Evolutionary learning driven load forecasting and demand response management model for smart grid

The smart meter data generated from the customers’ usage help utility providers to manage and control the electricity distribution among the customers for reliable services in the smart grid. Smart meter data provide an opportunity for utility providers to address these challenges by implementing demand response management (DRM), which enables customers to reduce consumption and balance load across the grid. For the demand response management in the smart grid, a multi-objective problem (MOP) is formulated considering three objectives, i.e., power scarcity, wastage and load factor. These objectives are computed by the newly proposed load forecasting neural network that precisely predicts upcoming loads on the demand side. The weights of this neural network are optimized using a hybrid learning approach that integrates evolutionary algorithms with Adam optimization. Additionally, the Multi-Objective Programming problem is efficiently addressed through the Non-dominated Sorting Genetic Algorithm (NSGA-III). Experimental results indicate that the proposed method surpasses state-of-the-art techniques in the terms of load forecasting. The proposed approach achieves a forecasting accuracy of up to 96.30% and a root mean squared error of 0.1367, based on the Open Energy Data Initiative (OEDI) dataset provided by the National Renewable Energy Laboratory of the United States. • A Machine Learning driven Demand Response and Load Forecasting (MLDR-LF) model is proposed for the demand and response management system with enhanced accuracy for the smart grid. • The weights of load forecasting neural network (LFNN) are optimized using the Hybrid adaptive Differential Evolution (HaDE) optimization algorithm, which combines the learning capabilities of evolutionary algorithm and Adam optimization. • A multiobjective problem (MOP) is formulated by considering the three objectives such as power scarcity, power wastage and load factor. These problems are solved by using the non-dominated sorting genetic algorithm III (NSGA-III) and solution of MOP provides the Pareto fronts. • Implementation and comprehensive experimental analysis demonstrate that the proposed work is effective and beneficial for the smart meter customer with a satisfying load factor. Numerical results of the MLDR-LF model over the benchmark dataset show the superiority to the state-of-the-art works for the smart grid.