A Hybrid Model For Short-term Photovoltaic Power Forecasting Based On EEMD-BP Combined Method

With more and more serious environmental pollution and energy consumption,the promotion of clean energy has become an urgent demand of all countries,and the introduction of relevant policies to make it develop rapidly.Increasing installed capacity of photovoltaic power station.In recent years,photovoltaic power generation in the world has gone from the primary stage to the stage of rapid development.When the wind power and photovoltaic power supply are connected to the power grid,the wind power or the lighting factor will cause the impact on the power grid,which will affect the stability of the power quality.Therefore,accurate prediction of solar power can enhance the competitiveness of solar power.There are many kinds of prediction methods that scholars have studied,each method has its advantages and disadvantages.For example:the traditional gray analysis model,considering the correlation between the various factors,the accuracy of the results need to be investigated.The drawback of the artificial neural network prediction model is that the training process is difficult to control.The output power of the photovoltaic system is very complex,and its change has a certain regularity,but it is affected by many external factors.The law is not very clear,before the use of the relevant temperature,humidity,solar irradiance and other factors to analyze the fitting,but the uncertainty is too complicated.Now the best way is to directly analyze the most intuitive historical data,the impact of all factors are reflected in the data.A new signal analysis method,the set of empirical mode decomposition(EEMD),can be used to deal with the non-stationary signal such as the output of the PV system.The impact of the waveform into a number of stable waveform.Ensemble empirical mode(EEMD)has been widely used in the short-term wind power forecasting,and it tends to be mature.In the field of photovoltaic power prediction,because of the influence of various meteorological factors,the output power history data is nonlinear,and the output power history data should be stabilized by EEMD.It is also used in the field of photovoltaic research and optimization.In this method,the data sequence is decomposed into several different sequences in order to weaken the influence of the mutation data.In this paper,a combination of empirical mode decomposition(EEMD)and BP neural network is proposed to predict the short-term output power of PV system.Firstly,EEMD is used to decompose the PV output power sequence into a series of relatively stable IMF sequences,and then the sequence components are combined and superimposed by the run length decision method.Then,the prediction model of BP neural network is established based on the superimposed component and meteorological data as input.Finally,the predicted results are added,and the PV power prediction value is obtained.In view of the actual research data show that the combination forecasting model proposed with EMD-BP neural network prediction model than in cloudy weather types under the error ratios were 11.4%and 12.1%,indicating that EEMD-BP neural network model is better.It is very common to use the traditional single BP neural network to predict the output power of the photovoltaic system,and the prediction result is poor due to the disadvantage of the algorithm itself.In this paper,we apply the method to the prediction of the output power of the PV system,and combine the advantages of the two algorithms,and overcome the defects of the single model prediction.