Predictive Modeling for Renewable Energy Optimization based on Non dominated Sorting Genetic Algorithm II

In modern times, predictive modelling has played a crucial role in managing and optimizing renewable energy integration in microgrids, particularly for sources like Photo Voltaic (PV) power and wind, to ensure economic efficiency and environmental sustainability. However, existing Seq2Seq based Recurrent Neural Networks (RNNs) combined with single objective weighted optimization faced several limitations. These limitations include, limited capability to capture long range temporal dependencies and inflexible optimization technique, where manual weight tuning is required and lacks the ability to balance cost and emissions. To overcome these challenges, an improved AI based predictive modelling and scheduling based on Multi scale Attention Temporal Network with Non dominated Sorting Genetic Algorithm II (MATNet-NSGA II) is proposed. Firstly, the historical power output data from wind turbines and PV panels is collected from microgrid components such as diesel generators and energy storage systems. Then, the collected data is preprocessed using K Nearest Neighbor (KNN) imputation to handle missing values, Functional Data Analysis (FDA) for smoothing the data and min max normalization to scale all the input values. Next, the preprocessed data is fed to the MATNet model, which performs prediction of renewable energy and NSGA-II performs optimization for scheduling and decision making. Experimental demonstrates that the proposed MATNet + NSGA-II achieves higher prediction accuracy with more cost effective and environment friendly dispatch schedules.