Detection of shading for solar power forecasting using machine learning techniques

This paper presents a machine learning based solar power forecast method that can take into account partial shading related fluctuations. Generated PV power is difficult to predict because there are various fluctuations. Such fluctuations can be weather related when a cloud passes over the array. But they can also occur due to shading caused by stationary obstacles. In this work an approach is presented that improves the forecast under such fluctuations caused by partial shading. This adaptation is necessary, because partial shading is usually not detected directly. Such shading occurs after the growth of trees or later built buildings. The presented algorithm can detect such effects itself and thus works self-learning. A correction of the prediction of a forecast model could successfully reduce forecast error due to partial shading. The model is evaluated on the basis of two months of recorded shading data in which shading was caused by a tree infront a PV array. The correction uses internal inverter data and irradiance values of the previous day to perform the correction and was able to reduce the RMSE of a 10 kWp under shading and thus improve the prediction accuracy by up to 40% depending on how strong the shading is. ... mehrThe model can detect how intense the shading is and correct the forecast by itself.