Learnings

Outcomes

1. Developed and tested performance improvement options for dispatch optimisation algorithm including 
   >Algorithm warm-starts and heuristics 
   >Enabling tighter formulations 
   >Tuning algorithm parameters 
2. An Explanations for Advice methodology (EfA) which enables analyzing and providing reasons for the output of the dispatch algorithm. The EfA module uses knowledge of the raw and processed unit data, the cconstraints that define the optimisation problem, to provide a reason why specific unit MW movements have been made. 
3. A Headroom and Footroom Interrogation (HaF) methodology which solves several optimisation scenarios to visualise the generation bounds available to the control room across the dispatch windows at different price limits. The resulting visualisation can help the control room prepare for forecast uncertainty. 

Lessons Learnt

We should anticipate that testing and improving control room operational algorithms is case and context specific. As we have seen in this project, approaches which in principle could offer an improvement do not always bring a significant benefit, or the benefits are realized under specific conditions – in this case in harder to solve cases which might not be easily identified in advance.  
 
The project provided a set of potential improvement options that could be assessed against any optimisation based decision support tool. To conclusively assess whether these options are beneficial, it is recommended to use a representative sample of production cases. 
 
Finally, the Explanations for Advice methodology can form the baseline for explaining the output of more complex problems. However, solutions affected by constraints coupling asset behavior over time can be challenging to fully interpret.