Improving the Accuracy of Short-term Forecasting of Electrical Loads Taking into Account Meteorological Factors Based on Support Vectors Method

声明

Abstract

摘要

Contents

1．Review and analysis of modem methods and mathematical models to predict electricity consumption

1．1 Classification of short-term load forecasting methods

1．2 Statistical methods of forecasting

1．2．1 Methods for regression

1．2．2 Time series methods

1．2．3 Methods based On wavelet transform of time series

1．3 Methods of artificial intelligence

1．3．1 Methods based on neural network models

1．3．3 Support vector method

1．4 Evolutionary algorithms

1．5 Requirements for short-term forecasting of electricity consumption

1．6．1 Accuracy of the input-output relationship hypothesis

1．6．2 Prediction of abnormal days

1．6．3 Inaccurate weather forecast data

1．7．1 Models of neural networks

1．7．2 Models of neuro-fuzzy networks

1．7．3 Model of wavelet transform

1．7．4 Regression models

1．8 Conclusions

2．Time series analysis of electricity consumption and its determinants

2．1 Characteristics of the electrical load diagrams of the power system

2．2 Time series of power consumption and influencing factors

2．3 Seasonal and meteorological factors affecting power consumption

2．4 Temperature and light：the analysis of their impact on power consumption in the control room operating area

2．5 Random disturbances

2．6 Conclusions

3．Modelling short term future energy consumption based on neuralnetworks and evolutionary algorithms

3．1 Short-term load forecasting using artificial neural network

3．2 Short-term load forecasting using artificial neural networks and particle swarm optimization algorithm

3．3 Short-term load forecasting using artificial neural networks and particle swarm optimization algorithm

3．3．1 Data analysis and pre-processing

3．3．2 The number of layers，neurons and transfer functions

3．3．3 Training of built neural networks

3．3．4 Architecture of the ANN for the operating zone

3．3．5 The choice of input variables

3．3．6 Building the structure of neural network

3．3．7 Selection of data for training，testing and validation

3．3．8 Simulation results

3．4 Training the ANN on the basis of self-organization

3．4．1 Dataset for the study

3．4．2 Training of self-organizing maps

3．4．3 The results of clustering and prediction

3．4．4 Performance criteria

3．4．5 Simulation results

3．5 Conclusions

4． Models of future energy consumption based on neural fuzzy network and support vector method

4．1 Predicting power consumption using adaptive neural fuzzy network

4．1．1 The architecture of neuro-fuzzy model

4．1．2 Hybrid algorithm for training neural networks

4．1．3 Simulation result

4．2 Energy consumption forecasting using support vector

4．2．1 Simulation results

4．3 Forecasting of power consumption based on the support vector method and particle swarm algorithm

4．3．1 Load forecasting steps and processes

4．3．2 A set of analysis parameters

4．3．3 Simulation results

4．4 Conclusions

Summarize

Acknowledgement

References

Research achievement during working for the degree