Undersea wireless power transmission system based on deep learning and dynamic control method

Abstract

The invention discloses an undersea wireless power transmission system based on deep learning, which comprises a pair of coupling coils, the pair of coupling coils are matched with an S-S energy resonance compensation network, the system adopts the S-S energy resonance compensation network to realize constant current output, and an impedance matching module is connected into the S-S energy resonance compensation network. The impedance matching module is sequentially connected with the LSTM control module and each sensor, the input end of the S-S energy resonance compensation network is further connected with a high-frequency inverter circuit, the high-frequency inverter circuit is connected with a power supply, the output end of the S-S energy resonance compensation network is further connected with a high-frequency rectifier module, and the high-frequency rectifier module is further connected with a load. The invention further discloses a dynamic control method of the undersea wireless power transmission system based on deep learning. Dynamic impedance matching and energy transmission optimization can be realized through a deep learning algorithm, and high-power, high-efficiency and high-stability energy transmission in a complex dynamic environment is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless power transmission, and specifically relates to an underwater wireless power transmission system based on deep learning. The present invention also relates to a dynamic control method of an underwater wireless power transmission system based on deep learning. Background Art

[0002] The ocean occupies more than 70% of the earth's surface area, and contains a wide variety of resources, which provide guarantees for human survival and development. However, the development of marine resources, deep-sea exploration, marine rescue, underwater facilities, etc. are inseparable from the use of underwater electrical equipment.

[0003] At present, underwater electrical equipment mainly includes tethered underwater robots (Remote Operated Vehicle, ROV), underwater unmanned vehicles (Underwater Unmanned Vehicle, UUV), autonomous underwater robots (Autonomous Underwater Vehicle, AUV), and composite submersibles (Autonomous Remotely Vehicle, ARV). However, in the process of performing marine energy exploration tasks, underwater electrical equipment mainly uses lithium batteries as a power source, and its endurance problem has always been a problem that restricts its long-term underwater operations.

[0004] However, lithium batteries need to be recharged to meet the continuous endurance requirements of underwater electrical equipment. Currently, common charging methods include wet plug charging and salvage battery replacement. Salvage battery replacement usually requires recovery to the mother ship or fixed platform; charging or replacing batteries for underwater electrical equipment through physical contact. Although this method is reliable, it is complicated and time-consuming to operate, especially under harsh sea conditions, which increases the risk and cost of operations. Wet plug charging refers to underwater contact power supply to underwater electrical equipment. The wet plug charging method is carried out underwater, and the sealing of the power transmission interface must be ensured, otherwise it may have serious consequences. At the same time, frequent squeezing and plugging operations at the power transmission interface will cause wear, thereby reducing its service life. In order to solve the above problems, wireless power supply technology is used to power underwater electrical equipment underwater.

[0005] Wireless Power Transfer (WPT) is a technology that uses electromagnetic waves, light waves, sound waves, microwaves and other carriers to transfer energy from the power supply to the load. The main current wireless power transmission technologies include: non-contact inductively coupled power transfer (ICPT), magnetic field resonance (ERPT), microwave radiation (MPT), radio frequency power transmission (RFPT), etc. Since the underwater environment has a great impact on microwave radiation, laser and ultrasonic radio transmission methods, it is not suitable for underwater wireless power transmission. Magnetic induction wireless power transmission is a current research hotspot.

[0006] However, in the dynamic environment under the sea (such as charging of underwater vehicles), there are problems such as eddy current loss caused by changes in the conductivity of seawater, drift of coupling coefficient caused by marine biological attachment, and the influence of mechanical deformation caused by dynamic ocean currents on the magnetic coupling mechanism. In order to achieve stable charging of underwater vehicles, higher control requirements are required for underwater wireless charging systems. Traditional underwater wireless power transmission systems mostly use fixed-frequency resonant coupling technology. However, fixed-frequency resonant coupling technology will generate high-frequency alternating strong electromagnetic fields when working underwater. Part of the electromagnetic field will be emitted outside the system, which is easy to interfere with the navigation, sonar, fuse and other electronic components of underwater vehicles, affecting their normal functions, and may even cause malfunction or damage. At present, the common PID control method has the problem of response lag. Therefore, an underwater wireless power transmission system and its control method for application in the marine environment are proposed. Dynamic impedance matching and energy transmission optimization are realized through deep learning algorithms, thereby realizing stable transmission of electric energy. Summary of the invention

[0007] The purpose of the present invention is to provide an underwater wireless power transmission system based on deep learning, which can realize dynamic impedance matching and energy transmission optimization through deep learning algorithms, and realize high-power, high-efficiency and high-stability energy transmission in complex dynamic environments.

[0008] Another object of the present invention is to provide a dynamic control method for an underwater wireless power transmission system based on deep learning.

[0009] The first technical solution adopted by the present invention is an underwater wireless power transmission system based on deep learning, including a pair of coupling coils, which cooperate with an SS energy resonance compensation network. The system adopts the SS energy resonance compensation network to achieve constant current output. The SS energy resonance compensation network is connected to an impedance matching module, which is connected to the LSTM control module and each sensor in turn. The input end of the SS energy resonance compensation network is also connected to a high-frequency inverter circuit, which is connected to a power supply. The output end of the SS energy resonance compensation network is also connected to a high-frequency rectifier module, which is connected to a load.

[0010] The first technical solution of the present invention is also characterized in that:

[0011] The primary side of the SS energy resonant compensation network is connected to a full-bridge high-frequency inverter circuit, the full-bridge inverter circuit is connected to the compensation capacitor in the resonant compensation network, the end of the compensation capacitor is connected to the primary coupling coil to form a primary S compensation network, the secondary coupling coil is connected to a secondary compensation capacitor to form a secondary compensation network, thereby forming an SS resonant network, the end of the secondary compensation network is connected to a high-frequency rectifier circuit, and then connected to the load to form a loop.

[0012] The specific structure of the impedance matching module is: it consists of a sensor, a wireless communication module, an LSTM control module and a variable capacitor. The sensor is connected to the wireless communication module and converts information into data and transmits it to the wireless communication module receiving end connected to the LSTM control module. The receiving end transmits the information to the LSTM control module, and the LSTM control module controls the capacitance change to achieve impedance matching, wherein the variable capacitor is the compensation capacitor in the SS energy resonance compensation network and is connected to the coupled inductor.

[0013] The specific structure of the high-frequency inverter circuit is as follows: it is composed of four N-type silicon carbide MOSFET switch tubes and four reverse freewheeling diodes connected in parallel at both ends of the switch tubes. The four switch tubes are divided into two groups, namely the left half bridge and the right half bridge. The upper and lower arms of the switch tubes Q1 and Q3 are complementary to each other to form the left half bridge, and the upper and lower arms of the switch tubes Q2 and Q4 are complementary to each other to form the right half bridge. The D poles of the switch tubes Q1 and Q2 are connected and connected to the positive pole of the DC power supply. The S poles of the switch tubes Q3 and Q4 are connected and then connected to the negative pole of the DC power supply. The S pole of the switch tube Q1 is connected to the D pole of the switch tube Q3 and then connected to the compensation capacitor of the compensation network. The S pole of the switch tube Q2 is connected to the D pole of the switch tube Q4 and then connected to the coupling inductor of the compensation network.

[0014] The specific structure of the high-frequency rectifier module is as follows: it consists of a filter capacitor and four fast recovery diodes, wherein the cathodes of diode D1 and diode D2 are connected to one end of the filter capacitor and connected to the positive electrode of the load, the anodes of diode D1 and diode D2 are connected to the cathodes of diode D3 and diode D4 respectively, the anode of diode D1 is connected to the secondary compensation capacitor in the secondary compensation network, the cathode of diode D4 is connected to the secondary coupling coil in the secondary compensation network, and the anodes of diode D3 and diode D4 are connected to the other end of the filter capacitor and the negative electrode of the load.

[0015] The second technical solution adopted by the present invention is a dynamic control method of an underwater wireless power transmission system based on deep learning, which is specifically implemented according to the following steps:

[0016] The primary side uses SiC-MOSFET to build a high-frequency inverter circuit to convert DC power into high-frequency AC power for wireless power transmission;

[0017] The high-frequency AC power enters the compensation network and transfers energy to the secondary side through the primary-secondary coupling coil, and converts the AC power into the required DC power through high-frequency rectification for use by the load;

[0018] When the docking position of the spacecraft fluctuates and the external environment changes, the charging platform detects the parameter changes and transmits the information to the LSTM control module. After calculation, the variable capacitor module in the impedance matching module is controlled to adjust the capacitance to achieve impedance matching and power maximization.

[0019] The state of the vehicle during each charging and the corresponding capacitance value of the variable capacitor are saved as data for analysis, and the next charging is optimized and adjusted based on the data, thereby achieving deep learning.

[0020] The second technical solution of the present invention is also characterized in that:

[0021] The high frequency inverter circuit works as follows:

[0022] The switch tubes Q1, Q2, Q3 and Q4 are all N-type MOSFET switch tubes constituting a full-bridge inverter;

[0023] (1) When the switch tubes Q1 and Q4 are closed and the switch tubes Q2 and Q3 are opened, the current flows through Q1 and Q4 to form a loop;

[0024] (2) When the switch tubes Q2 and Q3 are closed and the switch tubes Q2 and Q3 are opened, the switch tubes Q2 and Q3 cannot be closed immediately, and the direction of the inductor current cannot change suddenly at this moment. At this time, the current flows through the anti-parallel diodes of the switch tubes Q2 and Q3 for freewheeling;

[0025] (3) After the inductor current passes through zero, the switch tubes Q2 and Q3 are closed, and the inductor current flows through the switch tubes Q2 and Q3 in the reverse direction;

[0026] (4) When the switch tubes Q2 and Q3 are disconnected and the switch tubes Q1 and Q4 are closed again, the switch tubes Q1 and Q4 cannot be closed immediately. As analyzed above, the direction of the current cannot change suddenly. The current flows through the anti-parallel diodes of the switch tubes Q1 and Q4, and the above process is repeated in the subsequent cycles.

[0027] Step (2) and step (4) are energy feedback processes, in which the diode provides a feedback energy channel, and this process is also a freewheeling process of the load current.

[0028] In the SS energy resonance compensation network, U in is the input voltage. The full-bridge inverter composed of switch tubes Q1 to Q4 provides the high-frequency AC power required by the system. The uncontrolled rectifier is composed of diodes D1 to D4. Lp and Ls are the self-inductance of the transmitting coil and the receiving coil respectively. Cp and Cs are the series compensation capacitors on the transmitting side and the receiving side respectively. M is the mutual inductance between the coils. C1 is the filter capacitor. R L is the load resistance, and the output voltage is Uout;

[0029] The output voltage of the inverter is U AB And the input equivalent resistance Req of the rectifier satisfies the following conditions:

[0030]

[0031] According to Kirchhoff's law, the resonance conditions of the transmitting circuit and the receiving circuit are derived and expressed as:

[0032]

[0033] Without considering the losses of the inverter and rectifier modules, the output voltage of the system is expressed as:

[0034]

[0035] The output power is expressed as:

[0036]

[0037] In the impedance matching module, U AB It is expressed as the power supply voltage, and the load impedance is set to R L The output impedance of the power supply is Rs, L P , R 1 are the series equivalent inductance and equivalent resistance of the transmitting resonant coil, C P is the transmitting end series resonant capacitor; LS , R 2 are the equivalent inductance and equivalent resistance of the receiving resonant coil in series, C S is the resonant capacitor at the receiving end;

[0038] In the magnetic resonance wireless energy transfer system, the input and load reflection coefficients are expressed as:

[0039]

[0040] The output and load reflection coefficients are expressed as:

[0041]

[0042] When the condition Γ is met s =Γ in *, the power output is maximum;

[0043] When the condition Γ is met out =Γ L *, the load absorbs the maximum power, that is:

[0044]

[0045] have to:

[0046] R s =Z in * ,R L =Z out * (8)

[0047] Output impedance Rs and input impedance Z in And load impedance R L With output impedance Z out They are conjugate to each other. From this analysis, we know that when the condition Γ is satisfied s =Γ in * and Γ out =Γ L *, that is, the power supply output impedance Rs and input impedance Z in The real part is equal to the output impedance Z out With load impedance R L The real part of is equal, at this time the power supply has the maximum output power, and the power delivered to the load is also the maximum;

[0048] First, assume that the output impedance of the power supply is Z 0 is a fixed value, the equivalent impedance of the wireless power transmitter is: Z in =R in +jX in When R in >Z 0When using an L-type impedance matching network, according to the basic circuit series-parallel law, if impedance matching is to be performed, the impedance of the matching network followed by the load impedance is equal to Z 0 :

[0049]

[0050] Further:

[0051]

[0052] When R in <Z 0 When using an inverse L-type impedance matching network, the formula for the inverse L-type impedance matching network is derived as follows:

[0053]

[0054] Separate the imaginary and real parts,

[0055]

[0056] Further:

[0057]

[0058] The core formula of the long short-term memory network is as follows:

[0059]

[0060] Among them, f t represents the output of the forget gate, indicating which data needs to be forgotten, i.e. eliminated, f t The value ranges from 0 to 1, where 0 means completely forgotten and 1 means completely retained. f represents the weight matrix of the forget gate; h t-1 Indicates the hidden state at the last moment, that is, the data state of the last charge, including voltage, current, efficiency, power, and the adjustable capacitance value in the compensation capacitor, x t Represents the information input at the current moment, including the current charging voltage, current, efficiency, power of the spacecraft, the adjustable capacitance value in the compensation capacitor, and the bias term b of the forget gate f is a learnable parameter used to adjust the activation threshold of the forget gate, i t Represents the output of the input gate, indicating whether to record the information of this charging and the recording ratio. Indicates temporary storage of new information; C t Indicates the current cell state, that is, after multiple charging and recording of data, the charging state with the most times is recorded as long-term memory. trepresents the output of the output gate, that is, the impact of the most common charging data on the current charging state, h t Indicates that the data has been recorded and used as the data for next comparison;

[0061] By analyzing historical data, LSTM can predict the position changes, attitude adjustments and changes in electromagnetic coupling efficiency of the spacecraft during the charging process, so as to adjust the charging parameters in advance. During the wireless charging process, LSTM can dynamically adjust the impedance parameters based on historical parameters and model predictions to achieve efficient charging. LSTM can also be used to monitor the operating status of the wireless charging system. By analyzing the time series data of current and voltage parameters during the charging process, abnormal situations can be detected in time and early warnings can be issued.

[0062] The beneficial effects of the present invention are that the underwater wireless power transmission system and dynamic control method based on deep learning, (1) the control method involved in the present invention performs impedance matching according to the underwater environment, which not only makes up for the problem that the fixed frequency resonant coupling technology cannot cope with the dynamic environment (seawater eddy current, biological attachment, etc.) when the resonant conditions change, but also improves the anti-interference ability of the system, increases the transmission distance of the system, and is more suitable for the underwater dynamic environment. (2) Compared with the traditional PID control method, the control method involved in the present invention has rapid response, stable control, can achieve fast and accurate control, and the system response time is shortened to 50ms, which greatly reduces the impact of the dynamic environment on wireless charging and improves the stability of the system charging. (3) The impedance matching involved in the present invention is a control method based on deep learning. Deep learning can automatically learn complex patterns and features from a large amount of data by simulating the neuron structure and information processing method of the human brain. The LSTM control module is adopted, which enables it to handle highly complex tasks, greatly improves the speed of processing information in a dynamic environment, can adapt to the range of seawater conductivity changes of 0.5-5S / m, and saves volume and cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of the overall architecture of the system involved in the present invention;

[0064] Figure 2 It is a structural diagram of the SS primary and secondary side compensation network involved in the present invention;

[0065] Figure 3 is a system equivalent circuit diagram of the SS involved in the present invention;

[0066] Figure 4 It is a schematic diagram of an L-type dynamic impedance matching circuit involved in the present invention;

[0067] Figure 5It is a schematic diagram of the reverse L-type dynamic impedance matching circuit involved in the present invention. DETAILED DESCRIPTION

[0068] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0069] The present invention proposes an underwater wireless power transmission system and dynamic control method based on deep learning. Based on the principle of dynamic impedance matching and energy transmission optimization realized by deep learning algorithm, the parameters of the transmitting device and the receiving device are reasonably set, and multiple sensors are set to collect parameters in real time, and the parameters are transmitted to the controller for analysis and control. The controller controls the variable capacitor module so that it has dynamic impedance in a dynamic environment, so that the wireless charging system is always in an "electromagnetic resonance" state, thereby realizing efficient energy transfer between the transmitting end and the receiving end. The SS type topology structure adopted by the magnetic coupling resonant circuit.

[0070] The primary series resonant compensation network is a coupling coil and a coupling capacitor that form a resonant network, and the secondary series resonant compensation network is just a coupling coil connected in series with a compensation capacitor. The transmission medium between the two coupling coils is water.

[0071] The underwater wireless power transmission system based on deep learning of the present invention has a structure as follows Figure 1 As shown, it includes a pair of coupling coils, and the pair of coupling coils cooperates with the SS energy resonance compensation network. The system adopts the SS energy resonance compensation network to achieve constant current output. The SS energy resonance compensation network is connected to an impedance matching module, and the impedance matching module is connected to the LSTM control module and each sensor in turn. The input end of the SS energy resonance compensation network is also connected to a high-frequency inverter circuit, and the high-frequency inverter circuit is connected to a power supply. The output end of the SS energy resonance compensation network is also connected to a high-frequency rectifier module, and the high-frequency rectifier module is connected to a load.

[0072] Combination Figure 2 , Figure 3 The primary side of the SS energy resonant compensation network is connected to a full-bridge high-frequency inverter circuit, the full-bridge inverter circuit is connected to the compensation capacitor in the resonant compensation network, the end of the compensation capacitor is connected to the primary coupling coil to form a primary S compensation network, and the secondary coupling coil is connected to a secondary compensation capacitor to form a secondary compensation network, thereby forming an SS resonant network. The end of the secondary compensation network is connected to a high-frequency rectifier circuit, and then connected to the load to form a loop.

[0073] The specific structure of the impedance matching module is: it consists of a sensor, a wireless communication module, an LSTM control module and a variable capacitor. The sensor is connected to the wireless communication module and converts information into data and transmits it to the wireless communication module receiving end connected to the LSTM control module. The receiving end transmits the information to the LSTM control module, and the LSTM control module controls the capacitance change to achieve impedance matching, wherein the variable capacitor is the compensation capacitor in the SS energy resonance compensation network and is connected to the coupled inductor.

[0074] The specific structure of the high-frequency inverter circuit is as follows: it is composed of four N-type silicon carbide MOSFET switch tubes and four reverse freewheeling diodes connected in parallel at both ends of the switch tubes. The four switch tubes are divided into two groups, namely the left half bridge and the right half bridge. The upper and lower arms of the switch tubes Q1 and Q3 are complementary to each other to form the left half bridge, and the upper and lower arms of the switch tubes Q2 and Q4 are complementary to each other to form the right half bridge. The D poles of the switch tubes Q1 and Q2 are connected and connected to the positive pole of the DC power supply. The S poles of the switch tubes Q3 and Q4 are connected and then connected to the negative pole of the DC power supply. The S pole of the switch tube Q1 is connected to the D pole of the switch tube Q3 and then connected to the compensation capacitor of the compensation network. The S pole of the switch tube Q2 is connected to the D pole of the switch tube Q4 and then connected to the coupling inductor of the compensation network. The high-frequency AC circuit is connected to the DC input, and the gate drive circuit controls the power switch to be turned on and off. The DC power passes through the power switch tube and then passes through the diode to realize the conversion of DC power to AC power.

[0075] The specific structure of the high-frequency rectifier module is as follows: it is composed of a filter capacitor and four fast recovery diodes, wherein the cathodes of diodes D1 and D2 are connected to one end of the filter capacitor and connected to the positive pole of the load, the cathodes of diodes D1 and D2 are connected to the cathodes of diodes D3 and D4 respectively, the anode of diode D1 is connected to the secondary compensation capacitor in the secondary compensation network, the cathode of diode D4 is connected to the secondary coupling coil in the secondary compensation network, and the anodes of diodes D3 and D4 are connected to the other end of the filter capacitor and the negative pole of the load. The compensation network transmits high-frequency AC to the high-frequency rectifier module, and the high-frequency rectifier module converts AC into DC.

[0076] The dynamic control method of the underwater wireless power transmission system based on deep learning of the present invention is specifically implemented according to the following steps:

[0077] The primary side uses SiC-MOSFET to build a high-frequency inverter circuit to convert DC power into high-frequency AC power for wireless power transmission;

[0078] The high-frequency AC power enters the compensation network and transfers energy to the secondary side through the primary-secondary coupling coil, and converts the AC power into the required DC power through high-frequency rectification for use by the load;

[0079] When the docking position of the spacecraft fluctuates and the external environment changes, the charging platform detects the parameter changes and transmits the information to the LSTM control module. After calculation, the variable capacitor module in the impedance matching module is controlled to adjust the capacitance to achieve impedance matching and power maximization.

[0080] The state of the vehicle during each charging and the corresponding capacitance value of the variable capacitor are saved as data for analysis, and the next charging is optimized and adjusted based on the data, thereby achieving deep learning.

[0081] In the circuit of the magnetically coupled resonant wireless power system involved in the present invention, a DC power supply is set to be converted into AC power after passing through a full-bridge inverter network, and a high-frequency inverter circuit is constructed using SiC-MOSFET (silicon carbide metal oxide semiconductor field effect transistor). SiC-MOSFET is a third-generation semiconductor material with the characteristics of high breakdown field strength and low on-resistance. Even in a high-temperature environment, its on-resistance rise rate is much lower than that of Si-MOSFET, and it can maintain low conduction loss in a wide current range. It has the advantages of extremely fast switching speed, high power density, high temperature resistance and high thermal conductivity. The high-frequency inverter circuit constructed thereby has the advantages of higher power density, smaller size, higher temperature resistance, high voltage and large current resistance and fast switching.

[0082] The high frequency inverter circuit works as follows:

[0083] The switch tubes Q1, Q2, Q3 and Q4 are all N-type MOSFET switch tubes constituting a full-bridge inverter;

[0084] (1) When the switch tubes Q1 and Q4 are closed and the switch tubes Q2 and Q3 are opened, the current flows through Q1 and Q4 to form a loop;

[0085] (2) When the switch tubes Q2 and Q3 are closed and the switch tubes Q2 and Q3 are opened, the switch tubes Q2 and Q3 cannot be closed immediately, and the direction of the inductor current cannot change suddenly at this moment. At this time, the current flows through the anti-parallel diodes of the switch tubes Q2 and Q3 for freewheeling;

[0086] (3) After the inductor current passes through zero, the switch tubes Q2 and Q3 are closed, and the inductor current flows through the switch tubes Q2 and Q3 in the reverse direction;

[0087] (4) When the switch tubes Q2 and Q3 are disconnected and the switch tubes Q1 and Q4 are closed again, the switch tubes Q1 and Q4 cannot be closed immediately. As analyzed above, the direction of the current cannot change suddenly. The current flows through the anti-parallel diodes of the switch tubes Q1 and Q4, and the above process is repeated in the subsequent cycles.

[0088] Step (2) and step (4) are energy feedback processes, in which the diode provides a feedback energy channel, and this process is also a freewheeling process of the load current.

[0089] Combination Figure 4 , Figure 5 , SS energy resonance compensation network, U in is the input voltage. The full-bridge inverter composed of switch tubes Q1 to Q4 provides the high-frequency AC power required by the system. The uncontrolled rectifier is composed of diodes D1 to D4. Lp and Ls are the self-inductance of the transmitting coil and the receiving coil respectively. Cp and Cs are the series compensation capacitors on the transmitting side and the receiving side respectively. M is the mutual inductance between the coils. C1 is the filter capacitor. R L is the load resistance, and the output voltage is Uout;

[0090] The output voltage of the inverter is U AB And the input equivalent resistance Req of the rectifier satisfies the following conditions:

[0091]

[0092] exist Figure 2 In the structure shown, according to Kirchhoff's law, the resonance conditions of the transmitting circuit and the receiving circuit are derived and expressed as:

[0093]

[0094] Without considering the losses of the inverter and rectifier modules, the output voltage of the system is expressed as:

[0095]

[0096] The output power is expressed as:

[0097]

[0098] In the impedance matching module, U AB It is expressed as the power supply voltage, and the load impedance is set to R L The output impedance of the power supply is Rs, L P , R 1 are the series equivalent inductance and equivalent resistance of the transmitting resonant coil, C P L is the series resonant capacitor at the transmitting end; S , R 2 are the equivalent inductance and equivalent resistance of the receiving resonant coil in series, C S is the resonant capacitor at the receiving end;

[0099] In the magnetic resonance wireless energy transfer system, the input and load reflection coefficients are expressed as:

[0100]

[0101] The output and load reflection coefficients are expressed as:

[0102]

[0103] When the condition Γ is met s =Γ in *, the power output is maximum;

[0104] When the condition Γ is met out =Γ L *, the load absorbs the maximum power, that is:

[0105]

[0106] have to:

[0107] R s =Z in * ,R L =Z out * (8)

[0108] Output impedance Rs and input impedance Z in And load impedance R L With output impedance Z out They are conjugate to each other. From this analysis, we know that when the condition Γ is satisfied s =Γ in * and Γ out =Γ L *, that is, the power supply output impedance Rs and input impedance Z in The real part is equal to the output impedance Z out With load impedance R L The real part of is equal, at this time the power supply has the maximum output power, and the power delivered to the load is also the maximum;

[0109] First, assume that the output impedance of the power supply is Z 0 is a fixed value, the equivalent impedance of the wireless power transmitter is: Z in =R in +jX in When R in >Z 0 When using an L-type impedance matching network, according to the basic circuit series-parallel law, if impedance matching is to be performed, the impedance of the matching network followed by the load impedance is equal to Z 0 :

[0110]

[0111] Further:

[0112]

[0113] When R in <Z 0 When using an inverse L-type impedance matching network, the formula for the inverse L-type impedance matching network is derived as follows:

[0114]

[0115] Separate the imaginary and real parts,

[0116]

[0117] Further:

[0118]

[0119] Based on the analysis of wireless power transmission of the above L-type impedance transformation network, it is concluded that by adding a suitable impedance matching network, any impedance can be transformed into a target impedance. Based on this, this function of the impedance transformation network can be used to achieve some design goals of the magnetically coupled resonant wireless power transmission system. When the high-frequency power supply internal resistance Z is known, 0 , power supply operating frequency f and load resistance R L , by adding an appropriate impedance matching network at the energy transmitting end of the system, the equivalent input impedance Z of the system after adding the impedance matching network can be in Equal to the internal resistance Z of the high-frequency power supply 0 . This achieves the maximum power output of the high-frequency power supply, and the corresponding system also achieves higher transmission efficiency.

[0120] The core formula of the long short-term memory network is as follows:

[0121]

[0122] Among them, f t represents the output of the forget gate, indicating which data needs to be forgotten, i.e. eliminated, f t The value ranges from 0 to 1, where 0 means completely forgotten and 1 means completely retained. f represents the weight matrix of the forget gate; h t-1 Indicates the hidden state at the last moment, that is, the data state of the last charge, including voltage, current, efficiency, power, and the adjustable capacitance value in the compensation capacitor, x t Represents the information input at the current moment, including the current charging voltage, current, efficiency, power of the spacecraft, the adjustable capacitance value in the compensation capacitor, and the bias term b of the forget gate fIt is a learnable parameter used to adjust the activation threshold of the forget gate. The initial value is large, so that the forget gate tends to retain information at the beginning, avoiding learning difficulties caused by excessive forgetting in the early stage of training. That is, the probability of charging information at the beginning avoids too little initial information. t Represents the output of the input gate, indicating whether to record the information of this charging and the recording ratio. Indicates temporary storage of new information; C t Indicates the current cell state, that is, after multiple charging and recording of data, the charging state with the most times is recorded as long-term memory. t represents the output of the output gate (values ​​range from 0 to 1, controlling the contribution of the cell state to the current output), that is, the impact of the most common charging data on the current charging state, h t Indicates that the data has been recorded and will be used as the data for the next comparison.

[0123] By analyzing historical data, LSTM can predict the position changes, attitude adjustments and changes in electromagnetic coupling efficiency of the spacecraft during the charging process, so as to adjust the charging parameters in advance. During the wireless charging process, LSTM can dynamically adjust the impedance parameters based on historical parameters and model predictions to achieve efficient charging. LSTM can also be used to monitor the operating status of the wireless charging system. By analyzing the time series data of current and voltage parameters during the charging process, abnormal situations can be detected in time and early warnings can be issued.

[0124] Example 1

[0125] The underwater wireless power transmission system based on deep learning of the present invention has a structure as follows Figure 1 As shown, it includes a pair of coupling coils, and the pair of coupling coils cooperates with the SS energy resonance compensation network. The system adopts the SS energy resonance compensation network to achieve constant current output. The SS energy resonance compensation network is connected to an impedance matching module, and the impedance matching module is connected to the LSTM control module and each sensor in turn. The input end of the SS energy resonance compensation network is also connected to a high-frequency inverter circuit, and the high-frequency inverter circuit is connected to a power supply. The output end of the SS energy resonance compensation network is also connected to a high-frequency rectifier module, and the high-frequency rectifier module is connected to a load.

[0126] Example 2

[0127] The underwater wireless power transmission system based on deep learning of the present invention has a structure as follows Figure 1As shown, it includes a pair of coupling coils, and the pair of coupling coils cooperates with the SS energy resonance compensation network. The system adopts the SS energy resonance compensation network to achieve constant current output. The SS energy resonance compensation network is connected to an impedance matching module, and the impedance matching module is connected to the LSTM control module and each sensor in turn. The input end of the SS energy resonance compensation network is also connected to a high-frequency inverter circuit, and the high-frequency inverter circuit is connected to a power supply. The output end of the SS energy resonance compensation network is also connected to a high-frequency rectifier module, and the high-frequency rectifier module is connected to a load.

[0128] Combination Figure 2 , Figure 3 The primary side of the SS energy resonant compensation network is connected to a full-bridge high-frequency inverter circuit, the full-bridge inverter circuit is connected to the compensation capacitor in the resonant compensation network, the end of the compensation capacitor is connected to the primary coupling coil to form a primary S compensation network, and the secondary coupling coil is connected to a secondary compensation capacitor to form a secondary compensation network, thereby forming an SS resonant network. The end of the secondary compensation network is connected to a high-frequency rectifier circuit, and then connected to the load to form a loop.

[0129] Example 3

[0130] The underwater wireless power transmission system based on deep learning of the present invention has a structure as follows Figure 1 As shown, it includes a pair of coupling coils, and the pair of coupling coils cooperates with the SS energy resonance compensation network. The system adopts the SS energy resonance compensation network to achieve constant current output. The SS energy resonance compensation network is connected to an impedance matching module, and the impedance matching module is connected to the LSTM control module and each sensor in turn. The input end of the SS energy resonance compensation network is also connected to a high-frequency inverter circuit, and the high-frequency inverter circuit is connected to a power supply. The output end of the SS energy resonance compensation network is also connected to a high-frequency rectifier module, and the high-frequency rectifier module is connected to a load.

[0131] Combination Figure 2 , Figure 3 The primary side of the SS energy resonant compensation network is connected to a full-bridge high-frequency inverter circuit, the full-bridge inverter circuit is connected to the compensation capacitor in the resonant compensation network, the end of the compensation capacitor is connected to the primary coupling coil to form a primary S compensation network, and the secondary coupling coil is connected to a secondary compensation capacitor to form a secondary compensation network, thereby forming an SS resonant network. The end of the secondary compensation network is connected to a high-frequency rectifier circuit, and then connected to the load to form a loop.

[0132] The specific structure of the impedance matching module is: it consists of a sensor, a wireless communication module, an LSTM control module and a variable capacitor. The sensor is connected to the wireless communication module and converts information into data and transmits it to the wireless communication module receiving end connected to the LSTM control module. The receiving end transmits the information to the LSTM control module, and the LSTM control module controls the capacitance change to achieve impedance matching, wherein the variable capacitor is the compensation capacitor in the SS energy resonance compensation network and is connected to the coupled inductor.

[0133] The specific structure of the high-frequency inverter circuit is as follows: it is composed of four N-type silicon carbide MOSFET switch tubes and four reverse freewheeling diodes connected in parallel at both ends of the switch tubes. The four switch tubes are divided into two groups, namely the left half bridge and the right half bridge. The upper and lower arms of the switch tubes Q1 and Q3 are complementary to each other to form the left half bridge, and the upper and lower arms of the switch tubes Q2 and Q4 are complementary to each other to form the right half bridge. The D poles of the switch tubes Q1 and Q2 are connected and connected to the positive pole of the DC power supply. The S poles of the switch tubes Q3 and Q4 are connected and then connected to the negative pole of the DC power supply. The S pole of the switch tube Q1 is connected to the D pole of the switch tube Q3 and then connected to the compensation capacitor of the compensation network. The S pole of the switch tube Q2 is connected to the D pole of the switch tube Q4 and then connected to the coupling inductor of the compensation network. The high-frequency AC circuit is connected to the DC input, and the gate drive circuit controls the power switch to be turned on and off. The DC power passes through the power switch tube and then passes through the diode to realize the conversion of DC power to AC power.

[0134] Example 4

[0135] The underwater wireless power transmission system based on deep learning of the present invention has a structure as follows Figure 1 As shown, it includes a pair of coupling coils, and the pair of coupling coils cooperate with the SS energy resonance compensation network. The system adopts the SS energy resonance compensation network to achieve constant current output. The SS energy resonance compensation network is connected to an impedance matching module, and the impedance matching module is connected to the LSTM control module and each sensor in turn. The input end of the SS energy resonance compensation network is also connected to a high-frequency inverter circuit, and the high-frequency inverter circuit is connected to a power supply. The output end of the SS energy resonance compensation network is also connected to a high-frequency rectifier module, and the high-frequency rectifier module is connected to a load.

[0136] Combination Figure 2 , Figure 3 The primary side of the SS energy resonant compensation network is connected to a full-bridge high-frequency inverter circuit, the full-bridge inverter circuit is connected to the compensation capacitor in the resonant compensation network, the end of the compensation capacitor is connected to the primary coupling coil to form a primary S compensation network, and the secondary coupling coil is connected to a secondary compensation capacitor to form a secondary compensation network, thereby forming an SS resonant network. The end of the secondary compensation network is connected to a high-frequency rectifier circuit, and then connected to the load to form a loop.

[0137] The specific structure of the impedance matching module is: it consists of a sensor, a wireless communication module, an LSTM control module and a variable capacitor. The sensor is connected to the wireless communication module and converts information into data and transmits it to the wireless communication module receiving end connected to the LSTM control module. The receiving end transmits the information to the LSTM control module, and the LSTM control module controls the capacitance change to achieve impedance matching, wherein the variable capacitor is the compensation capacitor in the SS energy resonance compensation network and is connected to the coupled inductor.

[0138] The specific structure of the high-frequency rectifier module is as follows: it is composed of a filter capacitor and four fast recovery diodes, wherein the cathodes of diodes D1 and D2 are connected to one end of the filter capacitor and connected to the positive pole of the load, the cathodes of diodes D1 and D2 are connected to the cathodes of diodes D3 and D4 respectively, the anode of diode D1 is connected to the secondary compensation capacitor in the secondary compensation network, the cathode of diode D4 is connected to the secondary coupling coil in the secondary compensation network, and the anodes of diodes D3 and D4 are connected to the other end of the filter capacitor and the negative pole of the load. The compensation network transmits high-frequency AC to the high-frequency rectifier module, and the high-frequency rectifier module converts AC into DC.

[0139] Example 5

[0140] The dynamic control method of the underwater wireless power transmission system based on deep learning of the present invention is specifically implemented according to the following steps:

[0141] The primary side uses SiC-MOSFET to build a high-frequency inverter circuit to convert DC power into high-frequency AC power for wireless power transmission;

[0142] The high-frequency AC power enters the compensation network and transfers energy to the secondary side through the primary-secondary coupling coil, and converts the AC power into the required DC power through high-frequency rectification for use by the load;

[0143] When the docking position of the spacecraft fluctuates and the external environment changes, the charging platform detects the parameter changes and transmits the information to the LSTM control module. After calculation, the variable capacitor module in the impedance matching module is controlled to adjust the capacitance to achieve impedance matching and power maximization.

[0144] The state of the vehicle during each charging and the corresponding capacitance value of the variable capacitor are saved as data for analysis, and the next charging is optimized and adjusted based on the data, thereby achieving deep learning.

[0145] Example 6

[0146] The dynamic control method of the underwater wireless power transmission system based on deep learning of the present invention is specifically implemented according to the following steps:

[0147] The primary side uses SiC-MOSFET to build a high-frequency inverter circuit to convert DC power into high-frequency AC power for wireless power transmission;

[0148] The high-frequency AC power enters the compensation network and transfers energy to the secondary side through the primary-secondary coupling coil, and converts the AC power into the required DC power through high-frequency rectification for use by the load;

[0149] When the docking position of the spacecraft fluctuates and the external environment changes, the charging platform detects the parameter changes and transmits the information to the LSTM control module. After calculation, the variable capacitor module in the impedance matching module is controlled to adjust the capacitance to achieve impedance matching and power maximization.

[0150] The state of the vehicle during each charging and the corresponding capacitance value of the variable capacitor are saved as data for analysis, and the next charging is optimized and adjusted based on the data, thereby achieving deep learning.

[0151] In the circuit of the magnetically coupled resonant wireless power system involved in the present invention, a DC power supply is set to be converted into AC power after passing through a full-bridge inverter network, and a high-frequency inverter circuit is constructed using SiC-MOSFET (silicon carbide metal oxide semiconductor field effect transistor). SiC-MOSFET is a third-generation semiconductor material with the characteristics of high breakdown field strength and low on-resistance. Even in a high-temperature environment, its on-resistance rise rate is much lower than that of Si-MOSFET, and it can maintain low conduction loss in a wide current range. It has the advantages of extremely fast switching speed, high power density, high temperature resistance and high thermal conductivity. The high-frequency inverter circuit constructed thereby has the advantages of higher power density, smaller size, higher temperature resistance, high voltage and large current resistance and fast switching.

[0152] The high frequency inverter circuit works as follows:

[0153] The switch tubes Q1, Q2, Q3 and Q4 are all N-type MOSFET switch tubes constituting a full-bridge inverter;

[0154] (1) When the switch tubes Q1 and Q4 are closed and the switch tubes Q2 and Q3 are opened, the current flows through Q1 and Q4 to form a loop;

[0155] (2) When the switch tubes Q2 and Q3 are closed and the switch tubes Q2 and Q3 are opened, the switch tubes Q2 and Q3 cannot be closed immediately, and the direction of the inductor current cannot change suddenly at this moment. At this time, the current flows through the anti-parallel diodes of the switch tubes Q2 and Q3 for freewheeling;

[0156] (3) After the inductor current passes through zero, the switch tubes Q2 and Q3 are closed, and the inductor current flows through the switch tubes Q2 and Q3 in the reverse direction;

[0157] (4) When the switch tubes Q2 and Q3 are disconnected and the switch tubes Q1 and Q4 are closed again, the switch tubes Q1 and Q4 cannot be closed immediately. As analyzed above, the direction of the current cannot change suddenly. The current flows through the anti-parallel diodes of the switch tubes Q1 and Q4, and the above process is repeated in the subsequent cycles.

[0158] Step (2) and step (4) are energy feedback processes, in which the diode provides a feedback energy channel, and this process is also a freewheeling process of the load current.

[0159] In the SS energy resonance compensation network, U in is the input voltage. The full-bridge inverter composed of switch tubes Q1 to Q4 provides the high-frequency AC power required by the system. The uncontrolled rectifier is composed of diodes D1 to D4. Lp and Ls are the self-inductance of the transmitting coil and the receiving coil respectively. Cp and Cs are the series compensation capacitors on the transmitting side and the receiving side respectively. M is the mutual inductance between the coils. C1 is the filter capacitor. R L is the load resistance, and the output voltage is Uout;

[0160] The output voltage of the inverter is U AB And the input equivalent resistance Req of the rectifier satisfies the following conditions:

[0161]

[0162] exist Figure 2 In the structure shown, according to Kirchhoff's law, the resonance conditions of the transmitting circuit and the receiving circuit are derived and expressed as:

[0163]

[0164] Without considering the losses of the inverter and rectifier modules, the output voltage of the system is expressed as:

[0165]

[0166] The output power is expressed as:

[0167]

[0168] In the impedance matching module, U AB It is expressed as the power supply voltage, and the load impedance is set to R L The output impedance of the power supply is Rs, L P , R 1 are the series equivalent inductance and equivalent resistance of the transmitting resonant coil, C P is the transmitting end series resonant capacitor; L S , R 2are the equivalent inductance and equivalent resistance of the receiving resonant coil in series, C S is the resonant capacitor at the receiving end;

[0169] In the magnetic resonance wireless energy transfer system, the input and load reflection coefficients are expressed as:

[0170]

[0171] The output and load reflection coefficients are expressed as:

[0172]

[0173] When the condition Γ is met s =Γ in *, the power output is maximum;

[0174] When the condition Γ is met out =Γ L *, the load absorbs the maximum power, that is:

[0175]

[0176] Get: R s =Z in * ,R L =Z out * (8)

[0177] Output impedance Rs and input impedance Z in And load impedance R L With output impedance Z out They are conjugate to each other. From this analysis, we know that when the condition Γ is satisfied s =Γ in * and Γ out =Γ L *, that is, the power supply output impedance Rs and input impedance Z in The real part is equal to the output impedance Z out With load impedance R L The real part of is equal, at this time the power supply has the maximum output power, and the power delivered to the load is also the maximum;

[0178] First, assume that the output impedance of the power supply is Z 0 is a fixed value, the equivalent impedance of the wireless power transmitter is: Z in =R in +jX in When R in >Z 0When using an L-type impedance matching network, according to the basic circuit series-parallel law, if impedance matching is to be performed, the impedance of the matching network followed by the load impedance is equal to Z 0 :

[0179]

[0180] Further:

[0181]

[0182] When R in <Z 0 When using an inverse L-type impedance matching network, the formula for the inverse L-type impedance matching network is derived as follows:

[0183]

[0184] Separate the imaginary and real parts,

[0185]

[0186] Further:

[0187]

[0188] Based on the analysis of wireless power transmission of the above L-type impedance transformation network, it is concluded that by adding a suitable impedance matching network, any impedance can be transformed into a target impedance. Based on this, this function of the impedance transformation network can be used to achieve some design goals of the magnetically coupled resonant wireless power transmission system. When the high-frequency power supply internal resistance Z is known, 0 , power supply operating frequency f and load resistance R L , by adding an appropriate impedance matching network at the energy transmitting end of the system, the equivalent input impedance Z of the system after adding the impedance matching network can be in Equal to the internal resistance Z of the high-frequency power supply 0 . This achieves the maximum power output of the high-frequency power supply, and the corresponding system also achieves higher transmission efficiency.

[0189] The core formula of the long short-term memory network is as follows:

[0190]

[0191] Among them, f t represents the output of the forget gate, indicating which data needs to be forgotten, i.e. eliminated, f t The value ranges from 0 to 1, where 0 means completely forgotten and 1 means completely retained. f represents the weight matrix of the forget gate; h t-1Indicates the hidden state at the last moment, that is, the data state of the last charge, including voltage, current, efficiency, power, and the adjustable capacitance value in the compensation capacitor, x t Represents the information input at the current moment, including the current charging voltage, current, efficiency, power of the spacecraft, the adjustable capacitance value in the compensation capacitor, and the bias term b of the forget gate f It is a learnable parameter used to adjust the activation threshold of the forget gate. The initial value is large, so that the forget gate tends to retain information at the beginning, avoiding learning difficulties caused by excessive forgetting in the early stage of training. That is, the probability of charging information at the beginning avoids too little initial information. t Represents the output of the input gate, indicating whether to record the information of this charging and the recording ratio. Indicates temporary storage of new information; C t Indicates the current cell state, that is, after multiple charging and recording of data, the charging state with the most times is recorded as long-term memory. t represents the output of the output gate (values ​​range from 0 to 1, controlling the contribution of the cell state to the current output), that is, the impact of the most common charging data on the current charging state, h t Indicates that the data has been recorded and will be used as the data for the next comparison.

[0192] LSTM networks can process time series data and predict these dynamic parameters. For example, by analyzing historical data, LSTM can predict the position change, attitude adjustment and electromagnetic coupling efficiency change of the spacecraft during the charging process, so as to adjust the charging parameters in advance. In the process of wireless charging, the real-time adjustment of dynamic parameters is the key to improving charging efficiency. LSTM can dynamically adjust parameters such as impedance, frequency and phase according to historical parameters and model predictions to achieve efficient charging. For example, after the spacecraft docks, LSTM analyzes historical data and quickly performs impedance transformation based on current data combined with historical data to achieve impedance matching. At the same time, it detects changes in spacecraft and surrounding environment parameters in real time and performs real-time transformation. The data is saved as historical reference data to predict the next charging, reduce reaction time, form "long and short-term memory", and make the system more stable. LSTM can also be used to monitor the operating status of the wireless charging system. By analyzing the time series data of parameters such as current and voltage during the charging process, abnormal conditions can be detected in time and early warnings can be issued. This helps to improve the reliability and safety of the system.

[0193] In laboratory tests, the following results were achieved: coupling coefficient identification accuracy: 98.2% (dynamic change of relative displacement 0-50cm); impedance identification response time: 8.7ms (load step change scenario); maximum resonant frequency tracking error: ±2.1kHz (temperature fluctuation ±20℃). In actual deployment, it is recommended to adopt an edge-cloud collaborative architecture, run a lightweight identification model on a local device, and upload abnormal data to a surface relay station through underwater acoustic communication for model iteration and update.

[0194] Implementation of control strategy based on LSTM: By deeply combining the timing prediction capability of LSTM with the physical characteristics of the UWPT system, the timing prediction capability based on LSTM can achieve better dynamic response speed (increased by about 40%) and environmental adaptability (efficiency fluctuation reduced by 30%) than traditional PID control.

Claims

1. The underwater wireless power transmission system based on deep learning is characterized by: The system comprises a pair of coupling coils, which cooperate with the SS energy resonance compensation network. The system adopts the SS energy resonance compensation network to realize constant current output. The SS energy resonance compensation network is connected with an impedance matching module, which is connected with the LSTM control module and each sensor in turn. The input end of the SS energy resonance compensation network is also connected with a high-frequency inverter circuit, which is connected with a power supply. The output end of the SS energy resonance compensation network is also connected with a high-frequency rectifier module, which is connected with a load.

2. The underwater wireless power transmission system based on deep learning according to claim 1, characterized in that: The primary side of the SS energy resonance compensation network is connected to a full-bridge high-frequency inverter circuit, the full-bridge inverter circuit is connected to the compensation capacitor in the resonance compensation network, the end of the compensation capacitor is connected to the primary coupling coil to form a primary S compensation network, the secondary coupling coil is connected to a secondary compensation capacitor to form a secondary compensation network, thereby forming an SS resonance network, the end of the secondary compensation network is connected to a high-frequency rectifier circuit, and then connected to a load to form a loop.

3. The underwater wireless power transmission system based on deep learning according to claim 2 is characterized in that: The specific structure of the impedance matching module is: it consists of a sensor, a wireless communication module, an LSTM control module and a variable capacitor. The sensor is connected to the wireless communication module and converts information into data and transmits it to the receiving end of the wireless communication module connected to the LSTM control module. The receiving end transmits the information to the LSTM control module, and the LSTM control module controls the capacitance change to achieve impedance matching, wherein the variable capacitor is the compensation capacitor in the SS energy resonance compensation network and is connected to the coupled inductor.

4. The underwater wireless power transmission system based on deep learning according to claim 3 is characterized in that: The specific structure of the high-frequency inverter circuit is as follows: it is composed of four N-type silicon carbide MOSFET switch tubes and four reverse freewheeling diodes connected in parallel at both ends of the switch tubes. The four switch tubes are divided into two groups, namely a left half bridge and a right half bridge. The upper and lower arms of the switch tubes Q1 and Q3 are complementary to each other to form the left half bridge, and the upper and lower arms of the switch tubes Q2 and Q4 are complementary to each other to form the right half bridge. The D poles of the switch tubes Q1 and Q2 are connected and connected to the positive pole of the DC power supply. The S poles of the switch tubes Q3 and Q4 are connected and then connected to the negative pole of the DC power supply. The S pole of the switch tube Q1 is connected to the D pole of the switch tube Q3 and then connected to the compensation capacitor of the compensation network. The S pole of the switch tube Q2 is connected to the D pole of the switch tube Q4 and then connected to the coupling inductor of the compensation network.

5. The underwater wireless power transmission system based on deep learning according to claim 4 is characterized in that: The specific structure of the high-frequency rectifier module is as follows: it is composed of a filter capacitor and four fast recovery diodes, wherein the cathodes of diodes D1 and D2 are connected to one end of the filter capacitor and connected to the positive electrode of the load, the anodes of diodes D1 and D2 are connected to the cathodes of diodes D3 and D4 respectively, the anode of diode D1 is connected to the secondary compensation capacitor in the secondary compensation network, the cathode of diode D4 is connected to the secondary coupling coil in the secondary compensation network, and the anodes of diodes D3 and D4 are connected to the other end of the filter capacitor and the negative electrode of the load.

6. A dynamic control method for an underwater wireless power transmission system based on deep learning, characterized in that: The underwater wireless power transmission system based on deep learning according to claim 5 is implemented specifically according to the following steps: The primary side uses SiC-MOSFET to build a high-frequency inverter circuit to convert DC power into high-frequency AC power for wireless power transmission; The high-frequency AC power enters the compensation network and transfers energy to the secondary side through the primary-secondary coupling coil, and converts the AC power into the required DC power through high-frequency rectification for use by the load; When the docking position of the spacecraft fluctuates and the external environment changes, the charging platform detects the parameter changes and transmits the information to the LSTM control module. After calculation, the variable capacitor module in the impedance matching module is controlled to adjust the capacitance to achieve impedance matching and power maximization. The state of the vehicle during each charging and the corresponding capacitance value of the variable capacitor are saved as data for analysis, and the next charging is optimized and adjusted based on the data, thereby achieving deep learning.

7. The dynamic control method of the underwater wireless power transmission system based on deep learning according to claim 6 is characterized in that: The high frequency inverter circuit works as follows: The switch tubes Q1, Q2, Q3 and Q4 are all N-type MOSFET switch tubes constituting a full-bridge inverter; (1) When the switch tubes Q1 and Q4 are closed and the switch tubes Q2 and Q3 are opened, the current flows through Q1 and Q4 to form a loop; (2) When the switch tubes Q2 and Q3 are closed and the switch tubes Q2 and Q3 are opened, the switch tubes Q2 and Q3 cannot be closed immediately, and the direction of the inductor current cannot change suddenly at this moment. At this time, the current flows through the anti-parallel diodes of the switch tubes Q2 and Q3 for freewheeling; (3) After the inductor current passes through zero, the switch tubes Q2 and Q3 are closed, and the inductor current flows through the switch tubes Q2 and Q3 in the reverse direction; (4) When the switch tubes Q2 and Q3 are disconnected and the switch tubes Q1 and Q4 are closed again, the switch tubes Q1 and Q4 cannot be closed immediately. As analyzed above, the direction of the current cannot change suddenly. The current flows through the anti-parallel diodes of the switch tubes Q1 and Q4, and the above process is repeated in the subsequent cycles. Step (2) and step (4) are energy feedback processes, in which the diode provides a feedback energy channel, and this process is also a freewheeling process of the load current.

8. The dynamic control method of the underwater wireless power transmission system based on deep learning according to claim 7 is characterized in that: In the SS energy resonance compensation network, U in is the input voltage. The full-bridge inverter composed of switch tubes Q1 to Q4 provides the high-frequency AC power required by the system. The uncontrolled rectifier is composed of diodes D1 to D4. Lp and Ls are the self-inductance of the transmitting coil and the receiving coil respectively. Cp and Cs are the series compensation capacitors on the transmitting side and the receiving side respectively. M is the mutual inductance between the coils. C1 is the filter capacitor. R L is the load resistance, and the output voltage is Uout; The output voltage of the inverter is U AB And the input equivalent resistance Req of the rectifier satisfies the following conditions: According to Kirchhoff's law, the resonance conditions of the transmitting circuit and the receiving circuit are derived and expressed as: Without considering the losses of the inverter and rectifier modules, the output voltage of the system is expressed as: The output power is expressed as:

9. The dynamic control method of the underwater wireless power transmission system based on deep learning according to claim 8 is characterized in that: In the impedance matching module, U AB It is expressed as the power supply voltage, and the load impedance is set to R L The output impedance of the power supply is Rs, L P R1 and R2 are the series equivalent inductance and equivalent resistance of the transmitting resonant coil, respectively. P is the transmitting end series resonant capacitor; L S R2 and R3 are the equivalent inductance and resistance of the receiving resonant coil in series, respectively. S is the resonant capacitor at the receiving end; In the magnetic resonance wireless energy transfer system, the input and load reflection coefficients are expressed as: The output and load reflection coefficients are expressed as: When the condition Γ is met s =Γ in * When , the power output is maximum; When the condition Γ is met out =Γ L * When , the load absorbs the maximum power, that is: have to: R s =Z in * ,R L =Z out * (8) Output impedance Rs and input impedance Z in And load impedance R L With output impedance Z out They are conjugate to each other. From this analysis, we know that when the condition Γ is satisfied s =Γ in * and Γ out =Γ L *, that is, the power supply output impedance Rs and input impedance Z in The real part is equal to the output impedance Z out With load impedance R L The real part of is equal, at this time the power supply has the maximum output power, and the power delivered to the load is also the maximum; First, assuming that the output impedance Z0 of the power supply is a fixed value, the equivalent impedance of the wireless power transmitter is: Z in =R in +jX in , when R in When >Z0, an L-type impedance matching network is used. For the L-type impedance matching network, according to the basic circuit series-parallel law, if impedance matching is to be performed, the impedance of the matching network followed by the load impedance is equal to Z0: Further: When R in <is less than 0, an inverted L-shaped impedance matching network is adopted. The formula for the inverted L-shaped impedance matching network is derived as follows: Separate the imaginary and real parts, Further:

10. The dynamic control method of the underwater wireless power transmission system based on deep learning according to claim 9, characterized in that: The core formula of the long short-term memory network is as follows: Among them, f t represents the output of the forget gate, indicating which data needs to be forgotten, i.e. eliminated, f t The value ranges from 0 to 1, where 0 means completely forgotten and 1 means completely retained. f represents the weight matrix of the forget gate; h t-1 Indicates the hidden state at the last moment, that is, the data state of the last charge, including voltage, current, efficiency, power, and the adjustable capacitance value in the compensation capacitor, x t Represents the information input at the current moment, including the current charging voltage, current, efficiency, power of the spacecraft, the adjustable capacitance value in the compensation capacitor, and the bias term b of the forget gate f is a learnable parameter used to adjust the activation threshold of the forget gate, i t Represents the output of the input gate, indicating whether to record the information of this charging and the recording ratio. Indicates temporary storage of new information; C t Indicates the current cell state, that is, after multiple charging and recording of data, the charging state with the most times is recorded as long-term memory. t represents the output of the output gate, that is, the impact of the most common charging data on the current charging state, h t Indicates that the data has been recorded and used as the data for next comparison; By analyzing historical data, LSTM can predict the position changes, attitude adjustments and changes in electromagnetic coupling efficiency of the spacecraft during the charging process, so as to adjust the charging parameters in advance. During the wireless charging process, LSTM can dynamically adjust the impedance parameters based on historical parameters and model predictions to achieve efficient charging. LSTM can also be used to monitor the operating status of the wireless charging system. By analyzing the time series data of current and voltage parameters during the charging process, abnormal situations can be detected in time and early warnings can be issued.

Images