Enhancing solar irradiance forecasting

Solar irradiance forecasting is crucial for maximizing the accuracy and dependability of solar energy systems. This chapter presents an innovative strategy integrating clustering techniques with multilayer perceptron (MLP) neural networks to predict solar irradiation. Clustering algorithms are utilized to detect patterns and divide the data into segments, which are subsequently input into an MLP neural network for precise prediction. By incorporating clustering into MLP, prediction accuracy is enhanced as it effectively captures the non-linear relationships present in the data. The chapter covers a comprehensive analysis of the technique, implementation details, and evaluation findings of the suggested strategy, showcasing its efficacy in enhancing solar irradiance predictions. Also covers different clustering methodologies, viz., K-means, K-means++, self-organizing maps (SOM), K-modes, agglomerative Clustering, mini-batch K-means, partial mixture model (PMM), BIRCH algorithm, hierarchical clustering, affinity propagation, spectral clustering, and several others. Efficiency of all these models is compared against standard metrices, namely coefficient of determination (R 2 score), mean squared error (MSE), and root mean square error (RMSE). The simulations are carried out using a publicly available dataset collected from Solcast website. Through systematic experimentation and analysis, it can be found out that the hybridized framework with CLIQUE outperforms all the other models.