Combination Forecast Method Based On EMD-ESN Research

The prediction methods have been widely used in data collection, industrial control, intelligent processing and so on. The mean prediction methods have been used are include Kalman filtering method, chaos theory method, artificial neural network, support vector machine(SVM) method and so on. The development of these methods have made the accuracy of prediction, forecasting efficiency improved to a certain extent. But as far as the current situation, the accuracy and efficiency of these methods need still need to be improved and also its forecast speed need to be improved.According to the research status of various forecasting methods and aiming to improve these problems, this paper put up a new intelligent forecasting method based on empirical mode decomposition(EMD) with the echo state network(ESN). Here are the main content of my work:1.Discusses the significance and the current develop situation of the forecast methods research. Introduce the commonly used combination models. Determine two kinds of test methods used in this article. They are empirical mode decomposition(EMD) and echo state network(ESN), then combine these two methods and get the EMD-ESN prediction model. The model use EMD to decompose the data and establish ESN network respectively for identification, then overlay each prediction result to get the final prediction value.2.Applies the new EMD-ESN model to short-term load forecasting for univariate forecast. Using EMD method to decompose a load sequence get the IMFs and RES as the input and output of echo state network. Then train the network and use the trained network to predict the component values next and get the final power load forecast.3.Expends the upper EMD-ESN prediction method and predict multivariate. Apply EMD-ESN model to hot roller temperature control system. Each value should be decomposed by EMD, during this should consider whether the decomposed value numbers of each variable are equal or not equal, in different case, the input of echo state network should be kept separate. 4.Combines the prediction result with fuzzy PID controller and use it in hot roller temperature prediction and control, establishes the fuzzy PID control model based on prediction and gets the hot roller temperature control system.The innovation point of this paper is to use two kinds of new methods EMD and ESN, base combination forecast model to forecast the single variable and multiple variables respectively, and use it in industrial control field. The concept of the industrial control system provide a new thought for industrial control.