Stochastic model predictive control for central HVAC plants

Abstract

We present a stochastic model predictive control (MPC) framework for central heating, ventilation, and air conditioning (HVAC) plants. The framework uses real data to forecast and quantify uncertainty of disturbances affecting the system over multiple timescales (electrical loads, heating/cooling loads, and energy prices). We conduct detailed closed-loop simulations and systematic benchmarks for the central HVAC plant of a typical university campus. Results demonstrate that deterministic MPC fails to properly capture disturbances and that this translates into economic penalties associated with peak demand charges and constraint violations in thermal storage capacity (overflow and/or depletion). Our results also demonstrate that stochastic MPC provides a more systematic approach to mitigate uncertainties and that this ultimately leads to cost savings of up to 7.5% and to mitigation of storage constraint violations. Benchmark results also indicate that these savings are close to ideal savings (9.6%) obtained under MPC with perfect information.

MPC Controller Tuning using Bayesian Optimization Techniques

“…In this case study, we leverage the MPC formulation proposed in [11] and build a BO framework for tuning TES back-off terms. In the HVAC plant, a chiller subplant produces chilled water and a heat recovery (HR) chiller subplant produces both chilled water and hot water; a hot water generator produces hot water; cooling towers are used to decrease temperature of water purchased from the market; a dump heat exchanger (dump HX) rejects heat from the hot water; and storage tanks (one for chilled water and one for hot water) are used as the TES.…”

“…The prediction horizon of MPC is chosen to be 168 hours (1 week) to reflect the weekly periodicity of loads and electricity prices. The optimization problem solved at each hour is a linear program with 168,450 variables and 143,750 constraints [11]. The problems were implemented in Julia 0.6.4 and were solved with Gurobi 8.1 on a computing server with 188 GB RAM, 32-core Intel Xeon 2.30 GHz CPU.…”

“…Errors in disturbance forecasts result in frequent constraint violations of thermal storage levels (overfilling or dry-up) that ultimately translate in decreased economic performance. Adding back-off terms to the storage levels has been shown to provide an effective strategy to deal with these issues [11]. This approach resembles constraint backoff approaches recently explored in the MPC literature [10].…”

Diffusing-Horizon Model Predictive Control

“…Constraints (1c)-(1d) allow us to accommodate models with dynamic and algebraic constraints and initial and terminal conditions. Clearly, ( 1) is a linear program and this been widely studied in the context of economic MPC [5], [8], [29], [30]. Problem (1) can be written in the following compact form:…”

“…The central plant seeks to satisfy time-varying energy demands from a university campus by manipulating chillers, storage tanks, and transactions with utility companies. The problem details are provided in [29]. The system under study is illustrated in Fig.…”