A real-time continuous monitoring system for long-term voltage stability with sliding 3D convolutional neural network

A real-time continuous monitoring system (CMS) for long-term voltage stability assessment with sliding three-dimensional convolutional neural network (3D-CNN) is proposed in this paper. Given a power system, its dynamic responses and topological information are strategically fused and converted into sequential state images, which constitute the learning input of the sliding 3D-CNN. In this way, both spatial and temporal correlation are considered in the CMS. The bus indices are reordered to strengthen the topology information in these images. With the localized weight-shared convolution operations, 3D-CNN can be conducted more efficiently with fewer parameters, and extract features independently of spatial locations. The translation invariance characteristic makes 3D-CNN highly generic to unknown scenarios, which is further enhanced after bus indices reordering. The sliding window of the 3D-CNN enables the CMS to handle varying degrees of disturbance and performs continuous online monitoring. Numerical tests are carried out on the New England 10-generator-39-bus system to demonstrate the effectiveness. Test results also show the potential of the CMS in robustness to PMU measurement errors and information losses.