Stochastic Modeling Method and Heavy Overload Analysis of Source Loads in Distribution Substations

Distributed photovoltaic (PV), charging piles, and residential electricity use have an increasing problem of heavy overloading of distribution transformers due to stochastic uncertainty in operation. To address this problem, a historical data-driven stochastic modeling of source loads in distribution substations is proposed. The method divides the peak and valley hours based on historical data, and describes the behavioral state of typical residential electricity consumption using the Gaussian mixture distribution and expectation maximization algorithm. It divides the cloud state by establishing a cloud coverage index, divides the weather type using a clustering algorithm, and describes the behavioral state of PV cluster output in combination with the beta distribution. Mines the historical operation data of source-load clusters using the Markov chain, establishes the system’s operation state transfer matrix at each moment, and adopts the sequential Monte Carlo modeling method. Matrix, and adopt sequential Monte Carlo simulation method to sample the behavioral state duration and establish a stochastic model of the system state; describe the charging load by establishing the probability distributions of battery capacity, daily mileage, and starting charging moment of electric vehicles, and adopt Monte Carlo sampling to calculate the charging load of electric vehicles. Finally, the reasonableness and effectiveness of the proposed method are verified by simulation statistics, and the risk characteristics of the heave overload probability of the station under different scenarios are analyzed.