Page 88 - 2023-Vol19-Issue2

84 | Swide & Marhoon

forecast is accurate, the spinning reserve can quickly make (a)
up for any errors. The load profile forecast over longer time
horizons is used to determine the capacity to contribute to a (b)
whole network in order to avoid an emergency. Forecasting Fig. 1. (a) electrical energy consumption distribution (b) Site
load electricity demand is crucial for modern smart energy location of Denmark
management systems [[5]. In both the long-term planning
for new transmission and generation facilities as well as the presented two deep RNN techniques to forecasting the usage
management of short-term load, it is crucial. Improved cost of electricity over a medium- to long-term horizon. Also,
and energy efficiency decisions are also made possible by an they presented a method for dealing with missing data. As
accurate projection. compared to the traditional Multi-Layer Perceptron (MLP)
neural network, the results showed that the proposed models
The number of apps for load forecasting is increasing daily. had a lower relative error [12]. Another well-liked approach
This is why the field of forecasting electricity consumption is random forest (RF), which relies on training [[13], [14]].
has received a lot of attention in the literature [6], [7]. Some of The improvement in RF comes from its nonlinear estimating
the literature predicts that weather variables like temperature, suitability and reduced sensitivity to parameter values. All
humidity, rainfall, or season may help to influence energy AI-based methodologies call for the best possible architectural
consumption [8]. While other studies made their estimates layout and parameter optimization, which hybridization can
using socioeconomic and population factors [9]. Although successfully handle. ANN was recently used by to forecast the
generic characteristics can be used to estimate power use energy analysis certificates of residential structures in Italy.
a different prediction model can lead to a better outcome.
Based on the original data acquired from Data from Kaggel Ahmad et al. (2017) anticipated energy consumption us-
Western Europe Power Consumption forecasting model is ing the weather, time, and building consumption[15]. Lee et
created for a specific Denmark power consumption. Fig. 1 al. (2017) used a large data analytics technology to calculate
displays the site’s location The National Aeronautics and the environmental consumption level by country[16]. Li et al.
Space Administration (NASA) released the Power data access (2017) used Autoencoder to anticipate future energy consump-
viewer data and how electrical energy is consumed. tions and extract the building’s energy demand [17]. Wang et
al. (2018) Studied a recent RF short-term electricity estimate
The structure of this essay’s content is as follows. The for commercial buildings with regard to their envelope, cli-
background material for various load consumption prediction mate and time when forecasting the hourly electricity load in
models is given in Section II. The approach is explained in buildings, the study shown that RF outperformed regression
Section III. The computational findings and discussion are
covered in Section IV. The work is concluded and future work
is discussed in section V.

II. RELATED WORK

The challenge of predicting power usage has been addressed
in earlier studies using a variety of prediction techniques,
ranging from statistical to machine learning-based methods
[10]. The majority of this research make the more predictable
assumption that electricity usage follows a regular pattern[11].
In reality, however, customer behavior influences the how
actual usage pattern evolves over time. Because there are so
many variables at play, including the temperature, occupancy
rates, and capabilities of the heating system, this behavior is
too unpredictable to predict. In order to address this problem,
a fresh research trend has emerged that uses Deep Learning
(DL) approaches for prediction and demand forecasting. In
[11], a brand-new pooling-based DL model was put forth with
the goal of employing deep learning to uncover the uncertainty
of the household load forecasting model. In comparison to the
RNN, Support Vector Regression (SVR), and Autoregressive
Integrated Moving Average (ARIMA) models, the proposed
model produced better accurate results. Rahman at el. (2017)