Composite Error Learning Estimator for Wiener–Hammerstein Systems With Time-Delay Under Binary-Valued Measurements

System modeling and parameter estimation of quantized systems are crucial for economizing communication resources and reducing the cost of sensors in the current information age of general communication and digital technology. In the presence of noise, binary-valued sensor, and nonlinearity, such an estimation is even more challenging. To address this challenge, this study introduces an alternative composite error learning estimator for nonlinear Wiener–Hammerstein systems with a time delay that is subject to binary-valued observations. Unlike existing techniques that use conventional single-error learning pattern, the proposed scheme leverages composite error data, including both pure estimation error and initial value error data, to develop a new parameter estimator, which achieves a high estimation performance compared with some available algorithms. In particular, the proposed compensation variable is applied to obtain the pure estimation error data by compensating the quantized data matrix on the effect of the estimator, thus avoiding the inverse problem of the data matrix. In addition, the provided online validation mechanism for the persistent excitation condition not only ensures the convergence of the proposed method, but can also select the type of input signal. Moreover, the system parameters and time delay can be estimated simultaneously, and the parameter error convergence is rigorously proven. The introduced method is illustrated and validated on a binary-valued Wiener–Hammerstein system with time delay and a permanent magnet synchronous motor drive plant.