OCEAN WAVE POWER GENERATOR WITH ARTIFICIALLY INTELLIGENT CONTROLLER

Abstract

The ocean wave power generator with an artificially intelligent controller is a wave power generator based on a two-body mass-spring-damper system, including a first mass, a second mass, and a linear generator coupled to the second mass. A linear actuator is coupled to the second mass, and first and second motion sensors are positioned for detecting position and speed of the first and second masses. The maximum power output of the linear generator is determined based on the position and the speed of the first mass, and an ideal position and an ideal speed of the second mass, corresponding to the maximum power output of the linear generator and the position and the speed of the first mass, are determined. The position and the speed of the second mass are adjusted using a linear actuator to match the ideal position and the ideal speed of the second mass.

Claims

1. An ocean wave power generator with an artificially intelligent controller, comprising: a first mass; a second mass; a first spring resiliently coupling the first mass and the second mass; a first damper joining the first mass and the second mass for damping relative oscillation between the first mass and the second mass; a second spring resiliently coupling the second mass and a support surface; a second damper joining the second mass and the support surface for damping relative oscillation between the second mass and the support surface; a linear generator mounted on the support surface, the linear generator being coupled to the second mass, relative oscillation between the second mass and the support surface driving the linear generator to generate power; a linear actuator coupled to the second mass; a first motion sensor positioned for detecting position and speed of the first mass; a second motion sensor positioned for detecting position and speed of the second mass; and a controller connected to the first and second motion sensors for receiving the position and the speed of the first mass from the first motion sensor and the position and the speed of the second mass from the second motion sensor, the controller being configured to output a driving signal to the linear actuator to drive oscillatory motion of the second mass to continuously optimize power output of the linear generator based on the position and, the speed of the first mass and the position and the speed of the second mass.

2. A method of controlling an ocean wave power generator, comprising the steps of: providing an ocean wave power generator having a first mass, a second mass, a first spring resiliently coupling the first mass and the second mass, a first damper joining the first mass and the second mass for damping relative oscillation between the first mass and the second mass, a second spring resiliently coupling the second mass and a support surface, a second damper joining the second mass and the support surface for damping relative oscillation between the second mass and the support surface, a linear generator mounted on the support surface, the linear generator being coupled to the second mass, and a linear actuator coupled to the second mass; detecting position and speed of the first mass; detecting position and speed of the second mass; determining maximum power output of the linear generator based on the position and the speed of the first mass; determining an ideal position and an ideal speed of the second mass corresponding to the maximum power output of the linear generator and the position and the speed of the first mass; and continuously adjusting the position and the speed of the second mass to obtain the determined ideal position and ideal speed of the second mass,

3. The method of controlling an ocean wave power generator as recited in claim 2, wherein the step of determining, the maximum power output of the linear generator and the ideal position and the ideal speed of the second mass comprises determining the maximum power output.sub.;and the ideal position and the ideal speed from a lookup table.

4. The method of controlling an ocean wave power generator as recited in claim 3, further comprising the step of producing the lookup table using a nonlinear autoregressive exogenous neural network, modeling the ocean wave power generator.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0009] FIG. 1 is a block diagram of an ocean wave power generator with an artificially intelligent controller.

[0010] FIG. 2 is a diagram of a conventional prior art wave power generator based on a two-body mass-spring-damper system.

[0011] FIG. 3 is a waveform diagram showing the training results of a nonlinear autoregressive exogenous neural network (NARX-NN) of the ocean wave power generator with an artificially intelligent controller, particularly showing an output voltage response (top) and mean square error (MSE) performance (bottom).

[0012] FIG. 4 is a plot of voltage vs. frequency showing electromotive voltage results produced by a linear generator similar to that used in the ocean wave power generator with an artificially intelligent controller.

[0013] FIGS. 5A, 5B, 5C, and 5D show the NARX-NN regression maximum using a log scale.

[0014] FIG. 6 is a graph comparing the MSE of training, validation and testing data sets of the NARX-NN.

[0015] FIG. 7 is an error histogram of the training, validation and testing data sets at maximum epoch 1000.

[0016] FIG. 8 is waveform diagrams showing the NARK-NN training state reaching maximum epoch 1000.

[0017] FIG. 9 is a chart showing training auto-correlation with a time varying lag at maximum epoch 1000 for the NARX-NN.

[0018] FIG. 10 is a diagram showing training input-error cross-correlation with a time varying lag at maximum epoch 1000 for the NARX-NN.

[0019] FIG. 11 is a plot showing the training results of a closed-loop NARX-NN model of the ocean wave power generator with an artificially intelligent controller, including the output voltage response (top) and the MSE performance (bottom).

[0020] Similar reference characters denote corresponding features consistently throughout the attached drawings.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] As shown in FIG. 1, the ocean wave power generator with an artificially intelligent controller, designated generally as 10 in the drawings, is similar to the conventional wave power generator 100 of FIG. 2, including a first mass 12 having a mass a second mass 14 having, a mass m.sub.2, a first spring 16 having a spring constant k.sub.1 resiliently coupling the first mass 12 to the second mass 14, and a first damper 18 having a damping constant b.sub.1 joining the first mass 12 and the second mass 14 for damping relative oscillation between the two masses. Further, similar to the conventional wave power generator 100 of FIG. 2, the ocean wave power generator 10 of FIG. 1 also includes a second spring 20 having a spring constant k.sub.2 resiliently coupling the second mass 14 to a support surface 5, such as the ground or a floor, a second damper 22 having a damping constant b.sub.2 joining the second mass 14 and the support surface S for damping relative oscillation between the second mass 14 and the support surface S, and a linear generator 24 mounted on the support surface S and coupled to the second mass 14, such that relative oscillation between the second mass 14 and the support surface S drives the linear generator 24 to generate power. However, the ocean wave power generator 10 of FIG. 1 also includes a linear actuator 26 coupled to the second mass 14, a first motion sensor 28 for detecting the position and speed of the first mass 12, and a second motion sensor 30 for detecting the position and speed of the second mass 14.

[0022] A controller 32, which may be a personal computer, programmable logic controller, microprocessor or the like, receives the position and the speed of the first mass 12 from the first motion sensor 28 and the, position and the speed of the second mass 14 from the second motion sensor 30. The controller 32 is configured to output a driving signal to linear actuator 26 to drive oscillatory motion of the second mass 14 to optimize the power output of the linear generator 24 based on the position and the speed of the first mass 12 and the position and the speed of the second mass 14.

[0023] It should be understood that the linear generator 24 may be any suitable type of generator for converting oscillatory motion into usable electrical power. For example, the linear generator 24 may comprise or consist of a conductive coil mounted on the support, surface S with a magnetic rod secured to the second mass 14 traveling through the coil in an oscillatory manner as the second mass 14 oscillates. In such a case, calculating the power output of the linear generator 24 is relatively simple, since Faraday's law gives the electrical potential produced across the coil, V as V=NBL, where N is the number of coil loops, B is the magnetic field strength, is the instantaneous velocity of the magnetic rod (which would be equal, in this case, to the instantaneous velocity of the second mass 14, as measured by the motion sensor 30), and L is the distance traveled by the magnetic rod within the coil (which would be equal, in this case, to the vertical displacement of the second mass 14, also measured by the motion sensor 30).

[0024] Thus, using equations (1) and (2) above, as well as Faraday's law, it is possible to model the power output of the linear generator 24. Since equations (1) and (2) are unstable and highly nonlinear, an artificial intelligence, such as a neural network, may be used to produce a lookup table for all possible values of position and speed (or, equivalently, amplitude and frequency) of the first mass 12 and the second mass 14. For example, a nonlinear autoregressive exogenous neural network (NARX-NN) 34 may be used. Thus, using equations (1) and (2) above, as well as Faraday's law, the NARX-NN 34 produces a lookup table of modeled power outputs of the linear generator 24 for each possible value of position and speed of the first mass 12 and the second mass 14. In use, the first motion sensor 28 measures the real-time position and speed of the first mass 28, and the second motion sensor 30 measures the real-time position and speed of the second mass 30. These values are fed to the controller 32, which receives the lookup table from NARX-NN 34, and for the measured position and speed of first mass 12, the ideal position and speed of the second mass 14, which would produce the maximum power output, is determined.

[0025] With the ideal position and speed of second mass 14 determined from the lookup table, the controller 32 sends a driving signal to the linear actuator 26 to either augment or dampen the motion of the second mass 14 (i.e., to either add or subtract from the present position and speed of the second mass 14) to match the ideal position and speed of the second mass 14 from the lookup table. This process is continuous, continuously measuring the position and speed of the first and second masses 12, 14 to provide continuous optimizing augmentation or dampening of the second mass 14 to maximize the power output of the linear generator 24. It should be understood that the linear actuator 26 may be any suitable type of linear actuator capable of instantaneously controlling the position and speed (or, equivalently, the amplitude and frequency) of the second mass 14.

[0026] In order to validate the effectiveness of the NARX-NN in modeling and controlling the ocean wave power generator 10, two layers were used to fit the dynamical input-output relationship system, using one single hidden layer that contained ten neurons. Two tap delays were used, one for motion input and one for voltage output, using two time-steps. The training results of the NARX-NN, shown in FIG. 3, show a highly accurate matching response of the ocean wave power generator 10 compared to the actual voltage output. This shows that the MARX-NN is trained efficiently in terms of matching the generator output response, as well as mean-square-error and correlation error. Further, the trained NN of the mean squared error shows a very high error performance of 9.0910.sup.9.

[0027] FIG. 4 shows the actual electromotive voltage produced by a linear generator for N=1000, B=1 Tesla, a nominal speed of 0.0625 m/sec, and a magnetic rod displacement of L=0.0625 sin(t). Comparing the results of FIGS. 3 and 4, it is found that, the trained MARX-NN perfectly matches the modelled results against the actual measured results. Furthermore, a rapid drop in mean squared error performance of the trained NN is measured in a log scale, as displayed in FIGS. 5A-5D.

[0028] The neural network's performance for three sets of data (training, validation and testing) is shown in FIG. 6 The results show high accuracy training, with the best validation performance of 1.110.sup.8 at epoch 1000. Similarly, the NN training state and error histogram are shown reaching the maximum. epoch at 1000 in FIGS. 7 and 8. Moreover, the correlation of error of the trained NN of the ocean wave power generator 10 as a function of time, with errors over varying lags, are displayed in FIG. 9. The error correlation results with respect to the NN inputs and varying lag fall within the confidence limits, as illustrated in FIG. 10.

[0029] Training of the NA-NN in closed loop form may be performed given initial voltage outputs, so that the NN uses its own predicted voltages recursively to predict new values. The results shows a good fit between the predicted and actual responses, but with non-perfect errors, as shown in FIG. 11. As shown, it took the system more than 120 seconds before the good match starts to separate.

[0030] It is to be understood that the ocean wave power generator with an artificially intelligent controller and a method of operating the same is not limited to the specific embodiments described above, but encompasses any and all embodiments within the scope of the generic language of the following claims enabled by the embodiments described herein, or otherwise shown in the drawings or described above in terms sufficient to enable one of ordinary skill in the, art to make and use the claimed subject matter.