![]() ![]() Berrisch et al. (2022) combined generalized additive models and deep ANN to predict high-resolution minimum and maximum peak load given only lower resolution data and weather information. El-Attar et al. (2009) proposed a multivariate load forecasting approach by combining support vector regression with a KNN local prediction framework. For example, El Desouky and Elkateb (2000) developed a hybrid ANN with ARIMA for load forecasting. These approaches can also be combined to improve load forecasting. Examples of advanced nonlinear ML methods are K-nearest neighbors (KNN) ( El-Attar et al., 2009), fuzzy regression models ( Hong and Wang, 2014), support vector machine (SVM) ( Niu et al., 2010), gradient boosting machine (GBM) ( Massaoudi et al., 2021), random forest (RF) ( Cheng et al., 2012 Huang et al., 2016), and artificial neural networks (ANN) ( El Desouky and Elkateb, 2000 Ringwood et al., 2001 Saini, 2008). More advanced nonlinear machine learning (ML) methods have also been proposed for load forecasting with multivariate predictors, including weather, calendar, and economics ( Hong and Fan, 2016). ![]() Traditional methods, such as autoregressive integrated moving average (ARIMA), forecast load with univariate historical load. Predictions of the monthly peak hour can be derived from load forecast that spans the entire month. Therefore, advanced peak day and peak hour forecast methods are critical to maximizing benefits from BESS for peak demand reduction. In addition, due to limited energy capacity, BESS may not be able to discharge at the rated power in all high-load hours. Simply discharging BESS on all high-load days helps capture the peak hour and reduce the coincident demand, but causes unnecessary battery degradation and energy losses associated with charging/discharging. An LSE does not know exactly when the peak hour will occur. Battery energy storage systems (BESS) are promising for peak demand reduction because of their flexibility and instantaneous response capability. Effectively reducing the peak demand leads to significant economic and environment benefits, as well as improved power grid security and stability ( Dai et al., 2021). A capacity charge is paid based on the coincident demand during system peak hours. Many cooperatives, municipally owned utilities, and other types of load serving entities (LSE) purchase power from electricity markets or through power purchase contracts. ![]() On 90% of the peak days, the actual peak hour is among the 2 h with the highest probabilities. The proposed approach is applied to the Duke Energy Progress system and successfully captures 69 peak days out of 72 testing months with a 3% exceedance probability threshold. Guidance is provided on preparation and augmentation of data as well as selection of machine learning models and decision-making thresholds. In this study, we develop a supervised machine learning approach to generate 1) the probability of the next operation day containing the peak hour of the month and 2) the probability of an hour to be the peak hour of the day. Two practical challenges are 1) accurately determining the peak load days and hours and 2) quantifying and reducing uncertainties associated with the forecast in probabilistic risk measures for dispatch decision-making. Pacific Northwest National Laboratory, Richland, WA, United Statesīattery energy storage systems can be used for peak demand reduction in power systems, leading to significant economic benefits. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |