Methods, systems, and devices for wind turbine control based on reinforcement learning are disclosed. The method comprises receiving data indicative of a current environmental state of the wind turbine, determining one or more controlling actions of the wind turbine based on the current environmental state of the wind turbine and a reinforcement learning algorithm, and applying the determined one or more controlling actions to the wind turbine.

A method for controlling a wind turbine connected to an electrical grid includes receiving, via a controller, a state estimate of the wind turbine. The method also includes determining, via the controller, a current condition of the wind turbine using, at least, the state estimate, the current condition defining a set of condition parameters of the wind turbine. Further, the method includes receiving, via the controller, a control function from a supervisory controller, the control function defining a relationship of the set of condition parameters with at least one operational parameter of the wind turbine. Moreover, the method includes dynamically controlling, via the controller, the wind turbine based on the current condition and the control function for multiple dynamic control intervals.

A method of operating a wind turbine generator comprising a plurality of blades, the method comprising iterating the following steps: comparing an indicated rotor speed with a rotor speed target to determine a rotor speed error; generating a modified rotor speed error by applying a control factor to the rotor speed error; controlling the pitch angle of the blades via a pitch control system in accordance with the modified speed error; and altering the control factor in dependence on a size of the indicated rotor speed.

A method for controlling a wind turbine system connected to a power grid. The method comprises generating a wind turbine control signal based on a power control reference for controlling a power output of a wind turbine; monitoring an electrical frequency of the power grid; in response to detecting a change in the frequency in the power grid, activating a fast frequency support method comprising the steps of; adjusting the power control reference to cause an overproduction of power by the wind turbine; the overproduction of power causing a transfer of inertial kinetic energy from the wind turbine to electrical power; wherein the power control reference is determined by applying an adaptive gain function to a measurement of a difference in grid frequency from a nominal level.

The invention provides a wind turbine method that includes receiving alarm state data indicating that the wind turbine has entered an alarm state in which operation of the wind turbine has stopped, and receiving sensor data from a plurality of sensors of the wind turbine indicative of operating conditions associated with the wind turbine. When the alarm state data is received, the method includes executing a trained machine learning model based on the received sensor data and the alarm state to obtain an output, where the machine learning model is trained based on historical data associated with a plurality of wind turbines, the historical data being indicative of the plurality of wind turbines previously being in the alarm state. The method includes providing, based on the obtained output, an action to be performed to allow the wind turbine to resume operation.

A computer-implemented reinforcement learning method includes determining, based on a target probability of satisfaction of a constraint condition related to a state of a control object and a specific time within which a controller causes the state of the control object not satisfying the constraint condition to be the state of the control object satisfying the constraint condition, a parameter of a reinforcement learner that causes, in a specific probability, the state of the control object to satisfy the constraint condition at a first timing following a second timing at which the state of control object satisfies the constraint condition; and determining a control input to the control object by either the reinforcement learner or the controller, based on whether the state of the control object satisfies the constraint condition at a specific timing.

A method for monitoring a wind turbine (10) is disclosed. The method comprises: collecting data that is associated with an abnormal behaviour of the wind turbine; comparing the collected data with anonymized data from other wind turbines; matching a fault condition with the abnormal behaviour through the comparison; and outputting the fault condition to the wind turbine.

A method for identifying underperforming agents in a multi-agent cooperative system includes receiving information relating to the performance of each agent in the multi-agent system, calculating an estimated extracted resource value of each agent based on the received information, comparing the estimated extracted resource value of each agent to a threshold value, calculating a performance index based on the comparison and identifying an agent as an under-performing agent based on the performance index. A system for identifying under-performing agents in a plurality of agents in a multi-agent cooperative system includes a performance analyzing processor, a communications port for receiving state information for each agent and control information for each agent, a classifier for identifying a subset of agents in the plurality of agents that are performance comparable and an optimizer configured to identify an under-performing agent of performance comparable agents and generate updated control information for the identified under-performing agent.

A system and method are provided for controlling a wind turbine of a wind farm. Accordingly, a controller implements a first model to determine a modeled performance parameter for the first wind turbine. The modeled performance parameter is based, at least in part, on an operation of a designated grouping of wind turbines of the plurality of wind turbines, which is exclusive of the first wind turbine. The controller then determines a performance parameter differential for the first wind turbine at multiple sampling intervals. The performance parameter differential is indicative of a difference between the modeled performance parameter and a monitored performance parameter for the first wind turbine. A second model is implemented to determine a predicted performance parameter of the first wind turbine at each of a plurality of setpoint combinations based, at least in part, on the performance parameter differential the first wind turbine. A setpoint combination is then selected based on the predicted performance parameter and an operating state of the first wind turbine is changed based on the setpoint combination.

A control system for a dynamic system including at least one measurement sensor. The system includes at least one computing device configured to generate and transmit at least one regulation device command signal to at least one regulation device to regulate operation of the dynamic system based upon at least one inferred characteristic.