Direct load control by distributed imperialist competitive algorithm

Fengji Luo, Junhua Zhao, Haiming Wang, Xiaojiao Tong, Yingying Chen, Zhao Yang DongJournal of Modern Power Systems and Clean Energy 2014 · DOI

Abstract

Demand side management techniques have drawn significant attention along with the development of smart grids. This paper proposes a new direct load control (DLC) model for scheduling interruptible air conditioner loads. The model is coordinated with unit commitment and economic dispatch to minimize the total operation cost over the whole dispatch horizon. Network constraints are also considered in the model. To ensure the thermal comfort of the occupants, we incorporate an advanced two-parameter thermal inertia dynamical model of customer houses into the DLC model to calculate indoor temperature variation. This paper also proposes a distributed imperialist competitive algorithm (DICA) to effectively solve the model. Simulation studies prove the efficiency of the proposed methodology.

Introduction

As renewable energy sources like wind and solar become more prevalent, the power grid faces a "stochastic" challenge: supply and demand are becoming increasingly unpredictable. Demand Side Management (DSM)—specifically Direct Load Control (DLC)—allows utilities to remotely cycle appliances like air conditioners (ACLs) to "shave" peak demand.

However, existing DLC research often suffers from three major gaps: oversimplified physics that ignore the thermal mass of walls, a lack of consideration for physical grid limits (transmission line constraints), and high computational complexity when scheduling thousands of units alongside traditional power plants. This paper addresses these challenges by integrating realistic thermal modeling with a novel distributed optimization metaheuristic.

Methodology

The proposed framework utilizes a three-pronged approach to optimize grid operations while maintaining consumer comfort.

1. Integrated Optimization Model

The authors formulate an objective function that minimizes the sum of generation costs (fuel and startup) and the "interruption cost" (the compensation paid to customers for load shedding). Crucially, the model includes AC power flow constraints to ensure the grid remains stable and within voltage limits during load adjustments.

2. Two-Parameter Thermal Inertia Model

Unlike standard models that only monitor indoor air temperature, this model accounts for the thermal capacitance of the walls. By treating the physical structure of the house as a form of thermal storage, the model provides a more accurate picture of how a house retains coolness. This "thermal inertia" allows for more flexible scheduling because the indoor temperature does not spike immediately when the AC is toggled off.

3. Distributed Imperialist Competitive Algorithm (DICA)

The scheduling problem is a non-convex, high-dimensional challenge. The authors utilize the Imperialist Competitive Algorithm (ICA), a socio-politically inspired metaheuristic where solutions ("empires") compete to "assimilate" colonies. To handle the massive scale of the problem, a distributed version (DICA) was developed to run across a network of processors. A coordination node manages the migration of "colonies" between empires on different processors, ensuring a balanced workload and a fast, global search for the optimal solution.

Experimental Validation

The study uses the IEEE 39-bus system (New England power grid) as a benchmark, modified to include 10 generating units and 8 distinct groups of air conditioner loads. To simulate a heterogeneous real-world environment, the authors defined 8 different types of buildings using materials like brick, wood, and concrete with varying dimensions and wall thicknesses.

Key Results and Insights

  • Peak Shaving: The DLC model effectively curtailed 285.25 MWh of load during the peak period (14:00 to 20:00), representing approximately 27.43% of the total interruptible capacity.
  • Cost Efficiency: The total system operation cost was optimized to $53,927 for the dispatch horizon.
  • Thermal Comfort: Despite the cycling, indoor temperatures for all customer groups were successfully maintained within their preferred "dead bands" (typically between 22°C and 28°C).
  • Computational Speedup: The distributed architecture showed near-linear speedup. Using 28 computing processors, the simulation time dropped significantly compared to a single-processor run, proving the method is scalable for large-scale smart grids.

Conclusion

This paper demonstrates that incorporating advanced thermal physics and distributed metaheuristics can significantly improve the efficacy of Direct Load Control. By treating building structures as thermal batteries and accounting for grid constraints, utilities can reduce costs and peak demand without compromising resident comfort. Future work will explore more elaborate load-balancing strategies for the DICA and address potential communication latencies in wide-area smart meter deployments.

Summary updated

Abstract

Demand side management techniques have drawn significant attentions along with the development of smart grid. This paper proposes a new direct load control (DLC) model for scheduling interruptible air conditioner loads. The model is coordinated with the unit commitment and economic dispatch to minimize the total operation cost over the whole dispatch horizon. The network constraints are also considered in the model. To ensure the thermal comfort of the occupants, we are among the first to incorporate the advanced two-parameter thermal inertia dynamical model of customer houses into the DLC model to calculate the indoor temperature variation. This paper also proposes a distributed imperialist competitive algorithm to effectively solve the model. The simulation studies prove the efficiency of the proposed methodology.
BibTeX
@article{luo2014direct,
  author = {Fengji Luo and Junhua Zhao and Haiming Wang and Xiaojiao Tong and Yingying Chen and Zhao Yang Dong},
  title = {{Direct load control by distributed imperialist competitive algorithm}},
  journal = {Journal of Modern Power Systems and Clean Energy},
  year = {2014},
  doi = {10.1007/s40565-014-0075-x},
}
Figure