Enhancing Power Flow Through Advanced Application of Extreme Learning Machine Generator Capability Curve

Abstract

Power flow analysis is a key component in system evaluation. Among the various methods available, the Newton-Raphson approach is particularly effective. However, this method typically represents the generator capability curve (GCC) using a quadrilateral limit, defined by Pmin-Pmax and Qmin-Qmax constraints. This often results in certain parts of the GCC being disregarded, which can lead to less optimal performance during power flow analysis. To address this issue, this study introduces a new method: the Extreme Learning Machine Generator Capability Curve (ELMGCC), which aims to more accurately represent the shape of the GCC. ELMGCC replaces the traditional rectangular limits to better constrain the generator's operating point during power flow calculations. The study applies ELM-GCC to both the Newton-Raphson (NR) and Fast Decoupled (FD) power flow methods to assess the effectiveness of this new approach. Simulation results using IEEE 30 Bus data with slight modifications show that the proposed method can maintain PV Bus performance up to 83.33% and reduce losses by 0.108216368053700 MW and 0.872096049537200 MVar for the NR method, and by 0.108236277781099 MW and 0.872099845833500 MVar for the FD method.