Applying Artificial Intelligence in Forecasting the Output of Industrial Solar Power Plant in Vietnam

Ninh Quang,Binh Van,Quang Dinh,Linh Duy

doi:10.4108/eai.29-3-2021.169166

Abstract

This paper uses recurrent neural network (Long Short – Term Memory - LSTM network) to build a model to forecast short-term generation capacity of Phong Dien solar power plant, (48 MWp – 35 MWAC) located in Thua Thien Hue Province, Viet Nam, with input fac

Highlights

Vietnam is considered to have high solar potential with a lot of sunshine hours a year as mentioned in [1]
Due to the uncertainty of solar source, in operation, both electricity system operators and the owner of industrial PV power plants need to know how many electric powers will be generated in hour, day
The authors in [22] compared forecasting models based on other neural network algorithms to the Long - short term memory model (LSTM) model, the results show that LSTM always give more accurate and stable results

Summary

Introduction

Vietnam is considered to have high solar potential with a lot of sunshine hours a year as mentioned in [1]. Solar power has had a strong boom in over the world. In Vietnam, since 2019, with strong incentives from the government, total installed capacity of PV power plants has increased rapidly and reached 4500 MWp [2]. A forecasting method that could predict the output of the PV power plants based on influencing factors concerned as input data, will solve the problem. Many forecasting techniques for generating capacity for solar PV systems have been developed and published. The results of Math processing machine learning SVR method (Support Vector Regression) and artificial neural network NAR method (Nonlinear Autoregressive) have been compared with the classical model. Jang in [4] has developed a new forecasting technique based on satellite images and SVM (Support Vector Machine). The results are not good enough due to the sporadic and random nature of the output power

Methods

Results

Conclusion