Power load forecasting of LSSVM based on real time price and chaotic characteristic of loads

With the rapid development of the electricity market and the smart grid, the variation of the price of electricity will influence user's power consumption pattern, which will challenge the existing load forecasting methods that are applied to a fixed-price environment. In regard to the influence of real-time price on load, this paper introduces the real-time price as an influencing factor when building the load forecasting model. Meanwhile, it verifies the chaotic characteristics of the load and electricity price time series by extracting the fractal dimension and Lyapunov exponent, and then combining the phase space reconstruction theory of chaotic time series and the least square support vector machine (LSSVM) to establish a load forecasting model. Also, the adaptive particle swarm optimizer (PSO) is used to optimize the model parameters, and a real-time power load forecasting method is obtained. Finally, the real data of PJM power system are taken as an example for simulation studies, and the prediction results are compared with those obtained by the radial basis function neural network (RBF-NN). The results show that the proposed method has high accuracy and stability.