An Optimizing Method Based On Combining Edge Computer And Long Short-Term Memory Networks Applied To Solar Power Forecasting

Along with the rapid development of solar energy sources on the global power system, the challenges related to the weather instability of this power type are also of increasing concern. Intending to solve these challenges, the study of solar power forecasting has become one of the most important problems in the power system field. Most of the published forecast algorithms aim to minimize prediction errors. However, in reality, convenience and data processing time are two necessary goals to commercialize a forecasting software. Recently, IoT is emerging as an effective solution to solve the above multi-objective problem. Especially, in the current trend of real-time forecasting, the application of Edge Computer in the IoT environment is bringing new opportunities for the goal of reducing data processing time while ensuring acceptable forecast errors. Moreover, Edge Computer also reduces the risk of network congestion due to its distributed processing characteristics. This paper introduces a new methodology using Edge Computer combined with the LSTM algorithm applied to the solar power forecasting problem. Real-time calculation results show that the accuracy can be improved by the good choice of the converted data type while the execution time is halved compared to without using Edge Computer.