Abstract
Conventional Power Distribution Networks (PDNs) are passive in nature. With the incorporation of distributed generations (DGs) in electric power systems, such distinctive feature of a traditional PDN is being distorted. DG penetration into the power distribution networks leads to serious technical problems in their operations. In order to effectively control a modern PDN, it is imperative to ascertain the state of that network. Existing works on state estimation as applied to PDNs are mostly hindered by the unbalanced nature of the network and inadequate real-time measurements which may lead to poor estimation of the network. In this work, the states of PDNs are estimated using a Deep Feed-Forward Neural Network technique which was then compared with two estimators in Ahmad et al., 2019 using Mean Absolute Deviation (MAD) as well as Mean Square Error (MSE) performance metrics for testing on a local network. The proposed estimator was tested on the 33-bus and 69-bus IEEE standard networks as well as the Zaria local distribution network under Normal and Dynamic conditions. The Simulation was implemented in MATLAB 2019a environment. As it was clearly observed from the obtained results, FFNNSE outperformed ANNSE for both 69-bus and 33-bus Networks with 7.41% and 12.0% MAD reduction respectively. It was however, performed excellently than WLSSE with the reduction of 66.0% and 78.0% MAD for both Networks. The Performance of the FFNNSE was tested on a 50-bus local distribution network under normal and dynamic conditions (Bad data and Load Variation) the performance was good for all conditions with minimal MAD of 0.0045, 0.0049 and 0.0051 for normal, Bad Data and Load Variation conditions respectively. However, MSE for all cases was computed as 0.000176, 0.000202 and 0.000215for normal and two dynamical operations respectively. The State Estimation approach results show the viability of the FFNNSE for real-time distribution networks.

Key words: state estimation; power distribution network; feed-forward neural network; MAD; MSE; WLSSE; FFNNSE