Smart-Grid Stability Prediction for Decentral Smart Grid Control Using Machine Learning

In Decentral Smart Grid Control (DSGC) every producer and consumer adapts its power to the grid frequency, which acts as a real-time price signal. Whether the grid stays stable depends on how fast the participants react (reaction time tau), how much power they produce or consume (p) and how strongly they respond to the price (elasticity g). This project implements the 4-node star DSGC model (one producer, three consumers) with the classical stability method - the rightmost root of the characteristic equation of the linearised delay model - and a time-domain simulation of the nonlinear swing equations, re-derives every label of the UCI "Electrical Grid Stability Simulated Data" (10,000 parameter sets) with its own solver (100 % identical labels, maximum difference 3e-9), and trains logistic regression, Random Forest, LightGBM and an MLP that predict stable / unstable and the stability margin directly from the 12 parameters. THE DATA ARE SIMULATED (by the dataset authors and by this project). On a 2,000-row hold-out test the deployed MLP reaches 97.40 % (ROC-AUC 0.998), LightGBM 96.60 %, Random Forest 93.30 % and logistic regression 82.45 %. The MLP is about 11,000 times faster than root finding in batch mode but only about 20 times for a single call, and it fails outside the simulated parameter ranges. A GNU Octave cross-check (run, difference 5e-16), a MATLAB version and an OpenModelica model (written, not run) and a web application with five databases are included. Everything runs on a student laptop CPU.

Category: Electrical and Electronics Engineering (Smart Grid, Power-System Dynamics, Decentral Smart Grid Control, Demand Response, Swing Equation, Delay-Differential Equations, Small-Signal Stability, Characteristic Roots, Machine Learning, Random Forest, LightGBM, Neural Network, GNU Octave, MATLAB, OpenModelica, Python)

Electrical & Electronics Engineering Project EE11 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.