimination (RFE) algorithms based on least absolute shrinkage and selection operator (Lasso) and random forest (RF) is performed to obtain two different low-dimensional subsets from high-dimensional simulated temperature features. Control groups with the same number of features are established. Finally, the performance of three typical models, namely, linear regression (LR), K-nearest neighbors (KNN), and light gradient boosting machine (LightGBM), for predicting the HRR are compared using different feature subsets. The results show that using the subset obtained by RFE based on RF, the LightGBM model demonstrates the lowest root mean square error (RMSE) and mean absolute error (MAE) values of 23.89 kW and 15.49 kW, respectively, indicating the least difference between its predicted and observed values. Furthermore, regarding the coefficient of determination, the LightGBM model reaches the highest value of 0.991 6, close to 1, thereby demonstrating its superior fitting capability. This is primarily due to the complex nonlinear relationship between the trained temperature dataset and HRR. Compared to the KNN and LR models, the histogram algorithm and the leaf-wise growth strategy with a depth limit enable the LightGBM model to give full play to its advantages. Tree-based models, such as LightGBM and XGBoost, can also be used for algorithm model-level improvements in the future. Additionally, deep learning models, such as multilayer fully connected and convolutional neural networks, can be utilized for fitting complex mappings. Compared with LightGBM models trained with manual feature subsets, RFE based on RF decreases the prediction errors (RMSE and MAE values decrease by 46.54 % and 50.66 %, respectively). The coefficient of determination also increases by 2.1 %, validating that this feature selection method can considerably improve prediction accuracy over manual feature selection. This paper proposes a comprehensive machine learning framework to obtain thermocouple temperature through the FDS fire simulation and then combines the feature selection and prediction models for HRR prediction, substantially improving prediction accuracy over manual feature selection. This comprehensive framework has theoretical and practical significance, thereby opening new pathes for HRR prediction. Subsequent research studies will construct additional comprehensive fire databases, increase the number of feature parameters, or explore algorithm combinations to improve HRR prediction.