Stay Informed
Follow us on social media accounts to stay up to date with REHVA actualities
Aristotelis Ntafalias | Panagiotis Papadopoulos | Alfonso P. Ramallo-González | Antonio F. Skarmeta-Gómez | Juan Sánchez-Valverde |
Motor Oil Hellas S.A, Greece | Motor Oil Hellas S.A, Greece | Faculty of Computer Science, Campus de Espinardo, Universidad de Murcia, Spain alfonsop.ramallo@um.es | Faculty of Computer Science, Campus de Espinardo, Universidad de Murcia, Spain | Faculty of Computer Science, Campus de Espinardo, Universidad de Murcia, Spain |
Maria C. Vlachou | Rafael Marín-Pérez | Alfredo Quesada-Sánchez | Fergal Purcell | Stephen Wright |
Kataskevastiki Makedonias, 54646 Thessaloniki, Greece | Department of Research and Innovation, Odin Solutions, Spain | Department of Research and Innovation, Odin Solutions, Spain | Arden Energy LTD, Dublin, Ireland | Arden Energy LTD, Dublin, Ireland |
Addressing energy cost reduction has become increasingly vital in the building sector due to the escalating demand for sustainable and efficient energy management [1], [2]. Conventional strategies for reducing energy costs often involve replacing energy-intensive appliances with more efficient alternatives, incorporating renewable energy systems, or enhancing the building's insulation and structure [3], [4]. Although these methods are effective and can lead to substantial energy savings, they usually require significant investment [5]. Moreover, concentrating on improving individual systems or components might not fully capitalise on the potential for energy optimisation, as it may reduce the benefits that arise from the interaction between various systems [6].
To maximise energy efficiency and achieve greater cost savings, more sophisticated approaches can be adopted such as shifting energy consumption from periods of high electricity prices to times when prices are lower. These strategies typically use advanced technologies and data analytics, including the prediction of energy use and electricity prices, to enhance overall energy efficiency [7], [8].
In this paper we show how PHOENIX solution integrates existing legacy equipment—such as heating, ventilation, and air conditioning (HVAC), or water boilers, and combined heat and power (CHP) units—with an advanced Internet of Things (IoT) platform. The proposed solution employs cutting-edge machine learning (ML) algorithms to precisely predict consumption trends and forecast day-ahead electricity prices, allowing users to benefit from the platform's flexibility services, thereby lowering their energy costs. The study presents specific case studies from Ireland and Greece to evaluate the practicality and effectiveness of this solution in different regional contexts. Additionally, it discusses the potential for scalability and adaptation of this approach to other geographic areas, considering the unique challenges and energy market conditions of each location.
This paper presents the PHOENIX platform, a cutting-edge ICT solution that uses the integration of legacy equipment to an IoT environment to integrate and optimise energy consumption in both residential and commercial buildings. The platform incorporates a suite of algorithms within the Pycaret tool, which are specifically developed for time series forecasting. These algorithms identify patterns and select the most suitable algorithm or combination of algorithms, such as ARIMA, ANN, or curve fitting, to make accurate predictions. The key contributions to optimizing energy use in buildings with existing legacy equipment includes the development of a platform that effectively integrates legacy systems, employs machine learning for data analysis and trend recognition, and offers innovative services to building occupants aimed at reducing energy usage and associated costs. The PHOENIX platform is introduced in Section II, with its implementation detailed in Section III, the results discussed in Section IV, and the conclusions outlined in Section V.
PHOENIX is an advanced ICT solution designed to make buildings smart, connected environments by integrating older systems, smart devices, and building management systems (BMS) with external sources like energy market prices and weather forecasts. This integration aims to improve energy efficiency, reduce costs, enhance occupant well-being, and support grid stability.
The use cases illustrate how extensively buildings are integrated within the PHOENIX platform, providing valuable insights into energy performance, environmental factors, equipment control, and relevant external data. The platform uses semantic data modelling and knowledge graphs for detailed data analysis and supports multiple communication protocols to efficiently manage data from various devices and sensors.
Figure 1. PHOENIX platform’s representation.
The PHOENIX architecture (Figure 1) emphasizes the integration of older devices within the platform. Pilot implementations have successfully demonstrated seamless access to various assets. This standardisation allows for uniform management of equipment from different manufacturers, enabling the development of smart services to control devices for both energy efficiency and comfort.
The Irish use case involves both domestic and commercial sites integrated within the PHOENIX platform. The domestic sites, located in southeast Dublin, were connected using a portal solution and an MQTT interface. The commercial site is the Rediscovery Centre, which features a BMS and various energy sources like solar thermal, solar PV, CHP, and a heat pump. The authors of [9] explored how CHP systems can optimize their usage based on day-ahead prices. Integration required connecting legacy equipment to the PHOENIX platform. Domestic site gateways with APIs for remote access were selected, and the BMS was upgraded for connectivity.
In the domestic site, the PHOENIX flexibility engine was tested with a domestic hot water boiler and an EV charger, utilizing dynamic pricing for demand shifting. By forecasting hot water demand and using day-ahead market pricing, the platform optimized activation times for the water heater. This resulted in an average load shift of 5.5 kWh/day (39% of total load), saving €0.19/day (equivalent to 12%).
In the commercial site, the Rediscovery Centre's CHP unit operation was optimized using the flexibility engine. Forecasting heat demand and market prices allowed the engine to determine optimal CHP operating times. Peak demand reduction was significant, with a 61% decrease in peak load. The load shifted was 5.5 kWh/day, representing 4% of the total daily consumption of 126 kWh, saving €0.30/day. Smart bills generated from these trials showed participants the cost benefits of optimized energy use and dynamic tariffs. Figure 2 shows the total active power input during the test day in the commercial building in kW.
Figure 2. Power consumption on commercial site during trial day.
The Greek pilot site tested PHOENIX grid flexibility services in both residential and commercial buildings, involving direct and indirect device control triggered by day-ahead dynamic energy prices. These prices do not yet exist for consumers in Greece [10], so realistic simulated prices were used. The residential building includes 8 apartments, each about 80 m² of floor area. Direct control involved a 5.1 kWh battery paired with a 6 kW hybrid inverter and a 4.95 kWp PV installation. PHOENIX simulated energy prices for the next day and identified the most expensive hour. Using energy production and consumption forecasts, the battery was charged either from the grid during low-priced periods or from solar energy if surplus was predicted. The battery then discharged during high-priced periods to reduce costs. Two tests showed energy savings of 3.66 kWh (86%) and 3.53 kWh (60%) during high-price periods, reducing costs by €2.48 (5.1%) and €1.64 (5.7%), respectively. Figure 3 depicts the typical hourly consumption patterns for the entire building over a period of five days.
The second service indirectly controlled high-consumption devices like heat pumps, kitchens, and washing machines by notifying residents to adjust usage during expensive periods. During the tests, consumption was shifted from high-prices hours to cheaper ones. Across the building, this service shifted 3.26 kWh of consumption, saving about €0.60 per day, or €0.09 per apartment per day, with savings ranging from 1.7% to 10.1% per apartment, or an average of 5.8 %.
The commercial site, a newly constructed office building of 800 m², tested direct control using a 10 kW hybrid inverter linked to a 10.92 kWp PV installation. The battery, charged from solar excess, was controlled the flexibility service and used during the highest-priced periods. This approach optimises energy use and reduces costs shifting consumption to more economical periods. On the first day, 6.91 kWh supplied during 17:00-18:00 saved €1.26 (27% of the day’s cost). On the second day, 6.91 kWh during 18:00-19:00 saved €1.00 (20%), as shown in Figure 3. On the third day, 6.84 kWh during 16:00-19:00 saved €1.08 (16%). The battery provided about 6.9 kWh of savings per day, significantly reducing costs by about €33 per month, or 16-27% of the total bill.
Figure 3. Building’s energy sources and battery’s state of charge during grid flexibility event.
The PHOENIX solution presents a promising approach with the potential to significantly reduce both energy consumption and costs. The findings from the pilot studies discussed in this paper illustrate the service's effectiveness across various contexts.
In the Irish pilot, the service was applied to a water heater and a CHP unit. Results indicated that, in the residential building, the service reduced energy consumption by up to 12% for the water heater, with an average load shift of approximately 39%. In the commercial building, the service's implementation on the CHP unit led to a 4% reduction in total daily energy consumption and a significant 61% decrease in peak load. These outcomes demonstrate the service's ability to optimize energy use for specific devices, resulting in notable cost savings for users.
In the Greek pilot, the service managed a battery system in both a residential and a commercial building. The residential application achieved up to an 86% reduction in energy consumption during peak hours and an overall reduction of up to 60%. Participants also experienced substantial cost savings, with daily savings averaging 10% of their monthly electricity bill. In the commercial building, the service provided an average daily energy savings of 6.9 kilowatt-hours (kWh), accounting for 4.6% of total consumption and leading to a 22% reduction in the monthly electricity bill. These results underscore the service's effectiveness in managing energy consumption during periods of high demand, yielding significant cost reductions for users
The approach used in this study emphasized the development of a highly organized platform, structured into multiple interconnected layers. A critical aspect of this approach was the adoption of a common information model, which has greatly simplified the integration and flow of data throughout the system. This methodology has been notably effective, allowing for the creation of innovative solutions like the demand flexibility service, which has been successfully implemented to achieve significant reductions in energy costs across various scenarios. Additionally, the incorporation of data analytics and advanced algorithms has been vital in processing the substantial amounts of data gathered by the platform.
The technology components created for the energy demand modification scenarios outlined in this paper are built on a modern, scalable design. The proposed architecture has proven its effectiveness by ensuring efficient resource management and quickly adapting to changing user demands. To ensure the developed components are both reliable and high-quality, comprehensive availability and performance tests were conducted. These tests were made possible through the use of the Jenkins framework, which streamlined the automation and management of the testing procedures.
Energy efficiency and sustainability are becoming increasingly crucial in the building industry. While replacing outdated, inefficient appliances can be expensive, advancements in AI and ML present new opportunities for optimising energy use. This paper introduces an IoT platform that leverages ML algorithms to forecast energy demands and predict day-ahead electricity prices, offering tenants flexibility services to lower energy costs while ensuring comfort. Pilot studies conducted in Ireland and Greece demonstrated notable reductions in energy consumption, particularly during peak times. The platform efficiently manages energy demand, optimizes the use of appliances, and addresses periods of high demand, leading to substantial cost savings. Although further large-scale research is necessary, these initial studies indicate that the PHOENIX solution holds significant promise for reducing energy expenses in smart buildings.
This research was funded by the European Commission’s Horizon 2020 Framework Programme for Research and Innovation in the project PHOENIX under grant agreement No 893079.
[1] Hafez, F. S. et al. (2023). Energy Efficiency in Sustainable Buildings: A Systematic Review with Taxonomy, Challenges, Motivations, Methodological Aspects, Recommendations, and Pathways for Future Research. Energy Strategy Reviews, vol. 45, p. 101013, Jan. 2023. https://doi.org/10.1016/J.ESR.2022.101013.
[2] Harsini, A. E. (2023). Resilience-oriented district energy system integrated with renewable energy and multi-level seasonal energy storage. J Energy Storage, vol. 72, p. 108645, Nov. 2023. https://doi.org/10.1016/J.EST.2023.108645.
[3] Anisah, I. et al. (2017). Identification of Existing Office Buildings Potential to Become Green Buildings in Energy Efficiency Aspect. Procedia Eng, vol. 170, pp. 320–324, 2017. https://doi.org/10.1016/J.PROENG.2017.03.040.
[4] Danish, M. S. S. et al. (2019). A managed framework for energy-efficient building. Journal of Building Engineering, vol. 21, pp. 120–128, Jan. 2019. https://doi.org/10.1016/J.JOBE.2018.10.013.
[5] Agrawal, R. et al. (2023). Adoption of green finance and green innovation for achieving circularity: An exploratory review and future directions,” Geoscience Frontiers, p. 101669, Jul. 2023. https://doi.org/10.1016/J.GSF.2023.101669.
[6] Badami, M. & Fambri, G. (2019). Optimising energy flows and synergies between energy networks. Energy, vol. 173, pp. 400–412, Apr. 2019. https://doi.org/10.1016/J.ENERGY.2019.02.007.
[7] Kim, I., Kim, B., & Sidorov, D. (2022). Machine Learning for Energy Systems Optimization. Energies 2022, vol. 15, Page 4116, vol. 15, no. 11, p. 4116, Jun. 2022. https://doi.org/10.3390/EN15114116.
[8] Wang, B. et al. (2023). Machine learning optimization model for reducing the electricity loads in residential energy forecasting. Sustainable Computing: Informatics and Systems, vol. 38, p. 100876, Apr. 2023. https://doi.org/10.1016/J.SUSCOM.2023.100876.
[9] Gao, S. et al. (2022). Potential benefits from participating in day-ahead and regulation markets for CHPs. Appl. Energy 306. 117974. Jan 2022. https://doi.org/10.1016/J.APENERGY.2021.117974.
[10] Makrygiorgou, J.J. et al. (2023). The electricity market in Greece: current status, identified challenges, and arranged reforms, vol. 15, Page 3767, Sustainability 15 (4) (2023) 3767. Jan 2023. https://doi.org/10.3390/SU15043767.
Follow us on social media accounts to stay up to date with REHVA actualities
0