Optimizing energy management of a dual-source renewable energy system with thermal storage for hot water production

Abstract

Healthcare facilities are one of the most energy-intensive buildings in the commercial sector due to the high energy consumption of space cooling and ventilation, as well as the high-water heating loads, continuous 24-hour operation for the majority of the facilities, and the high number of medical equipment. With South Africa relying almost exclusively on electricity for energy, a considerable increase in electricity prices has been noted in recent years which has placed a major strain on the country’s electricity supplier. Therefore, healthcare facilities tend to experience simultaneously high electricity demand, which exerts significant peak load pressure on the grid. This has, therefore, had a negative effect on both electricity suppliers and customers, resulting in financial and capacity challenges. Water heating is essential in high-capacity healthcare facilities for hygiene, medical operations, and space heating and is one of the high energy consuming processes in these buildings. Most water heating systems used are conventional, relying on fossil fuels or electricity as primary heat sources such as the electric storage tank water heating (ESTWH) systems. These systems are one of the highest contributors to the energy consumption during the high morning and evening peak periods, which may consume approximately 40-50% of the electricity bill, as well as the increase in the carbon dioxide (CO2) emissions. An effective energy conservation measure in the commercial sector is the implementation of demand-side management (DSM) strategies to reduce energy consumption and operating costs. One method explored involves using waste heat recovery (WHR) with thermal energy storage (TES) as a supplementary heat source to preheat water. As this practice is used for the purpose of load shifting under time-of-use (ToU) tariff price signals and optimal control, it guarantees to fulfil the above-mentioned purposes. Additionally, in high-capacity buildings, the high energy consumption may still be a challenge due to the continuous high demands of hot water supply. Incorporating energy-efficient and renewable energy systems such as solar thermal heating (STH) and heat pumps (HP) offers viable energy-saving solutions. These systems may be retrofitted as stand-alone supplementary thermal heat sources or be integrated as the solar-assisted heat pump (SAHP) system enabling energy efficient, stable, reliable and cost-effective heating system, working independently and simultaneously as heat sources. This study contributes to the field by developing an energy management and optimal control of the water heating processes, of WHR-TES tanks integrated with SAHP system, as a hybrid supplementary heat supply. The hybrid source-TES tanks are used to feed water to the 57 ESTWH systems in a high-capacity healthcare building. The system is developed through mathematical model using the analytical formulation approach, developing the system and formulating the optimal control problem for optimal control under the demand-side management (DSM) strategy. The baseline system and the proposed optimally controlled system was then simulated using MATLAB software to obtain the operation profiles of the system performance. Using the same formulation approach, the energy consumption and operational costs were determined for the economic analysis to evaluate energy and cost savings compared with the baseline system. In addition, the artificial neural networks (ANN) modelling was conducted for the performance validation of the proposed system, using some of the acquired data and the simulation results for variable selection, determining the input and output and training the model. The ANN model was developed and generated to train, validate and test the system under the summer and winter conditions using the Levenberg Marquardt (LM) backpropagation function. The ANN was modelled at the proportions of 70%, 15% and 15% of the training, validation and testing models, with each variable having 288 data points. Due to the tanks’ different parameters of the TWS tanks and the 57 ESTWH systems, with different sizes of 100L, 150L, 200L and 250L, the tanks were grouped by their parameters and different models were developed for summer and winter cases. The TWS tanks had 5 input variables and one output variable, which is the pre-heated water temperature inside the TWS tank and the ESTWH systems 4 input variables and one output variable, which is the hot water temperature inside the each ESTWH system. The data used for the ANN modelling was prepared, cleaned and processed using the MATLAB and Microsoft Excel worksheet. From the simulation results of the analytical models obtained, the optimally controlled system was able to shift the heating loads of the multifarious ESTWH systems to the off-peak periods of the ToU pricing structure. Additionally, the SAHP system, heating the preheated water in the TWS tanks and supplying the makeup hot water to the multifarious ESTWH systems has resulted in reducing, and even eliminating, the use of the electric resistive elements for most of the multifarious ESTWH systems. Consequently, the hot water temperatures inside the multifarious ESTWH systems were maintained within the range of 50 ℃ to 60 ℃, which is safe considering the health of the patients. The system was able to reduce the use of electrical power, during the operation, and shift the heating loads to the cheapest ToU pricing signals while maintaining the required heating loads, as well as the delivering the water at the safe temperature to the end users. For the economic analysis, the initial costs of the baseline case were lower as compared to the optimally controlled proposed case, which have however accumulated to be very high at the end of the project lifespan. This is due to the high cumulative costs incurred because of the continuous energy consumption throughout the day, even during the peak periods, where the energy costs are very high. Conversely, the optimally controlled proposed system accumulated lower costs at the end of the project lifespan due to the retrofitted renewable energy source water heating system, the optimal control and shifting the water heating loads to the ToU pricing signals. The analysis indicated the accumulated energy costs of 261.57 USD and 133.57 USD for the summer case and 625.34 USD and 145.09 USD for the winter case for the baseline system and optimally controlled proposed system, respectively, from the simulated results for a period of 24 hours. From the economic analysis calculations, in a typical day, the cumulative energy costs obtained for the baseline system and proposed optimally controlled system are 32.05 USD and 15.61 USD for the summer case and 83.96 USD and 18.55 USD for the winter case, respectively. For the energy and cost-saving analysis, the estimated potential annual energy savings were 15,001.93 kWh, equivalent to approximately 49.6% per annum. This amount of energy may equate to 15.93 metric tons of CO2 per year. At the beginning of the project, the costs of implementation of the baseline and the proposed systems are approximated to 40,464,82 USD and 93,335.37 USD, respectively. Over the estimated project duration of 20 years, based on the calculated results, the baseline system and the proposed optimally controlled system may achieve 943,559.91 USD and 341,860.80 USD of the cumulative energy costs, 74,817.81 USD and 94,755.08 USD of the cumulative replacement costs, 13,996.05 USD and 32,283.01 USD of the cumulative operation and maintenance costs and 8,092.96 USD and 18,667.07 USD of the salvage costs, respectively. Finally, at the end of the project span, comparing the proposed optimally controlled system with the baseline system, the total life-cycle costs may therefore be approximated to 1,064,745.62 USD and 543,567.18 USD, respectively. These estimates equate to the cost savings of 521,178.43 USD, which is about 48.95% of the costs of the baseline system saved. These results show significant economic and energy-saving potential, demonstrating a viable solution for reducing the energy consumption and operational costs in healthcare facilities. The ANN models indicated high prediction accuracy, with correlation coefficients (R-values) above 0.95 across training, validation, and test phases. TWS tank obtained higher R-values for the summer case of 0.99414, 0.99518, and 0.98063 and 0.99219 for training, validation and test, respectively. For the ESTWHS, the 100L and 150L size parameters obtaining higher R-values for the summer case and 200L and 250L size parameter models for the winter case. The 250L size parameter obtained the highest R-values of 0.99988, 0.99984 and 0.99988 for the summer case and 0.99986, 0.99984 and 99981 for the winter case, respectively, for training, validation, test results. For performance validation of the models, TWS tanks obtained the lowest MSE errors for the winter case, training for 16 epochs and indicating training phase MSE, cross-validation phase MSE and the testing phase MSE errors of 2.22e-05, 2.14e-05 and 1.97e-05, respectively, at 10th epoch. For the ESTWH systems, comparing the MSE errors between summer and winter cases for each season, 100L and 150L size parameters achieved lower MSE errors indicating more effective training for the winter case, whereas for the 200L and 250L size parameters it was for the summer case. When comparing between the size parameters, the 250L size parameters obtained the lowest errors for both cases, training for 21 epochs, obtaining 2.91e-08, 4.08e-08 and 3.08e-08 at 21st epoch for the summer case and training for 13 epochs, obtaining 3.26e-08, 4.54e-08 and 6.00e-08 at 13th epoch of training phase MSE, cross-validation phase MSE and the testing phase MSE results, respectively. For error distribution, TWS tanks indicate good concentration ranging between 70 – 80% around the zero for both summer and winter cases, indicating lowest maximum error range of ±0.015 for the winter case. For the ESTWH systems, the 200L and 250L size parameters show strong concentration of data points over 90% around the zero-error region, for both the summer and winter cases, with the lowest maximum error range obtained in the summer case for both size parameters, within ±0.05 and ±0.0007, respectively. For the 100L size parameter, good concentration ranges between 70 – 80% around the zero and the maximum error range obtained in the winter case within ±0.04. Whereas the 150L size parameter shows strong concentration over 90% of data plots around the zero-error region for the summer case and good concentration ranging between 70 – 80% around zero, where the lowest maximum error range is obtained in the winter case within ±0.04. These results confirm the robustness and adaptability of the developed ANN model across diverse operating conditions, making it a valuable tool for predicting the thermal performance of advanced hybrid water heating systems. Overall, this study contributes a practical, rapid, and reliable AI-based modelling approach for thermal heating systems, offering opportunities for improved energy efficiency, system responsiveness, and informed decision-making in building and industrial energy management contexts. For future research, the proposed water heating process of incorporating WHR-TES, SAHP system and energy management and optimal control may further be customized for various commercial buildings, from small clinics to large hospitals, hotels, etc., depending on their energy needs and available infrastructure. Additionally, other types of solar heating systems or heat pump setup may be explored in different climate conditions considering the geographical locations, for the system to work efficiently and assist in energy and costs savings. Furthermore, the integration of optimisation algorithms (e.g. GA, PSO) to the ANN model of the particular or similar system and also incorporating hybrid AI models (e.g. ANFIS or ANN-LSTM), may be explored.