Application and Comparison of Metaheuristic and New Metamodel Based Global Optimization Methods to the Optimal Operation of Active Distribution Networks

Abstract

As an imperative part of smart grids (SG) technology, the optimal operation of active distribution networks (ADNs) is critical to the best utilization of renewable energy and minimization of network power losses. However, the increasing penetration of distributed renewable energy sources with uncertain power generation and growing demands for higher quality power distribution are turning the optimal operation scheduling of ADN into complex and global optimization problems with non-unimodal, discontinuous and computation intensive objective functions that are difficult to solve, constituting a critical obstacle to the further advance of SG and ADN technology. In this work, power generation from renewable energy sources and network load demands are estimated using probability distribution models to capture the variation trends of load fluctuation, solar radiation and wind speed, and probability scenario generation and reduction methods are introduced to capture uncertainties and to reduce computation. The Open Distribution System Simulator (OpenDSS) is used in modeling the ADNs to support quick changes to network designs and configurations. The optimal operation of the ADN, is achieved by minimizing both network voltage deviation and power loss under the probability-based varying power supplies and loads. In solving the computation intensive ADN operation scheduling optimization problem, several novel metamodel-based global optimization (MBGO) methods have been introduced and applied. A comparative study has been carried out to compare the conventional metaheuristic global optimization (GO) and MBGO methods to better understand their advantages, drawbacks and limitations, and to provide guidelines for subsequent ADN and smart grid scheduling optimizations. Simulation studies have been carried out on the modified IEEE 13, 33 and 123 node networks to represent ADN test cases. The MBGO methods were found to be more suitable for small- and medium-scale ADN optimal operation scheduling problems, while the metaheuristic GO algorithms are more effective in the optimal operation scheduling of large-scale ADNs with relatively straightforward objective functions that require limited computational time. This research provides solution for ADN optimal operations, and forms the foundation for ADN design optimization.