The invention discloses a least square support vector machine-based power equipment fault detection method. An LSSVM regression model is trained by utilizing a large amount of power equipment experiment data, wherein the data includes an instrument gain, a measurement distance, an equipment voltage, a contamination degree, an environment air pressure, an environment temperature, environment humidity, a photon number, a local discharge quantity of power equipment, and the like; and two parameters, namely, a kernel width theta and a regularization coefficient gamma, of the model are optimized byadopting grid search and cross verification algorithms. The local discharge quantity of the power equipment is detected through the trained model; fault levels of the equipment are classified; and fault equipment is timely and effectively detected out, so that safe operation of a power grid is ensured and the property loss is reduced. The least square support vector machine-based power equipmentfault detection method provided by the invention can quickly and accurately detect out the fault equipment, and can be widely applied to the field of power equipment fault detection.