Optimization of Relay Protection Setting for Distribution Networks Based on a Multi-scale Combinatorial Kernel Limit Learning Machine

The conventional distribution network relay protection setting planning is generally fixed-point or distribution network target optimization, which is relatively limited, resulting in the increase of the final setting standard deviation. Therefore, the design and practical implementation of the distribution network relay protection setting optimization method based on multi-scale combined kernel limit learning machine are proposed. According to the current relay protection processing requirements, feature extraction is carried out first, and multi-level method is adopted to break the restricted conditions and realize multi-level linear setting planning. On this basis, the optimization model of distribution network relay protection setting based on multi-scale combined kernel limit learning machine is designed, and the adaptive scheduling method is adopted to finally realize the optimal setting processing. The test results show that the multi-scale joint kernel extreme learning machine method proposed in this paper can better reduce the calibration relative error. Whether before DG access, before optimization, or after optimization, the calibration relative error can be maintained at a level below 1.5, and the convergence time can be effectively controlled.