Intelligent optimization control method for wave power generation device based on improved MPC

Abstract

The invention discloses an intelligent optimization control method for a wave power generation device based on an improved MPC, and the method comprises the steps: collecting the wave height, period, flow velocity and device motion posture data in real time through a multi-source sensing module, constructing an exciting force prediction model based on a bidirectional long-short-term memory network, and achieving the precise prediction of the wave exciting force of n control periods in the future; prediction data and real-time feedback signals are input into an improved jellyfish search optimization algorithm, through a two-stage chaotic mapping perturbation mechanism, the convergence speed and the global optimization ability of the algorithm are remarkably improved, and finally optimal electric damping control parameters are output. Compared with a traditional MPC method, the method has the advantages that a feed-forward-feedback composite control framework is formed by fusing deep learning prediction and an intelligent optimization algorithm, so that the energy efficiency conversion efficiency of the power generation system is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent control of marine renewable energy power generation equipment, and specifically relates to an intelligent optimization control method for a wave energy power generation device based on improved model predictive control (MPC). Background Art

[0002] Wave power generation, a key approach to the development and utilization of marine renewable energy, boasts significant advantages due to its unique energy density distribution in distributed offshore scenarios, such as deep-sea energy refueling stations, intelligent monitoring buoy clusters, and offshore ecological aquaculture platforms. Current energy conversion devices primarily utilize multi-physics coupling designs, including phase-synchronized oscillating mechanisms, impeller arrays for fluid kinetic energy capture, and pressure-differential-driven pendulum structures. However, these devices are limited by frequency-domain mismatching in the power take-off (PTO) system under dynamic conditions, resulting in average energy conversion efficiencies of less than 35% in practical engineering applications, severely hindering their commercialization.

[0003] Existing frequency-domain optimization control theories are generally based on the linear regularized wave assumption, employing adaptive damping regulation and predictive model control strategies to achieve an approximate match between the natural frequency of the PTO system and the wave excitation frequency. However, waves in real ocean environments exhibit multidimensional non-steady-state characteristics, and the joint probability distribution of wave height and period exhibits significant spatiotemporal heterogeneity. Traditional model predictive control primarily relies on mathematical calculations, which are computationally time-consuming and difficult when faced with nonlinear data. To address the inaccuracy of the model prediction portion of current model predictive control for wave energy power generation devices, a new model predictive control method for wave energy power generation devices is proposed. Summary of the Invention

[0004] Purpose of the invention: In response to the problems existing in the prior art, the present invention provides an intelligent optimization control method for wave energy power generation devices based on improved MPC. It adopts a neural network solution method to solve the problems of long calculation time and high calculation difficulty when facing nonlinear data in traditional model predictive control, effectively reducing the calculation time, and uses an improved jellyfish search algorithm for rolling optimization to improve the optimization efficiency.

[0005] Technical solution: The intelligent optimization control method of a wave energy power generation device based on an improved MPC according to the present invention comprises the following steps:

[0006] (1) Obtain the wave height and period of external waves through sensors, and preprocess them to obtain equivalent wave height and equivalent period;

[0007] (2) Input the equivalent wave height, equivalent period and the electric power W(t-1) output by the control object at the previous moment into the nominal model to obtain the exciting force F under the current ideal state r (t), and the predicted output excitation force F p (t) comparison, the excitation force error F is obtained e (t);

[0008] (3) The obtained excitation force error is input into the improved jellyfish search algorithm and a rolling optimization is performed. The damping value u at the optimal power moment obtained after the rolling optimization is simultaneously output to the LSTM-based prediction model and the control object;

[0009] (4) The equivalent wave height, equivalent period and the optimal electrical damping value obtained by rolling optimization in step (3) are used as the input of the improved prediction model CNN-LSTM-GRU, and the model output excitation force F m (t) Output the data and obtain the predicted results;

[0010] (5) Compare the prediction result with the exciting force F of the control object at the current moment c (t) is compared to obtain the prediction model excitation force error, which is then input into the improved jellyfish optimization algorithm. The prediction error is reduced through iterative optimization, thus realizing intelligent optimization control of wave energy power generation device.

[0011] Furthermore, the preprocessing in step (1) is to divide the wave height and period data of the external waves into a different wave components through Fourier transform, and perform weighted processing on the different wave components; establish an energy weight coefficient:

[0012]

[0013] Among them, E i represents the energy density of the i-th component, and the processed wave height and period are obtained by the following formula:

[0014]

[0015] Among them, H eq represents the wave height, T eq Represents the cycle.

[0016] Furthermore, the improved jellyfish search algorithm in step (3) is:

[0017] S1: Assuming the population size is N, the maximum number of iterations is Max n , and set the upper and lower limits U of the search space for it i and D i ;

[0018] S2: Use the 2-level Logistic chaotic map to redistribute the fractal coefficients:

[0019] L n+2 =sL n+1 (1-L n+1 )+(4-s)L n (1-L n )

[0020] Where, L n ∈(0,1), s is the classification coefficient, when s=0 or 4, it satisfies the full mapping condition of Logistic chaotic mapping, L n is the Logistic chaotic value of the nth jellyfish position, L0 is the initial value of the Logistic chaotic map;

[0021] S3: The formula for updating the position of each jellyfish in the ocean current is as follows:

[0022]

[0023] Where, is the position of the i-th jellyfish at the t-th iteration, is the best position in the current jellyfish group, β is the distribution coefficient, which is 3; μ is the average position of the jellyfish in the population;

[0024] S4: In order to condition the movement of jellyfish in the population and the movement with the ocean current, a time control mechanism is introduced to switch the control. The formula is as follows:

[0025]

[0026] Where C(t) is a random value that fluctuates between 0 and 1 with the number of iterations; C0 = 0.5, t is the current number of iterations, Max n is the maximum number of iterations. When C(t)≥C0, the particle swarm algorithm is used. First, the initialized jellyfish position is input into the particle swarm, and the maximum number of particle swarm optimization times is set. It mainly searches for the optimal solution by updating two extreme values, which are the optimal solution of individual position p and best and the global optimal solution g best , the update formula is as follows:

[0027] v i =v i +c1×rand()×(pbest i -x i )+c2×rand()×(gbest i -x i )

[0028] Where x i and vi are the displacement and velocity at time i, respectively. By using the particle swarm algorithm, the problem of slow optimization speed of ocean current movement in the original jellyfish optimization algorithm is effectively alleviated;

[0029] S5: After the swarm is formed, most of the jellyfish move passively, i.e. they move around their own positions. The new positions are as follows:

[0030]

[0031] Where, Υ is the motion coefficient, which is 0.1;

[0032] S6: In order to find more food, the jellyfish moves towards its companions with more food. Its movement is active movement, and the resulting new position formula is as follows:

[0033]

[0034] in, is the position of the i-th jellyfish at the t-th iteration, is a jellyfish randomly selected from the population;

[0035] S7: When the jellyfish exceeds the bounded search range, return to the opposite bound;

[0036] S8: After multiple rounds of iterations, the improved jellyfish optimization algorithm can obtain the optimal jellyfish position, which is the electrical damping under the optimal power generation condition of the current wave energy power generation device.

[0037] Furthermore, the exciting force F in step (4) m (t) is:

[0038]

[0039] Among them, h e (t) is the impulse response function, η(t) is the wavefront equation, and the output exciting force is the convolution sum of the two.

[0040] Furthermore, the improved prediction model CNN-LSTM-GRU in step (4) uses a bidirectional long short-term memory network model combined with a gated recurrent unit model and a convolutional neural network model; the external wave information obtained by the sensor is input into the gated recurrent unit model for preliminary prediction, and then the result of the preliminary prediction is used as the input of the CNN-LSTM model to obtain the final model prediction result; through the combination and reconstruction of the three neural networks, CNN-LSTM-GRU is obtained, and the correlation of the wave information in the spatial dimension is improved by GRU, and the correlation of the wave information in the time dimension is improved by LSTM; finally, the results obtained by the two models are organically combined through the CNN model.

[0041] Beneficial effects: Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention adopts a neural network to replace the original least squares linear regression method, which effectively improves the prediction accuracy and reduces the required calculation amount. The wave conditions are decomposed through the preprocessing method, which further improves the prediction accuracy; 2. The present invention adopts a two-level Logistic chaotic mapping, which effectively solves the problem of generating different pseudo-random sequences due to different input parameters in the Logistic chaotic mapping, and further improves the problem of reduced randomness when the analysis coefficient is reduced; 3. The present invention adopts the improved jellyfish optimization algorithm to replace the quadratic programming solution method in the original model predictive control, which reduces the calculation amount when seeking the optimal solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of the present invention;

[0043] Figure 2 This is a schematic diagram of the structure of the improved jellyfish search algorithm of the present invention;

[0044] Figure 3 This is a schematic diagram of the improved neural network model structure;

[0045] Figure 4 These are the results obtained using MATLAB simulation. DETAILED DESCRIPTION

[0046] The present invention will be described in further detail below with reference to the accompanying drawings.

[0047] like Figure 1 As shown, the present invention provides an intelligent optimization control method for a wave energy power generation device based on an improved MPC, characterized in that it includes the following steps:

[0048] Step 1: Obtain the wave height and period of external waves through sensors, and preprocess them to obtain equivalent wave height and equivalent period.

[0049] The wave parameters and period are obtained by sensors on the wave energy power generation device, and then divided into a different wave components through Fourier transform. The different wave components are weighted to obtain equivalent wave height and period. The energy weight coefficient is established:

[0050]

[0051] Among them, E i represents the energy density of the i-th component, and the processed wave height and period are obtained by the following formula:

[0052]

[0053] Where H eq represents the wave height, T eq Represents the cycle.

[0054] Step 2: Input the equivalent wave height, equivalent period and the electric power W(t-1) output by the control object at the previous moment into the nominal model to obtain the exciting force F under the current ideal state. r (t), and the predicted output excitation force F p (t) comparison, the excitation force error F is obtained e (t).

[0055] Step 3: Input the obtained exciting force error into the improved jellyfish search algorithm and perform rolling optimization. The damping value u at the optimal power moment obtained after rolling optimization is output to the LSTM-based prediction model and control object at the same time.

[0056] The improved jellyfish search algorithm is used for rolling optimization. The jellyfish search algorithm is mainly based on a mathematical model that simulates the way jellyfish follow ocean currents and their movement within the population. It is mainly implemented through three idealized rules: (1) Jellyfish only follow ocean currents and move within the population. (2) In the ocean, when jellyfish search for food, the location with more food has greater attraction. (3) The amount of food found is related to the objective function corresponding to the location. Based on these three idealized rules, such as Figure 2 As shown, the improved jellyfish search algorithm is described as follows:

[0057] S1: Assuming the population size is N, the maximum number of iterations is Max n , and set the upper and lower limits U of the search space for it i and D i .

[0058] S2: Using the improved Logistic chaotic map, the original Logistic chaotic map mainly characterizes its chaotic behavior through a one-dimensional nonlinear iterative function. By adjusting the value of the input parameter, a completely different pseudo-random sequence can be generated. To solve this problem, this paper adopts a two-level Logistic chaotic map to redistribute the fractal coefficients. The formula is:

[0059] L n+2 =sL n+1 (1-L n+1 )+(4-s)L n (1-L n )

[0060] Where, L n ∈(0,1), it can be seen that when the classification coefficient s=0 or 4, the full mapping condition of the Logistic chaotic mapping is satisfied, Ln is the Logistic chaotic value of the nth jellyfish position, and L0 is the initial value of the Logistic chaotic map.

[0061] S3: Because the ocean current contains a large amount of nutrients, jellyfish will be attracted to it. The direction of the ocean current is determined by calculating the optimal jellyfish position and the average vector position of other jellyfish. The large amount of nutrients represents the optimal damping in wave energy control. In order to determine the value of the common solution, the position of the jellyfish in the ocean current represents the value of the common solution. The update formula of each jellyfish's position in the ocean current is as follows:

[0062]

[0063] Where, is the position of the i-th jellyfish at the t-th iteration, is the best position in the current jellyfish group, β is the distribution coefficient, which is taken as 3, and μ is the average position of jellyfish in the population.

[0064] S4: As time goes by, more and more jellyfish gather to form a swarm. In order to determine the movement of jellyfish in the swarm and the movement behavior with the ocean current, a time control mechanism is introduced to switch the control. The formula is as follows:

[0065]

[0066] Where C(t) is a random value that fluctuates between 0 and 1 with the number of iterations. C0=0.5, t is the current number of iterations, Max n is the maximum number of iterations. When C(t)≥C0, the particle swarm algorithm is used. First, the initialized jellyfish position is input into the particle swarm, and the maximum number of particle swarm optimization times is set. It mainly searches for the optimal solution by updating two extreme values, which are the optimal solution of individual position p and best and the global optimal solution g best , the update formula is as follows:

[0067] v i =v i +c1×rand()×(pbest i -x i )+c2×rand()×(gbest i -x i )

[0068] Where x i and v i are the displacement and velocity at time i respectively. By using the particle swarm algorithm, the problem of slow optimization speed of ocean current movement in the original jellyfish optimization algorithm is effectively alleviated.

[0069] S5: After the swarm is formed, most of the jellyfish move passively, i.e. they move around their own positions. The new positions are as follows:

[0070]

[0071] Where Υ is the motion coefficient, which is taken as 0.1.

[0072] S6: However, as time goes by, in order to find more food, the jellyfish moves towards its companions with more food. Its main movement is active movement, and the new position formula is as follows:

[0073]

[0074] in, is the position of the i-th jellyfish at the t-th iteration, is a jellyfish randomly selected from the population.

[0075] S7: The ocean is distributed everywhere in time, and the Earth is spherical, so when the jellyfish goes beyond the bounded search range, it will return to the opposite boundary.

[0076] S8: After multiple rounds of iterations, the improved jellyfish optimization algorithm can obtain the optimal jellyfish position, which is the electrical damping under the optimal power generation condition of the current wave energy power generation device.

[0077] Step 4: Take the equivalent wave height, equivalent period and the optimal electrical damping value obtained by rolling optimization in S3 as Figure 3 The input of the improved neural network model is shown, and the model output excitation force F m (t) The data is output and the prediction result formula is as follows:

[0078]

[0079] Among them, h e (t) is the impulse response function, η(t) is the wavefront equation, and the output exciting force is the convolution sum of the two.

[0080] like Figure 3As shown in the figure, the improved prediction model CNN-LSTM-GRU uses a bidirectional long short-term memory network model combined with a gated recurrent unit model and a convolutional neural network model; the external wave information obtained by the sensor is input into the gated recurrent unit model for preliminary prediction, and then the result of the preliminary prediction is used as the input of the CNN-LSTM model to obtain the final model prediction result. Through the combination and reconstruction of the three neural networks, CNN-LSTM-GRU is obtained. The GRU improves the correlation of the wave information in the spatial dimension, and the LSTM improves the correlation of the wave information in the time dimension. Finally, the CNN model is used to organically combine the results obtained by the two models, which can effectively improve the prediction speed and accuracy.

[0081] Step 5: Compare the predicted exciting force with the actual exciting force at the current moment, calculate the prediction error e, and input this error into the improved jellyfish optimization algorithm in order to reduce the prediction error through iterative optimization and provide more accurate parameter adjustment for the next prediction, thus forming a closed-loop prediction and optimization process.

[0082] MATLAB software was used to simulate the analysis, and the simulation parameters are shown in Table 1:

[0083] Table 1 Simulation parameters

[0084]

[0085]

[0086] The simulation results are as follows Figure 4 As shown, the speed and displacement of the improved wave energy power generation device are significantly increased. Since the power generation efficiency of the wave energy power generation device mainly depends on these two parameters, it is proved that the present invention effectively increases the wave energy power generation efficiency.

[0087] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made in accordance with the spirit of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. An intelligent optimization control method for a wave energy power generation device based on improved MPC, characterized in that: The steps include: (1) Obtain the wave height and period of external waves through sensors, and preprocess them to obtain equivalent wave height and equivalent period; (2) Input the equivalent wave height, equivalent period and the electric power W(t-1) output by the control object at the previous moment into the nominal model to obtain the exciting force F under the current ideal state r (t), and the predicted output excitation force F p (t) comparison, the excitation force error F is obtained e (t); (3) The obtained excitation force error is input into the improved jellyfish search algorithm and a rolling optimization is performed. The damping value u at the optimal power moment obtained after the rolling optimization is simultaneously output to the LSTM-based prediction model and the control object; (4) The equivalent wave height, equivalent period and the optimal electrical damping value obtained by rolling optimization in step (3) are used as the input of the improved prediction model CNN-LSTM-GRU, and the model output excitation force F m (t) Output the data and obtain the predicted results; (5) Compare the prediction result with the exciting force F of the control object at the current moment c (t) is compared to obtain the prediction model excitation force error, which is then input into the improved jellyfish optimization algorithm. The prediction error is reduced through iterative optimization, thus realizing intelligent optimization control of wave energy power generation device.

2. The intelligent optimization control method for wave energy power generation device based on improved MPC according to claim 1 is characterized in that: The preprocessing in step (1) is to divide the wave height and period data of the external waves into a different wave components through Fourier transform, and perform weighted processing on the different wave components; establish the energy weight coefficient: Among them, E i represents the energy density of the i-th component, and the processed wave height and period are obtained by the following formula: Among them, H eq represents the wave height, T eq Represents the cycle.

3. The intelligent optimization control method for wave energy power generation device based on improved MPC according to claim 1 is characterized in that: The improved jellyfish search algorithm in step (3) is: S1: Assuming the population size is N, the maximum number of iterations is Max n , and set the upper and lower limits U of the search space for it i and D i ; S2: Use the 2-level Logistic chaotic map to redistribute the fractal coefficients: L n+2 =sL n+1 (1-L n+1 )+(4-s)L n (1-L n ) Where, L n ∈(0,1), s is the classification coefficient, L n is the Logistic chaotic value of the nth jellyfish position, L0 is the initial value of the Logistic chaotic map; S3: The formula for updating the position of each jellyfish in the ocean current is as follows: Where, is the position of the i-th jellyfish at the t-th iteration, is the best position in the current jellyfish group, β is the distribution coefficient, and μ is the average position of jellyfish in the population; S4: In order to condition the movement of jellyfish in the population and the movement with the ocean current, a time control mechanism is introduced to switch the control. The formula is as follows: Where C(t) is a random value that fluctuates between 0 and 1 with the number of iterations; C0 = 0.5, t is the current number of iterations, Max n is the maximum number of iterations. When C(t)≥C0, the particle swarm algorithm is used. First, the initialized jellyfish position is input into the particle swarm, and the maximum number of particle swarm optimization times is set. It mainly searches for the optimal solution by updating two extreme values, which are the optimal solution of individual position p and best and the global optimal solution g best , the update formula is as follows: v i =v i +c1×rand()×(pbest i -x i )+c2×rand()×(gbest i -x i ) Where x i and v i are the displacement and velocity at time i, respectively. By using the particle swarm algorithm, the problem of slow optimization speed of ocean current movement in the original jellyfish optimization algorithm is effectively alleviated; S5: After the swarm is formed, most of the jellyfish move passively, i.e. they move around their own positions. The new positions are as follows: Where γ is the motion coefficient; S6: In order to find more food, the jellyfish moves towards its companions with more food. Its movement is active movement, and the resulting new position formula is as follows: in, is the position of the i-th jellyfish at the t-th iteration, is a jellyfish randomly selected from the population; S7: When the jellyfish exceeds the bounded search range, return to the opposite bound; S8: After multiple rounds of iterations, the improved jellyfish optimization algorithm can obtain the optimal jellyfish position, which is the electrical damping under the optimal power generation condition of the current wave energy power generation device.

4. The intelligent optimization control method for wave energy power generation device based on improved MPC according to claim 3 is characterized in that: When the classification coefficient s in step S2 is 0 or 4, the full mapping condition of the Logistic chaotic mapping is satisfied.

5. The intelligent optimization control method for wave energy power generation device based on improved MPC according to claim 3 is characterized in that: The distribution coefficient β in step S3 is 3.

6. The intelligent optimization control method for wave energy power generation device based on improved MPC according to claim 3 is characterized in that: The motion coefficient γ in step S5 is 0.

1.

7. The intelligent optimization control method for wave energy power generation device based on improved MPC according to claim 1 is characterized in that: The exciting force F in step (4) m (t) is: Among them, h e (t) is the impulse response function, η(t) is the wavefront equation, and the output exciting force is the convolution sum of the two.

8. The intelligent optimization control method for wave energy power generation device based on improved MPC according to claim 1 is characterized in that: The improved prediction model CNN-LSTM-GRU in step (4) uses a bidirectional long short-term memory network model combined with a gated recurrent unit model and a convolutional neural network model; the external wave information obtained by the sensor is input into the gated recurrent unit model for preliminary prediction, and then the result of the preliminary prediction is used as the input of the CNN-LSTM model to obtain the final model prediction result; through the combination and reconstruction of the three neural networks, CNN-LSTM-GRU is obtained, and the correlation of the wave information in the spatial dimension is improved by GRU, and the correlation of the wave information in the time dimension is improved by LSTM; finally, the results obtained by the two models are organically combined through the CNN model.

Images