To meet the requirement of continuous reactive power compensation with low cost, high reliability and large capacity, a hybrid dynamic reactive power compensation system was proposed based on TSC (thyristor switched capacitor in parallel) and SVG (static var compensator). This system consisting of discrete subsystem TSC and continuous subsystem SVG, coordinated the capacitor switching rating of TSC with dynamic compensation of SVG by using two hybrid control laws based on expert decision making, made full use of their respective advantages. The machine learning approach was applied to avoid repeated movements of capacitors and extend equipment life.