A short-term power forecasting model for photovoltaic plants based on data mining

Short - term power prediction is an important issue in the fine operation and maintenance of photovoltaic power plants. The estimation of power generation of the existing models is based on characteristics such as irradiance. The models are suitable for long-term power forecasting, and not suitable for short-term power forecasting. The irradiance can not be accurately predicted, it is generally obtained from previous years with large particle size such as annual irradiance and monthly irradiance. In this paper, a photovoltaic power generation prediction model - DaPoF model is established based on the data mining method, using the partial least squares regression method. The model uses features that are readily available in daily life, such as month, time, temperature, weather, temperature of the first three days, power generation in the first three days, and weather features within 7 days. The results show that the accuracy of the model reach 83.53% on 5 minute-prediction. So, the model shows advantages of high availability and high accuracy compared with the existing methods.