In this paper, we consider a method for detecting abnormal behavior in the operation of cyber-physical systems, the Internet of Things (IoT) and distributed control systems using the prediction and analysis of multidimensional time series using neuroevolutionary algorithms based on the development of the hypercube substrate. The method is based on identifying deviations between the current values of the state of the cyber-physical system and the predicted results. The results of studies of the described method are presented, demonstrating the correctness and accuracy of this approach