SNR Estimation in Wireless Control Systems for Electrical Microgrids Using Wavelet Neural Networks

This paper introduces a design for a Wavelet Neural Network (WNN) to estimate the link quality through the signal-to-noise ratio (SNR) between a wireless sensor network. The samples of the received signal strength indicator data are obtained from a wireless link between a low-cost ESP32 microcontroller and a WiFi access point. The SNR is described in a time-varying nonlinear component and a non-stationary random component. Each of these components is processed independently. The configuration and initialization parameters of two WNNs are defined based on the acquired data, and their training is performed. The configuration of the WNN allows for the estimation of the confidence interval in which the SNR values will be contained. Furthermore, a comparison of the algorithm's performance concerning a conventional neural network is presented, demonstrating that the WNN exhibits superior performance due to the estimated confidence interval covering 15 more values than its traditional counterpart.