Hybrid Virtual Collection Method of IMOWOA and LightGBM for Distributed PV Power Data
The large-scale access to the grid of distributed PV power stations with many points and wide,scattered,and disorderly is one of the crucial paths for the evolution of China's new power system towards low carbon.Therefore,low-cost and high-efficiency distributed PV power station operation data collection is the primary condition for its exemplary management and lean operation and maintenance.To this end,this paper proposes a distributed PV virtual collection scheme based on the improved multi-objective whale optimization algorithm(IMOWOA)and light gradient boosting machine(LightGBM).For the virtual collection region division problem,the scheme first proposes an autoencoder similarity analysis method under the premise of grided region division to obtain the PV power station set satisfying the similarity requirement.In order to solve the problem of reference power station set selection,an improved multi-objective whale optimization algorithm is proposed to improve the global search capability of the algorithm.The subset of reference power plants and LightGBM hyperparameters is optimized simultaneously based on the region's historical power data of photovoltaic power stations,thus achieving selection of only distributed photovoltaic power stations for installation of complete data collection devices to complete high-precision virtual collection for power data of all stations within the region.Finally,the feasibility and effectiveness of the proposed virtual collection method are verified by analyzing 29 distributed PV stations in a regional area of Jiangsu Province,China.
万方数据
维普
