Please use this identifier to cite or link to this item: http://krishikosh.egranth.ac.in/handle/1/5810143063
Authors: Singh, Anamika
Advisor: SRIVASTAVA, DR. MANISH KUMAR
Title: ELECTRIC FORECASTING OPTIMIZATION BASED ON ARTIFICIAL INTELLIGENCE TECHNIQUES
Publisher: DEPARTMENT OF ELECTRICAL ENGINEERING VAUGH INSTITUTE OF AGRICULTURAL ENGINEERING AND TECHNOLOGY SAM HIGGINBOTTOM UNIVERSITY OF AGRICULTURE, TECHNOLOGY AND SCIENCES
Language: en
Type: Thesis
Pages: 135p.
Agrotags: null
Keywords: Keywords: Short-Term Electric Forecasting, Power System Planning, Artificial Neural Network, Genetic Algorithm and Particle Swarm Optimization
Abstract: Recently, utilization of nonlinear gadgets like power electronics, continuous power supplies, flexible speed drives, and delicate loads like personal computers, etc has expanded. It is seen that nonlinearity in the electric load profile increases with the use of these devices. Therefore, an accurate load forecasting is required to improve quality and quantity of power services. A significant truth about the power is that it cannot be stored for quite a long time in AC form; it is conceivable to store it in DC form, but it is restricted to a less amount comparing to demand and that too at an extreme high cost. Therefore, an accurate load forecasting is required. Lower accuracy level can be accomplished by utilizing any conventional technique however for higher accuracy; improved models are to be created. Therefore, the need for accurate and robust load forecasting model is evident in the current scenario of non linear electric load profile forecasting. Electric forecasting (EF) is an important tool for power system operation, planning, and control for decisions such as load management, generation scheduling, and system security assessment, etc. Most of the research is performed for short-term electric forecasting (STEF). It shows the importance of the STEF. In the literature, several robust and accurate forecasting models were developed such as auto-regressive, autoregressive integrated moving average and moving average and found capable of forecasting stationary time–series data but real-time series is never stationary. These models were failed to provide the desired level of accuracy with the nonlinearity present in electric load profile. Therefore, time-series models are not suitable for accurate shortterm load forecasting. STEF is related to operational tasks such as economic dispatch, fuel arrangement, load scheduling, etc. Thus, it becomes necessary to develop forecasting models with enhanced accuracy. The application of Artificial Intelligence (AI) techniques has been explored to solve the above problem. Additionally, various papers for electric forecasting exhibit that Artificial Neural Network (ANN) has the capability of learning the nonlinear behavior and ability to generalize. Other advantages of an ANN are parallel data processing, adaptability, fault tolerant, etc. Therefore, ANN-based models can forecast electricity generation, load with higher accuracy. Feed-Forward Neural Networks (FFNN) is commonly used architecture of ANN. The result of FFNN has been analyzed and vii compared for accuracy. Result shows that accuracy of forecasted models is not as desired, learning rate is slow and time consuming. In order to remove above problems, ANN based forecasting models is optimized by two optimization tools, viz., Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). These optimized models provide more accurate forecasting results. The experimentation shows that PSO algorithms better than GA for optimizing these models. This way this thesis fulfills its aim to develop an improved, modern STEF model with reduced complexity.
Description: Ph. D. Thesis
Subject: Others
Theme: ELECTRIC FORECASTING OPTIMIZATION BASED ON ARTIFICIAL INTELLIGENCE TECHNIQUES
Research Problem: ELECTRIC FORECASTING OPTIMIZATION BASED ON ARTIFICIAL INTELLIGENCE TECHNIQUES
These Type: Ph.D
Issue Date: 2020
Appears in Collections:Thesis

Files in This Item:
File Description SizeFormat 
ANAMIKA SINGH 16PHEE102.pdf2.75 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.