An Evaluation Method of Running Status for Power Grid Dispatching Control System Based on Decision Graph and Pairwise Constraints

The running status evaluation of smart grid dispatching control system makes dispatcher easier to master the operating state in time ensuring the safety and reliable operation of the system. At present, most traditional evaluation methods are based on expert experience, which are subjective. Machine learning algorithms could improve the objectivity of evaluation. At the same time, considering that it is difficult for D5000 system to obtain a large number of sample labels of its running state, this paper proposes a running status evaluation method based on decision graph and pairwise constraint (DGPC). Firstly, the pairwise constraints set are generated using the small amount of labeled data obtained by AHP with the running status evaluation model of D5000 system. Secondly, the density and distance of labeled samples are considered to determine the initial clustering center through the decision graph, so as to improve the reliability of the clustering center. Then, on the basis of not violating the pairwise constraints set, k-means algorithm is used to cluster the data of the system operating state by using distance measurement and index weight, so as to obtain the running status of D5000 system. Finally, clustering experiments were carried out using three sets of data selected from the UCI data set. The results show that the proposed algorithm can effectively improve the clustering performance. In addition, the D5000 system monitoring data collected by a power company was used for experiments. The effectiveness of the running status evaluation method proposed in this paper was verified by comparison with the traditional typical methods, which provided a new idea for D5000 system running status evaluation.