Statistisk modellering av vindkraftsobalanser i Sveriges elområden

In a synchronous electric grid the consumption of electricity must always be met by an equal amount of generation. In the Nordic power system, this balance is first and foremost kept by the balance responsible parties in the electric markets. However, from one hour before delivery, it is the Swedish Transmission System Operator (TSO), Svenska kraftnät (Svk) together with its Nordic counterparts, who take over the responsibility. They achieve this by for example purchasing ancillary services such as Frequency Restoration Reserves (FRR) to compensate for frequency deviations. A way of explaining the frequency deviations that would have occurred without the TSO taking actions, is that they are caused by imbalances. Imbalances are the difference between measured and traded energy volumes in the bidding areas, where volumes equals HVDC-connections, consumption and different kinds of power production. In the future, these imbalances will be one of the dimensioning factors of FRR. The purpose of this thesis is to study the imbalances caused by wind power production and to create a model that can simulate future wind power imbalances. The long term goal is that the model will be part of a larger project whose purpose is predicting the future need of FRR. The model has been designed to use future market data, such as traded volumes and spot prices to make the predictions. The model has been developed using statistical methods in MATLAB together with another master student, who has studied consumption imbalances. Due to lack of deterministic correlations, the final model created was an Autoregressive-Moving-Average (ARMA) model together with a linear correlation between quarterly average traded volumes and quarterly standard deviations of the wind power imbalances. The model can recreate the historical autoregressive behaviour and the historical distribution of the imbalances to a satisfactory degree, as well as scaling up the imbalances with a correlation of 0.92. Applying future market data on the model, imbalances are expected to increase by 50\% to 180\% from today to the year 2023, depending on bidding area. However, there are uncertainties due to yearly variations in the wind power production. One conclusion is therefore that a windy year probably also will increase the required need of FRR. Before applying the model to evaluate the future need for FRR, the reliability used in the traded data for developing the model should be checked. A final validation of the total simulated imbalances, not just wind power imbalances, against historic data should also be performed. To develop the model further, a suggestion is to study possible spatial correlations of the imbalances between bidding areas