Shu Fan

Short-term load forecasting based on an adaptive hybrid method

This paper aims to develop a load forecasting method for short-term load forecasting, based on an adaptive two-stage hybrid network with self-organized map (SOM) and support vector machine (SVM). In the first stage, a SOM network is applied to cluster the input data set into several subsets in an unsupervised manner. Then, groups of 24 SVMs for the next days load profile are used to fit the training data of each subset in the second stage in a supervised way. The proposed structure is robust with different data types and can deal well with the nonstationarity of load series. In particular, our method has the ability to adapt to different models automatically for the regular days and anomalous days at the same time. With the trained network, we can straightforwardly predict the next-day hourly electricity load. To confirm the effectiveness, the proposed model has been trained and tested on the data of the historical energy load from New York Independent System Operator.

Short-Term Load Forecasting Based on a Semi-Parametric Additive Model

Short-term load forecasting is an essential instrument in power system planning, operation, and control. Many operating decisions are based on load forecasts, such as dispatch scheduling of generating capacity, reliability analysis, and maintenance planning for the generators. Overestimation of electricity demand will cause a conservative operation, which leads to the start-up of too many units or excessive energy purchase, thereby supplying an unnecessary level of reserve. On the other hand, underestimation may result in a risky operation, with insufficient preparation of spinning reserve, causing the system to operate in a vulnerable region to the disturbance. In this paper, semi-parametric additive models are proposed to estimate the relationships between demand and the driver variables. Specifically, the inputs for these models are calendar variables, lagged actual demand observations, and historical and forecast temperature traces for one or more sites in the target power system. In addition to point forecasts, prediction intervals are also estimated using a modified bootstrap method suitable for the complex seasonality seen in electricity demand data. The proposed methodology has been used to forecast the half-hourly electricity demand for up to seven days ahead for power systems in the Australian National Electricity Market. The performance of the methodology is validated via out-of-sample experiments with real data from the power system, as well as through on-site implementation by the system operator.

Forecasting the Wind Generation Using a Two-Stage Network Based on Meteorological Information

This paper proposes a practical and effective model for the generation forecasting of a wind farm with an emphasis on its scheduling and trading in a wholesale electricity market. A novel forecasting model is developed based on indepth investigations of meteorological information. This model adopts a two-stage hybrid network with Bayesian clustering by dynamics and support vector regression. The proposed structure is robust with different input data types and can deal with the nonstationarity of wind speed and generation series well. Once the network is trained, we can straightforward predict the 48-h ahead wind power generation. To demonstrate the effectiveness, the model is applied and tested on a 74-MW wind farm located in the southwest Oklahoma of the United States.

Density Forecasting for Long-Term Peak Electricity Demand

Long-term electricity demand forecasting plays an important role in planning for future generation facilities and transmission augmentation. In a long-term context, planners must adopt a probabilistic view of potential peak demand levels. Therefore density forecasts (providing estimates of the full probability distributions of the possible future values of the demand) are more helpful than point forecasts, and are necessary for utilities to evaluate and hedge the financial risk accrued by demand variability and forecasting uncertainty. This paper proposes a new methodology to forecast the density of long-term peak electricity demand. Peak electricity demand in a given season is subject to a range of uncertainties, including underlying population growth, changing technology, economic conditions, prevailing weather conditions (and the timing of those conditions), as well as the general randomness inherent in individual usage. It is also subject to some known calendar effects due to the time of day, day of week, time of year, and public holidays. A comprehensive forecasting solution is described in this paper. First, semi-parametric additive models are used to estimate the relationships between demand and the driver variables, including temperatures, calendar effects and some demographic and economic variables. Then the demand distributions are forecasted by using a mixture of temperature simulation, assumed future economic scenarios, and residual bootstrapping. The temperature simulation is implemented through a new seasonal bootstrapping method with variable blocks. The proposed methodology has been used to forecast the probability distribution of annual and weekly peak electricity demand for South Australia since 2007. The performance of the methodology is evaluated by comparing the forecast results with the actual demand of the summer 2007-2008.

Short-Term Load Forecasting Using Comprehensive Combination Based on Multimeteorological Information

Short-term load forecasting is always a popular topic in the electric power industry because of its essentiality in energy system planning and operation. In the deregulated power system, an improvement of a few percentages in the prediction accuracy would bring benefits worth of millions of dollars, which makes load forecasting become more important than ever before. This paper focuses on the short-term load forecasting for a power system in the U.S., where several alternative meteorological forecasts are available from different commercial weather services. To effectively take advantage of the alternative meteorological predictions in the load forecasting system, a new comprehensive forecasting methodology has been proposed in this paper. Specifically, combining forecasting using adaptive coefficients is applied to share the strength of the different temperature forecasts in the first stage, and then, ensemble neural networks have been used to improve the models generalization performance based on bagging. The proposed load forecasting system has been verified by using the real data from the utility. A range of comparisons with different forecasting models have been conducted. The forecasting results demonstrate the superiority of the proposed methodology.

Multiregion Load Forecasting for System With Large Geographical Area

In a power system covering a large geographical area, a single model for load forecasting of the entire area sometimes cannot guarantee satisfactory forecasting accuracy. One of the major reasons is because of the load diversity, usually caused by weather diversity, throughout the area. Multiregion load forecasting will be a feasible and effective solution to generate more accurate forecasting results, as well as provide regional forecasts for the utilities. However, a major challenge is how to optimally partition/merge the areas according to the regional load and weather conditions. This paper investigates the electricity demand and weather data from an electric utility in Midwest, U.S. Based on the data analysis, we demonstrate the existence of weather and load diversity within its control area and then develop a short-term multiregion load forecasting system based on support vector regression for day-ahead operation and market. The proposed multiregion forecasting system can find the optimal region partition under diverse weather and load conditions and finally achieve more accurate forecasts for aggregated system load. The proposed forecasting system has been tested by using the real data from the system. The numerical results obtained for different region partition schemes validate the effectiveness of the proposed multiregion forecasting system. The detailed discussions on the forecasting results have also been given in this paper.

Probabilistic short-term wind power forecast using componential Sparse Bayesian Learning

A practical approach for the probabilistic short-term generation forecast of a wind farm is proposed in this paper. Compared with deterministic wind generation forecast, probabilistic wind generation forecast can provide important wind generation distribution information for operation, trading, and some other applications. The proposed approach is based on a sparse Bayesian learning (SBL) algorithm, which produces probabilistic forecast results by estimating the probabilistic density of the weights of Gaussian kernel functions. Furthermore, since the wind generation time series exhibits strong nonstationary property, a componential forecast strategy is used to improve the forecast accuracy. According to the strategy, the wind generation series is decomposed into several more predictable series by discrete wavelet transform, and then, the resulted series are forecasted using the SBL algorithm. To fulfill multilook-ahead wind generation forecast, a multi-SBL forecast model is constructed in the context. Tests on a 74-MW wind farm located in southwest Oklahoma demonstrate the effectiveness of the proposed approach.

Peak load forecasting using the self-organizing map

This paper aims to study the short-term load forecasting of electricity by using an extended self-organizing map. We first adopt a traditional Kohonen self-organizing map (SOM) to learn time-series load data with weather information as parameters. Then, in order to improve the accuracy of the prediction, an extension of SOM algorithm based on error-correction learning rule is used, and the estimation of the peak load is achieved by averaging the output of all the neurons. Finally, as an implementation example, data of electricity demand from New York Independent System Operator (ISO) are used to verify the effectiveness of the learning and prediction for the proposed methods.

Comparative study on load forecasting technologies for different geographical distributed loads

For a power system covering large geographical area, a single forecasting model for the entire region cannot guarantee the satisfactory forecasting accuracy. One of the major reasons is because the load diversity and weather diversity throughout the region. For such a system, multi-region load forecasting will be a feasible and effective solution to generate more accurate forecasting results. However, some technical issues arise when performing the multi-region load forecasting, the major challenge is how to optimally partition/combine the regions to achieve better forecasting results, especially under transient weather conditions. On the other hand, load forecasting for small areas, especially for a distribution feeder or micro grid, is also difficult because load variation in local areas is larger than that of a large system. In addition, the correlation between weather variables and small area loads would be unstable. Therefore, a two-stage load forecasting module could be utilized to improve the forecasting accuracy, and the risk assessment of local load forecasting uncertainty could be studied. This paper discusses respectively a large geographical load forecasting in Midwest US and a small area load forecasting in a UK distribution feeder. For the load forecasting at the large geographical area, a multi-region forecasting system that can find the optimal region partition in both stationary and transient weather and load conditions is discussed. For the load forecasting at the small feeder, a two-stage combination module is discussed; furthermore, risk evaluation technologies based on time-domain and frequency-domain methods are also proposed to assess the uncertainty of load forecasting.

An Integrated Machine Learning Model for Day-Ahead Electricity Price Forecasting

This paper proposes a novel model for short-term electricity price forecasting based on an integration of two machine learning technologies: Bayesian clustering by dynamics (BCD) and support vector machine (SVM). The proposed forecasting system adopts an integrated architecture. Firstly, a BCD classifier is applied to cluster the input data set into several subsets in an unsupervised manner. Then, groups of 24 SVMs for the next days electricity price profile are used to fit the training data of each subset in a supervised way. To demonstrate the effectiveness, the proposed model has been trained and tested on the data of the historical energy prices from the New England electricity market