Harmonics Forecasting of Renewable Energy System Using Hybrid Model Based on LSTM and ANFIS

Abstract

Harmonics forecasting stands as a crucial approach in the development of devices aimed at minimizing harmonics disturbances. The primary objective of this study is to create a hybrid forecasting model that can deliver precise and dependable forecasts for harmonics in Renewable Energy Systems (RES). To achieve this goal, the Adaptive Neuro Fuzzy Inference System (ANFIS) with the Long Short-Term Memory Network (LSTM) are combined in two distinct structured models. In the first model, LSTM is employed in the initial stage and ANFIS in the subsequent one, while the second model follows the reverse order. Additionally, for the generation of harmonics, two renewable generator models are utilized. The first model encompasses a grid-connected Double-Fed Induction Generator (DFIG) driven by a wind turbine and integrated with a Solar Photovoltaic (PV)-based power generator. The second generator model combines a Solar-PV generator with a wind turbine-linked Permanent Magnet Synchronized Generator (PMSG) connected to a shared grid. The harmonics produced by these generator models are used to construct training and testing datasets, which are subsequently employed for generating forecasts using the proposed hybrid forecasting models. The accuracy of forecasting results is verified through a comparison with benchmark studies in the literature. The findings reveal that the model employing ANFIS in the initial stage and LSTM in the second stage (referred to as the ANFIS-LSTM model) consistently yields the best forecasts among all the models tested in this study with RMSE of 0.0287, 0.0372, 0.0396 and 0.0311 for THD, h7, h11 and h13 respectively. Moreover, it exhibits a significant improvement over any of the techniques used in previous literature. Ultimately, this research establishes that both hybrid models proposed outperform the individual forecasting techniques used as benchmarks in terms of accuracy and precision.