EXPERT SYSTEM FOR CONTROLLING OPERATING MODES OF SOLAR PANELS WITH NEURAL NETWORK-BASED OPTIMALITY ASSESSMENT OF DECISIONS

Abstract

The rapid growth of solar energy utilization necessitates increasing the efficiency of control systems for photovoltaic installations operating under conditions of variable solar irradiance, temperature fluctuations, and component degradation. Modern photovoltaic systems are characterized by nonlinear behavior, stochastic external influences, and dynamic load conditions. Under such circumstances, traditional Maximum Power Point Tracking (MPPT) algorithms, which are typically based on fixed logic and local optimization procedures, do not always ensure optimal system performance, especially in transient and rapidly changing environments.

Existing approaches to photovoltaic system control primarily rely on classical MPPT techniques, rule-based logic, or monitoring-oriented analytical modules. While these methods provide stability and acceptable efficiency under steady-state conditions, they are often limited in adaptability and do not adequately account for complex interdependencies between environmental and electrical parameters. In particular, conventional solutions lack mechanisms for self-learning, dynamic optimality evaluation, and real-time corre ction of control actions, which significantly reduces their effectiveness under uncertainty and nonstationary operating conditions.

A promising direction for overcoming these limitations is the development of cyber-physical control systems that integrate expert knowledge with adaptive data-driven models. In such systems, an expert subsystem generates control decisions based on formalized rules and domain knowledge, while a neural network module evaluates the quality of these decisions and performs their correction based on learned patterns. This hybrid approach enables combining interpretability and structural clarity of expert systems with the adaptability and approximation capabilities of artificial neural networks.

The use of neural networks allows modeling nonlinear relationships between system parameters, approximating complex objective functions, and adapting to changing operating conditions. At the same time, expert systems provide a transparent and logically structured mechanism for initial decision formation, ensuring reliability and compliance with operational constraints. The integration of these components creates conditions for building intelligent control systems capable of maintaining high efficiency, stability, and robustness of photovoltaic installations.

Therefore, the development of an expert system for controlling operating modes of solar panels with neural network-based optimality assessment of decisions represents a modern and relevant scientific and practical task. Such systems have significant potential for improving energy efficiency, reducing losses, and enhancing the adaptability of renewable energy sources within modern cyber-physical infrastructures.