Parameter Identification for a Power Distribution Network Based on MCMC Algorithm

The calculation and analysis of a power distribution network (PDN) require accurate device parameters. However, a PDN has many points, and the distribution area is very wide. The PDN parameters are influenced by manual entry, and most are relatively random. Additionally, these parameters are affected by the operating status. Thus, this paper proposes an algorithm that accurately identifies PDN parameters based on the Markov chain and Monte Carlo (MCMC) method. The algorithm assumes that the PDN parameters conform to a nonlinear probability space. The parameters are the line resistance R-L, line reactance X-L, short-circuit loss P-k, short-circuit voltage percentage U-k%, no-load loss P-0, no-load current percentage I-0%, etc. The algorithm in this paper uses the Monte Carlo method to provide parameter values that conform to the initial probability distribution and then combines the data collected from the actual feeder to perform power flow calculations to obtain the loss function. The data include the head and end voltages and active and reactive power on the low voltage side. The Markov chain and loss function update the initial parameter probability distribution. The low voltage side voltage of the power flow calculation is iteratively calculated under the new given parameters to obtain the new loss function, and finally, the PDN line and transformer parameter values are identified. Actual feeder data verification results show that this MCMC PDN parameter identification method can obtain high-precision parameter values without phase angle information; additionally, this method is insensitive to the initial values and exhibits fast convergence.

Automated summary

This paper proposes an algorithm that accurately identifies PDN parameters based on the Markov chain and Monte Carlo method. The algorithm provides parameter values that conform to the initial probability distribution through the Monte Carlo method and combines actual feeder data to perform power flow calculations to obtain the loss function. The algorithm shows high-precision parameter values without phase angle information, insensitivity to initial values, and fast convergence.