The Short-Term Load Forecasting Method Based On Eemd And Ann By Considering Grid-Connected Wind Power

With the wind power capacity increasing, its drawbacks have become more and more prominent. Wind power is an intermittent energy, and it has own characteristics, such as the stochastic volatility and intermittence. When wind power injected into the power grid, it will have a great disturbance on the voltage and frequency of the power system. And it will cause serious influence on power quality and the stable operation of the power system, and even make a threat to conventional power generation and lead to the breakdown of the power system. In addition, when the fluctuation of load and wind power superimposed on each other, it will greatly increase the load fluctuation of conventional units, which will have a great impact on the operation and safety of power system, and increase the difficulty of the work of electric power dispatching department. Therefore, it has important practical significance for the area containing the wind farm by establishing a short-term load forecasting model by considering grid-connected wind power to forecast the equivalent load.This paper mainly studies the short-term load forecasting and wind power forecasting based on multi-scale decomposition method. And it presents a short-term load forecasting model considering grid-connected wind based on ensemble empirical mode decomposition (EEMD) and neural network. This paper has mainly completed the following work:By comparing the decomposition results of short-term load data and wind power data by three kinds of multi-scale methods, the wavelet analysis, singular spectrum analysis and the ensemble empirical mode decomposition, it verifies that EEMD decomposition method can display the details of complex time series in the application of time series decomposition, the performance is better than the wavelet and singular spectrum. The high frequency components and periodic components obtained by EEMD are predicted using BP neural network, RBF neural network, wavelet neural network and Elman neural network model. And particle swarm optimization (PSO) is used to optimize the weight of linear combination model established by these four kinds of neural network. And the fitness function of PSO is the maximum degree of grey incidence between the result of the linear combination forecast and the true value. The subsequences with the trend characteristic are predicted by RBF neural network. While the prediction results of all subsequences are superimposed in the end.According to the new modeling idea of the new time series prediction above, the actual load data and wind output power are predicted. It can prove the effectiveness of the proposed model by comparing with other kinds of prediction method. Through the introduction of "equivalent load" at the end. the paper proposes a new prediction model of "equivalent load" by effective combination of the load forecasting model and wind power output forecasting model. The results show that the short-term load forecasting by considering grid-connected wind power based on EEMD and ANN has higher prediction accuracy by comparing with other prediction results. In addition, with the increasing of wind power capacity, the advantage of proposed in the paper will be more obvious.