Benchmarking stochastic and deterministic MPC: A case study in stationary battery systems

Abstract

We present a computational framework that integrates forecasting, uncertainty quantification, and model predictive control (MPC) to benchmark the performance of deterministic and stochastic MPC. By means of a battery management case study, we illustrate how off-the-shelf deterministic MPC implementations can suffer significant losses in performance and constraint violations due to their inability to handle disturbances that cannot be adequately represented by mean (most likely) forecasts. We also show that adding constraint back-off terms can help ameliorate these issues but this approach is ad-hoc and does not provide performance guarantees. Stochastic MPC provides a more systematic framework to handle these issues by directly capturing uncertainty descriptions of a wide range of disturbances.

Stochastic model predictive control for central HVAC plants

“…We extend the benchmarking procedure given in [14] for battery management systems. To distinguish the policies obtained from the three MPC schemes, we denote the policies obtained from the deterministic MPC as u…”

“…Model predictive control (MPC) is becoming a established automation technology in HVAC central plants [3,[9][10][11][12][13]. MPC can anticipate and counteract disturbances and accommodate complex models, constraints, and cost functions [3,14,15]. However, existing MPC implementations for HVAC central plants use deterministic representations of the disturbances.…”

“…In the context of energy systems, it has been recently reported that stochastic MPC can systematically mitigate constraint violations and improve economic performance [14,25]. The benefits of stochastic MPC have also been widely reported in the context of building climate (airside) control [26][27][28][29] and energy management [30][31][32][33][34].…”

Deep Learning-based Predictive Control of Battery Management for Frequency Regulation

“…Optimal strategies for battery participation in FR markets have received much attention and are still being investigated. [3][4][5][6][7] Especially, as a powerful control framework for optimizing control performance while guaranteeing system constraints, model predictive control (MPC) is utilized in designing optimal battery management strategies for FR. Kumar et al 6 proposed a stochastic MPC scheme combining uncertainty qualification (UQ) of uncertain signals, e.g., prices and FR signals, to optimize the economic benefits of the battery system for simultaneously participating FR markets and demand-side management.…”

“…Kumar et al 6 proposed a stochastic MPC scheme combining uncertainty qualification (UQ) of uncertain signals, e.g., prices and FR signals, to optimize the economic benefits of the battery system for simultaneously participating FR markets and demand-side management. Kumar et al 7 presented a stochastic MPC scheme to optimize the profit of participating FR markets, and provided a comprehensive comparison of control performance with deterministic and stochastic MPC. It should be pointed out that the above FR strategies only consider FR signals at low time resolution (only hourly averaged FR signals are considered in both strategy design and numerical simulation).…”