Power control method, system and equipment based on wave power generation of magnetic speed increasing device and medium
Through the combination of a robust controller and TD3 network, the control parameters of the magnetic speed increase device are optimized, and the problem of power output fluctuations when wave conditions are not ideal is solved, the efficiency and anti-interference ability of wave energy generation are improved, and the wave energy generation is promoted. Development of technology.
Patent Information
- Application Number
- CN202411207543.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-11-12
AI Technical Summary
The existing wave power generation technology of magnetic speed growth devices fluctuates when the wave conditions are not ideal, the power generation efficiency is low, and the composite system has insufficient anti-interference ability, so it needs to be improved to improve the wave energy conversion efficiency.
A robust controller is used to combine the TD3 network for linear analysis and control of the magnetic speed growth device. Through machine learning, the maximum power output of the magnetic speed growth device is achieved, and the decoupled vector control method is used to simplify the control strategy.
It improves the anti-interference ability and power generation efficiency of the magnetic speed-growing device, reduces the loss of the composite power generation system, and promotes the development of wave energy power generation technology.
Smart Images

Figure CN118934416A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wave energy power generation, and in particular to a power control method, system, equipment and medium based on wave power generation of a magnetic speed increasing device. Background Art
[0002] With the continuous growth of global energy demand and the depletion of fossil fuel resources, people are paying more and more attention to the development and utilization of renewable energy. As a renewable and clean energy, wave energy has great potential to mitigate climate change and meet the growing energy demand. At present, wave energy generation in my country is still in its initial stage, with high construction and maintenance costs, immature technology and limited wave energy conversion efficiency. The efficiency of wave energy conversion depends on the input of wave energy and the efficiency of the conversion equipment. The efficiency of current conversion technology is still limited and needs further research and optimization.
[0003] Magnetic speed-increasing device wave power generation is a new energy technology that uses wave energy to convert into electrical energy. Its basic principle is to use the up and down movement of the magnetic speed-increasing device in the waves to increase the relative speed of the wave power generation device and drive the generator to generate electricity. Due to the instability and complexity of wave energy itself, the existing magnetic speed-increasing device composite system is limited by wave conditions and has high requirements for wave conditions. It needs to have a certain wave height and frequency. Therefore, when the wave conditions are not ideal, the power output has a certain volatility and the power generation efficiency is low.
[0004] In order to optimize the power output of wave power generation with magnetic booster device, it is necessary to improve and optimize the traditional control method. Therefore, in the field of wave power generation with magnetic booster device, power optimization control technology is a very important link. By rationally applying these technologies, the stable and efficient output of wave power generation with magnetic booster device can be achieved, thereby promoting the further development of wave energy power generation technology and promoting the use of renewable energy. Summary of the invention
[0005] The present invention aims at the deficiencies in the prior art and provides a power control method, system, device and medium for wave power generation based on a magnetic speed-increasing device. The linear analysis and control of the operation state of the magnetic speed-increasing device are realized, the anti-interference ability of the magnetic speed-increasing device composite power generation system is improved and its performance is optimized, and the control parameters can be learned and optimized according to the actual situation to further improve the power generation efficiency.
[0006] To achieve the above object, the present invention adopts the following technical solutions:
[0007] A power control method for wave power generation based on a magnetic speed-increasing device, wherein the structure of the magnetic speed-increasing device includes a magnetic screw, a permanent magnet synchronous generator and a buoy located outside the permanent magnet synchronous generator, and a robust controller is used to control the magnetic speed-increasing device to output a maximum power. The power control method includes the following steps:
[0008] The stator current and rotor mechanical angular velocity of the magnetic speed-increasing device are collected and input into the robust controller. After removing the low-frequency wave, the stator current reference value and the rotor mechanical angular velocity reference value are obtained to make a preliminary adjustment to the output power of the magnetic speed-increasing device.
[0009] The stator current, rotor mechanical angular velocity, stator current reference value and rotor mechanical angular velocity reference value are input into the TD3 network as training samples. The TD3 network performs machine learning to obtain the expected voltage as an instruction, which is used to correct the stator d-axis equivalent current and the stator q-axis equivalent current input into the robust controller, so that the stator d-axis equivalent current and the stator q-axis equivalent current track the expected value, thereby enabling the magnetic speed increase device to output maximum power.
[0010] To optimize the above technical solutions, the specific measures taken also include:
[0011] Furthermore, the specific process of collecting the stator current and the rotor mechanical angular velocity of the magnetic speed increasing device and inputting them into the robust controller, removing the low-frequency wave, and obtaining the stator current reference value and the rotor mechanical angular velocity reference value includes:
[0012] The magnetic speed increasing device equation under the robust controller is linearized within a selected range of nominal values including the stator current and the rotor mechanical angular velocity, so as to obtain a stator current reference value and a rotor mechanical angular velocity reference value;
[0013] The magnetic speed increaser device equation under the robust controller is:
[0014]
[0015] In the formula, u q is the stator q-axis voltage; R is the stator equivalent resistance; i q is the stator q-axis current; L q is the stator q-axis inductance; ω e is the stator electromagnetic angular velocity; ψ f is the permanent magnet flux; u d is the stator d-axis voltage; i d is the stator d-axis current; L d is the stator d-axis inductance.
[0016] Furthermore, the relationship between the elements controlled by the robust controller is as follows:
[0017]
[0018] Where u represents the command provided by the robust controller K, r is the reference control element, y is the measured control element, and i dref is the stator d-axis current reference value, ω refis the reference value of rotor angular velocity; i d is the stator d-axis current collected, ω is the rotor angular velocity collected, u d is the stator d-axis voltage, u q is the stator q-axis voltage.
[0019] Furthermore, the state space of the TD3 network is all possible states s at time t t = the set of [X*; U*], X* is the state of the magnetic speed-increasing device, and U* is the instruction; the action space of the TD3 network is all possible actions a that can be executed at time t t =ω ref -ω t The rotor mechanical angular velocity ω at time t is t and the rotor mechanical angular velocity reference value ω ref The difference as action a t ; Set each action to select only two adjacent sampling times for adjustment; The reward function of the TD3 network is defined as R t =R1+R2+R3+R4; R t Indicates reward;
[0020] R1 is the error rate Z ω The feedback formula is as follows:
[0021]
[0022] Where, the error rate Z ω =ω ref -ω t ;
[0023] R2 is the stator d-axis current feedback, the formula is as follows:
[0024] R2=0.2Z id
[0025] R3 is the stator q-axis current feedback, the formula is as follows:
[0026]
[0027] In the formula, Z id is the stator d-axis current error rate, Z iq is the error rate of the stator q-axis current;
[0028] R4 is the feedback of the action, and the formula is as follows:
[0029]
[0030] Where U t-1 is the voltage value at time t-1;
[0031] The structure of the TD3 network includes a strategy network and two evaluation strategy networks. The strategy network includes an Actor network and an Actor target strategy network. Each evaluation strategy network includes a Critic network and a Critic target strategy network. The strategy network is used to read the spatial state s at the next moment. t+1 , generate the next moment action a t+1 ; The Critic network in the evaluation strategy network is used to evaluate the spatial state s at the current moment t and action a t Evaluation; Critic target strategy network is used to evaluate the spatial state s at the next moment t+1 and action a t+1 Make an evaluation.
[0032] Furthermore, the TD3 network performs machine learning by adopting double Q value learning, and the specific process is as follows:
[0033] The Actor target policy network is updated using a deterministic policy gradient algorithm, as shown in the following formula:
[0034]
[0035] In the formula, Q π (s,a) refers to the expected return of the magnetic speed-increasing device taking action a in the state space s in accordance with the strategy π; Represents the gradient of expected return, derived from the Actor network, π φ represents the optimal strategy under parameter φ, R t Indicates reward, represents the gradient of the policy; Indicates Q π The gradient of (s,a), a = π(s) represents the action a taken by the magnetic acceleration device in state s following the π strategy, represents the expected return of action a made by the magnetic speed increasing device in state s following the π strategy;
[0036] The evaluation functions of the Critic network and the Critic target policy network are based on the Bellman equation for delayed learning:
[0037] Q π (s,a)=r+γE s',a' [Q π (s',a')],a'~π(s')
[0038] Where r is the reward value of the training sample at the current moment; (s', a') represents the state and action of the magnetic acceleration device at the next moment; γ∈[0,1] is the discount coefficient; Q π(s,a) represents the expected return of action a made by the magnetic acceleration device in state s at the current moment following the π strategy, a'~π(s') represents the action a' made by the magnetic acceleration device in state s' at the next moment following the π strategy, Q π (s',a') represents the expected return of action a made by the magnetic speed increasing device in state s at the next moment following the π strategy, E s',a' [·] also represents the expected benefit of the action a' taken by the magnetic speed increasing device in the next moment in state s' following the π strategy;
[0039] The specific content of dual Q value learning is:
[0040]
[0041] In the formula, y k is the target Q value of the kth training sample, θ k is the kth parameter of the Actor target strategy network, φ k is the kth evaluation strategy network parameter; γ∈[0,1] is the discount coefficient, r is the reward value of the training sample at the current moment, Indicates that the state s follows the parameter φ k The evaluation value of the π strategy, Indicates that the parameter φ is followed in state s k π strategy;
[0042] The network with the less overestimation among the two evaluation strategies is selected as the estimated value, and the estimated value is minQ θk (s,a),Q θk (s,a) represents the value of the k-th evaluation strategy network evaluation state-action pair (s, a);
[0043] The loss functions of the two evaluation policy networks are as follows:
[0044]
[0045] Where, L Q represents the loss of the evaluation policy network, M represents the sample capacity of the experience pool, and Q k (s j ,a j ) represents the evaluation value of the kth evaluation π strategy, representing the state s j The value of
[0046] Gaussian random noise is added to the target strategy, and the parameters of the two evaluation strategy networks are updated through the back propagation of the loss function to update the target Q value. The formula is as follows:
[0047] y'=r+γmin(Q θ1 ,Q θ2 )(s',πθ' (s')+ε)
[0048] In the formula, y' represents the updated target Q value, r is the reward value of the training sample at the current moment, γ is the discount coefficient, min(Q θ1 ,Q θ2 ) means taking Q θ1 and Q θ2 The minimum value of the two, Q θ1 represents the evaluation value under the θ1 parameter, Q θ2 represents the evaluation value under the θ2 parameter, s' represents the state at the next moment, π θ' (s') means that the π strategy is followed in the next state s', and ε represents Gaussian noise;
[0049] ε~clip(N(0,σ),-c,c)
[0050] Among them, ε represents Gaussian noise; c represents the upper and lower boundaries of the strategy smoothing noise; clip function represents the truncation function, and N(0,σ) represents a Gaussian distribution with an expected value of 0 and a variance of σ;
[0051] Use soft update to update the policy network and the evaluation policy network, so that the update frequency of the policy network is lower than that of the evaluation policy network.
[0052] Furthermore, the method of updating the policy network and evaluating the policy network using a soft update method is specifically as follows:
[0053] Update the Actor target network parameters and evaluation strategy network parameters using the following formula:
[0054] θ k =ζθ k +(1-ζ)θ k '
[0055] φ=ζφ+(1-ζ)φ'
[0056] Where ζ is the software update coefficient, θ k is the kth parameter of the Actor target strategy network, φ is the evaluation strategy network parameter, k = 1, 2; θ k ' represents the updated Actor target network parameters, and φ' represents the updated evaluation strategy network parameters.
[0057] The present invention also proposes a power control system for wave power generation based on a magnetic speed-increasing device. The structure of the magnetic speed-increasing device includes a magnetic screw, a permanent magnet synchronous generator and a buoy located outside the permanent magnet synchronous generator, including:
[0058] A data acquisition module is used to collect the stator current and rotor mechanical angular velocity of the magnetic speed increasing device;
[0059] A preliminary adjustment module is used to input the stator current and the rotor mechanical angular velocity into the robust controller, and after removing the low-frequency wave, obtain the stator current reference value and the rotor mechanical angular velocity reference value, and perform preliminary adjustment on the output power of the magnetic speed increasing device;
[0060] The formal regulation module is used to input the stator current, rotor mechanical angular velocity, stator current reference value and rotor mechanical angular velocity reference value as training samples into the TD3 network. The TD3 network performs machine learning to obtain the expected voltage as an instruction, which is used to correct the stator d-axis equivalent current and the stator θ-axis equivalent current input to the robust controller, so that the stator d-axis equivalent current and the stator q-axis equivalent current track the expected value, thereby enabling the magnetic speed increase device to output maximum power.
[0061] The present invention also proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the power control method based on wave power generation of a magnetic speed increasing device as described above are implemented.
[0062] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the steps of the power control method for wave power generation based on a magnetic speed increasing device as described above.
[0063] The beneficial effects of the present invention are:
[0064] The present invention provides a framework for the power control strategy of wave power generation of magnetic speed-increasing device. The TD3 proxy algorithm is combined with a robust controller to linearize the equations of the magnetic speed-increasing device near the reference point close to the rated parameters, so that the power generation device can learn and optimize parameters according to the actual situation, and a decoupled vector control method is also used to simplify the control strategy. At the same time, the control strategy can reduce the loss of the composite power generation system, further improve the power generation efficiency, and promote the further development of wave power generation technology. BRIEF DESCRIPTION OF THE DRAWINGS
[0065] Figure 1 This is a schematic diagram of a model of a robust controller based on the TD3 proxy algorithm combined with a magnetic speed increasing device in this application.
[0066] Figure 2 Schematic diagram of the robust controller principle.
[0067] Figure 3 This is a TD3 network structure diagram based on the operating environment of the magnetic speed increaser. DETAILED DESCRIPTION
[0068] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0069] Embodiment 1
[0070] The present invention proposes a power control method for wave power generation based on a magnetic speed increasing device, such as Figure 1 As shown, a robust controller is used to control the maximum power output of the magnetic speed-increasing device. The entire motor composed of a magnetic screw and a permanent magnet synchronous generator is a magnetic speed-increasing device, also called a composite motor. The buoy is located outside the generator and the iron core is located inside the magnetic screw. The power control method includes the following steps:
[0071] The stator current of the magnetic speed-increasing device obtained by the current sampler and the rotor mechanical angular velocity obtained by the float sampler are input into the robust controller. After removing the low-frequency wave, the stator current reference value and the rotor mechanical angular velocity reference value are obtained to preliminarily adjust the output power of the magnetic speed-increasing device.
[0072] The stator current, rotor mechanical angular velocity, stator current reference value and rotor mechanical angular velocity reference value are input into the TD3 network as training samples. The TD3 network performs machine learning to obtain the expected voltage as an instruction, which is used to correct the stator d-axis equivalent current and the stator q-axis equivalent current input into the robust controller, so that the stator d-axis equivalent current and the stator q-axis equivalent current track the expected value, thereby enabling the magnetic speed increase device to output maximum power.
[0073] The stator current and rotor mechanical angular velocity of the magnetic speed increasing device are collected and input into the robust controller. After removing the low-frequency wave, the specific process of obtaining the stator current reference value and the rotor mechanical angular velocity reference value includes:
[0074] The magnetic speed increasing device equation under the robust controller is linearized within a selected range of nominal values including the stator current and the rotor mechanical angular velocity, so as to obtain a stator current reference value and a rotor mechanical angular velocity reference value;
[0075] The magnetic speed increasing device equation under the robust controller is:
[0076]
[0077] In the formula, u q is the stator q-axis voltage; R is the stator equivalent resistance; i q is the stator q-axis current; L q is the stator q-axis inductance; ω eis the stator electromagnetic angular velocity; ψ f is the permanent magnet flux; u d is the stator d-axis voltage; i d is the stator d-axis current; L d is the stator d-axis inductance.
[0078] like Figure 2 As shown in the figure, a robust controller is a controller that can maintain stability and performance in the presence of system disturbances or parameter uncertainties. It has good robustness, adaptability and scalability, and can achieve a balance between fast system response, stability and anti-interference. The robust controller is combined with a magnetic speed-increasing device. First, the composite motor drives the magnet to rotate through the mechanical connection of the float. The magnetic speed-increasing device consists of an internally threaded iron core, a coil and a magnet. When the magnet rotates, the magnetic field changes, and the current in the coil also changes accordingly. The robust controller detects the current change in the coil and adjusts the frequency and magnitude of the current according to the preset control algorithm to control the movement speed and electromagnetic force. The internal coil of the magnetic speed-increasing device will feel the magnetic field change and generate an anti-magnetic thrust proportional to the rate of change of the magnetic field. The generated force will move the magnetic speed-increasing device and convert the rotational motion of the motor into linear motion. The robust controller can accurately control the magnetic speed-increasing device by continuously adjusting the current, thereby accurately controlling the position of the platform or piston. Therefore, the robust controller controls the movement speed and electromagnetic force of the magnetic speed-increasing device by adjusting the current in the coil, thereby converting the rotational motion into linear motion. The wave energy captured by the outer buoy can be converted into mechanical energy and input into the generator, which is further outputted as electrical energy by the generator, thus realizing the conversion of wave energy into electrical energy.
[0079] The relationship between the elements controlled by the robust controller is as follows:
[0080]
[0081] Where u represents the command provided by the robust controller K, r is the reference control element, y is the measured control element, and i dref is the stator d-axis current reference value, ω ref is the rotor angular velocity reference value; i d is the stator d-axis current collected, ω is the rotor angular velocity collected, u d is the stator d-axis voltage, u q is the stator q-axis voltage.
[0082] like Figure 3 The figure shows the TD3 network structure diagram based on the operating environment of the magnetic speed increasing device. The state space of the TD3 network is all possible states s at time t. t = the set of [X*; U*], X* is the state of the magnetic speed-increasing device, and U* is the instruction; the action space of the TD3 network is all possible actions a that can be executed at time tt =ω ref -ω t The rotor mechanical angular velocity ω at time t is t and the rotor mechanical angular velocity reference value ω ref The difference as action a t ; Set each action to select only two adjacent sampling times for adjustment; The reward function of the TD3 network is defined as R t =R1+R2+R3+R4; R t Indicates reward;
[0083] R1 is the feedback of the error rate Zω, and the formula is as follows:
[0084]
[0085] Where, the error rate Z ω =ω ref -ω t ;
[0086] R2 is the stator d-axis current feedback, the formula is as follows:
[0087] R2=0.2Z id
[0088] R3 is the stator q-axis current feedback, the formula is as follows:
[0089]
[0090] In the formula, Z id is the stator d-axis current error rate, Z iq is the error rate of the stator q-axis current;
[0091] R4 is the feedback of the action, and the formula is as follows:
[0092]
[0093] Where U t-1 is the voltage value at time t-1;
[0094] The structure of the TD3 network includes a strategy network and two evaluation strategy networks. The strategy network includes an Actor network and an Actor target strategy network. Each evaluation strategy network includes a Critic network and a Critic target strategy network. The strategy network is used to read the spatial state s at the next moment. t+1 , generate the next moment action a t+1 ; The Critic network in the evaluation strategy network is used to evaluate the spatial state s at the current moment t and action a tEvaluation; Critic target strategy network is used to evaluate the spatial state s at the next moment t+1 and action a t+1 Make an evaluation.
[0095] The TD3 network performs machine learning by using double Q value learning, and the specific process is as follows:
[0096] The Actor target policy network is updated using a deterministic policy gradient algorithm, as shown in the following formula:
[0097]
[0098] In the formula, Q π (s,a) refers to the expected return of the magnetic speed-increasing device taking action a in the state space s in accordance with the strategy π; Represents the gradient of expected return, derived from the Actor network, π φ represents the optimal strategy under parameter φ, R t Indicates reward, represents the gradient of the policy; Indicates Q π The gradient of (s,a), a = π(s) represents the action a taken by the magnetic acceleration device in state s following the π strategy, represents the expected return of action a made by the magnetic speed increasing device in state s following the π strategy;
[0099] The evaluation functions of the Critic network and the Critic target policy network are based on the Bellman equation for delayed learning:
[0100] Q π (s,a)=r+γE s',a' [Q π (s',a')],a'~π(s')
[0101] Where r is the reward value of the training sample at the current moment; (s', a') represents the state and action of the magnetic acceleration device at the next moment; γ∈[0,1] is the discount coefficient; Q π (s,a) represents the expected return of action a made by the magnetic acceleration device in state s at the current moment following the π strategy, a'~π(s') represents the action a' made by the magnetic acceleration device in state s' at the next moment following the π strategy, Q π (s',a') represents the expected return of action a made by the magnetic speed increasing device in state s at the next moment following the π strategy, E s',a' [·] also represents the expected benefit of the action a' taken by the magnetic speed increasing device in the next moment in state s' following the π strategy;
[0102] The specific content of dual Q value learning is:
[0103]
[0104] In the formula, y k is the target Q value of the kth training sample, θ k is the kth parameter of the Actor target strategy network, φ k is the kth evaluation strategy network parameter; γ∈[0,1] is the discount coefficient, r is the reward value of the training sample at the current moment, Indicates that the state s follows the parameter φ k The evaluation value of the π strategy, Indicates that the parameter φ is followed in state s k π strategy;
[0105] In order to reduce the overestimation bias problem in the dual Q value learning of the network, the less overestimated value in the two evaluation strategies is selected as the estimated value, and the estimated value is minQ θk (s,a),Q θk (s,a) represents the value of the k-th evaluation strategy network evaluation state-action pair (s, a);
[0106] The loss functions of the two evaluation policy networks are as follows:
[0107]
[0108] Where, L Q represents the loss of the evaluation policy network, M represents the sample capacity of the experience pool, and Q k (s j ,a j ) represents the evaluation value of the kth evaluation π strategy, representing the state s j The value of
[0109] Gaussian random noise is added to the target strategy, and the parameters of the two evaluation strategy networks are updated through the back propagation of the loss function to update the target Q value. The formula is as follows:
[0110] y'=r+γmin(Q θ1 ,Q θ2 )(s',π θ' (s')+ε)
[0111] The time difference (TD) error is:
[0112] TD err =y′-Q θk (s t ,a t |θ t )
[0113] In the formula, y' represents the updated target Q value, r is the reward value of the training sample at the current moment, γ is the discount coefficient, min(Q θ1 ,Q θ2 ) means taking Q θ1 and Q θ2 The minimum value of the two, Q θ1 represents the evaluation value under the θ1 parameter, Q θ2 represents the evaluation value under the θ2 parameter, s' represents the state at the next moment, π θ' (s') means that the π strategy is followed in the next state s', and ε represents Gaussian noise;
[0114] ε~clip(N(0,σ),-c,c)
[0115] Among them, ε represents Gaussian noise; c represents the upper and lower boundaries of the strategy smoothing noise; clip function represents the truncation function, and N(0,σ) represents a Gaussian distribution with an expected value of 0 and a variance of σ;
[0116] In order to ensure that the estimation error is reduced before the policy is updated and to improve the performance and response speed of the system, the policy network and the evaluation policy network are updated using a soft update method, so that the update frequency of the policy network is lower than that of the evaluation policy network. Specifically:
[0117] Update the Actor target network parameters and evaluation strategy network parameters using the following formula:
[0118] θ k =ζθ k +(1-ζ)θ k '
[0119] φ=ζφ+(1-ζ)φ'
[0120] Where ζ is the software update coefficient, θ k is the kth parameter of the Actor target strategy network, φ is the evaluation strategy network parameter, k = 1, 2; θ k ' represents the updated Actor target network parameters, and φ' represents the updated evaluation strategy network parameters.
[0121] According to the working principle of the magnetic screw and the mechanical angular velocity of the rotor of the composite generator with a magnetic speed increasing device The linear speed of the stator is obtained by the transmission ratio, which is:
[0122]
[0123] Furthermore, under the premise that the motor magnetic circuit is not saturated and the hysteresis loss and eddy current loss of the motor are ignored, in order to reduce the stator winding loss of the motor, simplify the control strategy and verify the feasibility of the control strategy, i d= 0 system decoupling vector control method, i is decomposed into dq;
[0124] The formula is as follows:
[0125]
[0126] In the formula, i dref is the stator d-axis equivalent current, i qref is the stator q-axis equivalent current; C is the damping coefficient of the active power generated by the magnetic speed-increasing device composite motor, v is the linear speed of the stator at time t, ψ f is the magnetic flux of the permanent magnet; p is the number of pole pairs, λ is the pole width of the magnetic speed-increasing device composite generator, and t is the magnetic field period of the magnetic speed-increasing device composite motor, that is, the lead.
[0127] Then the output power P of the compound motor is
[0128] P=F g v = β g v 2
[0129] Among them, the hydrodynamic damping resistance effect coefficient
[0130] In the formula, i q is the stator q-axis current; L q is the stator q-axis inductance; i d is the stator d-axis current; L d is the stator d-axis inductance.
[0131] Embodiment 2
[0132] The present invention further proposes a power control system for wave power generation based on a magnetic speed increasing device corresponding to the method of the first embodiment, comprising:
[0133] A data acquisition module is used to collect the stator current and rotor mechanical angular velocity of the magnetic speed increasing device;
[0134] A preliminary adjustment module is used to input the stator current and the rotor mechanical angular velocity into the robust controller, and after removing the low-frequency wave, obtain the stator current reference value and the rotor mechanical angular velocity reference value, and perform preliminary adjustment on the output power of the magnetic speed increasing device;
[0135] The formal regulation module is used to input the stator current, rotor mechanical angular velocity, stator current reference value and rotor mechanical angular velocity reference value as training samples into the TD3 network. The TD3 network performs machine learning to obtain the expected voltage as an instruction, which is used to correct the stator d-axis equivalent current and the stator q-axis equivalent current input to the robust controller, so that the stator d-axis equivalent current and the stator q-axis equivalent current track the expected value, thereby enabling the magnetic speed increase device to output maximum power.
[0136] The implementation methods of each module and module function in the system are completely consistent with the steps of the method in Example 1, so they will not be repeated here.
[0137] Embodiment 3
[0138] The present invention also proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the power control method for wave power generation based on a magnetic speed increasing device as described in Example 1 are implemented.
[0139] Embodiment 4
[0140] The present invention also proposes a computer-readable storage medium storing a computer program, characterized in that the computer program enables a computer to execute the steps of the power control method for wave power generation based on a magnetic speed increasing device as described in Example 1.
[0141] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.
[0142] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0143] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.
Claims
1. A power control method for wave power generation based on a magnetic speed-increasing device, wherein the structure of the magnetic speed-increasing device includes a magnetic screw, a permanent magnet synchronous generator and a buoy located outside the permanent magnet synchronous generator. It is characterized in that A robust controller is used to control the magnetic speed increasing device to output maximum power. The power control method comprises the following steps: The stator current and rotor mechanical angular velocity of the magnetic speed-increasing device are collected and input into the robust controller. After removing the low-frequency wave, the stator current reference value and the rotor mechanical angular velocity reference value are obtained to make a preliminary adjustment to the output power of the magnetic speed-increasing device. The stator current, rotor mechanical angular velocity, stator current reference value and rotor mechanical angular velocity reference value are input into the TD3 network as training samples. The TD3 network performs machine learning to obtain the expected voltage as an instruction, which is used to correct the stator d-axis equivalent current and the stator q-axis equivalent current input into the robust controller, so that the stator d-axis equivalent current and the stator q-axis equivalent current track the expected value, thereby enabling the magnetic speed increase device to output maximum power.
2. The power control method of wave power generation based on a magnetic speed increasing device as claimed in claim 1, It is characterized in that The specific process of collecting the stator current and the rotor mechanical angular velocity of the magnetic speed increasing device, inputting them into the robust controller, removing the low-frequency wave, and obtaining the stator current reference value and the rotor mechanical angular velocity reference value includes: The magnetic speed increasing device equation under the robust controller is linearized within a selected range of nominal values including the stator current and the rotor mechanical angular velocity, so as to obtain a stator current reference value and a rotor mechanical angular velocity reference value; The magnetic speed increasing device equation under the robust controller is: In the formula, u q is the stator q-axis voltage; R is the stator equivalent resistance; i q is the stator q-axis current; L q is the stator q-axis inductance; ω e is the stator electromagnetic angular velocity; ψ f is the permanent magnet flux; u d is the stator d-axis voltage; i d is the stator d-axis current; L d is the stator d-axis inductance.
3. The power control method of wave power generation based on a magnetic speed increasing device as claimed in claim 1, It is characterized in that The relationship between the elements controlled by the robust controller is as follows: Where u represents the command provided by the robust controller K, r is the reference control element, y is the measured control element, and i dref is the stator d-axis current reference value, ω ref is the rotor angular velocity reference value; i d is the stator d-axis current collected, ω is the rotor angular velocity collected, u d is the stator d-axis voltage, u q is the stator q-axis voltage.
4. The power control method of wave power generation based on a magnetic speed increasing device as claimed in claim 1, It is characterized in that The state space of the TD3 network is all possible states s at time t. t = the set of [X*; U*], X* is the state of the magnetic speed-increasing device, and U* is the instruction; the action space of the TD3 network is all possible actions a that can be executed at time t t =ω ref -ω t The rotor mechanical angular velocity ω at time t is t and the rotor mechanical angular velocity reference value ω ref The difference as action a t ; Set each action to select only two adjacent sampling times for adjustment; The reward function of the TD3 network is defined as R t =R 1 +R 2 +R 3 +R 4 ; R t Indicates reward; R 1 is the error rate Z ω The feedback formula is as follows: Where, the error rate Z ω =ω ref -ω t ; R 2 is the stator d-axis current feedback, the formula is as follows: R 2 =0.2Z id R 3 is the stator q-axis current feedback, the formula is as follows: In the formula, Z id is the error rate of the stator d-axis current, Z iq is the error rate of the stator q-axis current; R 4 is the feedback of the action, and the formula is as follows: Where U t-1 is the voltage value at time t-1; The structure of the TD3 network includes a strategy network and two evaluation strategy networks. The strategy network includes an Actor network and an Actor target strategy network. Each evaluation strategy network includes a Critic network and a Critic target strategy network. The strategy network is used to read the spatial state s at the next moment. t+1 , generate the next moment action a t+1 ; The Critic network in the evaluation strategy network is used to evaluate the spatial state s at the current moment t and action a t Evaluation; Critic target strategy network is used to evaluate the spatial state s at the next moment t+1 and action a t+1 Make an evaluation.
5. The power control method of wave power generation based on a magnetic speed increasing device as claimed in claim 4, It is characterized in that The TD3 network performs machine learning by using double Q value learning, and the specific process is as follows: The Actor target policy network is updated using a deterministic policy gradient algorithm, as shown in the following formula: In the formula, Q π (s,a) refers to the expected return of the magnetic speed-increasing device taking action a in the state space s in accordance with the strategy π; Represents the gradient of expected return, derived from the Actor network, π f represents the optimal strategy under parameter φ, R t Indicates reward, represents the gradient of the policy; Indicates Q π The gradient of (s,a), a = π(s) represents the action a made by the magnetic acceleration device in state s following the π strategy, represents the expected return of action a made by the magnetic speed increasing device in state s following the π strategy; The evaluation functions of the Critic network and the Critic target policy network are based on the Bellman equation for delayed learning: Q π (s,a)=r+γE s',a' [Q π (s',a')],a'~π(s') Where r is the reward value of the training sample at the current moment; (s', a') represents the state and action of the magnetic acceleration device at the next moment; γ∈[0,1] is the discount coefficient; Q π (s,a) represents the expected return of action a made by the magnetic acceleration device in state s at the current moment following the π strategy, a'~π(s') represents the action a' made by the magnetic acceleration device in state s' at the next moment following the π strategy, Q p (s',a') represents the expected return of action a made by the magnetic speed increasing device in the next moment under state s following the π strategy, E s',a' [·] also represents the expected benefit of the action a' taken by the magnetic speed increasing device in the next moment in state s' following the π strategy; The specific content of dual Q value learning is: In the formula, y k is the target Q value of the kth training sample, θ k is the kth parameter of the Actor target strategy network, φ k is the kth evaluation strategy network parameter; γ∈[0,1] is the discount coefficient, r is the reward value of the training sample at the current moment, Indicates that the state s follows the parameter φ k The evaluation value of the π strategy, Indicates that the parameter φ is followed in state s k π strategy; The network with the less overestimation among the two evaluation strategies is selected as the estimated value, and the estimated value is minQ θk (s,a),Q θk (s,a) represents the value of the k-th evaluation strategy network evaluation state-action pair (s, a); The loss functions of the two evaluation policy networks are as follows: Where, L Q represents the loss of the evaluation policy network, M represents the sample capacity of the experience pool, and Q k (s j ,a j ) represents the evaluation value of the kth evaluation π strategy, representing the state s j The value of Gaussian random noise is added to the target strategy, and the parameters of the two evaluation strategy networks are updated through the back propagation of the loss function to update the target Q value. The formula is as follows: y'=r+γmin(Q θ1 ,Q θ2 )(s',π θ' (s')+e) In the formula, y' represents the updated target Q value, r is the reward value of the training sample at the current moment, γ is the discount coefficient, min(Q θ1 ,Q θ2 ) means taking Q θ1 and Q θ2 The minimum value of the two, Q θ1 Represents θ 1 The evaluation value under the parameter, Q θ2 Represents θ 2 The evaluation value under the parameter, s' represents the state at the next moment, π θ' (s') means that the π strategy is followed in the next state s', and ε represents Gaussian noise; ε~clip(N(0,σ),-c,c) Among them, ε represents Gaussian noise; c represents the upper and lower cutoff boundaries of the strategy smoothing noise; clip function represents the truncation function, and N(0,σ) represents a Gaussian distribution with an expected value of 0 and a variance of σ; Use soft update to update the policy network and the evaluation policy network, so that the update frequency of the policy network is lower than that of the evaluation policy network.
6. The power control method of wave power generation based on a magnetic speed increasing device as claimed in claim 5, It is characterized in that The specific method of updating the policy network and evaluating the policy network using the soft update method is as follows: Update the Actor target network parameters and evaluation strategy network parameters using the following formula: i k =ζθ k +(1-ζ)θ k ' φ=ζφ+(1-ζ)φ' Where ζ is the software update coefficient, θ k is the kth parameter of the Actor target strategy network, φ is the evaluation strategy network parameter, k = 1, 2; θ k ' represents the updated Actor target network parameters, and φ' represents the updated evaluation strategy network parameters.
7. A power control system for wave power generation based on a magnetic speed-increasing device, wherein the structure of the magnetic speed-increasing device includes a magnetic screw, a permanent magnet synchronous generator and a buoy located outside the permanent magnet synchronous generator. It is characterized in that include: A data acquisition module is used to collect the stator current and rotor mechanical angular velocity of the magnetic speed increasing device; A preliminary adjustment module is used to input the stator current and the rotor mechanical angular velocity into the robust controller, and after removing the low-frequency wave, obtain the stator current reference value and the rotor mechanical angular velocity reference value, and perform preliminary adjustment on the output power of the magnetic speed increasing device; The formal regulation module is used to input the stator current, rotor mechanical angular velocity, stator current reference value and rotor mechanical angular velocity reference value as training samples into the TD3 network. The TD3 network performs machine learning to obtain the expected voltage as an instruction, which is used to correct the stator d-axis equivalent current and the stator q-axis equivalent current input to the robust controller, so that the stator d-axis equivalent current and the stator q-axis equivalent current track the expected value, thereby enabling the magnetic speed increase device to output maximum power.
8. An electronic device, It is characterized in that include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the power control method for wave power generation based on a magnetic speed increasing device as described in any one of claims 1 to 6 is implemented.
9. A computer-readable storage medium storing a computer program, It is characterized in that The computer program enables the computer to execute the steps of the power control method for wave power generation based on a magnetic speed increasing device as described in any one of claims 1 to 6.
