Mar 26, 2016
Comparison of Various Methods Used In Short Term Load Forecasting
Short term load forecasting (STLF) has become very important in the de-regulated environment. This point is highlighted by the fact that every year hundreds of papers are published on this topic which discuss or propose new methods and insights for STLF. You can also Subscribe to FINAL YEAR PROJECT'S by Email for more such projects and seminar.
Short term load forecasting has become very important in modern times due to the deregulation of the power sector. Private players have come in the fields of generation, transmission and distribution of electricity.
These organizations or companies are not charitable organizations and strive for maximizing their profits. One challenge before these companies is to ensure that a customer, who now has power connections from more than one provider, draws power from them only. Also to even out the load curves these organizations need to find out the valleys and the peaks in the demand and take subsequent load management steps to fill the valleys and reduce the peaks.
Recommended Project: Generation of Electricity from Wind Power
The importance of good Short Term Load Forecast (SLTF) can never be emphasized more in modern times and this is one of the pivotal reasons for existence of a large number of methods for STLF.
In this project, author has made a comparative study of these methods in this B.Tech dissertation and hence chose the following methods-
1. Feedforward Neural Networks
2. Radial Basis Function Networks
3. Discounted Multiple Regression
4. Adaptive Neuro Fuzzy Inference Systems (ANFIS)
Below are some of the conclusion of this project for your reference.
i) Forecasts rely heavily on the data and thus a detailed analysis of the data must be done prior to selection of the method for STLF. We should try to find out whether some correlation exists between the load profiles of different days of the week. If the correlation is high then day indices, if included, will not have much effect on the results. However, including them may actually give poor forecasts as the complexity of model will increase.
ii) Feed forward neural networks must be used with better learning algorithms because gradient descent method is very slow and gets stuck in local minima. Hence any second order learning algorithm should be used or validation procedures must be incorporated during the training time.
iii) Radial basis function networks are much better than feed-forward neural networks and gave much better results. They gave better testing results due to inherent generalization. However, caution must be observed in choosing the number of clusters since too many clusters make generalization poor and also a possibility of empty cluster formation exists.
There are few more conclusion which can be found in this project report on "comparison of various methods used in short term load forecasting". There are few limitations also which is discussed in this project. Use it for your study purpose and all thanks to author of this project.
Author:- Mr. Amandeep Singh, Mr. Divyanshu Negi, Mr. Himanshu Jain and Mr. Kritiman Bhattacharya
About : Final-yearproject.com
Final-yearproject.com is one stop for all your project needs. You can find projects and seminars related to Electronics, Electrical, Civil, Mechanical, MBA, Pharmacy, Computer Science, IT, Bio-Tech, Robotics, Telecom, Wireless, Automation, BCA, MATLAB etc. You can Like us at Facebook and follow us on Twitter and Google+.