Power Generation System

Abstract

To provide a power generating system in which a voltage superimposition device controls a voltage applied by following changes in an AC voltage output by a generator.SOLUTION: A power generating system 100 includes a wind power generation device 10, a voltage superimposition device 30, a control device 40, and a prediction device 90. The voltage superimposition device 30 applies a voltage with a waveform of an opposite phase to a waveform of a noise component out of an AC voltage output by a generator 18 of the wind power generation device 10. The prediction device 90 stores mapping data 94A that defines a mapping M that outputs an output variable that indicates a prediction waveform of the noise component after a predetermined constant time by inputting a plurality of input variables. The mapping M is learned by machine learning in advance. The prediction device 90 executes acquiring processing for acquiring input variables, and calculation processing for outputting a value of the output variable by inputting the plurality of input variables to the mapping. The control device 40 controls a voltage applied by the voltage superimposition device 30 based on the prediction waveform shown by the output variable.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a power generation system. [Background technology]

[0002] The power generation system of Patent Document 1 includes a wind power generation device, a sound collector, a speaker, and a control circuit. The wind power generation device includes a wind turbine and a generator. The sound collector acquires noise generated by the wind power generation device. The control circuit converts the waveform of the noise acquired by the sound collector into a waveform with an opposite phase. The control circuit then outputs the sound with the opposite phase waveform from the speaker toward the wind power generation device. As a result, part of the noise generated by the wind power generation device is cancelled out by the sound output from the speaker. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2014-227977 A Summary of the Invention [Problem to be solved by the invention]

[0004] In a power generation system such as that described in Patent Document 1, when the speed of the wind striking the windmill of the wind power generation device changes, the rotation speed of the windmill and the rotation speed of the rotating shaft of the generator change. The waveform of the noise generated by the wind power generation device changes according to the rotation speed of the rotating shaft. If the waveform conversion process by the control circuit cannot keep up with such changes in the noise waveform, there is a risk that the noise generated by the wind power generation device cannot be sufficiently suppressed. [Means for solving the problem]

[0005] A power generation system for solving the above problem includes a wind power generation device having a wind turbine and a generator, a voltage superposition device that applies a voltage with a waveform in opposite phase to the waveform of a noise component of the AC voltage output by the generator, a control device that controls the voltage superposition device, and a prediction device that predicts the waveform of the noise component, wherein the prediction device includes an execution device and a memory device, and the memory device stores mapping data that specifies a mapping that outputs an output variable indicating a predicted waveform of the noise component after a predetermined fixed time by inputting a plurality of input variables, the mapping being learned in advance by machine learning, and the mapping includes a variable indicating a wind speed of wind colliding with the wind turbine as one of the input variables, the execution device executes an acquisition process that acquires the input variables, and a calculation process that outputs a value of the output variable by inputting the plurality of input variables acquired by the acquisition process into the mapping, and the control device controls the voltage applied by the voltage superposition device based on the predicted waveform indicated by the output variable.

[0006] According to the above configuration, by applying a voltage having a waveform in opposite phase to the waveform of the noise component of the AC voltage output by the generator, the AC voltage output by the generator approaches an ideal sine wave. Therefore, the rotating shaft of the generator rotates smoothly, and the noise generated by the generator can be suppressed. The voltage having the waveform in opposite phase is generated based on the predicted waveform of the noise component after a certain time. In this way, if the predicted waveform of the noise component after a certain time can be grasped, the voltage applied by the voltage superposition device can be controlled in accordance with the change in the AC voltage output by the generator, even in a situation where a control delay or the like may occur in the control device. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a schematic diagram of a power generation system. [Diagram 2] 13 is a flowchart showing prediction control. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] <Outline of power generation system configuration> Hereinafter, one embodiment of the present invention will be described with reference to Figures 1 and 2. First, a schematic configuration of a power generation system 100 will be described.

[0009] As shown in FIG. 1, the power generation system 100 includes a wind power generation device 10, a voltage superposition device 30, and a control device 40. The wind power generation device 10 includes a tower 11, a nacelle 12, and a wind turbine 13. The wind power generation device 10 also includes a speed increaser 16, a brake device 17, a generator 18, and a power line 19. The wind turbine 13 can rotate relative to the nacelle 12 by wind colliding with the wind turbine 13. The nacelle 12 houses the speed increaser 16, the brake device 17, and the generator 18. The generator 18 includes a rotating shaft 18A. The rotating shaft 18A is connected to the wind turbine 13 via the speed increaser 16 and the brake device 17. Therefore, when the wind turbine 13 rotates, the rotation is accelerated by the speed increaser 16 and transmitted to the rotating shaft 18A. The power line 19 supplies the power generated by the generator 18 to electrical equipment (not shown). The brake device 17 decelerates the wind turbine 13 when the rotation speed of the wind turbine 13 becomes excessive.

[0010] The voltage superposition device 30 is located midway along the power line 19. The voltage superposition device 30 applies a voltage having a waveform in opposite phase to the waveform of a noise component in the AC voltage output by the generator 18. The control device 40 and the voltage superposition device 30 can communicate with each other. The control device 40 controls the voltage superposition device 30 by outputting a control signal to the voltage superposition device 30. Specifically, the control device 40 controls the voltage output by the voltage superposition device 30 based on a predicted waveform of the noise component calculated by a prediction device 90 described later. The control device 40 can detect the AC voltage in the power line 19 via the voltage superposition device 30.

[0011] The power generation system 100 includes a wind speed sensor 51, a temperature sensor 52, a humidity sensor 53, a rotation shaft angle sensor 54, and a communication device 57. The wind speed sensor 51 detects the wind speed SW, which is the flow speed of the wind that hits the wind turbine 13. The temperature sensor 52 detects the internal temperature TE, which is the temperature inside the nacelle 12. As an example, the temperature sensor 52 detects the temperature inside the nacelle 12 near the generator 18 as the internal temperature TE. The humidity sensor 53 detects the internal humidity LH, which is the humidity inside the nacelle 12. As an example, the humidity sensor 53 detects the humidity inside the nacelle 12 near the generator 18 as the internal humidity LH. The rotation shaft angle sensor 54 detects the rotation shaft angle SR, which is the angular position of the rotation shaft 18A of the generator 18. The communication device 57 can acquire date information ID and time information IT by communicating with an external device of the power generation system 100. Here, the date information ID is information including the "month" and "day" as the current date. The time information IT is information including the "hour" and "minute" as the current time.

[0012] The power generation system 100 includes a prediction device 90. The prediction device 90 acquires signals indicating various values ​​from the wind speed sensor 51, the temperature sensor 52, the humidity sensor 53, the rotation shaft angle sensor 54, and the communication device 57. The prediction device 90 can detect the AC voltage in the power line 19 via the control device 40 and the voltage superposition device 30. The prediction device 90 calculates a generator rotation speed NR, which is the rotation speed of the rotation shaft 18A, based on the rotation shaft angle SR.

[0013] The prediction device 90 predicts noise components in the AC voltage output by the generator 18. Specifically, the prediction device 90 includes a CPU 91, a peripheral circuit 92, a ROM 93, a storage device 94, and a bus 95. The ROM 93 stores various programs in advance for the CPU 91 to execute various processes. The storage device 94 stores the above-mentioned various values ​​acquired by the prediction device 90. The storage device 94 stores mapping data 94A that defines a mapping M learned by machine learning in advance. The mapping M outputs an output variable that indicates a predicted waveform of the above-mentioned noise component after a predetermined fixed time by inputting a plurality of input variables. A specific description of the mapping M will be given later. In this embodiment, the CPU 91 and the ROM 93 are execution devices. The storage device 94 is a storage device. The peripheral circuit 92 includes a circuit that generates a clock signal that defines the internal operation, a power supply circuit, a reset circuit, and the like.

[0014] <Predictive Control> Next, a description will be given of the predictive control executed by the predicting device 90. The predicting device 90 repeatedly executes predictive control on the condition that, for example, the generator rotation speed NR is greater than zero.

[0015] 2, when the CPU 91 starts predictive control, it executes the process of step S11. In step S11, the CPU 91 accesses the storage device 94 to acquire various latest values ​​at the time of the process of step S11. Specifically, the CPU 91 acquires the wind speed SW, the internal temperature TE, the internal humidity LH, the date information ID, the time information IT, and the generator rotation speed NR. In this embodiment, the process of step S11 is the acquisition process. After step S11, the CPU 91 advances the process to step S12.

[0016] In step S12, the CPU 91 generates various values ​​acquired in the processing of step S11 as input variables x(1) to x(6) for the mapping M. Specifically, the CPU 91 assigns the wind speed SW to the input variable x(1). The CPU 91 assigns the internal temperature TE to the input variable x(2). The CPU 91 assigns the internal humidity LH to the input variable x(3). The CPU 91 assigns the date information ID to the input variable x(4). The CPU 91 assigns the time information IT to the input variable x(5). The CPU 91 assigns the generator rotation speed NR to the input variable x(6). In this embodiment, the input variable x(1) is a variable indicating the wind speed SW, which is the flow speed of the wind that hits the wind turbine 13. After step S12, the CPU 91 advances the processing to step S13.

[0017] In step S13, the CPU 91 calculates the value of the output variable y(i) by inputting the input variables x(1) to x(6) generated in the processing of step S12 and the input variable x(0) as a bias parameter to the mapping M. The CPU 91 also calculates a predicted waveform indicated by the value of the output variable y(i). This predicted waveform is a predicted waveform of the noise component of the AC voltage output by the generator 18.

[0018] An example of the mapping M defined by the mapping data 94A is a function approximator, which is a fully connected forward propagation type neural network with one intermediate layer. Specifically, in the mapping M, the input variables x(1) to x(6) and the input variable x(0) as a bias parameter are transformed by a linear mapping defined by coefficients wFjk (j=1 to m, k=0 to 6), and each of the "m" values ​​is substituted into the activation function f. As a result, the values ​​of the nodes in the intermediate layer are determined. Also, the values ​​of the nodes in the intermediate layer transformed by the linear mapping defined by coefficients wSij (i=1 to P) are substituted into the activation function g, and the output variables y(1) to y(P) are determined. The sum of the output variables in step S13 is described as "P". This "P" is determined, for example, as follows. First, the lower limit and upper limit of the noise component that may be generated in the generator 18 are obtained by experiments and simulations. Furthermore, the frequency band from the lower limit to the upper limit of the obtained noise component is divided into a plurality of frequency bands without overlapping. The number of divided frequency bands is the sum of the output variables, "P". An example of the sum of the output variables, "P", is "10". Each of the output variables y(1) to y(P) is an output variable indicating a predicted waveform of the above noise component after a predetermined fixed time. Here, "after a certain time" refers to, for example, several milliseconds to several tens of milliseconds from the current time point. In this embodiment, an example of the activation function f is the ReLU function. An example of the activation function g is the sigmoid function.

[0019] In this embodiment, the map M is generated, for example, as follows. As a premise, in the wind power generation device 10, when the wind speed SW colliding with the wind turbine 13 changes, the generator rotation speed NR changes thereafter. Then, the AC voltage output by the generator 18 changes. That is, when the wind speed SW changes, the noise component of the AC voltage output by the generator 18 changes after a certain time from the change. Therefore, in learning the map M, a plurality of wind power generation devices 10 for testing are prepared. Also, while operating each wind power generation device 10 under various conditions, the wind speed SW, the internal temperature TE, the internal humidity LH, the date information ID, the time information IT, the generator rotation speed NR, and the above noise components are continuously acquired. At this time, the wind speed SW, the internal temperature TE, the internal humidity LH, the date information ID, the time information IT, and the generator rotation speed NR for a specific time are associated with the noise components after a certain time from the specific time as a set of data. Then, for each set of data, input variables x(1) through x(6) are generated based on the wind speed SW, etc., and output variables y(1) through y(P) are generated based on the noise components. Furthermore, by inputting the input variables x(1) through x(6) as training data and the output variables y(1) through y(P) as teacher data to mapping M, mapping M is trained by machine learning.

[0020] In this embodiment, the processes of steps S12 and S13 are calculation processes that output an output variable indicating a predicted waveform of the above-mentioned noise component after a predetermined fixed time. After step S13, the CPU 91 ends the current prediction control.

[0021] <Action of this embodiment> In the power generation system 100, for each frequency band corresponding to "P", which is the sum of the output variables, the voltage superimposing device 30 applies a voltage having a waveform in opposite phase to the waveform of the noise component in the AC voltage output by the generator 18. Therefore, the AC voltage output by the generator 18 approaches an ideal sine wave. This allows the rotating shaft 18A of the generator 18 to rotate smoothly, thereby suppressing the noise generated by the generator 18.

[0022] <Effects of this embodiment> (1) In this embodiment, the voltage with the opposite phase waveform is generated based on the predicted waveform of the noise component after a certain time predicted by the prediction device 90. If the predicted waveform of the noise component after a certain time can be grasped in this way, the voltage applied by the voltage superposition device 30 is controlled in accordance with the predicted waveform of the noise component after a certain time, for example, even in a situation where the noise component changes due to a change in the wind speed SW. As a result, even in a situation where a control delay or the like may occur in the control device 40, the voltage applied by the voltage superposition device 30 can be controlled in accordance with the change in the AC voltage output by the generator 18.

[0023] (2) In this embodiment, the prediction device 90 predicts the predicted waveform of the noise component after a certain time period by taking into account the internal temperature TE, the internal humidity LH, the date information ID, the time information IT, and the generator rotation speed NR in addition to the wind speed SW. This improves the prediction accuracy of the predicted waveform of the noise component after a certain time period.

[0024] <Example of change> This embodiment can be modified as follows: This embodiment and the following modifications can be combined with each other to the extent that there is no technical contradiction.

[0025] In the above embodiment, as long as the mapping M includes the wind speed SW as an input variable, other input variables may be omitted or changed. For example, the input variables of the internal temperature TE, the internal humidity LH, the date information ID, the time information IT, and the generator rotation speed NR may be omitted.

[0026] In the above embodiment, further learning of the mapping M may be performed. For example, when the wind power generation device 10 is operating, the prediction device 90 acquires the AC voltage in the power line 19 via the control device 40 and the voltage superposition device 30. Also, when the wind power generation device 10 is operating, the prediction device 90 acquires a signal indicating the sound pressure of the noise from a sound collection device such as a microphone. Then, the prediction device 90 extracts a noise component when the sound pressure of the noise is equal to or higher than a predetermined threshold. Furthermore, the prediction device 90 identifies the wind speed SW etc. a certain time before the point in time when the noise component was extracted. Then, the prediction device 90 may further learn the mapping M in the same procedure as the above embodiment by associating the extracted noise component and the identified wind speed SW etc. as a set of data. With this configuration, it is expected that the prediction accuracy of the noise component that causes the noise can be improved.

[0027] For example, when the wind power generation device 10 is operating, the prediction device 90 acquires the AC voltage in the power line 19 via the control device 40 and the voltage superposition device 30. Also, when the wind power generation device 10 is operating, the prediction device 90 acquires a signal indicating vibration from a vibration sensor. Then, the prediction device 90 extracts a noise component when the vibration is equal to or greater than a predetermined threshold. Furthermore, the prediction device 90 identifies the wind speed SW etc. a certain time before the point in time when the noise component was extracted. Then, the prediction device 90 may further learn the mapping M in the same procedure as in the above embodiment by associating the extracted noise component with the identified wind speed SW etc. as a set of data. With this configuration, it is expected that the prediction accuracy of the noise component causing the vibration can be improved. [Explanation of symbols]

[0028] M...mapping, 10...wind power generation device, 30...voltage superposition device, 40...control device, 90...prediction device, 91...CPU, 92...peripheral circuit, 93...ROM, 94...storage device, 94A...mapping data, 95...bus, 100...power generation system.

Claims

[Claim 1] A wind power generation device having a wind turbine and a generator; a voltage superposition device that applies a voltage having a waveform inversely phase to a waveform of a noise component in the AC voltage output by the generator; A control device for controlling the voltage superposition device; A power generation system including a prediction device that predicts a waveform of the noise component, The prediction device includes an execution device and a storage device, the storage device stores mapping data that defines a mapping for outputting an output variable indicating a predicted waveform of the noise component after a predetermined fixed time period in response to a plurality of input variables; The mapping is learned in advance by machine learning, the mapping includes, as one of the input variables, a variable indicating a wind speed of wind colliding with the wind turbine; The execution device is An acquisition process for acquiring the input variables; a calculation process for inputting the plurality of input variables acquired by the acquisition process into the mapping, thereby outputting a value of the output variable; The control device controls the voltage applied by the voltage superimposition device based on the predicted waveform represented by the output variable. Power generation system.

Images