Convolution Neural Network and Long Short-Term Memory Network to Mitigate High Power Ramp Rates in Utility Grid Integrated Wind–Solar System
Abstract
On integration of the wind and solar based renewable energy systems to supply large loads of industries, it leads to high power ramp rates due to power grid stability issues raised by wind gust in deployed area of the system. Traditionally many control strategies has been designed for power converters using machine learning. In this paper, a new hybrid deep learning approach integrating Convolution Neural Network and Long Short-Term Memory Network is applied to power converter of Utility Grid Integrated Wind–Solar System as it is highly efficient in mitigating high power ramp rates. CNN Model extracts spatial features such as wind speed, solar irradiance etc. An extracted feature is employed to Long Short-Term Memory to identify complex relationships and long-term dependencies as it is highly efficient in processing nonlinear relationships. Finally, dependency map in the processed further to forecast the power generation the wind and solar system for efficient management of the load in the industries through other conventional energy backups as conventional generators to compensate the power fluctuations. Especially forecasting of the wind and solar based integrated energy system is performed to provide smooth overall power profile to industrial loads. Simulation results demonstrate that the proposed CNN–LSTM controller achieves an RMSE of 0.038, MAE of 0.026, and forecasting accuracy of 98.2%, thereby improving grid stability and mitigating high power ramp-rate fluctuations.