Potential Maximum Power Estimation for Wind Farms Based on LSTM Neural Network

When a wind farm is operating in a state of non-maximum power generation, grid operators often need to know its potential maximum power to provide a reference when formulating active power scheduling strategies or frequency regulation reserve capacity allocation strategies. At present, the potential maximum power is generally obtained through the measured wind speed calculation. However, because of wake effect and the low accuracy of wind speed measurement, the directly calculated potential maximum power is usually inaccurate. Therefore, this paper proposes a potential maximum power estimation method for wind farms based on a long short-term memory neural network. This method makes the estimated potential maximum power more accurate by learning the relationship between wind turbines' states, wind speed distribution, wake effect, and potential maximum power. The results of simulation prove the usefulness of the approach that was proposed.