Accurate and affordable forecasting is essential for effective and widespread utilization of renewable energy sources such as wind power and solar energy. The objective of this research is to develop a Forecast-as-a-service (FaaS) framework that (1) generates accurate forecasts that integrate prediction results from different models using data from different sources; and (2) provides on-demand delivery of different types of forecasts at different levels of details for different prices. Forecasting activities and business workflows are modeled by applying the concept of services, composite services, and the principles of service-oriented architecture. Implemented by using the Azure platform, the FaaS framework involves the orchestration of service activities performed by a Forecast Generator Framework, an Internal Data Retrieval Framework and an External Data Collection Framework. Potential users of the FaaS framework include renewable energy users and providers, power system operators, and potential renewable energy users/producers that are at different stages of planning for new facilities. The FaaS framework can contribute to the filling of two national needs ? energy independence and environmental stewardship. It can help power companies in many states to meet their respective state mandates in renewable portfolio standards. The FaaS framework can become a part of a national forecast cyber-infrastructure.
The goal of this project is to develop the cloud-based Forecast-as-a-Service (FaaS) framework that enables the use of different types of data from different sources to generate different kinds of renewable energy forecasts on demand at different levels of details for different prices. This goal has been achieved. The FaaS framework has been developed by using the Azure cloud computing platform provided by Microsoft. It provides on-demand forecasts of solar or wind power at locations specified by the users. Forecasts can be long-term forecasts useful for prospecting or planning by potential investors, or short-term forecasts suitable for operational decision making by operators of existing facilities. This project shows that the costs of prospecting forecasts are in the range of $60-70 while the costs for operational forecasts are in the range of $10-20. Additional services such as uncertainty quantification can be requested for additional fees in the order of a few dollars. The FaaS framework provides a more flexible and affordable alternative to the subscription model provided by current forecast service providers. An intellectual merit of this project is that the FaaS framework demonstrates a way to address both the technical and economic aspects of forecasting. The FaaS framework has been developed based on the service oriented architecture (SOA). Software entities called services and composite services have been developed and organized in different ways to implement the workflow processes for different kinds of forecasts. In the FaaS framework, a software service has two endpoints - one for the technical aspect and the other for the economic aspect. Based on this design, the same workflow can be developed to handle both the technical aspect and the economic aspect of forecasting. This finding is important because it provides the basis for the future development of multidisciplinary services that have impact in several disciplines simultaneously. Multidisciplinary SOA applications, especially in the context of cloud computing, will expand the full potential of SOA and cloud computing. Another intellectual merit is that the concept of the probability of persistence (POP) has been developed so that the quality of solar irradiance can be forecasted together with the quantity. POP represents the chance that the solar irradiance received at the ground level remains within a certain range for a certain time period. As an indicator for the extent of fluctuation, POP is useful for the design and operation of the backup systems for renewable energy sources. Using POP and the sky clearness index, a new classification method for daily sky conditions has been developed. Previously unknown patterns of daily sky conditions haven been observed and the understanding of how to forecast solar power based on the forecasts of sky conditions has been deepened. Another intellectual merit is that this project advances the knowledge of how to utilize cloud computing to automate the processes that extract information from large volumes of historical data. Developing the statistical forecasting model used in the FaaS framework and discovering concepts such as POP were data intensive processes. Historical data useful for model development had been archived in certain formats that could be decoded only by using special software. The process to extract information from data collected over a number of years could take weeks or months. Using the computing resources available from the cloud, this process has been automated and results could be obtained in hours, at most days. The FaaS framework could bring broad impact to society by enhancing more widespread and effective utilization of renewable energy. A major drawback of solar and wind power is their intermittent and fluctuating nature. This drawback can be countered by making appropriate preparation based on accurate forecast to accommodate the intermittency and fluctuation. Availability of affordable and Internet-accessible forecasts such as those provided by the FaaS framework would not only enable renewable energy to be used more effectively but also motivate more people to consider the use of renewable energy. The FaaS framework is especially meaningful to individuals and small companies that lack the computing resources to obtain forecast information that are pertinent to their particular situations. By design, the FaaS framework can function both as a SaaS (Software-as-a-Service) and as a PaaS (Platform-as-a-Service). When it is used as a SaaS, a user can utilize the developed features to specify and obtain forecast information. When it is used as a PaaS, a user may develop a new application by reusing, revising or reorganizing some of the developed features to work with the newly developed features. Forecast researchers in different areas can use the FaaS framework as a research tool to test out new ideas. The work accomplished in this project has the potential to be developed into a national cyber-infrastructure for forecasting activities, especially in the area of renewable energy.
