Objective and Background
The high penetration of intermittent renewable energy in electricity grids brings the role of balancing production and demand for energy systems with flexible dispatch. Due to their thermal storage capability, concentrated solar power (CSP) plants provide flexibility between collecting thermal energy from the sun and transforming it into electricity by means of a power cycle.
The dispatchability of CSP plants is based on the electricity demand and weather forecasts, to accurately schedule the production over the following days. Since weather forecasts include uncertainties, the dispatch planning needs to take them into account. Modification of energy deliveries once scheduled is limited and usually associated with financial drawbacks. Therefore, the uncertainty treatment in the delivery schedule is essential to ensure the optimization of dispatch.
As chance is not involved in any future state of the traditional deterministic weather forecasting method, improvements to produce more accurate forecasts have been made . The ensemble prediction system captures the sources of uncertainty by producing ensemble forecasts. With that, uncertainties can be considered to plan improved energy dispatch schedules.
Method and Principal Findings
A dispatch optimization algorithm was developed to derive a CSP plant operation schedule for the day-ahead market. It considers weather and electricity pricing forecasts with a special focus in the incorporation of uncertainty information, by the use of probabilistic weather forecasts.
A partitioned calculation is done between an optimization algorithm and an uncertainty processing. The optimization algorithm runs directly with each weather forecast ensemble member, based on a heuristic rule-based approach, resulting in an ensemble of predicted power schedules. Then, the uncertainty is dealt as a post-processing, using a machine learning system based on a fuzzy decision tree. The UPP analyzes and improves the predicted schedules, leading to the selection of one final delivery profile for the next day (Figure 1).
This partitioned approach brings the possibility of considering not only the accuracy of the weather data but also the market and plant constraints in the uncertainty treatment. For the dispatch problem, uncertainty is of interest during hours when scheduling is desired. Therefore, the use of probabilistic forecasts brings more possibilities of evaluating dispatch scheduling, treating the decision making criteria under uncertainty as a combination of factors.
Discussion and Conclusion
This research shows as benefit the possibility of treating uncertainty as an asset for the development of energy dispatch schedule, by dealing with probabilistic weather forecasts. An improved energy delivery schedule is obtained by an innovative dispatch optimization methodology. That enables the evaluation of weather forecast data quality and its efficiency for optimizing the dispatch.
Recent investigations on operation and financial benefits will be presented. It is expected that further results relating weather forecast accuracy and financial income expected from the CSP plant can be of great value for a better understanding on how to treat uncertainties for energy generation. This enables the participation of such plants in the balancing power market, ensuring the importance of such plants in the pathway to a highly renewable energy mix.