An optimized deep learning model for estimating load variation type in power quality disturbances

Power quality is one of the most important fields of energy study in the modern period (PQ). It is important to detect harmonics in the energy as well as any sharp voltage changes. When there are significant or rapid changes in the electrical load, i.e. load variations, it can lead to several issues affecting power quality, including voltage fluctuations, harmonic distortion, frequency variations, and transient disturbances. Estimating load variation is a difficult task. The main aim of this work is to design and develop an Improved Lion Optimization algorithm to tune the CNN classifier. It involves the estimation of the type of load variation. Initially, the time series features are taken from the input data in such a way to find the type of load with enhanced accuracy. To estimate load variation, a Convolutional Neural Network (CNN) is used, and its weights are optimally modified using the Improved Lion Algorithm, a proposed optimization algorithm (ILA). The proposed method was simulated in MATLAB and the result of the ILA-CNN method is generated based on error analysis based on the indices, such as MSRE, RMSE, MAPE, RMSRE, MARE, MAE, RMSPE, and MSE. The proposed work examines load variations ranging from 40×10⁶$\Omega$to 130×10⁶$\Omega$while considering different learning rates of 50 %, 60 %, and 70 %. The result demonstrates that at learning percentage 50, the MAE of the proposed ILA-CNN method is 7.06 %, 62.98 %, 41.13 % and 54.63 % better than the CNN, DF+CNN, PSO+CNN and LA+CNN methods.