A bottom-up intra-hour proactive scheduling of thermal appliances for household peak avoiding based on model predictive control

Abstract

Residential-level peak shaving is beneficial for the supply–demand balance of buildings with a limited power capacity of renewable energy sources like building-integrated photovoltaics. Thermostatically controlled loads (TCLs) are recognized as flexible resources for peak shaving. In practice, most thermal appliances are controlled by local on–off controllers without any communication and coordination. Multiple appliances may operate simultaneously and produce load peaks. Existing studies assume the existence of an electricity market and time-varying tariffs, which are not always available. This paper developed a model predictive control (MPC) based bottom-up proactive method to leverage the thermal inertia of residential thermal appliances for avoiding household peak load without relying on time-varying tariffs. First, we used a physics-based modeling approach to model the thermal dynamics of several residential thermal devices. Then, we designed a centralized MPC controller to minimize the appliance concurrent operations considering occupant comfort zones and appliance rate powers. Consequently, the house peak load is reduced by minimizing concurrent operations. Simulations over homogenous and heterogenous appliance set validated that the MPC-based proactive scheduling method can effectively reduce household peak load without compromising user comfort too much. Finally, we discussed the impact of control interval, appliance rated powers, and penalty factors on peak avoiding and the computational feasibility of the MPC controller. We concluded that the MPC-based proactive scheduling is feasible and effective for a limited number of thermal appliances with short horizons (i.e., six-step). Replacing a mixed-integer nonlinear program solver with a mixed-integer quadratic program solver could significantly reduce the computational burden. However, the application of the MPC controller to multiple households at the system level is still impractical. Further research efforts are emphasized, like mining and forecasting occupant demand and appliance usage patterns in households and designing scalable control strategies for system-level peak shaving.