Graph-Based Prediction on the Available Capacity of 10kV Distribution Transformers for DGs Connection

In China's modern power system, the proportion of renewable energy generation is increasing rapidly. The level of electrification in end-use energy consumption is continuously growing, and electricity plays an increasingly important role in achieving carbon peak and carbon neutrality strategies. With the large-scale integration of distributed generation and electric vehicles, the distribution grid undergoes fundamental changes in its structure and configuration due to the characteristics of randomness and decentralization. Currently, power grid companies analyze and forecast the available capacity for Distributed Generations (DGs) integration into the distribution grid, often using rough calculations based on the previous year's peak load, which makes it difficult to accurately assess the DER integration capacity and test the related safety, reliability, and distributed resource accommodation capabilities. This paper Combines cluster analysis and long short-term memory (LSTM) time series prediction to engineer the forecasting of available capacity based on seasonal characteristics. The method consists of four modules: graph data modeling, user cluster analysis, load prediction, and DG capacity prediction. Instead of using relational databases, it employs a graph database, where the minimum granularity of the data system extends from discrete numbers to nodes and edges in the graph. The validation of the method is conducted by predicting and verifying the available capacity of four distribution transformers on a feeder line. The results demonstrate the superiority of the DG integration available capacity analysis and prediction method based on graph database technology regarding prediction accuracy, providing an effective decision support tool for power system planners and operators.