Advanced load modelling techniques for state estimation on distribution networks with multiple distributed generators

The paper compares a variety of modelling methods of different complexity to improve the accuracy of pseudo-measurements used for state estimation on distribution networks. The pseudo-measurements are required due to the lack of real-time measurements. However, pseudo-measurements of low accuracy increase the error of the resultant state estimates, degrading the confidence in their use for control applications. The solution is to adopt advanced modelling methods to produce pseudo-measurements with higher accuracy. The candidate methods adopted here are the method of assumed variance, normal distribution fitting, the correlation method and Gaussian mixture modelling. These methods are tested on a real distribution network incorporating distributed generation. The results show that using more advanced methods normally improves the accuracy of state estimates but on some occasions the improvement is not significant and sometimes the accuracy can become worse. Moreover, false large percentage errors in power flow estimates caused by distributed generation have been observed, and this would give wrong indication in ideal locations to add real-time measurements.