Abstract
Smart home appliances can time-shift and curtail their power demand to assist demand side management or allow operation with limited power, as in an off-grid application. This paper proposes a scheduling process to start appliances with time-varying deterministic load profiles. Self-triggered model predictive control is used to limit the household net power demand below a given threshold. Meanwhile, deterministic load profiles are more difficult to schedule compared to variable charging or thermal loads because system failure will occur once power demand is not satisfied. The proposed scheme formulates the decision of the load shifting time as a continuous optimization problem, and an inhomogeneous time grid system is introduced to handle the optimization of different appliances and their consensus at this resolution. The efficacy of the proposed scheme is studied by numerical comparison with a mixed-integer MPC controller and by a case study of three home appliances and an interruptible washing machine.
Introduction
In off-grid homes or microgrids, power is a finite resource. Scheduling appliances like washing machines or ovens—which have "deterministic load profiles" (fixed power needs over a set duration)—is challenging. Unlike a battery that can be charged at any variable rate, these appliances usually have a specific "shape" of power consumption once they start. Traditionally, the "start time" of an appliance is treated as an integer, turning scheduling into a Mixed-Integer Programming (MIP) problem. While accurate, MIP is NP-hard; as the number of appliances or the time resolution increases, the computational burden becomes prohibitive for real-time home automation.
Methodology
The authors propose a shift in perspective, drawing inspiration from self-triggered Model Predictive Control (MPC). Instead of evaluating start times at every discrete clock tick, the controller treats the start time ($t_s$) as a continuous decision variable.
Inhomogeneous Time Grids
A key innovation is the use of "Inhomogeneous Time Grids." The system employs a "Global Grid" to track total power limits and "Local Grids" for each specific appliance. This allows the controller to allocate computational resources where they are most needed—for example, high resolution during a complex wash cycle and low resolution when a machine is idle.
Smooth Interpolation and Constraint Handling
Standard load profiles are often "blocky" or discontinuous, making them difficult for continuous solvers. To address this, the authors use Radial Basis Function (RBF) kernels and piece-wise linear interpolation to "smooth" these profiles, making them compatible with gradient-based numerical solvers like Ipopt. The model explicitly accounts for load shifting (delaying a start) and curtailment (reducing power slightly) to ensure total household demand never exceeds a safety threshold ($P_{max}$).
Results and Discussion
The study validated its approach using synthetic benchmarks and real-world profiles for an oven, dishwasher, and washing machine.
Computational Efficiency
The continuous formulation demonstrated significant speed improvements. In benchmark comparisons, the proposed solver reached a solution in approximately 1.1 seconds. In contrast, Mixed-Integer Quadratic Programming (MIQP) solvers like Gurobi took between 9 and 36 seconds, while open-source Branch-and-Bound (BNB) solvers took nearly 200 seconds or failed at high resolutions.
Performance and Feasibility
Despite the speedup, the continuous solver achieved performance nearly identical to the global optimum. The "Total Loss" (measuring delay and curtailment) was 1.47 for the continuous solver, compared to 1.41 for the much slower integer solvers.
The system successfully managed complex scenarios, such as intelligently pre-charging a home battery and shifting a washing machine's dry cycle to avoid exceeding grid capacity. By treating time as a continuum, the problem becomes a standard nonlinear program (NLP), decoupling complexity and allowing the shifting decision to remain constant regardless of time resolution.
Conclusion
This research transforms the computationally heavy task of scheduling household appliances into a fast, continuous optimization challenge. While the continuous formulation is non-convex and carries a theoretical risk of local minima, practical results show performance comparable to global solvers. Future work may explore distributed optimization, where individual appliances negotiate schedules with a central hub, further enhancing the scalability of smart home energy management.
Summary updated