Optimal sizing and placement of independent power producers on MV, HV and EHV networks for minimum power line losses using neural networks

Lombard, James

Optimal sizing and placement of independent power producers on MV, HV and EHV networks for minimum power line losses using neural networks

Thesis / Dissertation

2024

Publisher

University of Cape Town

Department

Department of Electrical Engineering

Faculty

Faculty of Engineering and the Built Environment

Abstract

In June 2021, the South African president announced the adjustment of schedule 2 of the Electricity Regulation Act. The effect of this adjustment was to increase the National Energy Regulator of South Africa's (NERSA) licensing limit for renewable projects applying for direct connection to the national grid, having export capacities ranging from 1 MW to 100 MW. This increased the number of Independent Power Producers (IPPs) seeking connection to the South African electricity transmission system (TS) or distribution system (DS) – which, as a mandatory requirement, are to comply with the minimum technical and design grid connection regulations stipulated within the South African Grid Code (SAGC) covering renewable energy. The contribution of this thesis is therefore the design and testing of an ANN model that locates and sizes IPPs of Category B (1 MW to 20 MW) and Category C (> 20 MW) of the SAGC seeking connection to MV/HV/EHV backbone feeders from geographical locations far from the Point of Connection (POC). In the ANN model developed using MATLAB, the user is prompted to enter specific network parameters applicable to the grid connection study. These parameters include backbone voltage (kV), backbone length (km), interconnecting feeder length (km), maximum load seen at the receiving end (MW), distance from the IPP at the POC to the sending end busbar as a percentage of the total backbone length (km) and the load power factor at the receiving end. Based on these inputs, the model determines the most suitable IPP location, IPP size, interconnecting conductor and IPP power factor in order to achieve the lowest overall power line losses for the network. The algorithm consists of 7 ANN models, with each ANN model applied specifically to a unique nominal voltage network undergoing IPP interconnection. Seven cases are presented starting from the lowest voltage test case (11kV) to the highest voltage test case (400kV). Case 1 and Case 2 test the 11kV and 22kV ANN models on modified IEEE 13– bus systems, while Case 3 to Case 7 test 66kV, 132kV, 220kV, 275kV and 400kV ANN models on modified IEEE–14 bus systems. For the 11kV test case, the user enters input parameters: backbone conductor length of 5km, receiving end power of 4.05MVA and receiving end power factor of 0.85 (lagging). The 11kV ANN model returns an IPP size of 1.5MW operating at 0.975 (lagging) power factor, located 4.5km from the sending end substation using an ACSR Chickadee conductor. The 11kV ANN model also returns a total line loss value of 0.05351MW, while the true loss value is shown to be 0.05343MW (when compared to DIgSILENT Powerfactory simulations). This translates to an error of 0.1684%. The 22kV case is presented using the same network parameters as the 11kV case but uprated to nominal voltage of 22kV. The same trend is seen for the 22kV case but with total losses significantly less than the 11kV case due to the increased network voltage. For the 66kV test case, the backbone conductor considered is a 15.6 km ACSR Kingbird with receiving end power 30MW operating at 0.95 lagging power factor. The 66kV ANN model recommends an optimal IPP size of 12MW operating at 0.975 (lagging) power factor, located 14.04km from the sending end substation using an ACSR Kingbird conductor. The 66kV ANN model also returns a total line loss value of 0.0193MW, while the true loss value is shown to be 0.01868MW when compared to DIgSILENT Powerfactory. This translates to an error of 3.29%. The 132kV test case achieves a prediction error of 0.775% and returns an optimal IPP size of 67.5MW, located 31.5km from the sending ending busbar on the 35km backbone feeder, operating at 0.95 lagging power factor. For the EHV cases (220kV – 400kV), the same trend is seen. For the 220kV network, the lowest losses are seen for an IPP connected furthest away from the sending end (120.9km) along the 134km backbone with receiving end power 201MW at 0.95 lagging power factor. This requires a 110MW IPP at operating at 0.95 lagging pf resulting in 3.7MW of line losses using a Single ACSR Zebra interconnecting conductor. It is shown that for an IPP operating at 0.95 leading power factor, the total system losses increase to 5MW, indicating that the algorithm predicted correctly. The 275kV case has lowest losses for a 110MW IPP size operating at a lagging power factor of 0.95. This generates 1.8MW of losses (approximately 500kW lower than the capacitive case), but also is significantly lower than the 220kW case since a twin conductor Zebra bundle is used for the interconnecting feeder. v Since the 400kV network is modelled using quad Zebra backbone conductors, losses are significantly smaller than the 220kV and 275kV cases, which only used a Twin bundle conductor geometry per phase. This increased the geometric mean radius which increased the maximum power transfer of 150MW required at the receiving end. Since the power factor at the 400kV receiving end load is unity, the required reactive VAR support, in addition to the high voltage level (400kV at 1.04pu), saw an optimal IPP power factor setpoint of 0.95 (leading) resulting in a surplus of VARs. For a 138km backbone feeder with receiving end load of 150MW at unity power factor, the 400kV ANN model returns a total loss value of 179kW. The model developed can be used as a tool for providing additional support to network engineers and independent power producers (IPPs), especially for performing grid application studies. DIgSILENT PowerFactory power system simulation software is used to verify the accuracy of the algorithm. This tool is especially relevant for current needs and caters specifically to IPP units that fall under Category B and Category C of the SAGC, since these are rapidly growing in today's South African Energy Sector

Keywords

Engineering

Reference:

Collections

Masters

Full item page