JP7090440B2 - Control systems, learning devices, control devices, and control methods - Google Patents

Description

The present invention relates to a control system, a learning device, a control device, and a control method.

Conventionally, in a wind power generation system, the rotation speed of a wind turbine having the maximum generated power (regenerative power) with respect to the wind speed received by the wind turbine is uniquely determined as a characteristic value peculiar to each wind turbine. For example, the rotation speed of the wind turbine is controlled by adjusting the torque of the wind turbine according to the wind speed, and the control is performed so that the maximum generated power according to the wind speed can be obtained.
Wind conditions change depending on the location of the wind turbine and the season. Therefore, it is preferable that the control is performed according to the state in which the wind turbine is installed, the environment of the place where the wind turbine is installed, the season, and the like.
As a countermeasure, a machine learning engine is used to learn the relationship between the wind speed and the control parameter that maximizes the regenerative power output from each wind turbine, and according to the control parameter output from the learned machine learning engine. A technique for controlling an individual wind turbine is disclosed (for example, Patent Document 1). In addition, as a machine learning technique, there is also a method of learning the relationship between the control parameter that maximizes the regenerative power and the wind speed by using reinforcement learning. In this case, by giving a reward according to the magnitude of the regenerative power generated as a result of controlling the wind turbine, it is possible to learn the relationship between the wind speed and the control parameter so that the regenerative power is maximized.

Special Table 2013-524744

However, if you simply give a high reward when the regenerative power increases, even if the control parameters are inappropriate, a high reward will be given even if the regenerative power increases due to an appropriate change in the wind speed. It is possible that it will end up. Even if the control parameters are appropriate, if the regenerative power is reduced due to an unexpected change in the wind speed, the reward will be low. Therefore, there is a problem that the learning model is erroneously learned.

The present invention has been made in view of such circumstances, and an object of the present invention is to provide a control device for a wind power generation system capable of performing appropriate control according to the relationship between wind speed and regenerative power. ..

In order to solve the above-mentioned problems, one embodiment of the present invention is a control system that controls regenerative power generated by a wind power generation system, and wind speed information indicating the wind speed at the installation location of the wind turbine of the wind power generation system. Based on the rotation information regarding the rotation of the wind turbine, the power information regarding the regenerative power generated by the wind power generation system, the relationship information between the wind speed and the regenerative power, and the reward according to the power information for the wind speed. A learning unit that learns correspondence information between the wind speed and the regenerative power, a storage unit that stores the correspondence information between the wind speed and the regenerative power, and a power control parameter that controls the regenerative power are set in the wind power generation system. The power information is based on the state detection unit that detects the rotation information, the power information, and the wind speed information, the rotation information and the wind speed information detected by the state detection unit, and the corresponding information. A determination unit that determines the command value of the wind power, a control unit that controls the regenerative power based on the command value determined by the determination unit, and the wind speed information and the rotation information detected by the state detection unit. The reward calculation unit includes a reward calculation unit that calculates the reward based on the power information, and the reward calculation unit is shown in the rotation information when the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value. A control system characterized in that the reward is calculated according to the number of rotations of the wind turbine, and the reward is calculated according to the power information when the wind speed shown in the wind speed information is less than a predetermined strong wind threshold value . be.

Further, one embodiment of the present invention is the above-mentioned control system, and the reward calculation unit is described when the wind speed indicated in the wind speed information detected by the state detection unit is less than a predetermined strong wind threshold value. Based on the regenerative power shown in the power information detected by the state detection unit and the reference power stored in advance, the larger the regenerative power is, the higher the reward is calculated.

Further, one embodiment of the present invention is the above-mentioned control system, and the reward calculation unit is described when the wind speed indicated in the wind speed information detected by the state detection unit is equal to or higher than a predetermined strong wind threshold value. Based on the rotation speed of the wind turbine shown in the rotation information detected by the state detection unit and the rotation threshold value stored in advance, the rotation speed is smaller than the rotation threshold value and the rotation speed is smaller than the rotation threshold value. , Calculate the higher reward.

Further, one embodiment of the present invention is the above-mentioned control system, in which the reward calculation unit is the wind speed acquired by the state detection unit in the past from the time when the power information is detected by the state detection unit. The reward is calculated based on the information.

Further, in one embodiment of the present invention, wind speed information indicating the wind speed at the installation location of the wind turbine of the wind power generation system, rotation information regarding the rotation of the wind turbine, power information regarding the regenerative power generated by the wind power generation system, and the above-mentioned A learning unit for learning the correspondence information between the wind speed and the regenerative power based on the relationship information between the wind speed and the regenerative power and the reward corresponding to the power information for the wind speed is provided , and the reward is the regenerative power. The wind speed calculated based on the wind speed information, the rotation information, and the power information detected when the power control parameter for controlling the wind power generation system is set to the wind power generation system, and the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value. It is a learning device that is calculated according to the rotation speed of the wind turbine shown in the rotation information in a certain case, and is calculated according to the power information when the wind speed shown in the wind speed information is less than a predetermined strong wind threshold value . ..

Further, in one embodiment of the present invention, wind speed information indicating the wind speed at the installation location of the wind turbine of the wind power generation system, rotation information regarding the rotation of the wind turbine, power information regarding the regenerative power generated by the wind power generation system, and the above-mentioned A learning unit that learns correspondence information between the wind speed and the regenerated power based on the relationship information between the wind speed and the regenerated power and a reward corresponding to the power information for the wind speed, and a power control that controls the regenerated power. A state detection unit that detects the rotation information, the power information, and the wind speed information when the parameters are set in the wind power generation system, the rotation information and the wind speed information detected by the state detection unit, and the correspondence. It was detected by the determination unit that determines the command value of the power information based on the information, the control unit that controls the regenerated power based on the command value determined by the determination unit, and the state detection unit. A reward calculation unit for calculating the reward based on the wind speed information, the rotation information, and the power information is provided , and the reward calculation unit is used when the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value. The reward is calculated according to the rotation speed of the wind turbine shown in the rotation information, and the reward is calculated according to the power information when the wind speed shown in the wind speed information is less than a predetermined strong wind threshold. It is a control device.

Further, in one embodiment of the present invention, rotation information regarding the rotation of the wind turbine of the wind power generation system when the power control parameter for controlling the regenerated power generated by the wind power generation system is set in the wind power generation system, the wind power generation. A state detection unit that detects power information regarding regenerative power generated by the system and wind speed information indicating the wind speed at the installation location of the wind turbine, the rotation information and the wind speed information detected by the state detection unit, and the wind speed. A determination unit that determines a command value of the power information based on the correspondence information between the wind power and the regenerative power generation, and a control unit that controls the regenerative power generation based on the command value determined by the determination unit. The corresponding information is information learned based on the wind speed information, the rotation information, the power information, the relationship information between the wind speed and the regenerated power, and the reward corresponding to the power information for the wind speed. The reward is calculated based on the wind speed information, the rotation information, and the power information detected by the state detection unit, and the rotation is when the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value. It is a control device calculated according to the rotation speed of the wind turbine shown in the information, and calculated according to the power information when the wind speed shown in the wind speed information is less than a predetermined strong wind threshold value .

Further, one embodiment of the present invention is a control method for controlling the regenerated power generated by the wind power generation system, wherein the learning unit indicates the wind speed at the installation location of the wind turbine of the wind power generation system, the wind turbine. The wind speed is based on the rotation information regarding the rotation of the wind power generation system, the power information regarding the regenerative power generated by the wind power generation system, the relationship information between the wind speed and the regenerative power, and the reward corresponding to the power information for the wind speed. The correspondence information between the wind power and the regenerative power is learned, the storage unit stores the correspondence information between the wind speed and the rotation, and the state detection unit sets the power control parameter for controlling the regenerative power in the wind power generation system. In this case, the rotation information, the power information, and the wind speed information are detected, and the determination unit determines the power information based on the rotation information, the wind speed information, and the corresponding information detected by the state detection unit. The command value is determined, the control unit controls the power generation based on the command value determined by the determination unit, and the reward calculation unit determines the wind speed information detected by the state detection unit. The reward is calculated based on the rotation information and the power information, and when the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value, the reward is shown according to the rotation speed of the wind turbine shown in the rotation information. Is a control method for calculating the reward according to the power information when the wind speed shown in the wind speed information is less than a predetermined strong wind threshold value .

As described above, according to the present invention, appropriate control can be performed according to the relationship between the wind speed and the regenerative power.

It is a block diagram which shows an example of the schematic structure of the wind power generation system 1 which concerns on 1st Embodiment. It is a block diagram which shows an example of the structure of the control device 60 and the learning device 70 of the wind power generation system 1 which concerns on 1st Embodiment. It is a figure which shows the example of the reward condition at the time of a strong wind which concerns on 1st Embodiment. It is a flowchart which shows the operation example of the control device 60 which concerns on 1st Embodiment. It is a block diagram which shows an example of the structure of the control device 60A which concerns on 2nd Embodiment. It is a figure explaining the process of estimating the response time by the response time estimation unit 69 which concerns on 2nd Embodiment. It is a figure which shows the example of the reward condition which concerns on 3rd Embodiment. It is a figure explaining the process of calculating the reward by the reward calculation unit 63B which concerns on 3rd Embodiment. It is a flowchart which shows the operation example of the control system 50C which concerns on 3rd Embodiment. It is a block diagram which shows an example of the schematic structure of the wind power generation system 1D which concerns on the modification of 4th Embodiment. It is a block diagram which shows an example of the schematic structure of the wind power generation system 1E which concerns on the modification of 5th Embodiment.

Hereinafter, the control system, the learning device, and the control device of the embodiment will be described with reference to the drawings.

<First Embodiment>
FIG. 1 is a block diagram showing an example of a schematic configuration of the wind power generation system 1 according to the first embodiment. The wind power generation system 1 includes a wind power generator main body 10 and a control system 50. Various information is exchanged between the wind power generator main body 10 and the control system 50.
As shown in FIG. 1, for example, a control parameter for controlling the wind power generator main body 10 is output from the control system 50 to the wind power generator main body 10.
Further, for example, a state parameter indicating the state of the wind power generator main body 10 is output from the wind power generator main body 10 to the control system 50.

The control parameter is, for example, a power control parameter that controls the amount of regenerative power generated by the generator 30.
Further, the state parameters are, for example, wind speed information indicating the wind speed at the installation location of the wind turbine 20, rotation information indicating the rotation speed of the wind turbine 20 (hereinafter, also simply referred to as rotation speed), and regenerative power generated by the generator 30. It is power information indicating the amount of power.

The wind power generator main body 10 includes a wind turbine 20, a generator 30, a rectifying / boosting unit 31, a voltage detection unit 32, a current detection unit 33, a wind speed sensor 41, and a rotation speed sensor 42.
The wind turbine 20 is configured as, for example, a vertical axis type wind turbine, and is configured by a straight wing vertical axis wind turbine or the like in which a plurality of straight blades are integrally rotatably connected around a rotating shaft extending in the vertical direction.

For example, the wind turbine 20 is connected to the rotor of the generator 30 described later via a rotation shaft, and rotates integrally with the rotor of the generator 30. Here, the rotor of the generator 30 rotates at a rotation speed corresponding to the electric energy of the regenerated electric power generated by the generator 30. Further, the electric energy of the regenerative power is controlled by MPPT (Maximum Power Point Tracking) by the control system 50 described later. Therefore, the rotation speed of the wind turbine 20 is indirectly controlled by the MPPT control by the control system 50.

The generator 30 is a device that converts the rotational force of the wind turbine 20 to generate electric power. For example, the generator 30 is configured as a three-phase alternating current generator, and a rotor that rotates in conjunction with the rotation of the wind turbine 20 rotates the wind turbine 20. AC power is generated by being connected to a shaft and rotating. The generator 30 supplies the generated AC power to the rectifying / boosting unit 31. The generator 30 operates as a generator that supplies the generated power to the rectifying / boosting unit 31 side, and also operates as an electric motor to which AC power is supplied from the rectifying / boosting unit 31. The generator 30 operates as an electric motor, for example, when performing assist control for assisting rotation when the wind turbine 20 is started.

The rectifying / boosting unit 31 converts the AC power generated by the generator 30 into DC power, and converts (boosts) the voltage of the converted DC power. The rectifying / boosting unit 31 is, for example, a boosting chopper circuit. Here, the DC power output from the rectifying / boosting unit 31 is an example of “regenerative power”.

The rectifying / boosting unit 31 includes, for example, a PWM (Pulse Width Modulation) servo controller, and changes the voltage of DC power based on MPPT control from the control system 50. Specifically, the rectifying / boosting unit 31 performs switching according to the PWM signal from the control system 50, and outputs DC power having a voltage corresponding to the change in the duty ratio of the PWM signal.

The rectifying / boosting unit 31 operates as a boosting chopper circuit when the generator 30 is operated to generate power, and operates as an inverter when the generator 30 is operated as an electric motor during assist control or the like. It is a circuit. The electric power supplied during the assist control may be the electric power from the battery (not shown) of the wind power generation system 1.

The voltage detection unit 32 is composed of a known voltmeter, detects the output voltage output from the rectifying / boosting unit 31, and outputs the detected output voltage to the control system 50.
The current detection unit 33 is composed of a known ammeter, detects the output current output from the rectifying / boosting unit 31, and outputs the detected output current to the control system 50.

The wind speed sensor 41 is composed of a known wind speed sensor, and is provided at a predetermined position near the wind turbine 20 (for example, a portion other than the rotary blade in the wind turbine 20) to detect the wind speed of the wind received by the wind turbine. The wind speed sensor 41 outputs information indicating the detected wind speed to the control system 50.

The rotation speed sensor 42 detects the rotation speed of the wind turbine 20. The rotation speed sensor 42 may be any sensor that can detect the rotation speed of the rotation shaft portion (not shown) of the wind turbine 20, and various known rotation speed sensors can be used. The rotation speed sensor 42 outputs information indicating the detected rotation speed to the control system 50.

The control system 50 includes a control device 60 and a learning device 70.
The control device 60 learns based on the wind speed detected by the wind speed sensor 41, the rotation speed detected by the rotation speed sensor 42, and the electric energy of the regenerative power detected by the voltage detection unit 32 and the current detection unit 33. The power control parameters are determined using the device 70. The control device 60 controls the amount of regenerated electric power generated by the generator 30 by setting the determined electric power control parameter in the wind power generator main body 10, and also (indirectly) determines the rotation speed of the wind turbine 20. Control.

Here, the wind speed detected by the wind speed sensor 41 is an example of "wind speed information". The rotation speed detected by the rotation speed sensor 42 is an example of "rotation information". Further, the electric energy of the regenerative electric power detected by the voltage detection unit 32 and the current detection unit 33 is an example of "electric power information".

The learning device 70 is, for example, a device that performs reinforcement learning. In this case, the learning device 70 corresponds to an agent that is a learning subject in reinforcement learning, and is for more appropriately controlling the controlled object by interacting with the controlled object (in the present embodiment, the wind power generator main body 10). Advance learning.

Hereinafter, the case where the learning device 70 performs reinforcement learning will be described as an example, but the present invention is not limited to this. The learning device 70 may be one that learns so that the parameters for controlling the controlled object become more appropriate based on the state of the controlled object (wind power generator main body 10). The learning device 70 may perform supervised learning, unsupervised learning, or other learning. Here, the state related to the controlled object (wind power generator main body 10) is the state around the wind power generator main body 10 and the wind power generator main body 10, and for example, the wind speed in the wind turbine 20 and the wind turbine 20 set by the state parameter. It is a variable such as the rotation speed of the generator 30 and the amount of power generated by the generator 30. Further, the state here includes a state that changes from moment to moment such as the above-mentioned wind speed, and a predetermined state, for example, a limit value of the rotation speed of the wind turbine 20, an upper and lower limit of the rotation torque of the wind turbine 20. And the maximum amount of electric power that the generator 30 can generate.

In the present embodiment, the learning device 70 outputs a power control parameter that controls the amount of regenerated power generated by the generator 30, and the output power control parameter is set in the wind generator main body 10. Observe the change in the state and determine the next power control parameter according to the change in the state.

Further, the learning device 70 receives a reward according to a change in the state of the wind power generator main body 10 due to the power control parameter output by the learning device 70 being set in the wind power generator main body 10. As a result, the learning device 70 can proceed with learning by determining whether the power control parameter output by itself is good or bad by using the reward as a clue, and can output a more suitable power control parameter.

FIG. 2 is a block diagram showing an example of the configuration of the control device 60 of the wind power generation system 1 according to the embodiment of the present invention.
As shown in FIG. 2, the control device 60 includes a parameter acquisition unit 61, a state detection unit 62, a reward calculation unit 63, and a reward output unit 64. Further, the learning device 70 includes a reinforcement learning unit 71. Here, the reinforcement learning unit 71 is an example of the “learning unit”.

The parameter acquisition unit 61 acquires the power control parameter output from the reinforcement learning unit 71. The parameter acquisition unit 61 outputs the acquired power control parameters to the wind power generator main body 10.

The state detection unit 62 detects a state parameter indicating the state of the wind power generator main body 10. The state parameters are information about the wind turbine 20 and the generator 30 included in the wind power generator main body 10, and are, for example, the wind speed detected by the wind speed sensor 41, the rotation speed detected by the rotation speed sensor 42, and the generator 30. It is information indicating the regenerated electric power generated. The state detection unit 62 outputs the detected state parameter to the reward calculation unit 63.

The reward calculation unit 63 calculates the reward based on the wind speed information and the electric power information detected by the state detection unit 62. The reward calculation unit 63 calculates the reward according to a predetermined reward condition. The reward calculation unit 63 outputs the calculated reward to the reward output unit 64.

The reward output unit 64 outputs the reward acquired from the reward calculation unit 63 to the reinforcement learning unit 71.

Here, the reward condition is determined according to the electric energy of the regenerative power with respect to the wind speed, for example, in a normal time when the wind speed is not a strong wind. That is, the larger the amount of power generation, the higher the reward is calculated. Further, in a strong wind, it is set so that a higher reward can be obtained when it is determined that the rotation speed of the wind turbine 20 is controlled more appropriately according to the safety of the wind turbine 20. The safety of the wind turbine 20 is such that the wind turbine 20 can rotate safely when the mechanical durability limit and the like are taken into consideration. For example, in a normal state where the wind speed is appropriate, the rotation speed of the wind turbine 20 is high. It is safe because there is no danger that the wind turbine 20 will be damaged even if it increases a little, but when controlling in a strong wind state where the wind speed is strong, the wind turbine 20 may be damaged even if the rotation speed increases a little. It is the degree to which the safety of the wind turbine 20 is shown. The normal state means that the wind speed is less than the predetermined strong wind threshold value, and the strong wind state means that the wind speed is equal to or higher than the predetermined strong wind threshold value.

In wind power generation, it is necessary to suppress the rotation speed of the wind turbine 20 in particular in a strong wind to prevent the wind turbine 20 from being over-rotated and damaged. In most cases, safety measures are taken such as forcibly stopping the wind turbine 20 when the rotation speed of the wind turbine 20 becomes excessive. If the wind turbine 20 is forcibly stopped, it will not be possible to obtain regenerative power. For this reason. In a strong wind, it is desirable that the rotation speed of the wind turbine 20 is controlled so as not to be excessive and the wind turbine 20 is controlled so as not to reach a forced stop.

Therefore, in the present embodiment, the reward calculation unit 63 calculates a higher reward as the generated regenerative power is larger in a normal state where the wind speed is not a strong wind. Further, the reward calculation unit 63 calculates a high reward when the rotation speed of the wind turbine 20 is controlled so as not to be excessive in a strong wind state where the wind speed is a strong wind.

The reinforcement learning unit 71 acquires a state parameter from the wind power generator main body 10. Further, the reinforcement learning unit 71 acquires a reward from the reward output unit 64. Based on the acquired state parameters and information indicating the safety of the wind turbine 20 with respect to the wind speed, the reinforcement learning unit 71 advances learning so that a higher reward can be obtained by using the reward as a clue, and outputs more appropriate power control parameters. do.

FIG. 3 is a diagram showing an example of the reward condition according to the first embodiment. FIG. 3A shows an example of a reward condition when the wind speed is in a normal state, and FIG. 3B shows an example of a reward condition when the wind speed is in a strong wind state.
As shown in FIG. 3A, when the wind speed is in the normal state, the reference power indicating the reference generated power is determined in advance according to the wind speed. In this example, the reference power 1 corresponds to the wind speed 1 which is the normal wind speed, and the reference power 2 corresponds to the wind speed 2 which is the normal wind speed. In addition, the higher (larger) the regenerative power generated than the reference power according to the wind speed, the higher the reward.

Here, it is known that the electric energy of the regenerative electric power is proportional to the cube of the wind speed V (V ^ 3). Therefore, the reference power may be determined using, for example, P / V ^ 3 as an index.

In wind power generation, the maximum value of regenerative power with respect to the wind speed received by the wind turbine 20 is uniquely determined as a characteristic value peculiar to each wind turbine. Then, the maximum value of this regenerative power is set as a target value of the regenerative power according to the wind speed, and the regenerative power is controlled so as to approach the target value of the regenerative power. The reference power is set to, for example, a value obtained by subtracting a predetermined margin value from the maximum value of the regenerative power.

As shown in FIG. 3B, when the wind speed is in a strong wind state, a strong wind reference speed indicating a reference of the rotation speed of the wind turbine 20 as a reference in a strong wind is determined in advance. Further, the reward becomes higher as the rotation speed of the wind turbine 20 is lower than the strong wind reference rotation speed.
In this case, if the rotation speed of the wind turbine 20 is lower than the strong wind reference rotation speed, the higher the reward is given, and even if the rotation speed of the wind turbine becomes 0 (zero), the higher reward may be calculated. .. As a countermeasure for this, a lower limit value of the rotation speed of the wind turbine 20 may be set. In this case, the reward is higher when the rotation speed of the wind turbine 20 is lower than the strong wind reference rotation speed and the rotation speed of the wind turbine 20 is higher (larger) than the lower limit of the rotation speed of the wind turbine 20.

FIG. 4 is a flowchart showing an operation example of the control device 60 according to the first embodiment.
First, the state detection unit 62 acquires a state parameter (step S11). Specifically, the state detection unit 62 acquires power information, wind speed information, and rotation information. The state detection unit 62 outputs the detected state parameter to the reward calculation unit 63.

Next, the reward calculation unit 63 calculates the reward.
The reward calculation unit 63 determines whether or not the wind speed is normal (step S12). When the wind speed is normal, the reward calculation unit 63 acquires the reference power according to the wind speed (step S13). The reference power is stored, for example, in a power information storage unit (not shown) of the control device 60. The reward calculation unit 63 calculates the difference between the regenerative power and the reference power (step S14), and calculates the reward according to the difference (step S15). Specifically, the reward calculation unit 63 calculates a higher reward as the regenerative power is larger than the reference power.

On the other hand, in step S12, when the wind speed is not normal but strong wind, the reward calculation unit 63 acquires a strong wind reference speed which is a reference of the rotation speed of the wind turbine 20 at the time of strong wind. The strong wind reference speed is stored, for example, in a rotation information storage unit (not shown) of the control device 60. The reward calculation unit 63 calculates the difference between the rotation speed and the strong wind reference speed (step S17), and calculates the reward according to the difference (step S15). Specifically, the reward calculation unit 63 calculates a higher reward as the rotation speed is smaller than the strong wind reference speed.

<Second embodiment>
Next, the second embodiment will be described.
This embodiment differs from other embodiments in that the control device 60A of the control system 50A calculates a reward in consideration of the response time. Hereinafter, the points different from those of the above-described embodiment will be described, and the same reference numerals will be given to the configurations having the same or similar functions as those of the above-described embodiments, and the description thereof will be omitted.

FIG. 5 is a block diagram showing an example of the configuration of the wind power generation system 1A according to the second embodiment. As shown in FIG. 5, the control device 60A includes a response time estimation unit 69.
The response time estimation unit 69 estimates the response time. The response time is the time from the time when the power control parameter is set in the wind power generator main body 10 to the time when the regenerated electric power generated by the generator 30 responds. The response of the regenerative power means that the power value corresponding to the set power control parameter is output.

The response time estimation unit 69 acquires wind speed information and electric power information, and compares the time-series change of the acquired wind speed information with the time-series change of the electric power information. The response time estimation unit 69 extracts, for a time interval indicating a time-series change of a certain electric power information, a time interval in which the time-series change of the electric power information and the tendency of the change are the same or similar. ..

The response time estimation unit 69 calculates, for example, the degree of similarity between the time interval showing the time-series change of a certain electric power information and the time-series change of the electric power information in the time interval retroactive to a predetermined search unit time in the time interval. The response time estimation unit 69 sequentially compares the direction of change from the value at the beginning of both time intervals as the degree of similarity. The direction of change is, for example, whether the value is increasing or decreasing. The response time estimation unit 69 makes it similar when the directions of change match, and dissimilar when they do not match, and calculates the sum of similar values in the entire time interval as the degree of similarity.

The response time estimation unit 69 calculates the degree of similarity to the time section of the wind velocity information that traces the time interval of the power information by a predetermined search unit time, and calculates the difference from the time interval of the wind speed information having the highest degree of similarity. Estimated as response time. The response time estimation unit 69 outputs the estimated response time to the reward calculation unit 63.

Here, the response time is generated by the inertia of the wind turbine 20. The response time is a time corresponding to a change in wind speed. That is, a non-constant delay occurs in the rotation speed with respect to the change in the wind speed. Therefore, the response time estimation unit 69 estimates the response time according to the wind speed. That is, it is desirable that the response time estimation unit 69 performs a process of estimating the response time each time the wind speed changes. Since the actual wind speed changes from moment to moment, the response time estimation unit 69 may estimate the response time according to the acquired power information each time the power information is acquired.

The reward calculation unit 63 calculates the reward according to the response time estimated by the response time estimation unit 69. The reward calculation unit 63 calculates the reward based on the electric power information after the response time has elapsed from the time when the wind speed information is acquired. For example, when the power control parameter for increasing the regenerated power is set at the time when the wind speed accelerates, the reward calculation unit 63 increases the regenerated power actually generated after the response time elapses from that time. In some cases, a higher reward is calculated as appropriate control is taken. On the contrary, when the power control parameter that increases the regenerated power is set at the time when the wind speed accelerates, the reward calculation unit 63 reduces the regenerated power actually generated after the response time elapses from that time. If so, a lower reward is calculated for improper control.

FIG. 6 is a diagram illustrating a process in which the response time estimation unit 69 according to the second embodiment estimates the response time. The upper part of FIG. 6 shows the relationship between the regenerative power and time, and the lower part of FIG. 6 shows the relationship between the wind speed and time.
As shown in FIG. 6, a case where the response time estimation unit 69 acquires the power information P (t2) at a certain time t2 will be described as an example. First, when the response time estimation unit 69 acquires the power information P (t2), the response time estimation unit 69 acquires the time-series change of the power information in the time interval T1 including the time t2. Next, the response time estimation unit 69 determines the degree of similarity between the time-series change of the electric power information in the time section T1 and the time-series change of the wind power generation in the time section T2 retroactive by a predetermined search unit time α from the time section T1. calculate. Next, the response time estimation unit 69 determines the degree of similarity between the time-series change of the electric power information in the time section T1 and the time-series change of the wind power generation in the time section T3 that goes back by a predetermined search unit time 2α from the time section T1. calculate. In this way, the response time estimation unit 69 sequentially calculates the degree of similarity with the time-series change of the wind power generation in the time section traced back by a predetermined search unit time α, and the time section having the largest calculated similarity degree. The difference from the time interval T1 is estimated as the response time D. This example shows an example in which the degree of similarity between the time-series change of the electric power information in the time section T1 and the time-series change of the wind power information in the time section T3 retroactively from the time section T1 by the time 2α is the largest. In this case, the response time estimation unit 69 estimates the difference D (= 2α) between the time interval T1 and the time interval T3 as the response time.

<Third embodiment>
Here, the third embodiment will be described. This embodiment is different from other embodiments in that the reward conditions differ depending on the direction of change in wind speed. Hereinafter, the points different from those of the above-described embodiment will be described, and the same reference numerals will be given to the configurations having the same or similar functions as those of the above-described embodiments, and the description thereof will be omitted.

FIG. 7 is a diagram showing an example of the reward condition according to the third embodiment. As shown in FIG. 7, in the present embodiment, the reward conditions are different depending on whether the wind speed is increasing or decelerating.
When the wind speed is increasing, the reward calculation unit 63B, as in the other embodiment, receives a higher reward when the electric energy of the regenerative electric power is larger according to the electric energy of the reference power and the regenerative electric power according to the wind speed. Is calculated.
When the wind speed is decelerating, the reward calculation unit 63B does not calculate the reward according to the electric energy of the regenerative power at the time when the wind speed is decelerating, and the regenerative power considering the time when the wind speed accelerates thereafter. The reward is calculated according to the amount of electricity.
Specifically, when the wind speed is decelerating, the reward calculation unit 63B acquires the reference power according to the average wind speed from that time until the predetermined time elapses after the wind speed changes to acceleration. ..
Further, when the wind speed is decelerating, the reward calculation unit 63B acquires the average electric energy which is the average of the regenerative power from that time until the predetermined time elapses after the wind speed changes to acceleration.
Then, the reward calculation unit 63B calculates the reward based on the reference power and the average electric energy according to the average wind speed. The reward calculation unit 63B may calculate the reward by using the wind speed information or the electric power information in consideration of the response time, or may calculate the reward without considering the response time.

In the wind power generation system 1, when the wind speed is decelerated, if the wind turbine is continuously rotated at a rotation speed at which the maximum regenerative power corresponding to the wind speed can be obtained, the rotation of the wind turbine may stall due to the power generation load. Therefore, when the wind speed is decelerated, it is desirable to control the regenerative power so as to be maintained as much as possible in consideration of the degree of deceleration of the wind speed (the amount of change in the wind speed per unit time).
However, even though the wind speed is decelerating, if a high reward is calculated when the amount of power is large according to the amount of regenerative power at that time, the rotation of the wind turbine will stall and the amount of power will be large. Even if the control is performed so that the amount can be obtained, a high reward will be calculated. When such a reward is given, the reinforcement learning unit 71 is made to perform erroneous learning.
As a countermeasure, when the wind speed is decelerating, the reward calculation unit 63B of the present embodiment has a reference power according to the average wind speed from that point onward until the predetermined time elapses after the wind speed changes to acceleration. To get.
The reward calculation unit 63B calculates, for example, the difference between the wind speed information acquired this time and the wind speed information acquired last time, and if the calculated difference is positive, it is determined that the wind speed is accelerating, and the difference is negative. If, it is determined that the wind speed is decelerating. In this case, when the difference is 0 (zero), the reward calculation unit 63B may determine that the wind speed is accelerating or may determine that the wind speed is decelerating.

FIG. 8 is a diagram illustrating a process in which the reward calculation unit 63B according to the third embodiment calculates a reward. The upper part of FIG. 8 shows the relationship between the regenerative power and time, and the lower part of FIG. 8 shows the relationship between the wind speed and time. FIG. 8 illustrates a case where the reward calculation unit 63B calculates the reward in consideration of the response time D.

As shown in FIG. 8, a case where the reward calculation unit 63B acquires the wind speed information Vm at a certain time t3 will be described as an example. First, the reward calculation unit 63B calculates the direction of change in the wind speed at time t3. In this example, the time t3 shows the case where the time t3 is included in the deceleration section Vg at which the wind speed decelerates.
When the wind speed at time t3 is decelerating, the reward calculation unit 63B determines that the average wind speed during the time interval Wave from that time t3 until the predetermined time T4 elapses after the time t4 when the wind speed changes to acceleration. That is, the reference power corresponding to the average wind speed from the wind speed information Vm to Vm # is acquired. The reward calculation unit 63B may arbitrarily determine the predetermined time T4 according to the characteristics of the wind turbine.

Further, the reward calculation unit 63B acquires the average electric energy amount which is the average of the regenerated electric power in the time interval Pave corresponding to the time interval Wave, that is, the average electric energy from the electric power information Pm to Pm #. In this example, the case where the interval delayed by a predetermined response time D with respect to the time interval Wave corresponds to the time interval Pave is shown.
The reward calculation unit 63B calculates a higher reward as the average power amount in the time section Pave is larger than the reference power corresponding to the average wind speed in the time section Wave.

FIG. 9 is a flowchart showing an operation example of the control system 50C according to the third embodiment.
First, the state detection unit 62 acquires a state parameter (step S21). Specifically, the state detection unit 62 acquires power information, wind speed information, and rotation information. The state detection unit 62 outputs the detected state parameter to the reward calculation unit 63B.

Next, the reward calculation unit 63B calculates the reward.
The reward calculation unit 63B determines whether or not the wind speed is accelerating (step S22). When the wind speed is accelerating, the reward calculation unit 63B acquires the regenerative power corresponding to the wind speed (step S23). The reward calculation unit 63B calculates a reward according to the reference power of the wind speed and the regenerative power (step S24). Specifically, the reward calculation unit 63B calculates a higher reward as the regenerative power is larger than the reference power.

On the other hand, in step S22, when the wind speed is deceleration instead of acceleration, the reward calculation unit 63B acquires the average value of the wind speed until a predetermined time elapses after the wind speed changes to acceleration (step S25). .. The reward calculation unit 63B calculates the average electric energy of the regenerated electric power in the time interval corresponding to the time interval in which the average value of the wind speed is calculated (step S26). The reward calculation unit 63B calculates a reward according to the average value of the reference power and the regenerative power according to the average value of the wind speed (step S27).

(Fourth Embodiment)
Next, a fourth embodiment will be described.
In the present embodiment, the control device 60D (see FIG. 10) differs from the above-described embodiment in that the rotation speed of the wind turbine 20 is controlled by using the trained model.
FIG. 10 is a block diagram showing an example of a schematic configuration of the wind power generation system 1D according to the modified example of the fourth embodiment. As shown in FIG. 10, the control device 60D includes a learned model storage unit 65, a determination unit 66, and a control unit 67.

The trained model storage unit 65 stores the trained model. The trained model is a database (learned model) in which information (relationship information) showing the relationship between the state of the wind power generator main body 10 to be controlled and the control for the wind power generator main body 10 is stored. The trained model is a model that estimates a parameter (hereinafter, referred to as a control index parameter) indicating an index for controlling the wind power generator main body 10 corresponding to the state of the wind power generator main body 10.

Here, the control index parameter is information that is an index for controlling the wind power generator main body 10, and may be the control parameter itself or the information used for deriving the control parameter.

For example, when the control index parameter is information that is an index for controlling the regenerative power, the control index parameter may be the power control parameter itself, or may indicate a target value of the regenerative power. It may relatively indicate the control of the regenerative power such as increasing or decreasing the regenerative power.

The trained model may be, for example, a trained model created by learning by the reinforcement learning unit 71 in the above-described embodiment, or another wind turbine having a structure similar to that of the wind turbine 20. It may be a trained model that has learned the relationship between the state and control of the wind power generation system in a wind turbine installed in an area similar to the area where the wind turbine 20 is installed.

The determination unit 66 determines control information regarding control for the wind power generator main body 10 based on the acquired control index parameters. The control information here is information indicating control determined according to the control index parameter, and is, for example, a command value of power information regarding regenerative power. That is, the determination unit 66 determines the command value of the power information based on the power index parameter. The determination unit 66 outputs the command value of the determined power information to the control unit 67.

Here, the electric power information includes, for example, information indicating a change in the regenerative power such as whether to increase or decrease the regenerative power, and also to change the regenerative power such as whether to change it stepwise or immediately. Information indicating the degree is also included.

The control unit 67 determines the control parameters for controlling the wind power generator main body 10 based on the control information determined by the determination unit 66. The control unit 67 determines, for example, a power control parameter for controlling the regenerative power so that the regenerative power approaches the command value based on the power information determined by the determination unit 66. The control unit 67 outputs the determined control parameter to the wind power generator main body 10 via the parameter acquisition unit 61.

(Fifth Embodiment)
Next, a fifth embodiment will be described.
In the present embodiment, the rotation speed of the wind turbine 20 is determined by using either the control index parameter (hereinafter, simply referred to as a parameter) output by the control device 60E using the trained model or the parameter output by the learning device 70. It differs from the above-described embodiment in that it is controlled.
FIG. 11 is a block diagram showing an example of a schematic configuration of the wind power generation system 1E according to the modified example of the fifth embodiment. As shown in FIG. 11, the control device 60E includes a selection unit 68.

The selection unit 68 outputs either a parameter output from the trained model stored in the trained model storage unit 65 or a parameter output by the learning device 70 to the determination unit 66. The selection unit 68 may determine which one to select according to a predetermined phase. The selection unit 68 selects, for example, the parameters output by the learning device 70 in the learning phase in which the learning device 70 learns the control of the rotation speed of the wind turbine 20. On the other hand, in the selection unit 68, the learning model learned to control the rotation speed of the windmill 20 is stored in the learned model storage unit 65, and in the control phase in which the learning model is used to control the rotation speed of the windmill 20. Select the parameter output from the trained model stored in the trained model storage unit 65.

Further, in at least one embodiment described above, the content learned by the reinforcement learning unit 71 is stored in the trained model storage unit 65 and other storage units (not shown), and further learning is performed based on the stored content. You may try to proceed. As a result, the model learned about some basic control common to the wind turbine 20 can be used for wind conditions in the area where the wind turbine 20 is installed, seasonal wind conditions, differences in wind conditions depending on the time of day and night, weather, etc. It is possible to perform control according to the state.

As described above, the control system 50E of the fifth embodiment is a control system that controls the regenerated power generated by the wind power generation system 1, and indicates the wind speed at the installation location of the wind turbine 20 of the wind power generation system 1. Wind speed based on wind speed information, rotation information on the rotation of the wind turbine 20, power information on the regenerated power generated by the wind power generation system 1, information on the relationship between the wind speed and the regenerated power, and rewards according to the power information on the wind speed. The enhanced learning unit 71 that learns the correspondence information between the wind power and the regenerative power, the learned model storage unit 65 that stores the correspondence information between the wind speed and the regenerative power, and the power control parameter that controls the regenerative power are set in the wind power generation system 1. The command value of the power information is determined based on the state detection unit 62 that detects the rotation information, the power information, and the wind speed information in the case of the above, the rotation information and the wind speed information detected by the state detection unit 62, and the corresponding information. It includes a determination unit 66 and a control unit 67 that controls regenerative power based on a command value determined by the determination unit 66. As a result, the control system 50E of the fifth embodiment can make the reinforcement learning unit 71 learn based on the reward according to the power information for the wind speed, and performs appropriate control according to the relationship between the wind speed and the regenerative power. be able to.

Further, the control system 50 of the first embodiment further includes a reward calculation unit 63 that calculates a reward based on the wind speed information and the electric power information detected by the state detection unit 62. As a result, the control system 50 of the first embodiment can perform more appropriate control according to the relationship between the wind speed detected by the state detection unit 62 and the regenerative power.

Further, in the control system 50 of the first embodiment, the reward calculation unit 63 calculates the reward using different reward conditions based on whether or not the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value. do. As a result, the control system 50 of the first embodiment may cause erroneous learning such that it may be difficult to secure the safety of the wind turbine if the reward is calculated under the same reward conditions as in the normal time in a strong wind. In some cases, the reward conditions can be changed between normal times and strong winds, so more appropriate control can be performed.

Further, in the control system 50 of the first embodiment, when the wind speed indicated in the wind speed information is less than a predetermined strong wind threshold value, the reward calculation unit 63 determines that the rotation speed of the wind turbine 20 indicated in the rotation information is a predetermined rotation threshold value. When the above is the case, the reward of the first level is calculated, and when the rotation speed is less than the predetermined rotation threshold value, the reward of the second level lower than the first level is calculated. As a result, the control system 50 of the first embodiment can calculate a higher reward and can perform more appropriate control when the control system 50 is normally controlled so as to increase the amount of power generation.

Further, in the control system 50 of the first embodiment, when the wind speed indicated in the wind speed information is equal to or higher than a predetermined strong wind threshold value, the reward calculation unit 63 determines that the rotation speed of the wind turbine 20 indicated in the rotation information is a predetermined rotation threshold value. If the above is the case, the reward of the third level is calculated, and if the rotation speed is less than the predetermined rotation threshold value, the reward of the fourth level higher than the third level is calculated. As a result, the control system 50 of the first embodiment can calculate a higher reward when the rotation speed is suppressed and controlled safely in a strong wind, and more appropriate control can be performed.

Further, in the control system 50 of the first embodiment, the reward calculation unit 63 may calculate the reward based on the wind speed information acquired in the past from the time when the power information is detected. As a result, the control system 50 of the first embodiment has a past wind speed even if it may take a predetermined response time from the setting of the power control parameter to the actual setting value of the power. Since the reward can be calculated according to the above, more appropriate control can be performed.

Further, the learning device 70 of the first embodiment has wind speed information indicating the wind speed at the installation location of the wind turbine 20 of the wind power generation system 1, rotation information regarding the rotation of the wind turbine 20, and electric power related to the regenerative power generated by the wind power generation system 1. The enhanced learning unit 71 is provided to learn the correspondence information between the wind speed and the regenerated power based on the information, the relation information between the wind speed and the regenerated power, and the reward according to the power information for the wind speed. Therefore, the learning device 70 of the first embodiment can learn how to control the regenerative power by using the reward based on the power according to the wind speed as a clue, and therefore performs more appropriate control. be able to.

Further, the control device 60 of the first embodiment has wind speed information indicating the wind speed at the installation location of the wind turbine 20 of the wind power generation system 1, rotation information regarding the rotation of the wind turbine 20, and electric power related to the regenerative power generated by the wind power generation system 1. The enhanced learning unit 71 that learns the correspondence information between the wind speed and the regenerated power based on the information, the relation information between the wind speed and the regenerated power, and the reward according to the power information for the wind speed, and the power control that controls the regenerated power. Based on the state detection unit 62 that detects rotation information, power information, and wind speed information when the parameters are set in the wind power generation system 1, and the rotation information, wind speed information, and corresponding information detected by the state detection unit 62. It includes a determination unit 66 that determines a command value of power information, and a control unit 67 that controls regenerated power based on the command value determined by the determination unit 66. As a result, the control device 60 of the first embodiment can control the regenerated power by using the power control parameter output by the reinforcement learning unit 71 that learns based on the reward according to the power information for the wind speed. , More appropriate control can be performed.

Further, the control device 60 of the fourth embodiment has a state detection unit 62 for detecting rotation information, power information, and wind speed information when a power control parameter for controlling regenerated power is set in the wind power generation system 1, and a state. The command value of the power information is determined based on the rotation information and the wind speed information detected by the detection unit 62, and the correspondence information (for example, the correspondence information between the wind speed and the regenerative power stored in the learned model storage unit 65). It includes a determination unit 66 and a control unit 67 that controls regenerative power based on a command value determined by the determination unit 66. As a result, the control device 60 of the fourth embodiment can perform control according to the power control parameter based on the corresponding information, and more appropriate control can be performed.

As described above, the control system 50E of the fifth embodiment is a control system that controls the regenerated power generated by the wind power generation system 1, and indicates the wind speed at the installation location of the wind turbine 20 of the wind power generation system 1. Wind speed information, rotation information regarding the rotation of the wind turbine 20, power information regarding the regenerative power generated by the wind power generation system 1, information on the relationship between the wind speed and the regenerative power, and after the response time in the control of the regenerative power with respect to the wind speed has elapsed. The enhanced learning unit 71 that learns the correspondence information between the wind speed and the regenerative power based on the reward according to the power information, the learned model storage unit 65 that stores the correspondence information between the wind speed and the regenerative power, and the regenerative power. A state detection unit 62 that detects rotation information, power information, and wind speed information when the power control parameter to be controlled is set in the wind power generation system 1, and rotation information, wind speed information, and corresponding information detected by the state detection unit 62. A determination unit 66 that determines a command value of power information based on the above, and a control unit 67 that controls regenerated power based on the command value determined by the determination unit 66 are provided. As a result, the control system 50E of the fifth embodiment can make the reinforcement learning unit 71 learn based on the reward according to the power information in consideration of the response time to the wind speed, and controls according to the response time. Can be done.

Further, the control system 50A of the second embodiment further includes a response time estimation unit 69 that estimates the response time based on the time-series change of the wind velocity information and the time-series change of the power information, and the reward calculation unit 63A Based on the response time estimated by the response time estimation unit 69, the wind speed information detected by the state detection unit 62, and the power information, depending on the power information after the response time has elapsed from the time when the wind speed information was detected. Calculate the reward. As a result, in the control system 50A of the second embodiment, the response time is estimated by the response time estimation unit 69 based on the time-series changes of both the wind speed information and the power information, so that the response time depends on the state of the change in the wind speed. Even when the time changes, the response time according to the time-series change at that time can be estimated, and the control can be performed according to a more appropriate response time.

Further, in the control system 50B of the third embodiment, the reward calculation unit 63B calculates the reward using different reward conditions based on whether or not the wind speed shown in the wind speed information is in the acceleration section where the wind speed accelerates. .. As a result, the control system 50B of the third embodiment can be used for acceleration and deceleration even when the total regenerative power amount is calculated when the reward is calculated under the same reward conditions as for acceleration during deceleration. Since the reward conditions can be changed with, more appropriate control can be performed.

Further, in the control system 50B of the third embodiment, in the reward calculation unit 63B, when the wind speed shown in the wind speed information is in the acceleration section where the wind speed accelerates, the reference power value according to the wind speed and the power value shown in the power information are used. Calculate the reward according to the difference between. Thereby, the control system 50B of the third embodiment can learn the control to increase the power generation when there is almost no possibility that the wind turbine 20 stalls at the time of acceleration, and performs more appropriate control. be able to.

Further, in the control system 50B of the third embodiment, the reward calculation unit 63B corresponds to the wind speed including the wind speed information after the wind speed changes to acceleration when the wind speed is in the deceleration section in which the wind speed is decelerated. Calculate the reward according to the power value shown in the power information. As a result, the control system 50B of the third embodiment accelerates the wind speed without increasing the power generation when the wind turbine 20 may stall when the regenerative power is increased during deceleration. By increasing the regenerative power after turning, it is possible to learn the control that increases the total amount of power generation, and it is possible to perform more appropriate control.

Further, in the control system 50B of the third embodiment, when the reward calculation unit 63B is in the deceleration section in which the wind speed shown in the wind speed information decelerates, the average value of the wind speed from the wind speed to the elapse of a predetermined time and the wind speed are used. The reward according to the average value in the power value shown in the power information corresponding to the wind speed information until the predetermined time elapses is calculated. As a result, the control system 50B of the third embodiment does not increase the power generation when the wind turbine 20 may stall when the regenerative power is increased during deceleration, and then the regenerative power is subsequently increased. By increasing the number of units, it is possible to learn the control that increases the total amount of power generation, and it is possible to perform more appropriate control.

Further, the learning device 70 of the second embodiment has wind speed information indicating the wind speed at the installation location of the wind turbine 20 of the wind power generation system 1, rotation information regarding the rotation of the wind turbine 20, and electric power related to the regenerative power generated by the wind power generation system 1. Enhancement to learn the correspondence information between wind speed and regenerated power based on the information and the relation information between the wind speed and the regenerated power and the reward according to the power information after the response time in the control of the regenerated power with respect to the wind speed has elapsed. A learning unit 71 is provided. Therefore, the learning device 70 of the first embodiment can learn how to control the regenerative power by using the reward based on the power according to the wind speed as a clue, and therefore performs more appropriate control. be able to.

Further, the control device 60A of the second embodiment has wind speed information indicating the wind speed at the installation location of the wind turbine 20 of the wind power generation system 1, rotation information regarding the rotation of the wind turbine 20, and electric power related to the regenerative power generated by the wind power generation system 1. Enhancement to learn the correspondence information between wind speed and regenerated power based on the information and the relationship information between wind speed and regenerated power and the reward according to the power information after the response time in the control of the regenerated power to the wind speed has elapsed. The learning unit 71, the state detection unit 62 that detects rotation information, power information, and wind speed information when the power control parameter that controls the regenerative power is set in the wind power generation system 1, and the rotation detected by the state detection unit 62. It includes a determination unit 66 that determines a command value of power information based on information, wind speed information, and correspondence information, and a control unit 67 that controls regenerated power based on the command value determined by the determination unit 66. As a result, the control device 60 of the first embodiment can control the regenerated power by using the power control parameter output by the reinforcement learning unit 71 that learns based on the reward according to the power information for the wind speed. , More appropriate control can be performed.

All or part of the processing performed by each of the control system 50, the control device 60, and the learning device 70 in the above-described embodiment may be realized by a computer. In that case, a program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read by a computer system and executed. The term "computer system" as used herein includes hardware such as an OS and peripheral devices. Further, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, and a storage device such as a hard disk built in a computer system. Further, a "computer-readable recording medium" is a communication line for transmitting a program via a network such as the Internet or a communication line such as a telephone line, and dynamically holds the program for a short period of time. It may also include a program that holds a program for a certain period of time, such as a volatile memory inside a computer system that is a server or a client in that case. Further, the above program may be for realizing a part of the above-mentioned functions, and may be further realized for realizing the above-mentioned functions in combination with a program already recorded in the computer system. It may be realized by using a programmable logic device such as FPGA.

Although the embodiments of the present invention have been described in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs and the like within a range that does not deviate from the gist of the present invention.

1 Wind power generation system 10 Wind power generator body 20 Wind turbine 30 Generator 31 Rectification / boosting unit 32 Voltage detection unit 33 Current detection unit 41 Wind speed sensor 42 Rotation speed sensor 50 Control system 60 Control device 61 Parameter acquisition unit 62 State detection unit 63 Reward Calculation unit 64 Reward output unit 65 Learned model storage unit 66 Decision unit 67 Control unit 68 Selection unit 69 Response time estimation unit 70 Learning device 71 Enhanced learning unit

Claims (8)

It is a control system that controls the regenerative power generated by the wind power generation system.
Wind speed information indicating the wind speed at the installation location of the wind turbine of the wind power generation system, rotation information regarding the rotation of the wind turbine, power information regarding the regenerative power generated by the wind power generation system, and information on the relationship between the wind speed and the regenerative power. And a learning unit that learns correspondence information between the wind speed and the regenerated power based on the reward corresponding to the power information for the wind speed.
A storage unit that stores information on the correspondence between the wind speed and the regenerative power,
A state detection unit that detects the rotation information, the power information, and the wind speed information when the power control parameter for controlling the regenerative power is set in the wind power generation system.
A determination unit that determines a command value of the power information based on the rotation information, the wind speed information, and the corresponding information detected by the state detection unit.
A control unit that controls the regenerative power based on the command value determined by the determination unit, and
With the reward calculation unit that calculates the reward based on the wind speed information, the rotation information, and the power information detected by the state detection unit.
Equipped with
When the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value, the reward calculation unit calculates the reward according to the rotation speed of the wind turbine shown in the rotation information, and the wind speed shown in the wind speed information. Is less than a predetermined strong wind threshold, the reward is calculated according to the power information.
A control system characterized by that. When the wind speed indicated in the wind speed information detected by the state detection unit is less than a predetermined strong wind threshold value, the reward calculation unit has previously combined the regenerative power indicated in the power information detected by the state detection unit with the regenerative power. Based on the stored reference power, the higher the regenerative power is, the higher the reward is calculated.
The control system according to claim 1 . When the wind speed indicated in the wind speed information detected by the state detection unit is equal to or higher than a predetermined strong wind threshold value, the reward calculation unit has the number of rotations of the wind turbine indicated in the rotation information detected by the state detection unit. And the rotation threshold value stored in advance, the higher the rotation speed is calculated as the rotation speed is smaller than the rotation threshold value and the rotation speed is smaller than the rotation threshold value .
The control system according to claim 1 or 2 . The reward calculation unit calculates the reward based on the wind speed information acquired by the state detection unit in the past from the time when the power information is detected by the state detection unit.
The control system according to any one of claims 1 to 3 . Wind speed information indicating the wind speed at the installation location of the wind turbine of the wind power generation system, rotation information regarding the rotation of the wind turbine, power information regarding the regenerative power generated by the wind power generation system, and information on the relationship between the wind speed and the regenerative power. A learning unit for learning the correspondence information between the wind speed and the regenerated power based on the reward corresponding to the power information for the wind speed is provided .
The reward is calculated based on the wind speed information, the rotation information, and the power information detected when the power control parameter for controlling the regenerative power is set in the wind power generation system, and the wind speed shown in the wind speed information. Is calculated according to the rotation speed of the wind turbine shown in the rotation information when is equal to or higher than a predetermined strong wind threshold, and is calculated according to the power information when the wind speed indicated in the wind speed information is less than the predetermined strong wind threshold. Calculated
Learning device. Wind speed information indicating the wind speed at the installation location of the wind turbine of the wind power generation system, rotation information regarding the rotation of the wind turbine, power information regarding the regenerative power generated by the wind power generation system, and information on the relationship between the wind speed and the regenerative power. A learning unit that learns correspondence information between the wind speed and the regenerated power based on the reward corresponding to the power information for the wind speed.
A state detection unit that detects the rotation information, the power information, and the wind speed information when the power control parameter for controlling the regenerative power is set in the wind power generation system.
A determination unit that determines a command value of the power information based on the rotation information, the wind speed information, and the corresponding information detected by the state detection unit.
A control unit that controls the regenerative power based on the command value determined by the determination unit, and
With the reward calculation unit that calculates the reward based on the wind speed information, the rotation information, and the power information detected by the state detection unit.
Equipped with
When the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value, the reward calculation unit calculates the reward according to the rotation speed of the wind turbine shown in the rotation information, and the wind speed shown in the wind speed information. Is less than a predetermined strong wind threshold, the reward is calculated according to the power information.
Control device. Rotation information regarding the rotation of the wind turbine of the wind power generation system when the power control parameter for controlling the regenerative power generated by the wind power generation system is set in the wind power generation system, and power information regarding the regenerative power generated by the wind power generation system. , And a state detection unit that detects wind speed information indicating the wind speed at the installation location of the wind turbine.
A determination unit that determines a command value of the power information based on the rotation information and the wind speed information detected by the state detection unit, and the correspondence information between the wind speed and the regenerative power.
A control unit that controls the regenerative power based on the command value determined by the determination unit is provided .
The correspondence information is information learned based on the wind speed information, the rotation information, the power information, the relationship information between the wind speed and the regenerative power, and the reward corresponding to the power information for the wind speed. ,
The reward is calculated based on the wind speed information, the rotation information, and the power information detected by the state detection unit, and when the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value, the rotation information is used. It is calculated according to the rotation speed of the wind turbine shown, and is calculated according to the power information when the wind speed shown in the wind speed information is less than a predetermined strong wind threshold value.
Control device. It is a control method that controls the regenerative power generated by the wind power generation system.
The learning unit has wind speed information indicating the wind speed at the installation location of the wind turbine of the wind power generation system, rotation information regarding the rotation of the wind turbine, power information regarding the regenerative power generated by the wind power generation system, and the wind speed and the regenerative power. Based on the relationship information with the wind speed and the reward corresponding to the power information for the wind speed, the correspondence information between the wind speed and the regenerated power is learned.
The storage unit stores the correspondence information between the wind speed and the regenerative power .
The state detection unit detects the rotation information, the power information, and the wind speed information when the power control parameter for controlling the regenerative power is set in the wind power generation system.
The determination unit determines the command value of the power information based on the rotation information, the wind speed information, and the corresponding information detected by the state detection unit.
The control unit controls the regenerative power based on the command value determined by the determination unit.
The reward calculation unit calculates the reward based on the wind speed information, the rotation information, and the power information detected by the state detection unit, and the wind speed shown in the wind speed information is equal to or higher than a predetermined strong wind threshold value. In this case, the reward is calculated according to the rotation speed of the wind turbine shown in the rotation information, and the reward is calculated according to the power information when the wind speed shown in the wind speed information is less than a predetermined strong wind threshold . Control method.

Images