Abstract:
Peak load shaving is a method used to reduce the maximum demand of electricity on utility
grid. Promoting the use of distributed energy resource such as solar Photovoltaic (PV) can
reduce maximum demand of utility grid during peak time and can be used as ancillary
support for critical loads. Ethiopian Electric Utility (EEU) started its plan to fully energize
the whole country by 2030.To determine whether or not 15kV side of Bahir-Dar
distribution substation is capable to satisfy the peak load demand of EEU there need to
forecast the peak load value of the substation for the future years. The peak load value is
forecasted by using newly registered domestic customers and their sanction load as an input
and their corresponding 15kV distribution substation side peak load data as output. Based
on those data the next future 44 months peak load value is forecasted using Artificial Neural
Network (ANN). The ANN result depicts that starting from 2023 the substation will be
overloaded after 22 months. Since the peak load demand of Bahir-Dar distribution occurs
at day time and based the city has sufficient irradiation level for PV system, we can use
grid connected PV system as peak load shaving. For analysis the paper focuses to substitute
the two-diesel running gensets of EEU datacenter by a peak value of 100kW PV system.
Adaptive Neuro Fuzzy Inference System (ANFIS) type Boost converter is modelled to step
up the designed 437.6V to 730.04V DC-link voltage and tracks its desired gating pulse for
the converter under variable environmental condition. This paper uses five level Modular
Multi level converter (MMC) topology to convert the DC to AC and to inject power
generated by PV system to the point of common coupling (PCC) with the running grid at
lower total harmonic distortion (THD) value. The designed MMC has 3 legs with each has
upper and lower arms and each arm has 4 submodules (SM). To control the injected power
to the running grid this paper uses Voltage oriented control (VOC) loop with outer voltage
or power control loops and inner current control loop. The SM capacitor voltage balancing
is done by sorting algorithm. The circulating current control of MMC is done by properly
sizing of arm inductor and SM capacitor. The MMC output voltage has THD value of
16.27% and by properly sizing RL filter the line current of injected power has THD value
1.12% which is far below international standard of 5%.
Keywords: ANN, PV, peak load, ANFIS, Modular Multi level Converter, Submodule