Kriging surrogate model enabled heuristic algorithm for coordinated Volt/Var management in active distribution networks.

• An optimal VVC model considering diverse controllable resources including energy storage devices, distributed generators, step voltage regulators, static var compensators, and switchable capacitor banks in the ADN is proposed in this paper, in which the VVC problem is considered in a more complete energy management framework coupling with active power control. • Given that the active distribution network VVC problem is a computation-intensive MINLP problem with high-dimensional variables and strong non-convexity, the Kriging assisted surrogate model is trained to approximate the computationally expensive black-box nonlinear system, reducing the total number of actual time-consuming simulations, and moderating the computational burden without losing precision. • By expected improvement and dynamic updating techniques, the Kriging assisted modified adaptive particle swarm optimization (KA-MAPSO) is ultimately developed to solve the proposed VVC optimization problem, avoids the low convergence in non-convex and nonlinear power flow calculations. And the simulation result shows that the proposed method outperforms other referred algorithm and accelerate the solving process of traditional heuristic calculation. The increasing penetration of distributed energy resources exacerbates the risk of power loss increment and voltage violations in active distribution networks (ADN). The Volt/Var control (VVC) involving multiple controllable resources is widely considered to be an effective means of ensuring the secure and stable operation of ADN. Firstly, considering diverse controllable resources including energy storage devices, distributed generators, step voltage regulators, static var compensators and switchable capacitor banks in the ADN, a comprehensive/novel VVC optimization model is established as a mixed-integer non-linear programming problem with high-dimensional variables. In order to mitigate the computational burden with enormous networks and reduce the dependence on an initial physical model of the distribution network, a Kriging-assisted evolutionary computation method for VVC problem is proposed in this paper. The surrogate model is used to approximate the nonconvex and nonlinear network, reduce the frequency of complex simulations, assist optimization decision-making and avoid the complicated explicit analysis. Furthermore, the surrogate-assisted model enabled heuristic algorithm, owning promising performance in approximating the computationally expensive black-box system, can be obtained via continuous training and dynamic updating. Finally, the numerical results on the modified IEEE 33-bus and PG 69-bus systems verify the effectiveness of the proposed method. The simulation results indicate that the proposed Kriging-assisted evolutionary algorithm outperforms other algorithms in the aspect of calculation efficiency and robustness while accelerating the solving process of traditional evolutionary calculation. [ABSTRACT FROM AUTHOR]