Study On The Key Techniques Of Dynamic Data Driven Forest Fire Spreading System

Forest fire is one of the most destructive disasters. Especially, forest fire disasters showed an upward tendency because global warming since the 1980s. Forest fire simulation technique is one of the mean of fighting a forest fire and minimizing number of forest fire disasters. There are some questions such as forest fire model selection, forest fire model updating, verifying simulation accuracy and so on in traditional forest fire spreading system. Study purpose is to build an innovative forest fire spreading simulation system based on DDDAS in this paper. The system can increase precision of forest fire spreading simulation by model base, model selection, model updating and so on. It can provide a strategic decision for forest fire save and new technologies demonstration for other related field of research. The research encompasses several major topics:(1) This paper brings up the term of dynamic data driven forest fire spreading simulation system and supplies new research approach and thought for forest fire spreading simulation study. The system framework was brought. There are four ingredients in the system: dynamic data acquisition and data processing, model base, simulation and control technology, visualization and customer interface. There are many system characteristics: mutual co-ordination, symbiosis feedback, self-adapting, self learning function and continuously self-improving. The system includes five key technologies such as dynamic data acquisition and data processing, model base, model selection, model updating and verifying simulation accuracy.(2) Forest fire spreading model evaluation criterions were established and based on these criterions we selected four most powerful and most typical and representative forest fire models in 42 models currently. The four models were analyzed and normalized. We studied the procedure and method of building forest fire model base. Model base was build by object-oriented method based on category of model. Forest fire spreading models algorithm were implemented. We designed model dictionary base, model file base, and model knowledge base. Model base administration and maintenance functions were developed and implemented model file management, parameter management and knowledge management. The problems of interface between of model base and database were solved by means of model data control files. Model base ensured the availability of supporting for forest fire model suitability selection technology and forest fire model self-adapting updating technology.(3) Based on BP artificial neural network, we designed a frame construction of forest fire model selection of suitability. Forest fire model selection knowledge was produced through BP artificial neural network. The system implemented forest fire model selection automatically and intelligently. BP artificial neural network model of forest fire model selection was build by treating forest fire environment data as inputting variable and treating appropriate forest fire model as outputting variable. At the same time, we studied the methods of acquiring and calculating data of inputting and outputting. The system implemented machine of model selection automatically based on dynamic data driven technology. We selected 72 items experimental data from historical forest fire records in Beijing to experiment and confirm the validity of model selection. It turned out that the reliability of model selection is more than 80 percent.(4) Using the a hand-hold GPS and wireless transmitting unit to rapidly locate the fire and determine the rate of fire spread, which also offered an indispensable way of verifying the speed of forest fire appealing. Discussing the representation and Computing Method of the simulate calculation error based on the thorough investigation of the sources of simulate calculation error. Pointing out that the speed and directional error are the root of forest fire appealing simulates error. Thus increasing the parameter of error correction is an effect way of reducing systematicsimulationerrors.(5) Putting forward the Automatic Generation of simulate error corrected parameter based on neural networks technique. Discussing the simulate error self-adaptive revising mechanism on line and simulate error revising knowledge automatically acquiring mechanism on line, which realized the process of the simulate error self-adaptive revision on line. The process is tested by four historical fire items, three of them have error less then the predict value 0.2m/min which indicted that the precision of the result is reliable.(6) Designed the systemic hierarchy and network architecture of dynamic data-driving forest fire simulation with the fireproof infrastructure of Beijing. Building the database and the model-database and forming the prototype system, which is realizing the key algorithms of models-chosen and simulating error correction and the spatial spread of fires. Collecting the environmental data and fire spreading speed data, this solved the data-collection problem in process of traditional forest fire simulation. Discussing the process and methods of revision based on the study of"Wang Zhengfei"model, which indicted the feasibility of dynamic data-diving and self-adaptive simulate technique of forest fire.