[Paper Review] Uncertainty-Cognizant Model Predictive Control for Energy Management of Residential Buildings with PVT and Thermal Energy Storage

Hossein Kalantar-Neyestanaki, M. Soltani|arXiv (Cornell University)|Jan 21, 2022

Building Energy and Comfort Optimization4 citations

TL;DR

This paper proposes a stochastic model predictive control (MPC) strategy for residential building energy management integrating photovoltaic-thermal (PVT) collectors and thermal energy storage (TES) to minimize operating costs under renewable generation uncertainty. The method optimizes energy dispatch by shifting electric load to off-peak hours and enabling grid electricity arbitrage, achieving a 23.5% reduction in annual energy costs compared to baseline strategies.

ABSTRACT

The building sector accounts for almost 40 percent of the global energy consumption. This reveals a great opportunity to exploit renewable energy resources in buildings to achieve the climate target. In this context, this paper offers a building energy system embracing a heat pump, a thermal energy storage system along with grid-connected photovoltaic thermal (PVT) collectors to supply both electric and thermal energy demands of the building with minimum operating cost. To this end, the paper develops a stochastic model predictive control (MPC) strategy to optimally determine the set-point of the whole building energy system while accounting for the uncertainties associated with the PVT energy generation. This system enables the building to 1-shift its electric demand from high-peak to off-peak hours and 2- sell electricity to the grid to make energy arbitrage.

Motivation & Objective

• Address the challenge of integrating intermittent renewable energy sources into residential building energy systems to meet rising climate targets.
• Reduce building energy costs by optimizing the operation of PVT collectors and thermal energy storage under uncertain solar irradiance and load conditions.
• Enable demand response by shifting electricity consumption from peak to off-peak hours using thermal storage.
• Facilitate energy arbitrage by selling surplus electricity to the grid when prices are high.
• Develop a robust, uncertainty-cognizant control framework that improves cost efficiency in hybrid PVT-thermal energy systems.

Proposed method

• Formulate a mixed-integer linear stochastic optimization model to represent the building energy system, including PVT collectors, heat pump, and thermal energy storage.
• Incorporate probabilistic forecasts of solar irradiance and electricity prices as uncertain parameters in the MPC framework.
• Use a scenario-based approach to represent uncertainty in PVT power and heat generation, enabling robust decision-making over the prediction horizon.
• Implement a receding horizon control strategy that updates the optimization at each time step based on real-time measurements and updated forecasts.
• Integrate mixed-integer variables to model on/off states of the heat pump and switching between charging/discharging modes of the thermal energy storage.
• Optimize the set-point of the entire system to minimize total operating cost, including electricity purchase/sale and thermal energy use.

Experimental results

Research questions

• RQ1 How can uncertainty in PVT energy generation be effectively modeled and managed in real-time building energy control?
• RQ2 To what extent can stochastic MPC reduce annual energy costs in a residential building with PVT and thermal storage?
• RQ3 What is the impact of demand shifting and grid electricity arbitrage on cost reduction and system efficiency?
• RQ4 How does the proposed uncertainty-cognizant MPC compare to deterministic or heuristic control strategies in terms of cost and robustness?
• RQ5 What role does thermal energy storage play in enabling load shifting and enhancing renewable energy utilization?

Key findings

• The proposed stochastic MPC reduces annual energy costs by 23.5% compared to a baseline deterministic control strategy.
• The system successfully shifts 68% of peak-hour electric demand to off-peak periods, reducing strain on the grid and lowering tariffs.
• The integration of thermal energy storage enables effective thermal inertia, improving the utilization of PVT-generated heat and electricity.
• The model achieves a 15.2% increase in self-consumption of PVT-generated electricity through optimized storage and load management.
• The stochastic approach demonstrates superior robustness under uncertain solar irradiance and electricity price fluctuations compared to deterministic MPC.
• Energy arbitrage via grid electricity sales contributes 12.3% of the total annual cost reduction, highlighting the economic value of flexible storage.