Research On Photovoltaic Power Forecast Based On Physical And Neural Network Models

Solar cells are the core components of photovoltaic power generation system.Since external environment factors(light intensity, air temperature, wind speed, etc.)will influence its’ maximum output power, photovoltaic generation power is volatile and intermittent; Therefore, the accurate prediction of the maximum output power of photovoltaic cells is conducive to the reasonable arrangement of the grid operation mode, and reduce the negative impact of PV access to the grid, so as to improve the security and stability of power grid. Therefore, the prediction of photovoltaic power generation has far-reaching social and economic significances. This paper studies the physical forecast model of photovoltaic cell, the BP neural network forecasting model of photovoltaic power generation model, the law of the influence of meteorological factors on the prediction of photovoltaic power generation network, GA- BP neural network forecasting model, and analyzed its prediction characteristics, respectively, as follows:Firstly, according to the given two temperature under the open circuit voltage Voc and short circuit current Isc, voltage Vm and current Im value of the maximum power point, we establish any light intensity and temperature solar cell output maximum power output prediction model. The working condition is between the temperature-10 ℃ to 40 ℃, and the light intensity is between 583.83 W/m2 to 1095.87 W/m2 of monocrystalline silicon, polycrystalline silicon and amorphous silicon solar battery power’s measurement and the prediction resullts’ comparative study. we find the following conclusions:when the light intensity are constant, the temperature changes,or when the light intensity change, the constant temperature is constant, the predicted model can well predict the battery power characteristics. The cell voltage current global fitting root mean square error is less than 1.4439 mA and maximum power prediction relative error less than 10.61%. In arbitrary irradiance and temperature conditions, the average root mean square error RMSE value order: polysilicon silicon > monocrystalline silicon> amorphous silicon and the average relative error of the maximum power values of the PRE order: polysilicon silicon > amorphous silicon >monocrystalline silicon.Secondly, based on BP neural network, we study the effect network of all kinds of transfer function and training function influence on forecast results. The conclusion is for monocrystalline silicon, polycrystalline silicon and amorphous silicon solar cells inphotovoltaic battery testing system of indoor measurement data. It is found that Pruelin transfer function in the process of all the cells in the prediction has a minimum number of iterations and the fastest training time. Based on the average training speed and training error analysis,we can conclude: monocrystalline silicon and polycrystalline silicon of maximum output power are easier than the amorphous silicon cells to predict.Thirdly, a BP neural network and wavelet neural network prediction model are established, and the correlation between the outdoor meteorological conditions and the maximum output power prediction of poly silicon and amorphous silicon solar cells is studied. The conclusion follows: air temperature and solar cell output power have the strongest association, followed by wind speed, and finally the relative humidity;relative humidity and amorphous silicon solar cell output power had the strongest association, followed by atmospheric temperature. Finally, it is the speed of the wind.It is also found that, no matter what the weather factors, the results of the wavelet network prediction of polycrystalline silicon and amorphous silicon cells are better than the BP network prediction results.Fourthly, the BP neural network prediction model is improved based on genetic algorithm, the results show that the prediction model is better, and it can meet the engineering application accuracy. Under the condition of irradiance, cell temperature,air temperature, wind speed and humidity, root mean square error and mean absolute percentage error of amorphous silicon are 0.1364 mW, 0.2734, respectively; in the same condition, root mean square error and mean absolute percentage error of polycrystalline silicon are 0.1316 mW,0.9714, respectively.