Methods for aggregation and disaggregation

The players in the Nordic power market, i.e. producers, transmission system operators and regulators use computer models to plan for the best possible utilization of the power system. The calculation of the optimal operation strategy for the hydro storages in the system is the most important and complicated computation. Emptying the reservoirs may result in curtailment of electricity and too cautious operation may result in unnecessary spillage, which is a loss to the society. The goal is to find operation strategies for all hydro storages in the Nordic system that gives the best utilization and to simulate the consequences of the operation strategy on e.g. prices and reservoir operation for possible futures (inflows, temperatures and wind power production etc). Moreover, the huge increase in non-controllable renewable production and the stronger coupling to continental Europe cannot be handled properly by existing computation methods. Utilization of the hydro storages may be formulated as a mathematical optimization problem. The large problem size and complexity requires several simplifications to obtain a solution. One of the most important simplifications is aggregation of physical hydro storages and plants into an aggregated equivalent representation. The aggregated hydro model gives more flexibility than the physical system and disaggregation techniques are used to verify that reservoir operation strategies for the aggregated model is feasible for the physical system. The project builds knowledge, improves mathematical methods and computer tools for aggregation and disaggregation of hydro power systems in optimization models. Both the aggregated model structure and methods for aggregation and disaggregation is addressed. Existing methods has been unchanged for decades and the project re-visit and upgrade these to produce a new model adapted to analysis of the future electricity market. The project work started summer 2015 and the focus has been the following activities: - Literature review and knowledge building related to application of aggregation and disaggregation techniques in other parts of the world. - Knowledge building on existing disaggregation technique. This activity has been combined with finding a new method for disaggregation by studying parts of the existing disaggregation technique and combine it with formal optimization to give an improved problem solution. Results from this new model show it is better at utilizing price variations for pumped storage plants, utilizing hydropower in serial watercourses and utilizing non-controllable renewable production. The model computation time is long and considerable effort has been devoted to testing of methods that can reduce computation time. A prototype model is available to the users. A paper presenting results from the model and an analysis of a future European energy has been published in Energies. The models developed in the project relies on existing algorithms. These algorithms have recently been modernized. Therefore, effort has been put into upgrading the model to exploit the modern algorithms. A detailed report of the disaggregation method has been distributed to the project members. -We have implemented a new Stochastic Dynamic Programming type algorithm for a general aggregated model structure for strategy calculations where the focus has been on an aggregated two-reservoir model. Testing shows that the new implementation is more time consuming than existing method. Parallel processing may be utilized to reduce calculation time. -We have implemented a version of a Sampling Stochastic Dynamic Programming algorithm (SSDP). According to the literature and the properties of the algorithm, SSDP improves correlations in time and space compared to Stochastic Dynamic Programming, e.g. better represent dry/wet years. The SSDP does not give better results than SDP for tested cases. - The last activity is automatic generation of the aggregated structure for the strategy calculation. Work has been divided in two sub-activities. The goal of the first sub-activity is to make a method that aggregates one or more hydro systems into two aggregated parallel reservoirs. The calculation of minimum production is necessary to generate aggregated energy inflow time series for the strategy calculations. The calculation of minimum production with detailed hydropower production has been implemented. The automatic generation of aggregated structure is partly finished. In the second sub-activity, a general aggregation procedure has been implemented. This activity is finalized, and a paper is published in Energies. In the last reporting period we finialize project reporting.

Den viktigste effekten av prosjektet er den nye markedsmodellen som er bedre egnet til analyser av et fremtidig kraftsystem med en mye større andel sol og vindkraftproduksjon. Bedre analyser gir bedre investeringsbeslutninger som er nyttig både for prosjektdeltakerne og samfunnet for øvrig. Det finnes nå et alternativ til eksisterende disaggregeringsmetodikk. SINTEF har økt kompetansen på et viktig felt som også kan anvendes i utviklingen av de modeller som brukes for å planlegge eksisterende kraftsystem.

Hydro-thermal market optimization and simulation models are crucial decision support tools used for price forecasting, operation planning, analysis of security of supply and investment analysis. The large number of hydro plants and reservoirs in the Nordic system significantly adds to the complexity of these models. Therefore, all models that need to represent the whole Nordic market must use an aggregated representation of the hydro system. The models that will be addressed in this project, Vansimtap and Samkjøringsmodellen, both apply an aggregated model description. The present aggregation technique used in these models introduces simplifications that may give non-optimal model decisions, which in turn may lead to non-optimal operation and utilisation of resources, and possibly wrong investment decisions. The models also include disaggregation procedures. Disaggregation are used in simulation procedures to verify that decisions from an aggregated model is valid for the detailed physical system. Disaggregation and validation also ensures that the simplifications and consequently added unrealistic flexibility given by the aggregation is corrected for in the simulated results. Models that combine simulation techniques with (strategy) optimization at an aggregate level can effectively address non-linear and state dependent constraints that otherwise needs to be simplified or even left out in formal optimization methods due to problem size and resulting computational burden. The Nordic power market will in the future see a stronger coupling to the European power markets and experience increasing quantities of non-storable new renewables. The existing aggregation and disaggregation methods used in the EMPS models are not adapted to the new challenges that this introduces. The purpose of the project is to develop new methods for aggregation and disaggregation that will improve the model properties in general and specifically with regard to the new challenges.