Intelligent prediction method for power generation based on deep learning and cloud computing in big data networks

Abstract

This paper aims to elevate the precision and efficiency of prevailing photovoltaic prediction algorithms by integrating deep learning and cloud computing techniques. The emphasis lies in leveraging measured solar power generation data to simulate the model's predictive capabilities and determine optimal parameters. The study employs a hybrid approach, combining a multilayer perceptron-deep belief network (MLP-DBN) algorithm, and contrasts it with other methods like support vector machine (SVM), long short-term memory (LSTM), multilayer perception (MLP), and deep belief networks (DBN). Assessment of model performance encompasses root-mean-square deviation, mean absolute error and the decision coefficient metrics. Empirical results highlight the superiority of the MLP-DBN technique, showcasing reductions in root mean square error by 2.20%, 1.64%, 2.09%, and 4.83%, and mean absolute error by 0.67%, 0.11%, 1.12%, and 1.30%, respectively. The coefficient of determination (R2) exhibits notable increments of 2.96%, 2.05%, 2.77%, and 8.64%. These strides underscore substantial advancements in prediction accuracy and error mitigation. The findings underscore the efficacy of the proposed hybrid model in ameliorating existing photovoltaic forecast algorithms, effectively addressing their limitations, including inadequate accuracy and performance.