The short-term forecasting of renewable energy output has been demonstrated as a valuable tool for the reduction of costs associated with the management of the non-dispatachable variability of wind and solar generation. This is especially true for utilities that operate small island grid systems with high penetration levels of wind and solar generation and no interconnections to other grid systems. However, the value obtained from the forecasts depends on the degree of alignment between the information provided by the real-time forecast data stream and that required by the operational decision-making process. The value also depends on the attributes of forecast performance that are important for the decision-making process. These two factors are often not considered in the selection or formulation of a forecast solution for grid operation applications. This ultimately results in a reduction in the forecast value that is realized by the user relative to what could have been achieved with a forecast solution that is customized and optimized for a specific application.
This study investigates the benefits that can be obtained through the (1) identification of the attributes of intra-day forecast performance that are important for key operational decisions during a typical daily management cycle, (2) formulation of performance metrics that measure the key forecast attributes and (3) the optimization of the forecast procedure to maximize performance for those attributes. The impact of the forecast customization procedure is assessed by comparing the performance of customized forecasts to a baseline state-of-the-art forecast that is not customized for the targeted applications.
The host system for this project is operated by the Hawaii Electric Light Company (HELCO) on the island of Hawaii (i.e. the Big Island). The HELCO system has a minimum demand of about 90 MW and peak demand of about 180 MW. The system has two wind generation facilities with a total of 31 MW of capacity. In addition, during 2018, there was approximately 90 MW of distributed (“behind-the-meter”) of solar-based generation, which is mostly in the form of roof-top PV systems on residential and commercial buildings.
Three key operational decisions that are sensitive to short-term forecast information were identified in the daily operational cycle.The specific attributes of intra-day forecast information that is important to each combination were identified. This provided the basis for the formulation of customized performance metrics and a target for the forecast system optimization.
The presentation will provide (1) an overview of the HELCO system, (2) a summary of the key daily grid management decisions and associated time frames that are impacted by wind and solar forecasts, (3) a specification of the key forecast performance attributes and customized metrics, (4) the results of a comparison of the performance of standard and application-optimized forecasts as measured by the application-focused metrics and (5) case examples of the relative value of the standard and application-optimized forecasts in operational decision-making scenarios. The results indicate that considerable additional forecast value can be realized in operational decision-making when application-focused performance metrics and forecast optimization is employed.