The Research On Forecasting Method Of Photovoltaic System’s Power Output

Solar energy as clean and renewable energy, it is greater than any other renewable energy in the world and has always been showed highly attention. Photovoltaic(PV) power generation is the main way of solar energy application, whose power output is random, undulate and difficult to control. To reduce the uncertainty impact after the power output of PV system put into the power grid, improve the system reliability, maintenance power quality, and increase the penetration level of photovoltaic power generation, it is significant to predict the power output of PV system and improve the prediction accuracy.Through analyzing the factors which affect the power output of PV system, combining with the characteristic of power output of PV system which is random, non-stationary and greatly affected by weather conditions, different prediction models were established according to the weather type.Firstly, predict the power output of PV system respectively based on BP neural network, grey neural network and support vector machine. The result show that these methods above can preliminary predict the power output of PV system, but it is not accurate. There are two factors leading to this result. On one hand, it is affected by the characteristic of power output of PV system. On the other hand, these methods have its own limitations and can not achieve better prediction result.Considering the power output of PV system is nonlinear and non-stationary, a prediction method based on empirical mode decomposition(EMD) and least square support machine(LSSVM) was proposed to predict the power output of PV system. Use the EMD to smooth the original power output of PV system, decompose it to several intrinsic mode function, and then select the optimal kernel function of LSSVM, set up the prediction model, and finally the prediction results are summation of the predicted values for each component. The simulation results show that the prediction accuracy was improved compared with other algorithms.Considering the local mean decomposition is superior to the EMD in terms of the number of iterations and the effect of endpoint, a method base on LMD and LSSVM was proposed to predict the power output of PV system. Firstly, find the similar days of predict day by using Euclidean distance. Secondly, use LMD to decompose the power output of PV system and get the product function(PF) component which has certain physical meaning. Thirdly, use LSSVM to establish the prediction model. And finally the predicted results are summed up by the predictive value of each component. The experimental results show that the method base on LMD and LSSVM is a good way to predict the power output of PV system under different weather conditions. Compared with the method base on EMD and LSSVM, the prediction accuracy is higher and the error fluctuation is smaller.