A probabilistic equivalent method of DFIG-based wind farm based on offline calculations and online match

The equivalent model of wind farm was changing when randomness and intermittent were taken into consideration. A probabilistic equivalent method of the DFIG-based wind farm is proposed in this paper. It was based on offline calculations and online match. Firstly, the equivalent method of the wind farm was discussed. It was assumed that the equivalent model consists of three type of equivalent DFIG. The structure of the equivalent model keeps unvaried whereas the parameters keep changing with the drifting of scenarios. The historical recorded scenarios are analyzed to get samples of the equivalent model. The representative samples are extracted through statistics and combinations with the three kind of typical equivalent DFIGs. The probability distribution function of each representative sample was also got by investigation. Then, a BP neural network with two hidden layers and six inputs was used in the wind power prediction. It was used in online match. The output of next moment was used in calculating to which degree the scenarios in the sample set match with the next scenario. The matching degrees were sorted and the most matched scenarios were chosen. Finally, this method was applied in the augmented IEEE 3-generator and 9 bus test system with wind farm connected in. It was shown that the wind farm equivalent model can be predicted and its probability is also given. It can reduce computation tasks effectively and solve the difficulty of wind farm equivalent when working conditions are changing.