A sensor network modeling and fault detection method for large wind farms by using neural networks

Sensor networks have been widely utilized in various applications. In large wind farms, numerous sensor nodes are deployed across the field for monitoring purpose. They are required to work in harsh environment and usually undergo unexpected failures. This paper introduces a new method to model nodes of sensor networks by using two-layer neural networks (NN). Each node's dynamics and interconnections with other sensor network nodes are integrated into the model, whose accuracy is guaranteed by the NN's universal approximation property. Furthermore, the model can be subsequently employed to detect any incipient failures which can be modeled as a nonlinear function of state and input variables. An additional NN along with a novel simplified updating law is utilized for self-diagnostics, whose output can declare a failure alarm if it exceeds a certain threshold. Mathematical analysis is substantiated with simulation results.