In many domains, e.g. in networks for power generation and distribution, heating and ventilation of buildings, and in the processing industries, the area with the highest potential for economic and ecologic impacts of monitoring and control is the system‐wide coordination layer. In power distribution systems, this is also the area with the most pressing needs due to the increasing amount of stochastically varying generation.
In WP3, we aim at coordinating research efforts to devise novel automatic decision-making algorithms for spatially distributed and interconnected systems, with a focus on large-scale energy systems and large processing plants. In particular, we envision control strategies based on mathematical programming tools that:
- - optimize operational efficiency under resource and safety constraints,
- - ensure robustness to component/connectivity failures and to stochastic variations of problem data,
- - take into account that the communication between the subsystems is possibly not ideal and in particular may fail for certain periods of time in which the integrity of the system nonetheless has to be guaranteed.
Fig. 1: Need for system-wide coordination and control in large-scale distributed systems.
The proposed research activities address several strategies for optimal decision-making in large systems:
- - Hierarchical structures where a supervisory level computes decisions in a centralized manner at lower sampling rates than used on the lower layers. The models used on the higher level can be richer (in some aspects) than the local models, as it is currently the case in chemical process control, but they can also be more abstract, e.g. discrete event models, possibly stochastic. The supervisory layer decides based upon reduced information collected from the lower layers. The supervisor operating at the higher hierarchical level is assumed to have larger computation power and sampling time, so that it can handle relatively complex decision algorithms. These can explicitly take into account global performance, global constraints on available shared resources (power, budget, work-space, link capacity, etc.), stochastic data, and robustness issues. The lower level consists of simpler decision algorithms that are executed in a distributed manner, and reiterated after an exchange of information (Task 3.1)
- - Distributed structures where decentralized unit controllers exchange information on the variables of the subsystems that they control and on their planned actions in a fashion that leads to global stability and if possible also to global optimality (Task 3.2)
- - Price-based co-ordination mechanisms where the subsystems are controlled locally but the competition for resources or the coordination of production is achieved by market-alike mechanisms where the prices for the resources resp. the sales prices are increased if the demand is high such that the coordinated system achieves the optimum state after a number of bidding rounds (Task 3.3)
These approaches will at first be investigated and applied independently to the application examples mentioned in application domain AD2. At the end of HYCON 2, a comprehensive evaluation report will be produced where, if this turns out to be feasible, the pros and cons of the approaches are discussed and demonstrated for one of the application examples. Moreover, in this WP we will foster the coordination and exchange between on-going STREPs and investigate future challenges for theoretical research and potential application areas in AD1 and AD3.
Fig. 2: Structure of WP3.