Abstract
This paper investigates the optimization control method and state-of-charge (SOC) balancing strategy for distributed battery energy storage system (BESS). A cooperative control strategy based on SOC balancing is proposed. Utilizing the theory of multi-agent systems (MAS), the cooperative control of battery energy storage system is achieved. A distributed algorithm based on MAS is adopted to achieve adaptive allocation of power instructions, further realizing dynamic SOC balancing. To address the issue of slow convergence in traditional multi-agent algorithms, a model predictive control (MPC)-based distributed algorithm is proposed, which optimizes the traditional multi-agent algorithm using MPC, improving convergence speed. Finally, simulations using actual storage power data validate the effectiveness of the proposed strategy and the advantages of the algorithm in terms of convergence speed.
Keywords: battery energy storage system, model predictive control, distributed algorithm, cooperative control, state-of-charge
1. Introduction
Energy storage has become an integral part of power systems, with distributed battery energy storage systems (BESS) offering significant advantages in energy density and widespread deployment in power systems. While research has focused on optimizing the control of energy storage systems and balancing battery state-of-charge (SOC), these aspects are often studied separately. Traditional power optimization strategies have focused on minimizing power limits and avoiding SOC overshoot, but have overlooked the issues of SOC imbalance, which can lead to reduced battery capacity utilization and shortened battery life.
This paper aims to address this gap by proposing a cooperative control strategy based on SOC balancing for distributed battery energy storage system. Utilizing multi-agent systems (MAS) theory, the strategy achieves SOC balancing while maintaining efficient power allocation. Furthermore, a model predictive control (MPC)-based distributed algorithm is introduced to optimize traditional multi-agent algorithms, enhancing convergence speed.
2. Literature Review
2.1 Energy Storage System Control
Previous research on energy storage system control has focused on various optimization strategies. For instance, some studies utilize empirical mode decomposition principles to achieve optimal power allocation. Others have employed genetic-simulated annealing algorithms to minimize over-limit power during charging and discharging while avoiding SOC overshoot. Variable droop control strategies have also been proposed, adjusting factors to improve power response speed. However, these approaches generally focus on power allocation without considering SOC imbalance.
2.2 SOC Balancing Techniques
SOC balancing techniques can be broadly classified into two categories: energy dissipation and energy transfer. Energy dissipation methods involve using resistors to consume excess power, leading to energy waste. Energy transfer methods, on the other hand, utilize inductors and capacitors to transfer energy between batteries, but this can increase charging and discharging frequencies, accelerating battery aging. Improved droop control strategies have attempted to incorporate average SOC into the droop coefficient, but this can result in large bus voltage deviations. Centralized control strategies for SOC balancing, while achieving high precision, can introduce significant communication pressures.
3. Proposed SOC Grouping Balancing Strategy Based on MAS
To achieve cooperative control and SOC balancing in distributed battery energy storage system, a novel SOC grouping balancing strategy based on multi-agent systems (MAS) is proposed. This section details the MAS framework, control strategy, and the grouping method.
3.1 MAS Framework
In the proposed framework, each battery unit in the distributed battery energy storage system serves as an agent in the MAS. Agents communicate with their neighboring agents to coordinate power allocation and SOC balancing. illustrates the topology of a distributed battery energy storage system with 10 battery units, with solid lines representing physical connections and dashed lines representing communication links.
The power instruction (P_b) is received by one of the agents, and the MAS distributed algorithm ensures that this power is adaptively allocated among the agents, with higher SOC batteries contributing more during discharge and receiving less during charge, achieving dynamic SOC balancing.
3.2 SOC Grouping Strategy
To reduce unnecessary switching between charging and discharging states, the battery units are grouped into two subsystems: one for charging and the other for discharging. This approach minimizes switching frequencies, leading to less battery aging and improved capacity utilization.
4. Multi-Agent Distributed Algorithm
Based on the previous discussion and the reference material provided, I will continue in English to complete the overview of the control strategy and algorithm for distributed battery energy storage systems.
3.4 Analysis of System Communication and Scalability
The proposed control strategy relies on a distributed communication network among the battery units. The MAS framework ensures that each agent only communicates with its neighboring agents, significantly reducing the communication pressure on the central controller compared to centralized control schemes. This decentralized approach also enhances the scalability of the system, as additional battery units can be easily incorporated into the network without significant changes to the overall communication architecture.
Moreover, the use of virtual control variables in the MPC-based multi-agent distributed algorithm allows for efficient power allocation while maintaining a low communication overhead. The algorithm’s reliance on local information and neighboring agent states ensures that the control decisions can be made rapidly and reliably, even in large-scale distributed battery energy storage systems.
4. Comparison with Existing Strategies
Compared to traditional power allocation and SOC balancing strategies, the proposed approach offers several advantages:
- Dynamic SOC Balancing: The MAS-based SOC grouping strategy achieves dynamic balancing of SOC among battery units without the need for additional balancing circuits. This increases the overall utilization of battery capacity and extends battery life.
- Reduced Charging and Discharging Frequency: By grouping battery units into separate charging and discharging subsystems, unnecessary state switches are avoided, thus reducing the charging and discharging frequency and mitigating battery aging.
- Improved Convergence Speed: The MPC-based distributed algorithm significantly accelerates the convergence speed compared to traditional multi-agent algorithms, enabling real-time control of large-scale distributed battery energy storage systems.
- Lower Communication Pressure: The decentralized communication network based on local and neighboring agent information reduces the communication load on the central controller, enhancing the scalability and reliability of the system.
5. Practical Applications and Future Work
The proposed control strategy has practical applications in various scenarios, including microgrids, renewable energy integration, and power grid ancillary services. With further optimization and testing, the algorithm can be tailored to specific system requirements and operating conditions.
Future work could focus on:
- Experimental Validation: Implementing the proposed control strategy on real-world distributed battery energy storage systems to validate its performance under various operating conditions.
- Adaptive Control Strategies: Developing adaptive control strategies that can dynamically adjust the parameters of the MPC-based algorithm based on system states and external disturbances.
- Fault Tolerance and Resilience: Enhancing the fault tolerance and resilience of the control system to ensure reliable operation in the presence of hardware failures or communication disruptions.
6. Conclusion
In summary, this paper proposes a collaborative control strategy based on SOC balancing for distributed battery energy storage systems. By leveraging the multi-agent system (MAS) framework and model predictive control (MPC), the proposed approach achieves dynamic SOC balancing among battery units, reduces charging and discharging frequency, improves convergence speed, and reduces communication pressure. Simulation results demonstrate the effectiveness and advantages of the proposed control strategy, suggesting its potential for practical applications in modern power systems.