Windmill Power Management Using Online and Active Learning Approach.
Date of Submission
December 2017
Date of Award
Winter 12-12-2018
Institute Name (Publisher)
Indian Statistical Institute
Document Type
Master's Dissertation
Degree Name
Master of Technology
Subject Name
Computer Science
Department
Electronics and Communication Sciences Unit (ECSU-Kolkata)
Supervisor
Das, Swagatam (ECSU-Kolkata; ISI)
Abstract (Summary of the Work)
Present day we are looking for alternative energy source instead of fossil fuel. Renewable energies which are available in our reach now days are solar, hydro and wind. We are considering wind power generation scenario in this paper. Wind energy is converted to electrical energy with help of windmill at high wind blowing regions. But we need to differentiate between the wind speed which is normal flow and which is storm like violent.Whole process can be divided in two parts. First, collecting labelled training set from huge collection of unlabelled dataset. Second, training with those data.Problem we face is that whole year we receive most usual and normal wind speed reading but storm like wind speed reading is very rare in kind. So it is imbalanced data problem. We know using imbalanced dataset we can get high accuracy but very poor precision and recall of minority class. Our problem is that we should shut our windmill power production when certain wind speed is crossed to save from over power generation and protectingwindmill from high speed rotation. Sudden or abrupt power generation cut off can create problem in power distribution grid. Using weather forecast like NWP(Numerical Weather Prediction) [16] we can predict the probable storm or typhoon like high speed wind flow over larger area but we can not say exact hour of impact. If we get the prediction for individual windmill farm using some reading and shut down the windmill production before some time like 1 hour, 6 hours or 12 hours then we can take decision how to redistribute the power requirement[21].Different statistical approaches[16], [18], like time series analysis and ANN implementation[17],[19],[20] of wind speed prediction for usual operational speed limit are done previously with historical data. But new problem arises when we are doing rare high wind speed prediction.1.2 MotivationIf we are equipped to receive data in batch and data are unlabelled and imbalanced in nature we can use Support Vector like classifier. But using all data points or instances are not necessary all the time. We can reduce the dataset size using active learning like approach. On the other hand, if it is imbalanced data and data are not coming in bulk or batch, we look for predicting event with online learning option. We know Multilayer perceptron, RBF neural net for online leaning. Those have some inherent problems of training and not good performer of imbalanced dataset. In this report, SVM using LIBSVM, Random Vector Functional Link, NORMA, Stochastic Gradient Descent Primal L1-SVM are explored for this issue.1.2 MotivationIf we are equipped to receive data in batch and data are unlabelled and imbalanced in nature we can use Support Vector like classifier. But using all data points or instances are not necessary all the time. We can reduce the dataset size using active learning like approach. On the other hand, if it is imbalanced data and data are not coming in bulk or batch, we look for predicting event with online learning option.
Control Number
ISI-DISS-2017-375
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
DOI
http://dspace.isical.ac.in:8080/jspui/handle/10263/6849
Recommended Citation
Seal, Sankarsan, "Windmill Power Management Using Online and Active Learning Approach." (2018). Master’s Dissertations. 116.
https://digitalcommons.isical.ac.in/masters-dissertations/116
Comments
ProQuest Collection ID: http://gateway.proquest.com/openurl?url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&res_dat=xri:pqm&rft_dat=xri:pqdiss:28843132