Forecasting of steam load based on phase space reconstruction and improved LSSVM algorithm

Heated steam is one of the most important secondary energy source, so it is significant for thermal power plants to predict the required steam load in the future hours and render sufficient heat and steam. Since the representation of steam load in time series model is proven to have chaotic characteristics, reconstructing phase space of steam load time series according to Takens theorem is feasible. Calculating delay time with C-C method and embedding dimension with Cao method are preliminary. The steam load forecasting model is established based on Least Squares Support Vector Machine (LSSVM) in the phase space. LSSVM parameters are selected with Simulated Annealing (SA) and Improved Particle Swarm Optimization (WPSO) algorithm. The SA_WPSO_LSSVM method has global convergence and its Mean Relative Error is 1.43%. The simulation results show it can achieve excellent prediction efficiency.