基于改进 BO-XGBoost 的兴安落叶松腐殖质阴燃升温速率预测

Predicting smoldering heating rates in Larix gmelinii duff using an improved BO-XGBoost model

摘要

【目的】兴安落叶松林下腐殖质富含有机质，易发生隐蔽且持久的阴燃火灾。升温速率作为直接反映产热与散热动态平衡状态的指标，是量化阴燃过程热量累积特性的关键参量。目前针对该参量的系统研究较为有限，且现有预测模型在表征阴燃升温过程时，难以同时兼顾预测精度与算法收敛效率。本研究旨在探究多因子耦合作用下阴燃升温速率的响应规律，构建一种适配阴燃升温速率预测场景的改进 BO-XGBoost 预测模型，以期为林下阴燃火行为预测提供技术支撑。【方法】基于兴安落叶松腐殖质室内阴燃试验构建 750 组样本数据集，引入 5% 强度高斯噪声开展数据增强，模拟野外林地环境扰动引发的升温速率数据波动偏差；采用单因素方差分析结合 Pearson 相关性检验，量化含水率、粒径、风速对升温速率的独立效应与交互作用；模型核心思路在于引入 Chebyshev 映射完成非线性变换，均匀生成初始寻优样本点，以扩大初期参数搜索范围；构建多保真度优化策略，结合低保真快速搜索与高保真精准计算，兼顾超参数搜索效率与预测精度；采用动态自适应权重更新机制，依据权重变化动态调节平滑系数，加快贝叶斯超参数收敛速度。将改进模型与基准 BO-XGBoost、GA-XGBoost、SSA-XGBoost 开展多维度性能对比，并设置消融实验评估各模块的提升贡献。【结果】（1）单因子分析结果表明，升温速率与含水率呈显著负相关关系（ P < 0.05），与粒径目数呈极显著负相关关系（ P < 0.01），说明含水率升高、粒径减小均会抑制腐殖质阴燃升温；风速对阴燃升温存在双重调控作用，在风速 0 ~ 6 m/s 区间内，升温速率与风速呈显著正相关关系（ P < 0.05），风速增大可促进阴燃升温；当风速超过 6 m/s 时，阴燃升温受到抑制。（2）3 类因子并非独立影响阴燃升温，含水率与粒径、含水率与风速交互作用均达到极显著水平（ P < 0.01），粒径与风速交互作用呈显著水平（ P < 0.05），共同调控兴安落叶松腐殖质阴燃升温进程。（3）模型性能对比结果表明，相比 3 种对照模型，改进的 BO-XGBoost 模型 R² 至少提升 2.14%，MAE 至少降低 24.54%， S² 至少提升 2.09%，RMSE 至少降低 12.08%，具有更高的拟合精度与稳定性。（4）消融实验表明 Chebyshev 映射、多保真策略、自适应权重可分别均衡模型样本分布，降低模型预测误差，加快模型迭代收敛。【结论】研究表明兴安落叶松腐殖质阴燃升温速率受含水率、粒径、风速及其交互作用共同调控，低含水率、大粒径、中等风速的工况条件可有效促进腐殖质热量积累与释放，加快阴燃升温进程。本研究构建的改进 BO-XGBoost 预测模型可模拟室内可控条件下腐殖质阴燃升温速率，为林下腐殖质阴燃火行为预测模型的算法优化与精度提升提供技术方案。

Abstract

[Objective] The humus beneath Larix gmelinii forests is rich in organic matter, making it prone to concealed and persistent smoldering fires. As an indicator directly reflecting the dynamic equilibrium between heat generation and dissipation, the heating rate is a key parameter for quantifying heat accumulation characteristics during smoldering. Current systematic research on this parameter remains limited. Furthermore, existing prediction models struggle to balance prediction accuracy with algorithmic convergence efficiency when characterizing the smoldering heating process. This study aims to investigate the response patterns of smoldering heating rates under multi-factor coupling effects. We constructed an improved BO-XGBoost prediction model tailored for predicting smoldering heating rates, providing technical support for forecasting smoldering fire behavior in forest understories. [Method] We constructed a dataset of 750 samples based on laboratory smoldering tests of Larix gmelinii humus. Data augmentation was performed by introducing 5% high-intensity Gaussian noise to simulate fluctuations in heating rate data caused by environmental disturbances in wild forest settings. We employed one-way analysis of variance combined with Pearson correlation tests to quantify the independent effects and interactions of moisture content, particle size, and wind speed on heating rates. The core strategy of the model involves introducing Chebyshev mapping for nonlinear transformation to uniformly generate initial optimization sample points, thereby expanding the initial parameter search space. We developed a multi-fidelity optimization strategy that combines low-fidelity rapid search with high-fidelity precise calculation, balancing hyperparameter search efficiency with prediction accuracy. Additionally, a dynamic adaptive weight update mechanism was adopted to dynamically adjust the smoothing coefficient based on weight changes, accelerating Bayesian hyperparameter convergence. We conducted multi-dimensional performance comparisons between the improved model and baseline models (BO-XGBoost, GA-XGBoost, and SSA-XGBoost). Ablation studies were also performed to evaluate the contribution of each module. [Result] (1) Single-factor analysis indicated that the heating rate was significantly negatively correlated with moisture content (P < 0.05) and extremely significantly negatively correlated with particle size mesh number (P < 0.01). This suggests that increased moisture content and reduced particle size inhibit smoldering heating in humus. Wind speed exhibited a dual effect on smoldering heating rates: it promoted heating at 0–6 m/s (P < 0.05), but inhibited the process once wind speeds surpassed 6 m/s. (3) Model performance comparisons showed that, compared to the three control models, the improved BO-XGBoost model increased R ² by at least 2.14%, reduced MAE by at least 24.54%, increased S ² by at least 2.09%, and reduced RMSE by at least 12.08%. This demonstrates higher fitting accuracy and stability. (4) Ablation studies indicated that Chebyshev mapping, the multi-fidelity strategy, and adaptive weights balanced sample distribution, reduced prediction errors, and accelerated model iteration convergence, respectively. [Conclusion] The study demonstrates that the smoldering heating rate of Larix gmelinii humus is jointly regulated by moisture content, particle size, wind speed, and their interactions. Conditions characterized by low moisture content, large particle size, and moderate wind speed effectively promote heat accumulation and release in humus, thereby accelerating the smoldering heating process. The improved BO-XGBoost prediction model constructed in this study can simulate the smoldering heating rate of humus under controlled laboratory conditions. It provides a technical solution for optimizing algorithms and improving accuracy in prediction models for smoldering fire behavior in forest understories.

关键词

兴安落叶松 / 腐殖质 / 阴燃 / 升温速率 / 多因子耦合 / 贝叶斯优化 / XGBoost / 火灾预测

Key words

Larix gmelinii / duff / smoldering combustion / heating rate / multi-factor coupling / Bayesian optimization / XGBoost / wildfire behavior prediction

参考文献

[1] 李维伟, 杨雪清, 张一鸣, 等. 基于小班尺度的北京市密云区森林火灾危险性评价[J].北京林业大学学报, 2024, 46(2): 75-86.

[2] Li W W, Yang X Q, Zhang Y M, et al. Hazard assessment of forest fire in Miyun District of Beijing based on the subcompartment scale[J].Journal of Beijing Forestry University, 2024, 46(2): 75-86.

[3] 王烁, 韩大校, 王千雪, 等. 不同林火烈度对大兴安岭兴安落叶松天然林林木空间分布格局的影响[J].北京林业大学学报, 2023, 45(2): 87-95.

[4] Wang S, Han D X, Wang Q X, et al. Effects of different forest fire intensities on the spatial distribution pattern of natural Larix gmelinii forests in the Great Xing’an Mountains of northeastern China[J].Journal of Beijing Forestry University, 2023, 45(2): 87-95.

[5] 刘洋鹏, 吴泽鹏, 周宇飞. 新形势下森林火灾综合防治管理初探[J].森林防火, 2024, 42(3): 38-41.

[6] Liu Y P, Wu Z P, Zhou Y F. A preliminary study of integrated forest fire prevention and control management in the new situation[J].Journal of Wildland Fire Science, 2024, 42(3): 38-41.

[7] 杨玖玲, 黄蔚, 王皓良, 等. 四川凉山州林区腐殖质阴燃传播特性的试验研究[J].安全与环境学报, 2024, 24(7): 2565-2573.

[8] Yang J L, Huang W, Wang H L, et al. Experimental study on smouldering propagation characteristics of humus in Liangshan Prefecture, Sichuan Province[J].Journal of Safety and Environment, 2024, 24(7): 2565-2573.

[9] 王越, 何诚, 张思玉, 等. 大兴安岭落叶松林腐殖质的热解特性及烟气成分[J].林业工程学报, 2020, 5(3): 155-160.

[10] Wang Y, He C, Zhang S Y, et al. Pyrolysis characteristics and smoke composition of humus of Larix gmelinii in Da Hinggan Mountains[J].China Forestry Science and Technology, 2020, 5(3): 155-160.

[11] 王立轩, 杨光, 高佳琪, 等. 兴安落叶松火烧迹地林褥层可溶性有机质的时间变化特征[J].生态学杂志, 2020, 39(12): 3913-3923.

[12] Wang L X, Yang G, Gao J Q, et al. Temporal variation of dissolved organic matter in post-fire forest floor of Larix gmelinii [J].Chinese Journal of Ecology, 2020, 39(12): 3913-3923.

[13] 于渤, 王明霞, 崔晨曦, 等. 我国北方森林地下火燃烧温度预测[J].北华大学学报 (自然科学版), 2023, 24(2): 244-251.

[14] Yu B, Wang M X, Cui C X, et al. Prediction of combustion temperature of forest sub-surface fires in northern China[J].Journal of Beihua University (Natural Science), 2023, 24(2): 244-251.

[15] 高博, 尹赛男, 韩喜越, 等. 腐殖质粒径和燃烧深度对兴安落叶松人工林地下火燃烧温度的影响[J].生态学报, 2022, 42(17): 7234-7243.

[16] Gao B, Yin S N, Han X Y, et al. Effects of humus particle and combustion depth on combustion temperature of underground fires in Larix gmelinii plantations[J].Acta Ecologica Sinica, 2022, 42(17): 7234-7243.

[17] 池泓颖, 辛颖, 朱梦龙, 等. 森林腐殖质热解动力学模型对比与分析[J].消防科学与技术, 2022, 41(12): 1741-1745.

[18] Chi H Y, Xin Y, Zhu M L, et al. Comparison and analysis of kinetic models of pyrolysis of forest humus[J].Fire Science and Technology, 2022, 41(12): 1741-1745.

[19] Sudhoff P, Krause U. A numerical model for predicting the smoldering behavior of bio-based insulation materials: model theory and validation[J].Fire Safety Journal, 2025, 152: 104351.

[20] 辛颖, 李禹洁, 邢美净. 基于红外图像的森林凋落物阴燃火灾探测[J].消防科学与技术, 2018, 37(1): 61-64.

[21] Xin Y, Li Y J, Xing M J. Forest litter smoldering fire detection based on infrared image[J].Fire Science and Technology, 2018, 37(1): 61-64.

[22] Pang Y, Li Y, Feng Z, et al. Forest fire occurrence prediction in China based on machine learning methods[J].Remote Sensing, 2022, 14(21): 5546.

[23] 朱龙祥, 王自法, 张昕, 等. 基于 GIS 和多种机器学习算法的广东省森林火灾预测模型[J].林业工程学报, 2024, 9(3): 159-167.

[24] Zhu L X, Wang Z F, Zhang X, et al. Forest fire prediction models for Guangdong Province based onGIS and multiple machine learning algorithms[J].Journal of Forestry Engineering, 2024, 9(3): 159-167.

[25] Song Q, Wang X, Zhang W, et al. Digital twin technology for sewage sludge smoldering process and CO/NO_x emissions based on back propagation neural network: A laboratory experimental study[J].Process Safety and Environmental Protection, 2024, 191: 1883-1895.

[26] 徐志霞, 张雅倩, 陶月, 等. 不同分解程度木麻黄凋落物的养分特征及微生物功能多样性分析[J].南京林业大学学报 (自然科学版), 2020, 44(2): 197-205.

[27] Xu Z X, Zhang Y Q, Tao Y, et al. Nutrient composition of litters and functional diversity of different microorganisms in various decomposition stages of Casuarina equisetifolia plantations[J].Journal of Nanjing Forestry University (Natural Sciences Edition), 2020, 44(2): 197-205.

[28] Geng Y, Ji Y, Wang D, et al. Strength prediction of recycled concrete using hybrid artificial intelligence models with Gaussian noise addition[J].Engineering Applications of Artificial Intelligence, 2025, 149: 110566.

[29] Lei Y, Shen Z, Tian F, et al. Fire risk level prediction of timber heritage buildings based on entropy and XGBoost[J].Journal of Cultural Heritage, 2023, 63: 11-22.

[30] Hou Y, Liu S. Predictive modeling and validation of carbon emissions from China’s coastal construction industry: a BO-XGBoost ensemble approach[J].Sustainability, 2024, 16(10): 4215.

[31] Zaheri M, Li C, Ranjbarnia M, et al. Predicting long-term displacements of deep tunnels using an artificial neural network optimized by sand cat swarm optimization with Chebyshev map[J].Environmental Earth Sciences, 2024, 83(8): 228.

[32] 尹赛男, 单延龙, 陈响, 等. 不同恢复程度的长白山风灾区地下火阴燃特征和发生概率模拟[J].林业科学, 2023, 59(9): 117-126.

[33] Yin S N, Shan Y L, Chen X, et al. Simulation of the smoldering characteristics and occurrence probability of sub-surface fires in the typhoon-caused disaster areas of Changbaishan Mountain with different recovery degrees[J].Scientia Silvae Sinicae, 2023, 59(9): 117-126.

[34] 王明霞, 王雅钧, 汪凤琴, 等. 基于模拟点烧不同加热时间和腐殖质粒径对森林地下火垂直燃烧的影响[J].北京林业大学学报, 2021, 43(3): 66-72.

[35] Wang M X, Wang Y J, Wang F Q, et al. Effects of different heating times and humus particle sizes on vertical combustion of forest underground fire based on simulated spot burning[J].Journal of Beijing Forestry University, 2021, 43(3): 66-72.

[36] 池泓颖. 兴安落叶松腐殖质热解动力学与阴燃蔓延研究 [D]. 哈尔滨: 东北林业大学,2024.

[37] Chi H Y. Study on pyrolysis kinetics and spreading of smoldering of Larix gmelinii humus[D]. Harbin: Northeast Forestry University,2024.

[38] Li W, Peng Y, Peng K. Diabetes prediction model based on GA-XGBoost and stacking ensemble algorithm[J].PLoS One, 2024, 19(9): e0311222.

[39] Xie H, Lin P, Kang J, et al. Prediction method of rock uniaxial compressive strength based on feature optimization and SSA-XGBoost[J].Sustainability, 2024, 16(19): 8460.

[40] Christensen E G, Hu Y, Purnomo D M J, et al. Influence of wind and slope on multidimensional smouldering peat fires[J].Proceedings of the Combustion Institute, 2021, 38(3): 5033-5041.

[41] 何诚, 舒立福, 刘柯珍. 大兴安岭地区夏季森林火灾环境因子特征分析[J].西南林业大学学报, 2021, 41(3): 87-93.

[42] He C, Shu L F, Liu K Z. Analysis on environmental factors' characteristics of summer forest fire in Daxing’an Mountains[J].Journal of Southwest Forestry University, 2021, 41(3): 87-93.

基金资助

中央高校基本科研业务费专项(2572026DG22)

国家重点研发计划(2023YFF1304001-02)