Predictive economic optimization of energy storage microgrids

Energy management systems currently being used to operate energy storage systems in commercial applications mostly focus on providing resilient backup power but are generally unable to produce the predicted utility bill savings as modeled by energy optimization tools developed by DOE labs. This gap between actual and optimal economic performance is due to a lack of intelligence and prediction capabilities in the control algorithms. This results in energy storage underperforming financially which is an impediment to deploying storage at scale in the small-to-midsize commercial market. The objective of this proposal is the development of a low-cost, adaptive, intelligent energy management system for use in small-to-midsize commercial energy storage applications. The proposed controller will use mathematical optimization combined with artificial intelligence and machine learning techniques to produce near-optimal economic performance in grid-connected mode across multiple stacked revenue streams including utility bill savings, grid participation programs, and controllable site loads. We developed a prototype of our controller which included enhancing our solar production and load prediction capabilities, integrating these modules into our optimization engine, and building a software simulation environment. We identified a demonstration site and calculated the optimal savings. We next deployed a prototype of our controller at the demonstration site which began operations in January 2024. We then compared the actual operational results to the theoretical performance and found that, despite various integration challenges, we still achieved 48% of the theoretical results. We compared these results to those of a battery operated by an industry-leading controller and found that our controller was already achieving 24x the savings, thus proving the viability of the concept. We will further demonstrate the feasibility of the controller at multiple pilot sites and then productize the controller to enable deployment at scale. We will refine our prediction capabilities to further close the gap between modeled and actual performance. There are currently 5.9 million commercial buildings in the US. Each of these is a candidate for distributed energy resources such as solar and storage, but such a system will require an intelligent energy management system to unlock the projected economic and resilience benefits. Improving the financial performance of these systems will speed their deployment and improve the resilience of our nationĺs electrical grid and critical infrastructure. Further, integrating intelligent microgrid systems with flexible loads such as electric vehicle charging will offset the cost of rapidly expanding electrification.