National Conference on Emerging Trends in Automation,

Data Engineering and Communication

(NCETinADC2021)

Review on Load forecasting Techniques

Authors:
Mr. Mohit Choubey Dr. Rahul
Chaurasiya
Department of ECE Department of ECE
MANIT, Bhopal
MANIT, Bhopal

Dr. Dr. Jigyendra Yadav

Department of ECE
MANIT, Bhopal
Contents

 Abstract
 Introduction
 Types of Loads
 Types of load forecasting
 Types of Data variable
 Standard data sets
 Methods of load forecasting
 Conclusion and future scope
Abstract
In the modern world any nation’s economic growth is related to its electrical energy availability and
development. As demand for electricity is increased drastically worldwide and so prices are also
increased because still there is a lack of demand and generation. In order to solve this problem many
researchers proposed different solutions for predicting and efficiently generating, distributing and
supplying electrical energy among residential and industrial load. Load forecasting is one of the important
aspects related to electricity development and maintenance and this study shows you a review on all
available techniques related to load forecasting and what we can do for in future to achieve forecasting
with good accuracy. While reviewing previous paper we look for what type of algorithm used and how
well it fit, what accuracy achieved, error calculation and time of forecasting. While analyzing all paper we
find that model with good accuracy used ANN for load forecasting, and RMSE and MAPE parameters are
used for error calculation. And finally, we proposed challenges and possibility for future research.
 Introduction
Electrical Load Forecasting is the estimation for future load by an industry or
utility company
Load forecasting is a difficult task.
First, because the load series is complex and exhibits several levels of seasonality:
the load at a given hour is dependent not only on the load at the previous hour, but
also on the load at the same hour on the previous day, and on the load at the same
hour on the day with the same denomination in the previous week.
Secondly, there are many important exogenous variables that must be considered,
especially weather- related variables.

Ref :- Short term electricity load forecasting using a hybrid model; Energy, 2018;158:774-781
 Types of Loads
Five broad categories:
• Domestic
• Commercial
• Industrial
• Agriculture
• Other Loads

Commercial and agricultural loads are characterized by seasonal variations.

Industrial loads are base loads and are little weather dependent.

Ref :-Probabilistic electric load forecasting: A tutorial review. International Journal of Forecasting,2016;3:article 13
Types of Load Forecasting
 Short term load forecasting
Short term electric load forecast spans the period from one hour up to one week
and it is mainly utilized for power system operation studies, losses reduction,
voltage regulations, unit commitment and maximizing the utility revenues in the
deregulated environment.
High forecasting accuracy and speed are the two most important requirements of
short-term load forecasting and it is important to analyze the load characteristics
and identify the main factors affecting the load. In electricity markets, the
traditional load affecting factors such as season, day type and weather, electricity
price.

Ref :- Electricity load forecasting: a systematic review. Journal of Electrical Systems and Inf Technol ,2020;7:article 13
 Medium Term Load Forecasting

Medium term electric load forecast spans the period from one week to
several weeks, it is mainly utilized for predicting the necessary power to
purchase or sell from other neighboring networks (inter-tie exchanged
power) and also the fuel required by the utility in the near future.
 Long Term Load Forecasting
Long term electric load forecast spans the period from one year to 20 years
and this forecasting in long-term horizon is important for electric utilities
and planners in terms of grid expansion planning, future investments and
revenue analysis for long-term decision making process
 Input data sources for Load forecasting
 Population of area
 Historic load data
 Building permits
 Employment of area
 Ventilation
 No. of AC’s and Heating Equipment's
 Weather data
 Building Structures
Ref :- Electricity load forecasting: a systematic review. Journal of Electrical Systems and Inf
Technol ,2020;7:article 13
Standard data sets

 Electricity load of New South Wales (NSW)

 French grid operator RTE
 Australian Energy Market Operator
 New York independent system operator
 Bureau of Meteorology Australia
 National weather service USA
 Andhra Pradesh Southern Power Distribution Corporation Limited
 Short term load forecasting methods
In general, Short term load forecasting models can be classified into three major
categories:

Time series Models.

Artificial intelligence models.
Hybrid models.

Ref :- Short term electricity load forecasting using a hybrid model; Energy, 2018;158:774-781
Time series analysis
Time series forecasting is based on the idea that reliable predictions can be achieved by
modeling patterns in a time series plot, and then extrapolating those patterns to the future.
Using historical data as input, time series analysis fits a model according to seasonality
and trend.
Time series approaches are not widely used for energy industry forecasting. Because they
typically do not take into account other key factor, such as weather forecasts, but we can
use it with other algorithm to remove its limitation
Time series models have been used for electricity load forecasting such as:-
 Linear regression.
 Exponential Smoothening.
 Autoregressive moving average (ARMA) .
 Seasonal autoregressive integrate moving average (SARIMA).
 Kalman filtering.
Ref :- A seasonal direct optimal hybrid model of computational intelligence and soft computing
techniques for electricity load forecasting. Energy 2017;140:988-1004
 Linear regression.

This approach is based on searching historical data for days within one,
two, or three years with similar characteristics to the forecast day.
Similar characteristics include weather, day of the week, and the date.

Accurate holiday, Temperature modelling.

Ref:- A regression-based approach to short term system load forecasting. IEEE Trans Power Syst 1990;
4:1535-1547
 Exponential Smoothening.

The observed load x(t) is represented as a linear combination of known

functions of time and a noise component, the estimates of the coefficients
are revised hourly, the estimates of the coefficients are computed using a
weighted least squares criterion.

Not much accurate, holidays load not predicted with accuracy.

Ref:- Short-term load forecasting using general exponential smoothing. IEEE trans. on power apparatus and
systems, 1971;3:900-911
 Autoregressive moving average (ARMA)

The ARMA models represent a combination of an autoregressive models AR and a

moving average models MA . In the ARMA models, the current value is expressed
linearly in terms of its past values and in terms of current and previous values of the
noise. ARMA models have been a popular choice and extensively applied to load
forecasting researches due to their relative simplicity and effectiveness and this only be
used for stationary time series data.

Improvements to forecast accuracy without cost increase as no new facilities involve.

Ref:- Load forecasting via suboptimal seasonal autoregressive models and iteratively reweighted least-squares estimation.
IEEE Trans Power Syst 1993; 1:343-348
 Seasonal autoregressive integrate moving average (SARIMA)

Autoregressive Integrated Moving Average, or ARIMA, is a forecasting

method for univariate time series data i.e. problem with ARIMA is that it
does not support seasonal data. That is a time series with a repeating
cycle.

Seasonal Autoregressive Integrated Moving Average, SARIMA or

Seasonal ARIMA, is an extension of ARIMA that explicitly supports
univariate time series data with a seasonal component.

Ref:- Short-term load forecasting using a two-stage SARIMAX model. Energy 2017; 133:108-114
 Kalman filtering
The Kalman filter (KF) is a set of mathematical equations in the state space that can
provide an efficient computational (recursive) means to estimate the state of an observed
process. this filter is very powerful in several other aspects: it supports estimations of past,
present, and future states.
The factors like weather and load data can be injected as inputs into the Kalman filter,
based on state space vector satisfying the hidden state and observation vector output is
present, The KF’s mechanism works in a two-step process, that is, the predictor step (PS),
and the corrector step (CS). In the PS, the KF estimates the current load’s state on the
basis of its previous state, together with its covariance uncertainty. Once the new SMD’s
measurement is observed, the estimated state vector is updated by deploying a weighted
average, where the higher weight is given to the estimate with a higher certainty.

Ref:- Short-term electric load forecasting based on Kalman filtering algorithm with moving window
weather and load mode. Electric Power Systems Research 2004;68: 47-59
 Artificial Intelligence Model
AI methods for forecasting have shown an ability to give better performance in
dealing with non-linearity and other difficulties in modeling of the time series.
Their advantage lies mainly in the fact that they do not require any complex
mathematical formulations or quantitative correlation between inputs and outputs.
Artificial intelligence models have been presented to forecast load such as :-
 Support vector machines.
 Fuzzy logic.
 Artificial neural networks.
 Support vector regression.

Ref:-Artificial intelligence in short term electric load forecasting: a state-of-the-art survey for the
researcher. Energy Conversion and Management ; 2003;44:1525–1534
 Support vector machines.
In the SVM models, a nonlinear mapping of the input space into a higher
dimensional feature space is deployed, and afterward, an optimally separating
hyperplane is extracted.
Accordingly, the complexity and quality of the SVM solutions do not directly depend
on the input space. When designing an SVM model, the training process that is
comparable to solving of a linearly constrained quadratic programming problem is
carried out. Therefore, conversely to the other networks’ training, the SVM solutions
appear to be always globally optimal and unique.Less database and processing
needed

Ref:- Forecasting regional electricity load based on recurrent support vector machines with genetic algorithms.
Electrical Power Syst Res 2005; 74:417-425
 Fuzzy logic
In this paper, the ‘time’ and ‘temperature’ of the day are taken as inputs for the fuzzy
logic controller and the ‘forecasted load’ is the output.
The input variable ‘time’ has been divided into eight triangular membership functions.
The membership functions are Mid Night, Dawn, Morning, Fore Noon, After Noon,
Evening, Dusk and Night.
Similarly Another input variable ‘temperature’ has been divided into four triangular
membership functions and output variable ‘forecasted load’ is divide into four
triangular membership functions,
this technique implements experience and intuition into a computer program
System complexity reduced, development time goes down,
cost and economics getting reduced

Ref:- Fuzzy approach for short term load forecasting. Electrical Power Syst Res 2006; 76:541-548.
 Artificial neural networks.

The neural network is a non-linear circuit that is capable of doing non-linear curve fitting.
It represents an information processing paradigm that was inspired by the way the
biological systems of humans, such as the brain, are able to process a certain piece of
information. In this process, the ANNs try to recognize regularities and patterns in the
input data, learn from experience, and then provide generalized results based on their
known previous knowledge.
An ANN is composed of several interconnected processing elements (PE), called
neurons, which are changing their dynamic state response with respect to external inputs.
The simplest form of an artificial neural network containing input, hidden and output
layer, layers have weight associate with them and by changing the weights of the ANN’s,
the preferred output from a specific input can be achieved.

Ref:- A new cascade NN based method to short-term load forecast in deregulated electricity market. Energy Convers
Manage 2013; 71:76-83
ANN Types.
 Feed-forward (FF) neural networks
 Back-propagation (BP) neural networks
 Radial basis function (RBF) neural networks
 The random neural networks
 Recurrent neural networks
 LSTM
 CNN
 RNN
 Hybrid Models
Hybrid model integrating different models will be a good choice for improving the
forecasting accuracy.
The reason can be mainly attributed to two aspects:
• The model combined with suitable models can capture the different features of
electricity load.
• the model combined with other models can overcome its own defects
The hybrid model can be classified into two main categories.
• Electricity load was predicted separately by different models. Then, the weight of each model was
calculated by a suitable method. The final forecasting value was the sum of each model’s forecasting
value multiplied by its weight.
• Electricity load was decomposed into several components. Then each component was predicted by a
suitable model. The final forecasting results were the sum of each component’s forecasting results.

Ref:-Artificial intelligence in short term electric load forecasting: a state-of-the-art survey for the researcher.
Energy Conversion and Management ; 2003;44:1525–1534
 Accuracy comparison

Time series methods up to 95%

Artificial intelligence methods up to 98%

Hybrid methods more than 99%

Conclusion and Future Scope
Accurate load forecasting is very important for electric utilities in a competitive
environment created by the electric industry.
Although single model based artificial intelligence model can consider other influence
factors, Complex to handle and it still has its own shortcoming such as local minimum
and over training. Thus, hybrid model integrating different models will be a good choice
for improving the forecasting accuracy.
As hybrid models with more than two methods have considerable advantage compared to
single or double models, future research will be focused on the design and comparison of
hybrid models with three or more methods which will address the disadvantages of the
current forecasters and achieving a well organized power system utility where the
demand load can be predicted with the minimum error percentage.
Thank you