Optimizing supply and production management through energy storage strategies: A solar cold production approach using artificial neural networks

Abstract

The reliability of clean renewable energy hinges on robust energy systems, with storage serving a critical function. This paper investigates the influence of various storage types and configurations on thermal performance, with a focus on optimal sizing for economic and environmental cost reduction. To achieve this objective, we simulate a solar cooling facility with varied configurations of hot/cold storage installations. This study employs an ANN methodology with a multi-layer perceptron approach to forecast unit performance for each configuration based on data generated during the simulation process. In the pursuit of the most efficient and high-performance network, a comprehensive investigation is conducted on the number of neurons, activation functions, and training algorithms. Subsequently, the optimization process, conducted through a genetic algorithm, determines the Pareto fronts representing the best solution sets. The comparison shows that a system design with double hot and cold storage tanks shows superior techno-economic-environmental performance. Among possible optimum solution sets, a point with this specification is selected; flow rate ratio, minimum flow ratio, cooling capacity ratio, cold storage ratio, and hot storage ratio of 1.2, 0.4, 0.91, 3.4, and 3.8, respectively. This configuration anticipates a levelized cost of cooling at 341 USD/MWhr, representing a 13 % reduction compared to the benchmark.