Anomaly Detection Method For Interactive Data of Third-Party Load Aggregation Platform Based on Multidimensional Feature Information Fusion

With the development and using of clean energy, more and more distributed generations including photovoltaic panels, which can generate the power by consuming the new and renewable energy are connected to the system. However, the power grid system is vulnerable to attack due to the greater load pressure and security risks. This paper presents a third-party load aggregation platform interactive data anomaly detection method based on multi-dimensional feature information fusion and deep residual network analysis in a comprehensive energy scenario. The method we proposed can collect, extract and analyze the interactive data of the third-party load aggregation platform, and then analyze and detect the anomaly of the load data collected by the platform from the perspective of multi-dimensional feature fusion analysis. Specifically, by extracting the initial data features of the multi-dimensional third-party load platform, this paper adopts wavelet transform and spectral clustering technology to denoise, filter pseudo data features and perform feature clustering analysis due to the magnanimity and dynamic acquisition characteristics of power load data; Then, by using the cross layer direct connected edge characteristics of the depth residual network, the error back propagation attenuation in the depth learning is constructed, and the depth network model of abnormal data detection is trained to achieve the task of abnormal data detection of the third-party load aggregation platform interactive data. The main contribution of this paper is that the method of the third-party load aggregation platform interactive data anomaly detection based on multi-dimensional feature information fusion and deep residual network is presented, and the test results have shown the efficient of the method.