Operation rule extraction based on deep learning model with attention mechanism for wind-solar-hydro hybrid system under multiple uncertainties

Abstract

With the increasing environmental problems caused by the use of traditional fossil energy, renewable clean energy has gradually attracted attention. Owing to uncertain components such as solar radiation intensity, wind speed, and power load, it brings difficulties to short-term operation of wind-solar-hydro (WSH) hybrid system. Therefore, the focus of this study is to solve the problem of probabilistic optimal operation model of WSH hybrid system under multiple uncertainties. The two keys to solving the probabilistic optimal operation model are estimating the probability density function (PDF) of model state variables and extracting the model operation rules. In this study, probability prediction, kernel density estimation, and deep learning model with attention mechanism are used to quantify uncertainty, estimate probability density functions, and extract operation rules, respectively. The experimental results show that the estimated probability density function is very practical and can provide abundant decision-making information to the dispatcher, and the extracted rules are very effective and can guide the dispatcher to perform operation. At the same time, the results also show that the rules extracted by the deep learning model using the attention mechanism increase the accuracy by 15.33% on average compared with those without attention mechanism.