Machine Learning-Based State Monitoring and Regulation Characterization of Distribution Grid with High Percentage Distributed Resource Access

The large number of accesses of distributed power supplies changes the distribution network from a passive network to an active network with small and medium-sized power supplies all over the network, which brings important impacts on all aspects of the distribution network, such as planning, operation, and power quality.The study analyzes the types of distributed power nodes and the traditional trend model of distribution network, studies the changes of voltage and network loss after the integration of distributed power sources into the distribution network, and analyzes the impact of different numbers, capacities and access locations of distributed power sources on the reactive power optimization of the distribution network by means of IEEE33 nodes.Analyze the impact of distributed power supply on distribution network.Firstly, the characteristics of distributed power supply are analyzed, distributed photovoltaic and distributed wind power operation models are established, and the influencing factors of the two power supply outputs are analyzed to generalize the distributed power supply output model.The basic principle of weighted least squares state estimation and its algorithmic process are introduced, and on its basis, an equation-containing constrained state estimation model for dealing with zero-injection nodes in the distribution network is introduced, and finally, the feasibility and validity of the proposed constrained state estimation model's state estimation method for the distribution network are verified through the analysis of an example of the IEEE 33-node system.Combining the sequence quadratic programming method and the idea of trust domain, the trust domain sequence quadratic programming method is proposed, and the use of the effective set method to quickly solve the subquadratic programming problem after downsizing is the key that the algorithm in this paper can solve the optimization problem relatively quickly.