Multiobjective meter placement in active distribution system state estimation using inverse-model-based multilabel Gaussian classification with adaptive reference point method

This paper proposes a new inverse-model-based multiobjective evolutionary algorithm for meter placement in active distribution system state estimation. The meter placement is designed as a multiobjective problem with minimizing conflict objectives such as meter cost, the relative error of voltage magnitude, and voltage angle. The multiobjective framework utilizes inverse model as a reproduction operator and maps the nondominated solution from objective space to decision space and is realized using multilabel Gaussian classification. The additional solutions are generated by sampling from inverse model that improves the search efficiency and diversity of Pareto optimal solutions. The combinatorial nature of meter placement optimization may produce a discontinuous Pareto front. The performance of multiobjective evolutionary algorithm depends on the shape of Pareto front. Therefore, to improve the performance, the adaptive reference point method is employed to adjust the reference points such that they follow the Pareto front. The proposed method is tested under different real measurement uncertainties for passive and active distribution networks. Moreover, different types of renewable sources are considered in active distribution system. The inverse model and adaptive reference point method improve the performance of multiobjective evolutionary algorithm. Therefore, the results obtained from the proposed method outperform the other existing multiobjective evolutionary algorithms, and the obtained optimal number of meters are less with the minimum voltage magnitude error and voltage angle error. The proposed method is tested on PG&E 69-bus distribution system and Practical Indian 85-bus distribution system.