The Temperature Prediction Technology And Methods Based On Chaos Theory And BP Network

Accurately temperature prediction is of great significance to agricultural production and national defense construction. In atmosphere filed, there are lots of factors influencing temperature, variances in temperature possess long term indetermination, and it is of typical chaotic characteristics, but variances in temperature possess short term determination. However, the prediction of variances in temperature in short term is a chaotic prediction problem.In this paper, i introduce Lyapunov exponent which belong to chaotic and BP neural network. Then use temperature records in Lianyungang to time series test sample, study Lyapunov exponent which belong to chaotic and BP neural network's application in atmosphere. The main work of this paper are as follow:Firstly, Though studying in G-P algorithm we get a proper embedded dimension and time delay. Do phase space reconstruct to the 91 samples in Temperature time series using phase space theory, reconstruct the state space of temperature system, and build prediction model. Next, i use wolf prediction algorithm work out the maximal Lyapunov exponent and other prediction value, and achieved prediction of short term temperature. In the prediction experiment, i program based on a Virtual Instrument. Using Labview 8.6 development platform, i design and realize a predict system based on Lyapunov exponent. Though prediction calculation and the prediction daily maximum, minimum temperature, Experimental results demonstrate that the predict system based on Lyapunov is effective for temperature prediction.Secondly, i study that the application of neural network theory in temperature prediction. Through it's self- learning abilities and fit nonlinear function, we build a temperature data model to predict the time series of temperature. I can get the proper network construction parameter by network training. then make network construction, simulation, prediction. i contract the temperate prediction in the next 10 days to the real air temperature, it prove that BP neural network has a high prediction precision. It also say that the time series prediction model based on BP neural network which i build has good prediction ability, it is worth popularizing in the future.Thirdly, i build an optimized BP network model which merge of Lyapunov exponent prediction and BP network. i can use this model to predict the daily maximum and minimum temperature in the next 10 days. Then, compared and analyzed the three experiment results in this paper.Research and experiment show that the application in temperature time series prediction based on Lyapunov exponent and BP neural network is effective.