Machine-Learning N-1 Contingency Security Assessment of the IEEE 39-Bus System

A transmission grid is N-1 secure when the loss of any single line, transformer or generator leaves every bus voltage and every branch flow within its limits. Control rooms check this with a full AC contingency analysis (one power flow per outage - accurate but slow) or with a fast DC power flow + line outage distribution factor (LODF) screening that sees only thermal limits. This project generates 5,000 operating points of the IEEE 39-bus New England system (pandapower case39 = MATPOWER case39) with loads of 70-114 % and a security-constrained DC re-dispatch, labels every point by a full AC N-1 analysis (225,000 AC power flows with an own vectorised Newton-Raphson solver that equals pandapower runpp and MATPOWER runpf in GNU Octave - labels 100 % identical), and trains logistic regression, Random Forest, LightGBM and MLP models that decide N-1 security from the pre-contingency state in about a millisecond. THE DATA ARE SIMULATED. On 1,000 hold-out test points the classical DC-LODF screening reaches 85.1 % but misses 17.0 % of the insecure points (all voltage-only cases); the deployed LightGBM reaches 98.7 % with only 0.57 % missed insecure points, the MLP classifier 98.4 %, the Random Forest 97.8 % and logistic regression 93.8 %. A MATLAB version (written, not run), MATPOWER cross-checks in GNU Octave (run) and a web application with five databases are included. Everything runs on a standard laptop CPU (8 GB RAM, no GPU).

Category: Electrical and Electronics Engineering (Power Systems, Power-System Security, N-1 Contingency Analysis, AC Power Flow, Newton-Raphson, DC Power Flow, Line Outage Distribution Factors, Optimal Power Flow, IEEE 39-Bus New England System, Machine Learning, LightGBM, Random Forest, Neural Network, pandapower, MATPOWER, GNU Octave, MATLAB, Python)

Electrical & Electronics Engineering Project EE12 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.