Voltage prediction using a Cellular Network

Better identification tools are needed for power system voltage profile prediction. The power systems of the future will see an increase in both renewable energy sources and load demand increasing the need for quick estimation of bus voltages and line power flows for system security and contingency analysis. A Cellular Simultaneous Recurrent Neural Network (CSRN) to identify and predict bus voltage dynamics is presented in this paper. The benefit of using a cellular structure over traditional neural network architectures is that the network can represent a direct mapping of any power system allowing for easier scalability to large power systems. A comparison with a standard single SRN is provided to show the advantages of this cellular method. Two types of disturbance are evaluated including perturbations on the power system generators and on the least stable loads. The method is also evaluated for a case involving a transmission line outage.