Reducing curtailment of wind power generation

A method comprising:

• receiving, by a computer, historical electrical power output measurements of a wind turbine for a time period immediately preceding a specified past time, wherein the historical electrical power output measurements of the wind turbine for a time period immediately preceding a specified past time comprise a predefined number of measurements at equal time intervals, ending at the specified past time;
• receiving, by the computer, historical wind speed micro-forecasts for the geographic location of the wind turbine, for a time period immediately preceding the specified past time and for a time period immediately following the specified past time, wherein the historical wind speed micro-forecasts for the geographic location of the wind turbine for a time period immediately preceding the specified past time and for a time period immediately following the specified past time comprise a predefined number of wind speed micro-forecasts at equal time intervals, ending at the specified past time, and a predefined number of wind speed micro-forecasts at equal time intervals beginning with the specified past time, respectively;
• converting, by the computer, the historical wind speed micro-forecasts to wind power output values;
• generating, by the computer, based on the historical electrical power output measurements and the wind power output values, a trained machine learning model for predicting wind power output of the wind turbine;
• receiving real-time electrical power output measurements of the wind turbine and real-time wind speed micro-forecasts for the geographic location of the wind turbine, wherein the real-time electrical power output measurements of the wind turbine comprise a predefined number of measurements at equal time intervals, ending at the current time, and wherein the real-time wind speed micro-forecasts for the geographic location of the wind turbine comprise a predefined number of wind speed micro-forecasts at equal time intervals, ending at the current time, and a predefined number of wind speed micro-forecasts at equal time intervals, beginning with the current time;
• converting, by the computer, the real-time wind speed micro-forecasts to real-time wind power output values, wherein the historical wind speed micro-forecasts are converted to wind power output values by one of:
• a manufacturer's power curve for the wind turbine, or a machine learning model trained to convert wind speed to wind power, based on historical wind speed measurements at the wind turbine and historical electrical power output measurements of the wind turbine;
• outputting, by the computer, using the trained machine learning model with the real-time electrical power output measurements and the real-time wind power output values, a wind power output forecast for the wind turbine at a future time, wherein the machine learning model is one of:
• a regression neural network, a support vector regression (SVR) model, or a linear regression mode;
• determining, by the computer, that an oversupply of wind power would be generated, based in part on the predicted wind power output of the wind turbine;
• and
• curtailing, by the computer, wind power output of the wind turbine.