Short-term power load forecasting based on improved T-S fuzzy-neural network

摘要

At present, in terms of the influences of temperature and holidays on the power load, an improved T-S fuzzy-neural network is proposed to forecast the short-term power load. Considering the learning rate and smoothing factor should be adjusted dynamically for the higher performance of T-S fuzzy-neural network, the two optimal parameters must be found automatically when the parameters are changed. Thus a T-S fuzzy-neural network based on the Particle Swarm Optimization(PSO) to adjust parameters dynamically is proposed. Firstly, the stable intervals of the parameters must be found, so that parameters can be adjusted dynamically in the intervals. Secondly, the PSO is introduced to find the optimal parameters, therefore the system can converge to a stable state rapidly, and achieve an optimal control precision. Finally, combining with data of power load, temperature and holidays provided by the UNITE(the European Network of Excellence on Intelligent Technologies for Smart Adaptive Systems)web, and the MATLAB simulation results analysis, the improved T-S network has been proved a higher control precision in power load forecasting.