Conference Paper

Very short-term electricity load demand forecasting using support vector regression.

DOI: 10.1109/IJCNN.2009.5179063 Conference: International Joint Conference on Neural Networks, IJCNN 2009, Atlanta, Georgia, USA, 14-19 June 2009
Source: DBLP

ABSTRACT In this paper, we present a new approach for very short term electricity load demand forecasting. In particular, we apply support vector regression to predict the load demand every 5 minutes based on historical data from the Australian electricity operator NEMMCO for 2006-2008. The results show that support vector regression is a very promising approach, outperforming backpropagation neural networks, which is the most popular prediction model used by both industry forecasters and researchers. However, it is interesting to note that support vector regression gives similar results to the simpler linear regression and least means squares models. We also discuss the performance of four different feature sets with these prediction models and the application of a correlation-based sub-set feature selection method.

0 Bookmarks
 · 
46 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: Through combining Information and Communication Technologies (ICT), advanced instrumentation, system intelligence, and information on the end user, the smart grid will increase building energy efficiency and conservation. Demand Side Management (DSM) programs serve as an aid in energy conservation and management strategies as well as in the collection of real-time consumption information data. Specifically, as proposed in this paper, real-time, fine-grain consumption data at the point of use in buildings can be used by building energy managers, utilities, and the end user for planning, load forecasting, and feedback for providing information that may lead to end user behavior change. This paper focuses on the development of node-level very short time load forecasting (NL-VSTLF), and specifically on a framework for the utilization of real-time data at the point of use for advancing research in the area of building-very short term load forecasting. This proposed framework will be the foundation for variable one minute to hourly and daily load forecasting using an aggregate of all active nodes in real-time.
    Innovations in Information Technology (IIT), 2013 9th International Conference on; 01/2013
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Forecasting electricity prices plays a significant role in making optimal scheduling decisions in competitive electricity markets. Predominantly, price forecasting is performed from a “point forecasting” perspective, i.e., forecasting the exact values of future prices. However, in some applications, such as demand-side management, operation decisions are made based on certain price thresholds. It is, hence, desirable to obtain the “classes” of future prices, which can be cast as an electricity price classification problem. In this paper, we investigate the application and effectiveness of several data mining approaches for electricity market price classification. In addition, we propose a new data model for forming the initial data set for price classification. Simulation results for New York, Ontario, and Alberta electricity market prices are provided. Finally, the application of the generated numerical results to a demand-side management case study is demonstrated.
    IEEE Transactions on Smart Grid 01/2012; 3(2):808-817.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Very short-term load forecasting predicts the loads in electric power system one hour into the future in 5-min steps in a moving window manner. To quantify forecasting accuracy in real-time, the prediction interval estimates should also be produced online. Effective predictions with good prediction intervals are important for resource dispatch and area generation control, and help power market participants make prudent decisions. We previously presented a two level wavelet neural network method based on back propagation without estimating prediction intervals. This paper extends the previous work by using hybrid Kalman filters to produce forecasting with prediction interval estimates online. Based on data analysis, a neural network trained by an extended Kalman filter is used for the low-low frequency component to capture the near-linear relationship between the input load component and the output measurement, while neural networks trained by unscented Kalman filters are used for low-high and high frequency components to capture their nonlinear relationships. The overall variance estimate is then derived and evaluated for prediction interval estimation. Testing results demonstrate the effectiveness of hybrid Kalman filters for capturing different features of load components, and the accuracy of the overall variance estimate derived based on a data set from ISO New England.
    IEEE Transactions on Power Systems 01/2013; 28(4):3806-3817. · 2.92 Impact Factor