Application of Machine Learning Algorithm to Forecast Load and Development of a Battery Control Algorithm to Optimize PV System Performance in Phoenix, Arizona

Abstract

The paper presents the results of the research work funded by Salt River Project Agricultural Improvement and Power District (SRP) on maximizing the economic benefits to customers installing residential rooftop PV systems in SRP territory. The optimized discharge of the battery power which would help in the reduction of Demand Charge paid by the customer was the primary goal. Machine Learning algorithms were utilized as a better load forecasting technique to the ones already in place. The improved battery discharge algorithm would also reduce the battery charge-discharge cycles (cycling aging) thus, improving the battery life. The tests were performed in the state of Arizona, on a residential rooftop grid-tied PV with storage system installed at the Tempe campus of the Arizona State University.