International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014 49
ISSN 2229-5518
Electric Load Forecasting in Smart Grid
Environment and Classification of Methods
B. R. More
(Department of Electrical Engg, Government Polytechnic, Nanded (MS), India Email: brmore65@gmail.com)
—————————— ——————————
Nowadays smart grid technologies are actively under
research and they are planned to be actively used in the
Load forecasting is a central and integral process in planning and operating of electric utilities. It involves the accurate prediction of both the magnitudes and geographical locations of electric load over the different periodic usually hours of the planning horizon. The basic quantity of interest in load forecasting is typically the hourly total system load. There are three types of load forecasts, short – term, medium term and long term, Long term forecasts are made for the period up to ten years or up to several decades and they help companies to plan their investments, plan and develop generation, transmission and distribution systems. It is made to, forecast electrical energy consumption and maximum loads in a grid so that companies could make economical based decisions of installing or replacing electrical equipment in a grid to make it capable of generate, transmit and distribute. Forecasted energy and provide and acceptable level of power quality and reliability of power supply. Smart technologies are being developing very actively, they allow increasing power quality and power supply reliability, reduce energy flows in a grid. There are several types of smart technologies, smart and efficient end use devices, distributed energy sources, advanced communication system and smart home systems. All these new technologies will effect on future energy consumption and maximum power. Some of smart technologies improve energy efficiency and increase savings, reduce peak load and load demand.
Most of existing forecasting principles use statistical forecasting of load growth and changes of load types for decision. We can only create different scenarios of their future implementation and evaluate how it could effect on energy consumption and future loads. We can use load models to calculate future electricity consumption and peak loads.
near future. There is a variety of technologies based on
smart grid, smart end-use devices, distributed energy resources, highly advanced control and communication systems. All these technologies will influence on the future electricity load demand. Some of them improve energy efficiency; increase permanent energy savings, permanent demand reduction and temporary peak load reductions but some new electricity loads will appear in the future and they will increase electrical energy consumption.
There are four main types of smart grid technologies. a) Smart distributed energy resources.
b) Advanced whole-building control systems.
c) Integrated communications architecture.
These technologies allow to create a highly energy – efficient and automated system, to optimize its operation based on utility’s and consumer’s requirements and constraints and other variables.
Long term forecasting of electrical loads is the first
step in planning and development of generation, transmission and distribution system. Based on the results of such forecasts, companies co-ordinate their resources to meet the forecasted demand. Long term load forests are made for a period up to several decades. They are very important in planning and help to make an economical based decision of buying and installing new equipment and facilities to meet the consumers future electric demand and to provide an acceptable level of power quality and reliability. Methods used for long-term forecasting can be classified into two categories: Parametric methods and artificial intelligence based methods. Parametric load forecasting methods are based on statistical techniques and historical data of loads and factors affecting on loads. The parametric methods are described below: Trend analysis, end use modeling and economic modeling. The artificial intelligence methods are
IJSER © 2014 http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014 50
ISSN 2229-5518
classified into neural networks, genetic algorithms, wavelet networks and fuzzy logics methods.
As it was already said, parametric methods take into
account historical data of loads, factors affecting on load demand and with the use of statistic make a long term forecast.
Trend analysis focuses on information about changes
in electricity demand in past and based on this information predict changes in electricity demand by extending past rates of electricity demand in to the future.
4.2 End use models
The end-use approach makes load forecasts based on
the statistical information about customers along with dynamics of charge of loads. These models use extensive information on end users, for instance, customer size, size of houses and so on. End use models divide customers for several sections (Residential, commercial, and industrial) and they are based on the principle that electricity demand is derived from customer’s demand for light, heating, cooling etc.
The econometric approach estimates the relationship
between energy consumption and factors influencing consumption by the least square method or time series methods. To forecast electricity demand it uses economic theory and statistical techniques.
In the econometric models the dependent variable, in our case it is a demand for electricity, is expressed as a function of various parameters such as population income per capita or value added or output, price of power, prices or alternative fuels, proxies for penetration of appliances/equipment etc.
Thus we have a formula for electricity demand.
ED = f (Y, Pi, Pj, POP, T). Where,
ED is electricity demand, Y is output or income.
Pi is own price
Pj is price related to fuels.
POP population and T is technology
The advantage of using econometric models in load
forecasting is that a researched cannot only forecast an electricity demand, but also explain its behavior in the future, why is it increasing or not and what factors and how effect on it. A disadvantage of using econometric models in load forecasting is that during the
forecasting period the changes in electricity remain the same as in the past in order to get an accurate forecast.
Nowadays artificial intelligence methods became very
popular and these are also used in load forecasting. Some kinds of them are described in this section.
Methods based on artificial neural networks can solve various power system problems, such as design,
planning, control protection, security analysis, fault diagnosis and load forecasting. Artificial neural networks are very popular in load forecasting because
of their ability in mapping complex non-linear relationships.
In most cases ANN’S are used in short-term load
forecasting, but they are also used in long term forecasting.
To design a neural network it is needed to choose its architecture, method of training and also type, size and number of used neural.
The output of ANN is
n
Yi = ∑WiXi
i =1
Where Xi - is input
Wi – weight of network
Yi – one of the outputs. i = 1, 2 ……n
Types of ANN’S that can be used for long-term
forecasting are recurrent neural network (for peak load forecasting) and feed forward back probation (to forecast annual peak load).
Wavelet theory received wide attention in electrical
load forecasting. It has an improved forecasting accuracy comparing tradition load forecasting methods.
This theory provides powerful and flexible tool to decompose load data in to different frequency components, component and its characteristics
Genetic algorithm (GA) is a numerical optimization technique. It is a very popular optimization tool and
helps to solve problems in engineering, finance, economics science etc.
Fuzzy control system is a decision mechanism represented by a set of fuzzy rules. The fuzzy logic
model provides and algorithm to convert strategy based on expert knowledge into an automatic strategy.
IJSER © 2014 http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014 51
ISSN 2229-5518
[3] Tao Hong, Mingui and Mesut E. Baran, Modelling and Forecasting Hourly Electric Load by Muttiple Linear Regression with
interaction, Power and Energy Society
General Meeting, 2010 IEEE, 1-8.
[4] Nathan Kowahl and Anthony Kuh, Microscale
Smart Grid Optimization, IEEE-The 2010
International Joint Conference Neural
Networks (IJCNN), 1-8.
[5] Z.A. Vale, H. Morais, H. Khodr, Bruno
Carizes and Joao Soares, Technical and economic resources management in smart
A skilled human operator can be replaced with a fuzzy
rule based system. Fuzzy logic models can be combined with neural network to train ANN. This combined system provides better load forecasting result. It combines advantages of both methods and generalization capability of artificial neural network and the ability of fuzzy logic model to use forecaster’s knowledge and experience. This method gives quite accurate load forecasts.
Fig. 1. Schematic structure of ANN and fuzzy logic methods
Thus, there are a lot of various long-term load
forecasting methods existing. They are based on statistics, historical data, data on customers, expert knowledge and experience, artificial intelligence methods etc.
They all have advantages and disadvantages and for better forecasting result a combination of several methods can be used. In our case we need to forecast load behavior when new types of load coil appear and new technologies will be implemented. We can model behavior of different types of loads, different scenarios of existing equipment and technologies future usage load. We can also use trend analysis forest future usage
[1] Hesham K. Alfares and mohammad
Nazeeruddin, Electrical Load Forecasting Literature Survey and Classification of Methods, International Journal of Systems Science 33 (1), 2002, 23-24.
[2] Mahnoosh Alizadeh, Ama Scaglione and Zhifang Wang” on the impact of Smart Grid metering infrastructure load forecasting” (
2010 ) . Sept 29 – Oct 1, 2010.
grids using Heuristic optimization methods,
Power and Energy Society General Meeting,
2010 IEEE, 25-29 July 2010, 1-7.
[6] Z.A. Bashir and M.E. Ej. Hawary, Applying Walvelets to short-term forecasting using PSO-Based Neural Networks, IEEE Transactions on Power Systems, Vol. 24, No.1, Feb. 2009.
[7] Galina Baglaeva “ Load forcasting in Smart Grid environment “ Masters Thesis 2011. Lappeviranta University of Technology.
[8] QUN Zhou Weiguan and Weisun, Impact of Demand Response Contracts on Load Forecasting in a Smart Grid Environment, Power and Energy Society General Meeting,
2010 IEEE, 25-29 July 2010, 1-7.
[9] Wei Wei, Qiarg Yin, Yang Bai, Regional Load Forecasting M ethod based on the S-carve power Business expansion, TENCON2013-
2013 IEEE Region Conference, 22-25 oct.
2013, 1-4.
[10] Yannig Goude, Raphael Nedellec and Nicolas
Kong, Locel Short and middle Term Electricity Load forecasting with Semi- parametric Additive Models, IEEE Transactions on Smart Grid, Vol. 5, No.1, Jan-2014, 440-446.
[11] Ashton Webberley, David Wenzhong Gao “
Study of Artificial Nectial Network based short term Load forecasting” Power and Energy Society General Meeting (PES), 21-25
July 2013, 1-4.
[12] Amit Jain, m Baita Jain, Fuzzy modeling and Similarity based short term load Forecasting using Evolutionary Particle Swarm Optimization, Power and Energy Society General Meeting, 2012 IEEE, 22-26 July
2012, 1.
[13] Simone Paoletti, Amdrea Garulli, Antorio Vicino, Electrical Load Forecasting in Presence of active demand, 51st IEEE Conference on Decision and Control, Maui, Hawaii, USA, 10-13 Dec. 2012.
IJSER © 2014 http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014 52
ISSN 2229-5518