Study On Steam Load Predicting Based On Phase Space Reconstruction And Support Vector Machine

The heating steam is one of secondary energy the thermal power plant mainly produces. Currently, most thermal power plants take the way of cogeneration in which the amount of electric energy they produce is up to the thermal energy that users need. Therefore, how to predict the heat load in the future time especially the hourly load is becoming an important research subject, which not only can save the coals, reduce energy consumption and pollution, but also has instructive significance for the thermal power plant to operate securely, economically and efficiently. This subject is studying a steam load forecasting method, which we can use to predict the required heating steam load in the next 24 hours accurately and timely so as to provide scientific basis for the production scheduling of thermal power plant.Firstly, the significance and the current research situation of heat load predicting are introduced in this paper. The characteristics and inherent rules of heat load are described. The method based on chaos theory and support vector machine is proposed on the basis of analysis and comparison of the existing methods of load predicting.Secondly, the preconditioning of original load data from one thermal power plant is done, including the analysis and processing of the missing data and abnormal data. And then denoising of the sample data is carried out with the method of wavelet soft threshold denoising. The steam load time series proves to be with chaos characteristics through the analysis and study of preprocessed load data using the chaos analysis methods. According to Takens theorem, we reconstruct phase space of steam load time series, and establish the steam load predicting model respectively using support vector machine (SVM) and least squares support vector machine (LSSVM) in the phase space. According to the simulation results, the method based on phase space reconstruction and LSSVM is confirmed.Finally, in order to improve the comprehensive performance of the algorithm, an improved particle swarm optimization (PSO) algorithm is proposed to optimize the two parameters in LSSVM. The CV method, standard PSO algorithm and the improved PSO algorithm are simulated, and the results show that the method based on phase space reconstruction and LSSVM with parameter optimization proposed in this paper can achieve good optimization results and high prediction accuracy.