Optimal energy management of the ice thermal storage-based air conditioning system for commercial buildings with a solar photovoltaic system

Abstract

Heating, ventilation, and air conditioning (HVAC) systems constitute a substantial portion of the energy demand of the commercial sector; representing 40% of total building energy consumption and 60% of base building energy consumption. Despite this significant consumption, the implementation of energy-efficient strategies is lacking, which is crucial to mitigate associated costs. More studies need to explore effective energy management schemes for HVAC systems, particularly focusing on cooling loads in commercial buildings. Ice thermal energy storage (ITES) systems are often adopted, as one of several options for improving HVAC systems. HVAC-ITES presents opportunities for demand-side management (DSM) and allows cost savings by considering time-of-use (ToU) tariff regions. However, the use of ITES is not without challenges. Problems such as early depletion of stored ice and overcharging still occur in most facilities, which lead to the need for excessive energy to build ice, thereby reducing efficiency gains. Additionally, hybrid energy sources, such as photovoltaic (PV) arrays and solar collectors, which could contribute to further reductions in energy costs, are rare. In this study, HVAC-ITES, coupled with optimal control, presents opportunities for DSM and allows cost savings by considering ToU tariff regions. This study aimed to address the challenges identified in existing ITES systems, particularly in South Africa. This research emphasises the prediction of HVAC-ITES performance with respect to important variables and parameters using machine learning, optimal control with OPTI-tool applications, and economic analysis to evaluate operational efficiency and to minimise costs related to cooling purposes. The study proposes a comprehensive approach of integrating ITES and hybrid energy systems, using predictive and optimisation tools, and offering advantages presented by ToU tariffs and peak demand management. The study investigated previous research and existing HVAC-ITES plants and utilised the Performance, Operational, Equipment, and Technology (POET) framework to systematically review HVAC-ITES cooling plants and to highlight potential cost savings through optimal energy usage profile shifts. Secondly, a predictive model was developed using an artificial neural network (ANN) to forecast various configurations for performance and output indicators, such as the storage temperature, coefficient of performance, cooling thermal load, and chiller power consumption. The study focused on developing an ANN model using a shallow neural net fitting application (nftool) to predict the necessary ITES systems for non-linear input variables. A comprehensive dataset was generated and utilised to build and validate the ANN model, comprising input variables such as dry-bulb temperatures, components’ inlet and outlet temperatures, flow rates, and the state of the control valves. The performance metrics associated with forecasting include mean squared errors (MSEs) and regression. These metrics were considered for training, validation, and testing in the prediction and analysis of individual configurations. Thirdly, the cost-saving potential of energy for cooling loads in commercial buildings was analysed using an optimisation control strategy to efficiently manage the ITES operation of a building’s HVAC system. ITES is a thermal energy storage system that reduces the cost of energy in HVAC systems by shifting the cooling load from peak and standard periods to the off-peak period, with the aid of stored ice produced when the energy cost is affordable and discharging when the cooling demand is high, particularly at costly tariff rates. Simulations were implemented in the MATLAB optimisation toolbox with the Solving Constrained Integer Programs (SCIP) solver to address mixed-integer programming problems for non-linear optimisation problems. Furthermore, the results from various outcomes were considered to present the monetary benefits of the study. From the general review, it was clearly seen that the effective utilisation of ITES allows for further advantages in managing off-peak tariffs and consumer demands according to ToU and maximum demand. As a result, the conventional system should be optimally controlled to achieve higher accuracy. Moreover, the investigations revealed that better performance may be achieved with optimal control and optimisation tools, such as fuzzy logic, neural networks, closed-loop control, model predictive control, and SCIP. The choice of adopting the non-linear optimisation application SCIP was based on its capabilities. The findings from the POET analyses of energy efficiency activities for HVAC-ITES cooling systems emphasise the potential cost savings through optimal energy usage profile shifts. In addition, the introduction of renewable energy-sourced systems may further enhance cost reductions, with the POET framework projecting a potential 50% reduction in energy costs. Furthermore, from the ANN prediction results, the regression values were observed to range from 0.9 and above, while the MSE depended on the data range. The study demonstrates the application of machine learning to an ITES cooling system, and the resulting model serves as a benchmark to assess the effectiveness of ANN methodologies in prediction and provides insight into the impacts of parameter variation for ITES cooling systems. The effective utilisation of ITES allows for further advantages in managing off-peak tariffs and consumer demands according to ToU and maximum demand. The optimisation and economic analyses provide evidence on maintaining thermal cost and controlling the state of charge of an HVAC-ITES system. The optimal switching control model achieved approximately 33% energy cost savings during the summer months compared to the current system control (baseline). The ITES system maintained temperatures below 0 ˚C to meet thermal requirements, and the PV and battery options used to power the pumps accounted for 10% to 15% of the system’s overall energy usage. Optimal control of the system to utilise stored energy during peak pricing and high-demand periods significantly reduced dependence on grid energy. The study projects a return on investment within two years for the additional capital required for the controller, solar PV, and battery installations. Over a 20-year period, the estimated energy cost savings are 35%; assuming an inflation rate and annual electricity price increase of 5% and 10% respectively.