This article explores the use of machine learning algorithms to identify anomalies in a solar heating system. The developed solar heating system consists of several parts to simplify the process of description and modeling. The author propose a new neural network architecture based on ordinary differential equations. The idea is to apply the new architecture to practical problems of accident forecasting (the problem of extrapolation of time series) and classification (classification of accidents based on historical data). The developed machine learning algorithms, artificial intelligence methods, and the theory of differential equations - these areas allow us to build a model for predicting system failure. Database management theory (relational databases) - these systems allow you to establish optimal storage of large time series.

Keywords: flat solar collector, solar heating system, machine learning, algorithm

In this work, an automatic control controller for a two-circuit solar installation with thermosiphon circulation based on the STM32 platform has been developed. The system operates using six sensors (temperature sensor, water flow sensor, pressure sensor, coolant temperature sensor in the heater tank, coolant temperature sensor in the heat exchanger and outdoor temperature sensor.

Keywords: flat solar collector, heat pump, two-circuit solar installation with thermosiphon circulation, automation system, SCADA, solar heat supply system, PLC, control system