Cloud-based model-predictive-control of a battery storage system at a commercial site. (1st December 2022)

• Record Type: Journal Article
• Authors: Goldsworthy, M.; Moore, T.; Peristy, M.; Grimeland, M.
• Is Part Of: Applied energy. Volume 327(2022)
• Publication Date: 2022-12-01
• Subjects: Battery energy storage -- Demand response -- Commercial buildings -- Model predictive control -- Micro-grid; BMS Building Management System -- BOM Bureau of Meterology -- DCH Data Clearing House -- IoT Internet of Things -- MAPE Mean Absolute Percentage Error -- MASE Mean Absolute Scaled Error -- MPC Model Predictive Control -- RRP Regional Reference Price -- SOC State of Charge -- SOH State of Health; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2022.120038
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: A data-driving model predictive control algorithm is deployed at a commercial site. System architecture & algorithms are described in detail. New forecasting models are developed and forecast accuracy is evaluated. Electricity bill savings of 5.5 % are obtained mostly from reducing capacity charge. Lessons learned from the deployment are discussed. Abstract: Reducing costs for battery energy storage systems and the increasing availability of onsite generation sources are driving development of complex battery control algorithms principally aimed at minimising electricity costs. These algorithms combine forecasts of site consumption, generation and electricity costs and a model of the battery system in a solver that minimises a cost objective over some forecast horizon. Often they are trialled in simulation environments without the complexities introduced by real-world deployments such data quality and reliability issues, communication issues and forecast inaccuracies to name a few. This study reports on a trial demonstration of a cloud-based data-driven robust model predictive battery control algorithm controlling an existing 150kWh lithium-ion battery at an operational site housing 100 + office staff. Forecasting model and control performance are evaluated in situ. Despite two of the four battery inverters being non-functional, the algorithm delivered electricity bill cost savings of 5.5 %, of which two-thirds was the result of reducing the sites capacity charge from 358 kVa to 317 kVa. Throughout the trial multiple operational issues were encountered mostly related to data outages, equipment and communications reliability and the lessons learned from managing these occurrences are also discussed.