Power distribution network hybrid observation stationing method based on deep learning and decision tree driving

Abstract

The invention discloses a power distribution network hybrid observation stationing method based on deep learning and decision tree driving, and relates to the technical field of power system optimization. Comprising the following steps: firstly, constructing a decision tree model, and quantitatively analyzing the importance degree of influence of each node feature of the power distribution network on power distribution network topology observability; according to the node feature importance degrees, sequentially arranging from large to small; selecting part of node measurement features and measurement data with high importance to form a mixed measurement sample set for deep learning model training; and then, using a PCA-DBN coupling deep learning model to analyze topological observability change performance indexes of the power distribution network under different hybrid measurement schemes, and finally, according to the selected nodes and the hybrid measurement schemes when the power distribution network is completely observable, determining an optimal stationing scheme of the hybrid measurement device. The proposed method breaks through the planning thought of a traditional measurement device optimization stationing method, and adopts data driving methods such as deep learning and decision tree to realize an economical and efficient hybrid observation stationing planning scheme for active power distribution network topology identification.

Description

technical field [0001] The invention relates to the technical field of power system optimization, in particular to a distribution network hybrid observation and distribution method based on deep learning and decision tree drive. Background technique [0002] With the development of distribution network automation, more and more measuring devices such as AMI and PMU have been put into the distribution network. Data support is provided for complex nonlinear problems. However, this type of data-based data science method has relatively high requirements on the quality and quantity of sample data, especially in a large-scale distribution network, due to the high-precision AMI, PMU and other distribution automation measurement devices cost Expensive, the number of configurations of this type of measurement device is very limited, and the quantity and accuracy of real-time measurement data are still difficult to meet the data requirements of the current algorithm. In most scenario...

AI Technical Summary

Problems solved by technology

However, this type of data-based data science method has relatively high requirements on the quality and quantity of sample data, especially in a large-scale distribution network, due to the high-precision AMI, PMU and other distribution automation measurement devices cost Expensive, the number of configurations of this type of measurement device is very limited, and the quantity and accuracy of real-time measurement data are still difficult to meet the data requirements of the current algorithm. In most scenarios, pseudo-measurement data based on prediction has to be used; although some power grids have been put into There are many advanced measurement devices, but their location is not reasonable enough, the collected measurement data contains more redundant information, and the cost of various measurement devices is different, and the measurement data is different due to data delay, data cycle, The data incompatibility caused by the large difference in data accuracy is an important factor affecting the complete control of the distribution network.

Images