attractive power generation options in many nations. Even while high penetration seems likely, power quality

anomalies such as voltage fluctuation, harmonics, and frequency fluctuation associated with RES hinder

seamless integration. The variability and unpredictability of these sources create the most oddities. In grid-tied

renewable energy, monitoring power quality efficiently is crucial. Power grid monitoring solutions in related

literature use sensor-based cloud and edge computing techniques. The existing systems struggle with excessive

latency when delivering large amounts of generated data to the cloud. To fill this gap, a new approach for the

detection and localization of voltage fluctuation is proposed in this study. The approach integrated three

techniques namely; feed-forward neural network (FFNN), Stockwell transform, and anomaly-aware edge

computing to detect and locate voltage fluctuation in a GtRE. Using MATLAB/Simulink, virtual emulation of a

modified IEEE 33 Bus and a GtRE representing a section of Ado Ekiti (in Nigeria) low-voltage distribution

90% in latency when compared to cloud computing and conventional edge computing respectively.

Comparison of the proposed approach to two closely related solutions in literature also demonstrates a 50% and

92.5 % reduction in latency. The contribution of the study is the reduced latency and minimal bandwidth

utilization achieved by the implementation of the developed technique.

Key-Words: - Power grid, renewable energy, voltage fluctuation, neural network, edge computing, latency.

Received: May 21, 2024. Revised: August 12, 2024. Accepted: September 5, 2024. Published: October 22, 2024.

from fossil fuels have great influence on climate,

[2]. Apart from the effects of the conventional

power system generation on climate, the motivation

to consider renewable energy sources (RES) is

derived from other factors such as rising demand for

carbon foot-print elimination by the year 2050, [5],

[6].

considerable amount of power close to the

distribution network, with renewable generators as

typical examples, [9], [10], [11]. Distributed

generation has benefits that include: lower power

loss, greater voltage support, peak shaving,

increased system efficiency, stability, and

dependability, [10], [12]. Meanwhile, the technical

challenges of GtRE from certain sources such as

solar photovoltaic and wind turbines are critical

power quality issues, [11]. According to [13], [14],

attention lately due to their ability to address issues

like the rising need for electricity, air pollution, and

the subsequent difficulties caused by global

warming. The inherent characteristics of these

renewable energy sources namely, fluctuations in

wind speed and solar radiation have a big impact on

power quality, dependability, and safety. Low PQ

levels therefore run the risk of causing motor

failure, line overheating, imprecise metering, early

device aging, and disruptions in communication

that equipment's operation" is how the IEEE defines

PQ. A PQA is defined as any variation in voltage or

current over a certain period from its nominal

values. PQAs are defined as a temporary deviation

from the nominal magnitude and/or frequency

components. The voltage fluctuation caused by the

integration of renewable energy sources, including

solar photovoltaic, is the main focus of this study.

One of the main problems with power quality that

arises from RES integration with the grid is voltage

one of the two main categories of artificial neural

networks, [22], [23], distinguished by how

information is transmitted between its layers. The

flow of the model is characterized as unidirectional,

indicating that information within the model

progresses solely in one way. This progression

occurs from the input nodes, passing through any

hidden nodes, and ultimately reaching the output

nodes. Feed-forward networks are trained by the

was used to simulate voltage signal datasets for

model training. Also, MATLAB was used to

simulate grid-tied renewable energy (solar) where

the developed system is evaluated. The edge

computing devices are comprised of simulated

voltage sensors and microcontrollers managed by a

Raspberry pi 2 (b+). The Raspberry pi

microcontroller is responsible for running the output

weight of the edge-based neural network model and

sending the captured data to the developed cloud

grid-tied renewable energy was explored. This study

focused majorly on the most prominent power

quality anomaly in grid-tied renewable energy

(GtRE) which is voltage fluctuation. Training

datasets of voltage fluctuation were generated from

the simulated modified IEEE 33 Bus network using

algorithmic codes. Data of normal voltage signals

and fluctuated voltage signals were obtained from

the simulated grid. The measured line voltages were

collected and saved at the output side of the power

is worth noting that the S transform does not suffer

from the issue of cross-terms, making it a more

effective method for achieving signal clarity

3.5 Description of the Studied Power Grids

Due to the adverse effects of voltage fluctuations on

other loads connected to the same network as the

disruptive load, and the constraints of technical

feasibility, the only viable method for experimental

verification was through simulation. The IEEE 33-

bus test system is the first studied network in this

study, the network, adopted from [28] is depicted in

Figure 2. The network is modeled in

MATLAB/Simulink environment. The modified 33-

transformer at bus 1 has a capacity of 3 MW.

Fig. 2: Line diagram of IEEE 33-bus test system,

[28]

low-voltage networks, because voltage fluctuations

most often occur in them. The majority of the low-

voltage networks in Nigeria are radial topologies

fluctuation signals and subsequently transmit them

to the next layer within the network. It is imperative

that the feature or attribute numbers inside the

continuously changed to enhance its voltage

fluctuation predictive capabilities. The output layer

simply represents the predicted voltage fluctuation

events.

Fig. 3: The Simplified FFNN Architecture

3.7 Description of the Edge Computing

System

deployed on a raspberry pi microcontroller to run

detection in GtRE. The first computing device is

represent Edge Computing (EC). The third

computing device is configured to represent

Enhanced Edge Computing (EEC), which is the

solution that is being proposed in this study. Each

to work within three layers, the first is sensor layer,

the second is edge computing layer and the third is

cloud layer.

In EEC, a voltage fluctuation-aware algorithm is

introduced to capture and send only data considered

by the trained model to be anomalies. Data from the

sensors are processed within the edge computing

layer, anomaly events are transmitted to the cloud

layer which comprises of a dedicated Java-based

graphical user interface. The algorithm of the cloud

represented as two classes, ‘fluctuation’ and ‘no

fluctuation’. Performance metrics are crucial for

evaluating the effectiveness of the developed model.

150. The curve in Figure 9 illustrates the

progression of model training and validation loss

over multiple epochs. The decline in loss

commenced during the 25th epoch, and by the 150th

epoch, the loss had diminished to a near-zero level.

Table 3. Comparison of the three computing devices

Fig. 10: Bus 4 Cloud Computing Curve

Fig. 11: Bus 4 Edge Computing Curve

Fig. 12: Bus 4 Enhanced Edge Computing Curve