A hybrid machine learning-assisted optimization and rule-based energy monitoring of a green concept based on low-temperature heating and high-temperature cooling system

Abstract

This article aims to support the targeted worldwide green transition process by introducing and thoroughly analyzing a low-temperature heating and high-temperature cooling, smart building system. This concept allows for greater use of renewable energy while utilizing less input energy than conventional heating and cooling techniques. The proposed system consists of a reversible water-to-water heat pump driven by low-temperature geothermal energy. A rule-based control strategy is developed to establish an intelligent connection with the regional energy grids for peak shaving and compensating for the building's energy costs over the year. The dynamic simulation is carried out for a multi-family building complex in Stockholm, Sweden, using TRNSYS. The most favorable operating condition is determined via an artificial neural network-assisted tri-objective optimizer based on the grey wolf algorithm in MATLAB. The comparison of the proposed smart model with the conventional system in Sweden results in 332%, 203%, and 190% primary energy reduction, cost saving, and carbon dioxide emission mitigation, respectively. As indicated by the parametric results, the conflicting fluctuation between desirable and unfavorable indicators highlights the importance of multi-objective optimization. The grey wolf optimizer obtains 12% higher efficiency, 1.2 MWh lower annual bought energy, 24 $/MWh lower unit cost, and 5.1 MWh more yearly sold energy than the design condition. The scattered distribution reveals that tank volume and subcooling degree are sensitive parameters. According to the transient results, the suggested smart system can independently satisfy the building's heating, cooling, and electricity demands for more than 81% of the year, thanks to the two-way connection with the electricity and heating networks via the rule-based controller.