An Optimization Method For Simulating Forest Fire Spreading Driven By The Real-Time Dynamic Data From UAV

Forest fire is a sudden and destructive global natural disaster,which can bring huge losses and hazards to the ecosystem and human society.Real-time and accurate simulation of forest fire spread is of great importance in guiding the rational suppression of fires.Static forest fire spread simulation methods require input parameters that are usually subject to measurement errors or large uncertainties,which can lead to large discrepancies between predicted and actual fire scenarios.With the development and progress of the UAV forest fire monitoring program,the dynamic integration of fire data into the forest fire spread simulation system and the realization of real-time status updates and parameter optimization of the system can effectively improve the accuracy of forest fire prediction.Guided by this idea,this paper develops a real-time monitoring platform for forest fires by UAV,establishes an extreme learning machine forest fire spread rate model,and builds a dynamic data-driven forest fire spread simulation and optimization system,which is dedicated to achieving accurate prediction of forest fire spread trends.The main research works include:Development of a UAV forest fire real-time monitoring platform.According to the demand for forest fire real-time monitoring and prediction,the hardware architecture of the forest fire monitoring platform is determined,the software optimization method is designed,and a fire infrared data processing scheme based on image processing technology is proposed to realize fast extraction and continuous discrete processing of fire contours,which provides software and hardware basis for data-driven modeling of forest fire simulation.Establishment of forest fire spread rate model.The indoor experimental fire rate data were used to train the extreme learning machine to establish the forest fire spread rate model,and the hidden layer weights and biases randomly generated by the extreme learning machine were optimized by the gray wolf algorithm to further reduce the model training error,and finally,the prediction accuracy of this fire rate model was verified using the test set data.Establishment of a dynamic data-driven forest fire spread simulation and optimization system.Based on the forest fire spread rate model,the Huygens principle,and the spreading rules,a forward model of forest fire spread simulation is established to achieve fireline location prediction.A cost function is constructed based on the differences between the real-time observed fire area and the predicted fire area,and the optimal parameter estimation of the wind vector is performed by an intelligent optimization algorithm to establish a dynamic data-driven forest fire spread simulation optimization system,and the parameter correction capability of the system under different model input errors is tested through simulation experiments to prove the feasibility of the method.We designed field ignition experiments to verify the fire rate prediction ability of the forest fire spread rate model based on field data training under different environmental variables,verified the accuracy of the forward simulation method of forest fire spread,verified the ability of the dynamic data-driven forest fire spread simulation optimization system to reduce model input errors and improve fireline prediction accuracy.This paper focuses on a UAV real-time dynamic data-driven forest fire spread simulation optimization system,which can monitor the occurrence of forest fires in real time,improve the accuracy of forest fire spread simulation,and provide a basis for scientific fire decision-making,which has important theoretical and practical significance for forest fire prevention work.