Grid reliability assessment for short-term planning

With the increasing amount of renewable and difficult-to-forecast generation units, Transmission System Operators (TSO) face new challenges to operate the grid properly. The deterministic N-1 criterion is currently used to assess grid reliability. This criterion states that the loss of an active component should not trigger the violation of operational constraints. It has been chosen in the conventional context of electricity in which large units produced power transmitted to the consumer through the transmission and the distribution systems. The renewable energy sources, the distributed generations, and the liberalization of the electricity market led to a revolution in power systems. Renewable energies present intrinsic variability and limited predictability. Those variables are thus subject to forecasting errors. Distributed generations changed the structure of the power system to include smaller productions dispersed in the grid. The competitive electricity market led the consumers to react to electricity prices. The load is thus now subject to higher forecasting errors. An increasing share of the power system’s variables is thus now subject to errors that are likely to affect the operations and the planning of the grid. The N-1 criterion reaches limitations in considering the new characteristics of the power system. To account for this evolution, TSOs have to make a shift in paradigm. They must go from the N-1 criterion to a reliability-based approach, with risk management and integration of errors on forecasted values. This amounts to going from deterministic to probabilistic approaches capable of quantifying this risk.The purpose of this research project is to develop the basis of an industrial tool. The method considers thus the barriers to the use of a probabilistic method for grid planning. TSOs are indeed reluctant to give up the N-1 criterion. They fear that a probabilistic method would be too difficult to understand and apply and that the related computational time would be too long for the operational planning.The method proposed in this research project aims at overcoming those barriers. The method sets the basis of a decision support tool for the planners to make sound decisions. It is thus not a black-box and the planners are included in the assessment. The method is based on the current work of the planners and widens it to englobe probabilistic considerations. It offers thus a smooth evolution from a deterministic to a probabilistic method which will ease the industrial development of the tool using it. The method is called Discrete Forecast ERrors Scenarios method: DIFERS. It has been developed to be consistent with the operational planning in terms of time constraints and available information. It relies on three evolutions from the deterministic N-1 criterion: 1. Include possible variations from the best estimate of the forecasts.2. Enlarge the contingency list to higher N-k events.3. Consider the impact and the probability of the events to compute their risk.The contingency list evolves thus toward a risk-based classification of events. The planners’ work aims then at proposing actions to decrease the risk to an acceptable level. The first step of DIFERS is performed off-line to relax the time constraints. It aims at evaluating the contingency list for a set of situations. It performs an assessment on the most probable events, considering a larger group than the N-1 criterion. The assessment focuses on N-1 and nearby N-2 events. The nearby events have been selected based on a distance criterion defined in terms of number of components in the smallest path from one component to another. Some random N-3 and N-4 events are also analyzed to assess the evolution of the risk with regard to the number of failing components. Continuous variables are represented by their probability density functions (PDFs), which represent the variation range for the set of situations considered. Those PDFs are discretized to limit the computational time. The assessment of the contingencies is performed on each combination of those discrete points. The second step uses the contingency list developed in step 1 to assess the risk related to a specific situation: a grid plan. The PDFs used in this step represent the forecasting errors on the continuous variables for the grid plan considered. They are also discretized and each combination is tested with the events of the contingency list. At the end of the assessment, indicators are computed and provided to the planner. The planner can then propose actions to be tested by the tool to see their impact on the reliability indicators. The assessment stops once the reliability target is met.The final step aims at updating the information computed in step 2 with newly acquired forecasts. As real-time is closer, those forecasts are more reliable. The method has been tested on a plausible scenario and on a simplified version of the Belgian grid. The load and the offshore wind production have been considered as input variables for this implementations. The results show that there is an interest in evolving toward a risk-based assessment to capture the new characteristics of the evolving context of electricity supply. The implementation of the DIFERS method should continue on several scenarios. It should integrate all continuous variables such as solar and onshore productions. Moreover, all real-life considerations on the input variables, such as correlations, should be included to represent the power system as best as possible.This research project has been conducted in collaboration with the Belgian TSO Elia and it has been financed by the Doctiris grant of Innoviris.Doctorat en Sciences de l'ingénieur et technologie