Combustible temperature and humidity prediction method and system based on ai and spatial features
By introducing various machine learning models and spatial features in urban-rural fringe scenarios, the problems of model adaptability and prediction accuracy of fire risk monitoring systems in urban-rural fringe areas have been solved, achieving high-precision, real-time fire risk monitoring and early warning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING FORESTRY UNIVERSITY
- Filing Date
- 2026-02-11
- Publication Date
- 2026-06-05
AI Technical Summary
Existing fire risk monitoring and early warning systems suffer from insufficient model adaptability and unstable prediction accuracy in urban-rural fringe scenarios, and lack systematic analysis of spatial characteristics, making it difficult to meet the needs of high-precision, real-time fire risk monitoring.
By combining multiple machine learning prediction models with the spatial geographic information characteristics of the urban-rural boundary region, and by constructing multi-source datasets, feature vectors, and comprehensive evaluation, key factors affecting prediction accuracy are identified, and the optimal model is adaptively selected.
It improves the accuracy and stability of temperature and humidity prediction for combustibles, enhances the ability to characterize local environmental differences and spatial heterogeneity, and provides technical support for fire hazard monitoring and refined early warning.
Smart Images
Figure CN122153300A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of forest fire risk monitoring and prediction technology, and in particular to a method and system for predicting the temperature and humidity of combustibles based on artificial intelligence and spatial characteristics. Background Technology
[0002] Against the backdrop of intensifying climate change and the continuous evolution of land-use patterns, wildfires have become one of the most frequent and destructive natural disasters globally. In recent years, wildfire events have shown an upward trend in terms of frequency, scope of impact, and intensity, posing a serious threat to ecosystem stability, socio-economic development, and the safety of people's lives and property. As urban spaces continue to expand into natural areas, human settlements and natural vegetation cover areas are intertwined, gradually forming a special transitional zone at the forest-town boundary. This area has dense distribution of combustible vegetation, complex terrain conditions, and highly concentrated human activities, resulting in diverse fire-causing factors. The probability and risk level of fires in this area are significantly higher than in single urban areas or natural vegetation areas, making it a key area for current fire prevention and risk management.
[0003] The occurrence and spread of fires are influenced by a variety of factors, including the characteristics of combustibles, meteorological conditions, and topography. Among these, the temperature and humidity of combustibles directly affect their ignition probability, burning rate, and fire behavior characteristics, and are important physical parameters characterizing fire susceptibility and fire hazard levels. Therefore, accurate acquisition and dynamic prediction of combustible temperature and humidity are crucial foundations for forest fire risk assessment and early warning systems.
[0004] In urban-rural fringe areas, due to the urban heat island effect, the high heterogeneity of vegetation types and structures, and the combined effects of anthropogenic heat sources and disturbances, the temperature and humidity of combustibles exhibit significant nonlinear variations and strong fluctuations across both temporal and spatial scales. Traditional prediction methods based on regional average meteorological conditions or empirical statistical relationships are insufficient to effectively characterize the dynamic evolution of temperature and humidity in combustibles under these complex environmental conditions, thus limiting the ability to conduct refined fire risk assessments and accurate early warnings.
[0005] Existing fire risk assessment and early warning systems mostly rely on meteorological factors or empirical indices for indirect characterization, such as the fine fuel humidity code and initial spread index in forest fire weather index systems. These methods typically suffer from limited temporal resolution, insufficient response to local microclimate differences, and prediction lag. Furthermore, their actual deployment and operation often require significant equipment investment and maintenance costs, making it difficult to meet the demands for high-precision, real-time fire risk monitoring in urban-rural fringe areas.
[0006] With the development of artificial intelligence technology, methods for predicting the moisture content or temperature and humidity of combustibles based on historical observation data are gradually being applied in the field of fire hazard monitoring. Existing related technologies typically rely on a single machine learning model, with relatively fixed model structures, feature input formats, and parameter configurations. In practical applications, these models often require manual adjustments for specific areas, making it difficult to simultaneously adapt to prediction needs under different spatial locations and environmental conditions. Especially in urban-rural fringe areas, automatic fire hazard monitoring stations are distributed across different spatial locations such as urban built-up areas, transition zones, and natural vegetation areas, resulting in significant spatial differences in meteorological conditions and human interference affecting combustibles. Using only a single prediction model cannot fully reflect the impact of spatial location differences on the prediction results. For example, the existing method for predicting the moisture content of combustibles based on long short-term memory networks (authorization announcement number CN115169729B) mainly focuses on time series feature modeling, and its model structure is relatively fixed, leaving room for further improvement in its predictive adaptability under different spatial conditions.
[0007] Furthermore, existing technologies generally lack a mechanism for unified evaluation and comparison of various artificial intelligence prediction models, and fail to incorporate the spatial characteristics of urban-rural boundary areas into the prediction modeling process, nor to comprehensively analyze and discriminate key factors affecting prediction accuracy. In particular, the role of geographical features such as spatial distance in the prediction of combustible material temperature and humidity has not been fully quantified and verified, limiting the optimization and widespread application of artificial intelligence prediction models in complex urban-rural boundary scenarios.
[0008] Therefore, it is necessary to propose a method and system for predicting the temperature and humidity of combustibles based on AI and spatial characteristics. By constructing multiple prediction models and conducting comprehensive evaluation, the method can achieve adaptive selection of prediction models. At the same time, spatial characteristic analysis is introduced to identify key factors affecting the prediction accuracy, so as to improve the accuracy and stability of the prediction of the temperature and humidity of combustibles in urban-rural border areas and provide reliable technical support for fire risk monitoring and refined early warning. Summary of the Invention
[0009] To address the shortcomings of existing fire risk monitoring and prediction technologies in urban-rural fringe areas, such as insufficient model adaptability, unstable prediction accuracy, and difficulty in identifying key influencing factors, this invention provides a method and system for predicting the temperature and humidity of combustibles based on AI and spatial features. By introducing multiple machine learning prediction models and combining them with the spatial geographic information features of urban-rural fringe areas, the temperature and humidity of combustibles are comprehensively modeled and predicted to improve the accuracy and reliability of fire risk monitoring and early warning in complex environments. The artificial intelligence described in this invention includes, but is not limited to, machine learning methods and deep learning methods.
[0010] To achieve the above objectives, the present invention adopts the following technical solution.
[0011] This invention provides a method for predicting the temperature and humidity of combustible materials based on AI and spatial characteristics, comprising the following steps: Acquire meteorological time-series data and combustible material temperature and humidity observation data to construct a multi-source time-series dataset; Remote sensing data such as land use type and built-up area ratio are acquired to generate a spatial distribution map of the urban-rural boundary area. The spatial distance between the automatic fire risk monitoring station and the urban-rural boundary area is calculated, and the monitoring stations are grouped or classified according to the spatial distance. The meteorological time series data, combustible material temperature and humidity observation data, and spatial feature data are standardized and time-aligned to construct a feature vector for model input. Based on the feature vectors, various machine learning prediction models are constructed as AI implementation methods to obtain corresponding temperature and humidity prediction results for combustible materials. The prediction results are comprehensively evaluated based on multiple evaluation indicators, and the comprehensive prediction performance index of each prediction model is calculated. Based on the comprehensive prediction performance index, the optimal prediction model corresponding to the target area is determined, and the prediction results of the temperature and humidity of combustibles are output. The influence of the spatial distance factor in the prediction model is evaluated using feature importance analysis and sensitivity analysis methods to analyze its impact on the prediction results.
[0012] Furthermore, the meteorological time series data includes at least one of temperature, relative humidity, wind speed, precipitation, air pressure, and radiation.
[0013] Furthermore, the machine learning prediction models used as AI implementation methods include random forest models, extreme gradient boosting models, and long short-term memory network models.
[0014] Furthermore, the evaluation index includes at least one of the following: root mean square error, mean absolute error, coefficient of determination, maximum error, minimum error, and average error, preferably two or more.
[0015] Furthermore, the spatial characteristics of the urban-rural boundary area include the spatial distance from the automatic fire risk monitoring station to the nearest boundary of the urban-rural boundary area.
[0016] Furthermore, the spatial distance factor is used as one of the input features to evaluate its contribution to the accuracy of combustible material temperature and humidity prediction.
[0017] This invention also provides a combustible material temperature and humidity prediction system based on AI and spatial characteristics, the system comprising: The data acquisition module is used to acquire meteorological time series data, combustible material temperature and humidity observation data, and remote sensing spatial data; The spatial feature construction module is used to generate a spatial distribution map of the urban-rural boundary area and calculate the spatial distance between the automatic fire risk monitoring station and the urban-rural boundary area. The feature processing module is used to standardize and align temporal and spatial features over time to construct the model input feature vector. The model prediction module is used to construct various machine learning prediction models based on the feature vectors and output prediction results; The model evaluation and selection module is used to comprehensively evaluate the prediction model based on multiple evaluation indicators and determine the optimal prediction model, providing model output results for subsequent spatial factor influence analysis. The results output module is used to output the predicted results of temperature and humidity of combustible materials, and to output data for the spatial distance factor influence analysis.
[0018] Furthermore, the model prediction module includes a random forest model, an extreme gradient boosting model, and a long short-term memory network model. These modules work together to implement the aforementioned prediction methods.
[0019] Compared with the prior art, the present invention has the following beneficial effects: 1) By introducing multiple machine learning prediction models and conducting unified evaluation and adaptive selection, the problem of insufficient adaptability of a single model in the complex environment of the urban-rural fringe area is avoided, and the stability and reliability of the temperature and humidity prediction results of combustibles are improved. 2) By combining geospatial information of the urban-rural boundary area with meteorological and observational data, spatial factors are introduced to participate in modeling, which enhances the ability of the prediction model to characterize local environmental differences and spatial heterogeneity. 3) A comprehensive analysis of the predictive model and the degree of influence of spatial factors helps to identify key factors affecting the accuracy of fire risk prediction, providing technical support for fire risk monitoring and refined early warning in urban-rural border areas. Attached Figure Description
[0020] Figure 1 This is a flowchart illustrating the method for predicting the temperature and humidity of combustibles based on AI and spatial characteristics according to the present invention.
[0021] Figure 2 This is a structural block diagram of the combustible material temperature and humidity prediction system based on AI and spatial characteristics according to the present invention.
[0022] Figure 3 This is a schematic diagram showing the spatial distribution of urban and rural border areas and the spatial distance relationship between automatic fire risk monitoring stations. Detailed Implementation
[0023] To enable those skilled in the art to better understand the technical solution and implementation process of the present invention, the present invention will be further described below in conjunction with specific embodiments. It should be noted that the following embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art can make various modifications or equivalent substitutions without departing from the technical concept of the present invention. All equivalent changes made based on the technical solution of the present invention should fall within the scope of protection of the present invention.
[0024] Example 1 This embodiment takes the fire risk monitoring scenario at the urban-rural boundary as the application background and provides a method for predicting the temperature and humidity of combustibles based on AI and spatial characteristics, which specifically includes the following steps.
[0025] Step 1: Acquisition and construction of multi-source data.
[0026] First, meteorological time-series data and corresponding temperature and humidity observation data of combustibles from automatic fire hazard monitoring stations within the study area are acquired. The meteorological time-series data may include, but is not limited to, at least one of temperature, relative humidity, wind speed, precipitation, air pressure, and radiation. The temperature and humidity observation data of combustibles are used to characterize the thermal and humidity state of combustibles within the area where the monitoring stations are located.
[0027] Simultaneously, remote sensing spatial data of the study area, including land use type data and built-up area ratio data, are acquired to characterize the spatial structure features of the urban-rural boundary region.
[0028] The aforementioned multi-source data were uniformly organized in terms of time and space to form a multi-source time-series dataset for subsequent modeling and analysis.
[0029] Step 2: Extraction of spatial features of the urban-rural boundary area.
[0030] Based on the acquired remote sensing spatial data, a spatial distribution map of the urban-rural boundary area is generated. For example... Figure 3 As shown, the urban-rural boundary area 2 is located at the boundary between the urban built-up area 1 and the wild vegetation area 3. The spatial distance 5 between the automatic fire risk monitoring station 4 and the urban-rural boundary area 2 is introduced into the prediction model as one of the spatial features. The spatial distance can be calculated using Euclidean distance, shortest path distance, or other spatial measures. This invention does not limit the specific calculation method.
[0031] The monitoring stations are grouped or classified according to the spatial distance to reflect the differences in the temperature and humidity changes of combustibles under different spatial conditions.
[0032] Step 3: Feature processing and feature vector construction.
[0033] Preprocessing is performed on meteorological time series data, combustible material temperature and humidity observation data, and extracted spatial feature data. The preprocessing process includes, but is not limited to, handling missing values, removing outliers, standardization, and time alignment.
[0034] After preprocessing, meteorological time features, combustible material temperature and humidity features, and spatial features are fused to construct a unified model input feature vector, which serves as the input data for multi-model prediction.
[0035] Step 4: Multi-model prediction modeling.
[0036] Based on the aforementioned feature vectors, various machine learning prediction models are constructed under the same feature input conditions to predict the temperature and humidity of combustible materials. These prediction models may include random forest models, extreme gradient boosting models, and long short-term memory network models, among others.
[0037] By inputting the feature vectors into each prediction model, the corresponding temperature and humidity prediction results for combustible materials are obtained.
[0038] Step 5: Comprehensive evaluation of models and selection of the optimal model.
[0039] The prediction results of each prediction model are comprehensively evaluated based on multiple evaluation indicators, which may include at least one of the following: root mean square error, mean absolute error, coefficient of determination, maximum error, minimum error, and mean error.
[0040] Based on the comprehensive prediction performance index corresponding to each prediction model, the prediction capabilities of different models are compared and analyzed, and the optimal prediction model corresponding to the target area is adaptively determined.
[0041] In the actual implementation process, historical data can be selected as the training set according to the actual application needs, and independent data can be used as the validation set to evaluate the performance of the prediction model.
[0042] Step Six: Spatial Factor Influence Analysis.
[0043] Based on the determination of the optimal prediction model, the role of spatial distance factor in the model input features is evaluated through feature importance analysis and sensitivity analysis, and its influence on the accuracy of combustible material temperature and humidity prediction is analyzed.
[0044] The above analysis results can provide a quantitative reference for optimizing the prediction model and monitoring fire risk in urban-rural border areas.
[0045] Example 2 This embodiment provides a system for implementing the above prediction method, the system comprising: The data acquisition module is used to acquire meteorological time series data, combustible material temperature and humidity observation data, and remote sensing spatial data; The spatial feature construction module is used to generate a spatial distribution map of the urban-rural boundary area and calculate the spatial distance between the automatic fire risk monitoring station and the urban-rural boundary area. The feature processing module is used to standardize and align temporal and spatial features over time to construct the model input feature vector. The model prediction module is used to construct various machine learning prediction models based on the feature vectors and output prediction results; The model evaluation and selection module is used to comprehensively evaluate the prediction model based on multiple evaluation indicators and determine the optimal prediction model. The results output module is used to output the predicted temperature and humidity results of combustible materials.
[0046] The modules work together through data interaction to achieve the function of predicting and analyzing the temperature and humidity of combustibles in the urban-rural boundary area.
Claims
1. A method for predicting the temperature and humidity of combustibles based on AI and spatial characteristics, characterized in that, Includes the following steps: Acquire meteorological time-series data and combustible material temperature and humidity observation data to construct a multi-source time-series dataset; Remote sensing data such as land use type and built-up area ratio are acquired to generate a spatial distribution map of the urban-rural boundary area. The spatial distance between the automatic fire risk monitoring station and the urban-rural boundary area is calculated, and the monitoring stations are grouped or classified according to the spatial distance. The meteorological time series data, combustible material temperature and humidity observation data, and spatial feature data are standardized and time-aligned to construct a feature vector for model input. Based on the feature vectors, various machine learning prediction models are constructed as AI implementation methods to obtain corresponding temperature and humidity prediction results for combustible materials. The prediction results are comprehensively evaluated based on multiple evaluation indicators, and the comprehensive prediction performance index of each prediction model is calculated. Based on the comprehensive prediction performance index, the optimal prediction model corresponding to the target area is determined, and the prediction results of the temperature and humidity of combustibles are output. The influence of the spatial distance factor in the prediction model is evaluated using feature importance analysis and sensitivity analysis methods to analyze its impact on the prediction results.
2. The method for predicting the temperature and humidity of combustibles based on AI and spatial characteristics according to claim 1, characterized in that, The meteorological time series data includes at least one of the following: temperature, relative humidity, wind speed, precipitation, air pressure, and radiation.
3. The method for predicting the temperature and humidity of combustibles based on AI and spatial characteristics according to claim 1, characterized in that, The machine learning prediction models used as AI implementation methods include random forest models, extreme gradient boosting models, and long short-term memory network models.
4. The method for predicting the temperature and humidity of combustibles based on AI and spatial characteristics according to claim 1, characterized in that, The evaluation indicators include at least one of the following: root mean square error, mean absolute error, coefficient of determination, maximum error, minimum error, and mean error, preferably two or more.
5. The method for predicting the temperature and humidity of combustibles based on AI and spatial characteristics according to claim 1, characterized in that, The spatial characteristics of the urban-rural boundary area include the spatial distance from the automatic fire monitoring station to the nearest boundary of the urban-rural boundary area.
6. The method for predicting the temperature and humidity of combustibles based on AI and spatial characteristics according to claim 1, characterized in that, The spatial distance factor is used as one of the input features to evaluate its contribution to the accuracy of combustible material temperature and humidity prediction.
7. A combustible material temperature and humidity prediction system based on AI and spatial characteristics, characterized in that, Includes the following modules: The data acquisition module is used to acquire meteorological time series data, combustible material temperature and humidity observation data, and remote sensing spatial data; The spatial feature construction module is used to generate a spatial distribution map of the urban-rural boundary area and calculate the spatial distance between the automatic fire risk monitoring station and the urban-rural boundary area. The feature processing module is used to standardize and time-align temporal and spatial features to construct the model input feature vector. The model prediction module is used to construct various machine learning prediction models based on the feature vectors and output prediction results; The model evaluation and selection module is used to comprehensively evaluate the prediction model based on multiple evaluation indicators and determine the optimal prediction model, providing model output results for subsequent spatial factor influence analysis. The results output module is used to output the predicted results of temperature and humidity of combustible materials, and to output data for the spatial distance factor influence analysis.
8. The combustible material temperature and humidity prediction system based on AI and spatial characteristics according to claim 7, characterized in that, The model prediction module includes a random forest model, an extreme gradient boosting model, and a long short-term memory network model.