Distribution network failures can cause unintentional islanding of three-/single-phase multimicrogrids (MMGs). The transient impulse that occurs during the unintentional islanding period can affect the stability of the voltage and frequency in an MMG. To address this problem, this article proposes an unintentional islanding transition control strategy based on artificial emotional reinforcement learning (AERL) for a three-/single-phase MMG. First, to solve the three-phase unbalance problem that occurs during the unintentional islanding period, a three-phase combination method based on a merge sort is proposed to realize the combination optimization of the single-phase source-load-storage. Second, a load-shedding strategy based on AERL is proposed to deal with the tie-line power shortage caused by the distribution network failures. This strategy can quickly eliminate power shortages and ensure the uninterrupted power supply of critical loads. Finally, the performance of the proposed transition strategy is verified in a three-/single-phase MMG model based on a modified IEEE 37-bus system and a modified IEEE 118-bus system. During the unintentional islanding period, the proposed transition strategy reduces the frequency recovery time by 21.74%, 14.29%, and 10%, the voltage recovery time by 16.22%, 13.89%, and 8.82% compared with the mixed-integer second-order cone programming method (MISOCPM), the implicit enumeration method (IEM), and the compound storage-regulating and load-shedding method (CSLM) in the modified IEEE 37-bus system, respectively; the proposed transition strategy reduces the frequency recovery time by 25.64%, 21.62%, and 12.12%, the voltage recovery time by 19.15%, 15.56%, and 7.32% compared with the MISOCPM, IEM, and CSLM in the modified IEEE 118-bus system, respectively. The test results show that the proposed transition control strategy realizes the seamless transition of a three-/single-phase MMG and ensures the uninterrupted power supply of critical loads.

Supply and demand balance is required to assure the normal operation of microgrids. When unintentional islanding occurs in a microgrid, a load shedding (LS) control scheme is necessary to achieve the power balance and frequency stability of an islanded microgrid due to its limited power generation. However, the LS control of a microgrid faces two key challenges, i.e., how to determine the appropriate LS amount and LS objects in an islanded microgrid. To solve these problems, this paper proposes a coordinated LS control scheme based on Double-Q learning for an islanded microgrid. In the proposed coordinated LS control scheme, aiming at the LS amount problem in the process of unintentional islanding, this paper considers the relationship between the active power and frequency deviation of each distributed energy resource (DER) and uses the frequency deviation measured locally to calculate the LS amount and guide LS actions. Aiming at the problem of LS object selection in the process of unintentional islanding, this paper describes the LS control problem as a Markov decision process (MDP). The load priority is considered in a reward value function to ensure the uninterrupted power supply of critical loads in a microgrid. The two value functions of Double-Q learning are used to make the two Q value tables measure each other, which can reduce the overestimation of Q values and improve the accuracy of decision-making. The performance of the proposed coordinated LS control scheme is verified by a microgrid model based on a modified IEEE 13-bus system. The test results verify the feasibility and effectiveness of the proposed coordinated LS control scheme.

Reasonable scheduling is the basic guarantee for an integrated energy system (IES) to achieve coordinated and efficient operation of multi-energies. For an IES including electric and thermal loads, a demand response (DR) model based on a compensation mechanism is established in this paper, and scheduling elasticity (SE) of different types of loads is analyzed to guide users to use energies reasonably and economically. On this basis, an optimization model is established for an IES in accordance with the energy consumption and system operation characteristics. In accordance with the dynamic demands of multi-energies, this model aims at meeting all energy demands with the lowest operation cost. It performs the coordinated optimization for the device output power and the power transmission between multi-energies. To solve the problems of the complex solving process and long computation time, a global optimization algorithm based on a polynomial response surface (PRS) metamodel is proposed in this paper. The proposed algorithm adopts a response surface method to fit the optimization model and construct a PRS metamodel to estimate the function values instead of the optimization model, thereby avoiding repeatedly calling the original complex objective function and reducing the computation time. The test results verify the effectiveness of the proposed model and algorithm.