Method for fire early warning of mountain photovoltaic scene
By combining image data acquisition and 3D reconstruction with multimodal data, the problem of a single fire prediction model in mountain photovoltaic scenarios has been solved, enabling more comprehensive fire prediction and risk analysis.
Patent Information
- Application Number
- CN202511190100.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-25
- Publication Date
- 2025-11-21
AI Technical Summary
Existing technologies for fire prediction in mountainous photovoltaic scenarios rely on single-factor models, lack multi-factor integration, suffer from scarce data, and produce unconvincing prediction results, failing to effectively incorporate environmental factors into the modeling.
By acquiring image data and reconstructing 3D data, and combining drone inspections, meteorological station data and geostationary satellite data, multimodal modeling is performed to calculate fire risk scores and render a 3D fire prediction probability distribution map.
It enables more comprehensive fire prediction, better analysis of fire risks, and provides more convincing prediction results.
Smart Images
Figure CN120998003A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fire early warning technology in mountain photovoltaic scenarios, specifically to a method for fire early warning in mountain photovoltaic scenarios. Background Technology
[0002] In recent years, the increasing frequency and severity of extreme weather events triggered by global warming have led to a significant increase in the occurrence of fires. Particularly under arid and hot climate conditions, the risk of fires in forests, grasslands, and vegetated areas has risen significantly. Fires not only cause enormous ecological damage and economic losses but also threaten human lives and living environments. Simultaneously, industrialization and urbanization have brought a series of fire safety hazards, making fire monitoring in densely populated areas such as factories and warehouses increasingly urgent. Furthermore, the smoke and greenhouse gases released by fires have a profound impact on air quality and climate change; therefore, the demand for fire detection has expanded to multiple fields, including fire prevention, environmental protection, and public safety.
[0003] In large-scale mountainous photovoltaic (PV) projects, on the one hand, certain characteristics of the PV panels themselves may lead to localized overheating, thereby increasing the frequency of fires. On the other hand, fire prediction is a significant concern in mountainous environments. Current fire prediction methods for mountainous areas suffer from the following problems: 1. The modeling conditions are relatively simple. Many fire prediction algorithms use historical fire data of the region as input and model the fire prediction task as a time series prediction task. However, on the one hand, fires occur at low probability times and data is scarce. On the other hand, there are many dependent variables for fire occurrences, and prediction results based solely on historical fire data lack persuasiveness.
[0004] 2. The occurrence of fires is usually closely related to environmental factors. According to relevant research, the three necessary conditions for forest fires are combustible material, ignition source, and fire environment, with the fire environment being a crucial condition for forest combustion. Forests may accumulate large amounts of combustible material, but sometimes, even with an ignition source present, a fire may not occur because a suitable fire environment is lacking. High temperatures and drought create a dry environment, which is an important meteorological condition for forest fire formation. However, current research rarely models these conditions in conjunction with other factors. Summary of the Invention
[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for fire early warning in mountain photovoltaic scenarios.
[0006] To achieve the above objectives, the present invention provides a method for fire early warning in mountain photovoltaic scenarios, comprising: Image data was collected from the mountain photovoltaic scene to obtain a set of scene images; Based on the scene image set, a 3D reconstruction of the mountain photovoltaic scene is performed to obtain the mesh model of the mountain photovoltaic scene; The vegetation coverage area is segmented on the mesh model of the mountain photovoltaic scene to obtain a three-dimensional vegetation distribution model of the mountain photovoltaic scene; The mesh model of the mountain photovoltaic scene is discretized in three dimensions, and voxel mesh is divided to reduce the spatial resolution. The space is locally divided using voxel mesh to obtain M local scenes, and the number of point clouds representing vegetation in each local scene is used to describe the three-dimensional vegetation cover of the area. Multimodal data input is constructed, including UAV inspection information input, meteorological station data input, and geostationary satellite data input. The UAV inspection information is obtained by the UAV conducting regular inspections of each local scene to monitor the task activities within each local scene. The meteorological station data includes temperature, relative humidity, wind speed, weather code, and precipitation probability. The geostationary satellite data is high-temperature point data of the mountain photovoltaic scene obtained through input location information. Then, based on the UAV inspection information, meteorological station data, and geostationary satellite data, fire source modeling, meteorological modeling, and high-temperature point modeling are completed respectively. Then, based on the current forecast time and the number of point clouds representing vegetation in each local scene, seasonal modeling and combustible material modeling are completed respectively to calculate the weight of each influencing factor. Calculate the fire risk score for each local scene. for: ; in, Let be the vegetation coverage ratio of the i-th local scene. As for the weight of the fire source, Weights are assigned to high-temperature points. The fire risk assessment is based on climate and seasonal factors, as detailed below: ; in, Seasonal weighting, Temperature weighting, Assuming wind speed as the weighting, Weighted by relative humidity, As the precipitation probability weight, Weighting of weather codes; The fire risk score for each local scenario is normalized and converted into a probabilistic description, as follows: ; in, Let be the probability of a fire in the i-th local scene. and These are the fire risk scores for the i-th and j-th local scenarios, respectively. The probability of a fire in each local scene is mapped to a heatmap color, and the heatmap color obtained by the mapping is used to render a 3D model to obtain a 3D fire prediction probability distribution map.
[0007] Furthermore, the seasonal weights The calculation method is as follows: ; Where m represents the month.
[0008] Furthermore, the temperature weighting The calculation method is as follows: ; Here, t represents temperature, and the unit is °C.
[0009] Furthermore, the wind speed weight The calculation method is as follows: ; Where V represents wind speed.
[0010] Furthermore, the relative humidity weight The calculation method is as follows: ; Where h represents relative humidity, in percentages (%).
[0011] Furthermore, the precipitation probability weight The calculation method is as follows: ; Where p represents the probability of precipitation obtained from the weather station, in units of %.
[0012] Furthermore, the weather code weight The calculation method is as follows: ; Where 'c' represents the weather code value.
[0013] Furthermore, the fire source weight The calculation method is as follows: ; Where p represents the number of people in the current local scene.
[0014] Furthermore, the weight of the high-temperature point The calculation method is as follows: ; Where r represents the reliability attribute obtained from the query.
[0015] Furthermore, the image data was acquired using UAV oblique photography technology.
[0016] Beneficial effects: This invention achieves spatial block modeling by leveraging the results of 3D reconstruction, which can directly render the results onto the 3D model, making analysis more efficient; by utilizing UAV inspection data, meteorological station data, geostationary satellite data, vegetation cover data, etc., the scene is modeled in all aspects, thereby enabling better fire prediction. Attached Figure Description
[0017] Figure 1 This is a schematic flowchart of a method for fire early warning in mountain photovoltaic scenarios according to an embodiment of the present invention; Figure 2 A schematic diagram of mesh rendering visualization of a mountain photovoltaic scene according to an embodiment of the present invention; Figure 3 This is a three-dimensional fire prediction probability distribution map obtained using the method of the present invention. Detailed Implementation
[0018] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. These embodiments are implemented based on the technical solutions of the present invention, and it should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.
[0019] like Figure 1 As shown, this embodiment of the invention provides a method for fire early warning in mountain photovoltaic scenarios, including: Image data was collected from the mountain photovoltaic scene to obtain a set of scene images. Specifically, oblique photogrammetry technology using drones was used to collect data on the mountain photovoltaic scene, resulting in the following scene image set: ; in Represents an image set. Let i represent the i-th image in the image set. This indicates the number of images in the image set.
[0020] A 3D reconstruction of a mountain photovoltaic scene is performed based on a set of scene images to obtain a mesh model of the scene. Specifically, the captured scene images are input into 3D reconstruction software such as DJI Terra to obtain the scene's mesh model, as shown below. Figure 2 As shown.
[0021] The vegetation coverage area is segmented on the mesh model of the mountain photovoltaic scene to obtain a three-dimensional vegetation distribution model of the mountain photovoltaic scene.
[0022] The mesh model of the mountain photovoltaic scenario is discretized in 3D space, and voxel meshes are used to reduce the spatial resolution. The space is then locally divided using voxel meshes to obtain M local scenes. The number of point clouds representing vegetation in each local scene is used to describe the 3D vegetation cover of the area. Each local scene is the smallest spatial unit for fire prediction, and subsequent fire predictions are performed based on this smallest spatial unit.
[0023] Constructing multimodal data input specifically includes: Drone inspection information input: Drones are used to inspect the mountain photovoltaic scene at regular intervals to monitor human activities in each area.
[0024] Weather station inputs include data such as temperature, relative humidity, wind speed, weather codes, and precipitation probability.
[0025] Geostationary satellite data input: Using the near-real-time surface high temperature anomaly query service system provided by the Institute of Remote Sensing and Digital Earth of the Chinese Academy of Sciences, input the location information of the monitoring area to obtain high temperature point data for that area.
[0026] Then, based on UAV inspection information, meteorological station data, and geostationary satellite data, fire source modeling, meteorological modeling, and high-temperature point modeling were completed respectively. Next, based on the current forecast time and the number of point clouds representing vegetation in each local scene, seasonal modeling and combustible material modeling were completed to calculate the weights of each influencing factor. Specifically, this includes seasonal weighting. Temperature weighting Wind speed weight relative humidity weight Precipitation probability weighting Weather code weight And fire source weight The calculation method is as follows: First, seasonal information is obtained based on the predicted time, and each season is assigned a seasonal weight. Seasonal weighting The calculation method is as follows: ; Where m represents the month.
[0027] Temperature weighting The calculation method is as follows: ; Here, t represents temperature, and the unit is °C.
[0028] Wind speed weight The calculation method is as follows: ; Where V represents wind speed.
[0029] relative humidity weight The calculation method is as follows: ; Where h represents relative humidity, in percentages (%).
[0030] Precipitation probability weighting The calculation method is as follows: ; Where p represents the probability of precipitation obtained from the weather station, in units of %.
[0031] Weather code weight The calculation method is as follows: ; Where 'c' represents the weather code value.
[0032] Fire source weight The calculation method is as follows: ; Where p represents the number of people in the current local scene.
[0033] This invention uses a satellite fire monitoring API to query information on high-temperature points within M local scene ranges, and calculates the weight of the high-temperature points using the reliability attribute of the query return value. The specific calculation method is as follows: ; Here, 'r' represents the reliability attribute obtained from the query, which represents the reliability value that the satellite fire monitoring API can predict for a high-temperature point.
[0034] Calculate the fire risk score for each local scene. for: ; in, The vegetation coverage ratio for the i-th local scene is obtained by calculating the ratio of the number of vegetation point clouds to the total number of point clouds in the local scene, as follows: ; in, The number of vegetation points in the i-th local scene This represents the total number of point clouds in i local scenes, and this ratio is used to describe the proportion of combustible materials in the scene.
[0035] The fire risk assessment is based on climate and seasonal factors, as detailed below: .
[0036] The fire risk score for each local scenario is normalized and converted into a probabilistic description, as follows: ; in, Let be the probability of a fire in the i-th local scene. and These are the fire risk scores for the i-th and j-th local scenarios, respectively.
[0037] The fire probability of each local scene is mapped to a heatmap color, and the resulting heatmap color is used to render a 3D model, thus obtaining a 3D fire prediction probability distribution map. When mapping the fire probability of each local scene to the heatmap color, low probability is mapped to near-blue, and high probability is mapped to near-red. The resulting 3D fire prediction probability distribution map is shown below. Figure 3 As shown.
[0038] The above description is merely a preferred embodiment of the present invention. It should be noted that for those skilled in the art, other parts not specifically described are existing technology or common knowledge. Several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for fire early warning in mountain photovoltaic scenarios, characterized in that, include: Image data was collected from the mountain photovoltaic scene to obtain a set of scene images; Based on the scene image set, a 3D reconstruction of the mountain photovoltaic scene is performed to obtain the mesh model of the mountain photovoltaic scene; The vegetation coverage area is segmented on the mesh model of the mountain photovoltaic scene to obtain a three-dimensional vegetation distribution model of the mountain photovoltaic scene; The mesh model of the mountain photovoltaic scene is discretized in three dimensions, and voxel mesh is divided to reduce the spatial resolution. The space is locally divided using voxel mesh to obtain M local scenes, and the number of point clouds representing vegetation in each local scene is used to describe the three-dimensional vegetation cover of the area. Multimodal data input is constructed, including UAV inspection information input, meteorological station data input, and geostationary satellite data input. The UAV inspection information is obtained by the UAV conducting regular inspections of each local scene to monitor the task activities within each local scene. The meteorological station data includes temperature, relative humidity, wind speed, weather code, and precipitation probability. The geostationary satellite data is high-temperature point data of the mountain photovoltaic scene obtained through input location information. Then, based on the UAV inspection information, meteorological station data, and geostationary satellite data, fire source modeling, meteorological modeling, and high-temperature point modeling are completed respectively. Then, based on the current forecast time and the number of point clouds representing vegetation in each local scene, seasonal modeling and combustible material modeling are completed respectively to calculate the weight of each influencing factor. Calculate the fire risk score for each local scene. for: ； in, Let be the vegetation coverage ratio of the i-th local scene. As for the weight of the fire source, Weights are assigned to high-temperature points. The fire risk assessment is based on climate and seasonal factors, as detailed below: ； in, Seasonal weighting; Temperature weighting, Assuming wind speed as the weighting, Weighted by relative humidity, As the precipitation probability weight, Weighting of weather codes; The fire risk score for each local scenario is normalized and converted into a probabilistic description, as follows: ； in, Let be the probability of a fire in the i-th local scene. and These are the fire risk scores for the i-th and j-th local scenarios, respectively. The probability of a fire in each local scene is mapped to a heatmap color, and the heatmap color obtained by the mapping is used to render a 3D model to obtain a 3D fire prediction probability distribution map.
2. The method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, The seasonal weight The calculation method is as follows: ； Where m represents the month.
3. The method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, The temperature weight The calculation method is as follows: ； Here, t represents temperature, and the unit is °C.
4. The method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, The wind speed weight The calculation method is as follows: ； Where V represents wind speed.
5. A method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, relative humidity weight The calculation method is as follows: ； Where h represents relative humidity, in percentages (%).
6. The method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, The precipitation probability weight The calculation method is as follows: ； Where p represents the probability of precipitation obtained from the weather station, in units of %.
7. A method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, The weight of the weather code The calculation method is as follows: ； Where 'c' represents the weather code value.
8. A method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, The weight of the fire source The calculation method is as follows: ； Where p represents the number of people in the current local scene.
9. A method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, The weight of the high temperature point The calculation method is as follows: ； Where r represents the reliability attribute obtained from the query.
10. A method for fire early warning in mountain photovoltaic scenarios according to claim 1, characterized in that, The image data was acquired using UAV oblique photography technology.