Page 215 - 2024-Vol20-Issue2

211 |
casting. This includes the utilization of the 1D CNN BI LSTM
model, as well as traditional STLF models like ARIMA and
ANN.
Step 3: Input Parameter Determination: An assessment is con-
ducted to identify the relevant input parameters necessary for
accurate load forecasting. Special attention isgiven to weather
parameters known to have a significant impact on electricity
demand.
Step 4: Load Forecasting: Building upon the 1D CNN BI
LSTM architecture, a load forecast model is developed. This
model undergoes thorough training and testing to ensure opti-
mal performance in load forecasting tasks.
Step 5: Performance Evaluation: A hybrid deep learning
model, specifically the 1D CNN BI LSTM model, is created
and assessed for performance. Comparative analysisis con-
ducted using statistical error matrices and actual measurement
data to evaluate the model’s efficacy.
Step 6: Model Selection: Based on the findings obtained from
the previous steps, the most suitable machine learning model
is recommended. This selection takes into account accuracy
and overall performance in load forecasting applications.
By following this systematic six-step approach, researchers
can develop a robust and effective load forecasting strategy
that leverages the strengths of the 1D CNN BI LSTM model
alongside other relevant machine learning techniques.
1) Data Evaluation and Planning
To ensure the accuracy of the model, it is imperative to per-
form proper pre-processing of raw data prior to transformation.
The initial step in this process involves gathering and orga-
nizing data to establish meaningful input-output relationships.
According to previous research [1, 2], essential pre-processing
operations such as normalization, ranking, and correlation
analysis are necessary. By adhering to recommended data
collection and processing practices, we can guarantee that
our models are constructedupon a robust foundation of high-
quality data.
a: data gathering
The datasets being collected encompass a range of informa-
tion, including climate data, calendar data, and power con-
sumption data. Weather Data Collection: The meteorological
dataset used in this research is sourced from the NASA Power
website (https://power.larc.nasa.gov/data-accessviewer/). Prior
to being utilized in the load forecasting model, the raw weather Accurate prediction of electricity demand is vital for efficient
data undergoes pre-processing, including weighting and sta-
tistical analysis. Load forecasting models incorporate weather
forecasts and other factors to minimize operational expenses.
Weather conditions significantly influence load profiles, par-
ticularly for domestic and agricultural customers. The model
development process incorporates the collection and utiliza-
tion of six weather parameters. These parameters play a vital
Gochhait, Sharma & Bachute
role in the development of the model and contribute to its
accuracy and effectiveness. Through careful analysis and
integration of these weather parameters, the model is able
to capture the important influences of weather conditions on
the target variable, enhancing its forecasting capabilities. By
including these specific weather parameters in the model, we
can ensure a comprehensive and robust approach to load fore-
casting.
1. Wet Bulb Temperature: This parameter measures the adia-
batically measured saturated temperature at two meters above
the earth’s surface. It plays a crucial role inload forecasting
as it impacts the performance of cooling systems, which sub-
sequently affects energy consumption.
2. Frost Point: It represents the temperature at which the air
becomes saturated, leading to condensation. This parameter
is critical for load forecasting as it influences the performance
of heating systems and, consequently, energy consumption.
3. Temperature: This parameter refers to the average temper-
ature of the air attwo meters above the earth’s surface, also
known as the dry bulb temperature. It significantly affects
energy consumption, particularly in buildings and homes.
4. Relative Humidity: Expressed as a percentage, relative hu-
midity represents the ratio between the actual partial pressure
of water vapor and the pressure at saturation.It is crucial for
load forecasting as it impacts the performance of cooling and
heating systems, which directly affect energy consumption.
5. Specific Humidity at 2 Meters: Specific humidity is defined
as the ratio of watervapor mass to total air mass at two meters
above the earth’s surface. It provides an indication of the air’s
moisture content and influences the performance of cooling
and heating systems.
6. Wind Speed at 10 Meters: Measured at a height of 10
meters above the earth’s surface, wind speed is an essential
parameter in load forecasting, particularly for wind energy
systems. It affects the performance of wind turbines and di-
rectly impacts energygeneration.
In conclusion, weather parameters such as Wet Bulb Tempera-
ture at 2 Meters, Dew/Frost Point at 2 Meters, Temperature at
2 Meters, Relative Humidity at 2 Meters, Specific Humidity at
2 Meters, and Wind Speed at 10 Meters are crucial featuresin
load forecasting models. They have a direct impact on energy
consumption and generation, and accurate measurement and
analysis of these parameters are essentialfor ensuring a stable
and reliable power supply.
power system management. By considering both the tim-
ing and specific characteristics of electricity demand, power
managers can make informed decisions regarding resource
allocation and ensure effective and efficient energy utilization.
Time Indicators: In this research, the investigation of time
indicators is of paramount importance, specifically the date,