Learning for System Identification of NDAE-modeled Power Systems

Wenjie Mei,Muhammad Nadeem,MirSaleh Bahavarnia,Ahmad F. Taha

2023-12-13

Abstract:System identification through learning approaches is emerging as a promising strategy for understanding and simulating dynamical systems, which nevertheless faces considerable difficulty when confronted with power systems modeled by differential-algebraic equations (DAEs). This paper introduces a neural network (NN) framework for effectively learning and simulating solution trajectories of DAEs. The proposed framework leverages the synergy between Implicit Runge-Kutta (IRK) time-stepping schemes tailored for DAEs and NNs (including a differential NN (DNN)). The framework enforces an NN to cooperate with the algebraic equation of DAEs as hard constraints and is suitable for the identification of the ordinary differential equation (ODE)-modeled dynamic equation of DAEs using an existing penalty-based algorithm. Finally, the paper demonstrates the efficacy and precision of the proposed NN through the identification and simulation of solution trajectories for the considered DAE-modeled power system.

Systems and Control,Dynamical Systems

What problem does this paper attempt to address?

The problem that this paper attempts to solve is in the power system, how to effectively identify and simulate the solution trajectories of dynamic systems described by differential - algebraic equations (DAEs) through machine - learning methods. Specifically, the dynamic behaviors in power systems are usually modeled by differential - algebraic equations, which can capture the dynamic responses and the algebraic constraints among causal components. However, traditional system identification methods face relatively large difficulties when dealing with such equations, especially when facing nonlinear differential - algebraic equations (NDAEs) with infinite - stiffness characteristics. Therefore, this paper proposes a framework based on neural networks (NN), especially by introducing the difference neural network (DNN), to overcome these problems and effectively learn and simulate the solution trajectories of NDAEs.

The main contributions of the paper include:
1. Proposing a neural - network scheme containing DNN for identifying the parameters of power systems modeled by NDAE. This method can also adapt to the general form of NDAEs (with an index of 1), which is the inherent physical property of the considered power model.
2. Under some standard assumptions, analyzing the error dynamics between the real system and the proposed DNN, and presenting the conditions for limiting the identification error in an infinite - time range. The effectiveness of these conditions is verified through numerical simulations, and it is pointed out that these conditions are more relaxed than those in the existing literature.

Through this method, the paper demonstrates the effectiveness and accuracy of using neural networks for system identification and simulation in power systems, especially in dealing with complex dynamic behaviors. This provides new tools and techniques for improving the efficiency and resilience of power systems.