This paper describes an algorithm for detecting changes in the state of a complex technical system by comparing it to a set of reference states using a spiking neural network based on a compartmental neuron model. The study covers a generalized approach to encoding system parameter information, where the system’s state is represented as spikes, a method for constructing a set of normal operating states from recorded data of guaranteed fault-free operation, and an algorithm for identifying the reference state most similar to the current one. The anomaly detection algorithm is described in detail, including the identification of the specific channel where a change occurred and the outlier value. Experimental results are presented for both the reference state selection algorithm and the change detection algorithm, tested on anonymized measurements from a real-world system and data generated by a computer model. The model data was intentionally designed to produce complex data for research purposes. In both cases, the algorithm demonstrates an F1-score ranging from 0.81 to 0.85, with a Type I error rate (missed detection) close to zero.