Photovoltaic Generation Data Cleaning Method Based on Approximately Periodic Time Series

Data cleaning of photovoltaic (PV) power generation is an important step during data preprocessing for further utilization, such as PV power generation forecasting. The PV power generation data can be treated as a time series. An improved data cleaning method based on approximately periodic time series is proposed. First, the abnormal data in the PV data time series is classified with three types of the outliers. Then these three types of outliers are quantified based on the physical characters of PV power generation, and the effective corresponding cleaning implementations are described considering the rate capacity of PV station and period of PV data time series. Finally, the data cleaning method is tested on the PV generation data from a certain real power grid. The results show that this data cleaning method can effectively improve the PV data quality, and provide an effective support tool for the further application of PV data.