Method for training an electrical power model, method for obtaining a transmission power and related devices
By acquiring sample data of various variable parameters of the target circuit, initializing the learner, calculating the negative gradient to fit the decision tree, determining the loss function, and obtaining the power model, the problem of unstable power transmission is solved, and accurate power transmission prediction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SOUTHERN POWER GRID DIGITAL GRID RESEARCH INSTITUTE CO LTD
- Filing Date
- 2022-11-15
- Publication Date
- 2026-06-09
AI Technical Summary
In the prior art, the power transmission prediction algorithm of the circuit cannot accurately predict the power transmission instability problem caused by the influence of resonant frequency and load impedance.
By acquiring sample data of various variable parameters of the target circuit, multiple learners are initialized, negative gradients are calculated and decision trees are fitted, the target loss function with the minimum loss is determined, the power model is obtained, and the weighted sum is performed according to the sample data of adjacent time steps to adjust the weights to improve the prediction accuracy.
It enables accurate prediction of power transmission based on the actual circuit conditions, adapts to changes in circuit parameters, and improves the accuracy and stability of prediction.
Smart Images
Figure CN115796302B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of circuit technology, and in particular to a method for training an electrical power model, a method for obtaining transmission power, and related equipment. Background Technology
[0002] Wireless power transfer (WPT) technology is receiving increasing attention from research and industry, particularly in its application in wireless charging systems for new energy vehicles and ships. Pure electric ships using battery / capacitor packs have become the best solution to pollution problems in coastal and inland river ports, leading to the rapid development of new energy ships. Furthermore, the short lifespan of batteries has drastically increased the demand for wireless charging for mobile objects such as electric vehicles and ships. Magnetic resonant coupling is one of the most suitable methods for power transfer using resonant coils due to its high transmission efficiency and certain isolation distance. Predicting the power transfer power of circuits using magnetic resonant coupling is therefore particularly important.
[0003] Currently, in practical circuit systems, the circuit efficiency is affected by the resonant frequency and load impedance, resulting in unstable power transmission. This means that traditional power transmission prediction algorithms cannot accurately predict the power transmission of a circuit. Summary of the Invention
[0004] Therefore, it is necessary to provide a power model training method, a power acquisition method, and related equipment that can improve the accuracy of power transmission prediction in circuits, addressing the aforementioned technical problems.
[0005] Firstly, this application provides a method for training an efficiency power model. The method includes:
[0006] Obtain the first training sample and the maximum number of iterations. The first training sample includes sample data of various variable parameters of the target circuit.
[0007] Select at least one sample data from the first training sample to initialize the learner, and obtain multiple initialized learners. The types of sample data corresponding to the multiple initialized learners cover all types in the training sample.
[0008] Calculate the negative gradient of each initialized learner;
[0009] The decision tree is obtained by fitting the negative gradient of each initialized learner, and the loss function of each leaf node in the decision tree is calculated to determine the target loss function with the minimum loss.
[0010] Based on the same type of sample data at adjacent time points in the first training sample and the target loss function, the recursive relationship of sample data at adjacent time points in each type is obtained;
[0011] The power model of the target circuit is obtained by weighted summation based on the recursive relationships corresponding to each type.
[0012] In one embodiment, the method further includes:
[0013] Collect the current operating data of the target circuit and input the operating data into the power model to obtain the first power transmission power;
[0014] Determine the first deviation between the preset efficiency and the first power transmission power;
[0015] If the first deviation exceeds the preset deviation range, adjust the weights of each recursive relation to obtain the updated power model.
[0016] In one embodiment, after adjusting the weights of each recursive relation to obtain the updated power model if the first deviation exceeds a preset deviation range, the method further includes:
[0017] The updated power model is input using the current operating data of the target circuit to obtain the second power transmission power.
[0018] Determine the second deviation between the preset efficiency and the second power transmission power;
[0019] If the second deviation exceeds the preset deviation range, a second training sample is obtained again, and the power model is updated according to the second training sample, wherein the number of sample data in the second training sample is greater than the number of sample data in the first training sample.
[0020] In one embodiment, before determining the first deviation between the preset efficiency and the first power transmission power, the method includes:
[0021] Obtain the voltage, current, capacitance, and inductance values on the input side and the output side of the target circuit, and construct the node equations of the target circuit based on the capacitance and inductance values on the input side and the output side of the target circuit.
[0022] The angular frequency of the target circuit resonant point is obtained based on the nodal equation.
[0023] The preset efficiency is obtained based on the voltage and current on the input side and the voltage and current on the output side of the target circuit, as well as the angular frequency.
[0024] In one embodiment, the fitting based on the negative gradient of each initialized learner to obtain a corresponding decision tree, and the calculation of the loss function for each leaf node in the decision tree to determine the target loss function with the minimum loss, includes:
[0025] Discretize the continuous features in the decision tree.
[0026] Secondly, this application also provides a method for obtaining electrical power transmission power, the method comprising:
[0027] Acquire the current operating data of the target circuit, including the coil voltage, input power, output power, coil distance, receiving frequency value, and actual impedance change of the target circuit;
[0028] The working data is input into the power model to obtain the power transmission power of the target circuit. The power model is obtained by the method described in the first aspect.
[0029] In one embodiment, the method further includes:
[0030] When a change in the load of the target circuit is detected, the operating frequency of the target circuit is adjusted so that the power transmission power of the target circuit reaches the power transmission power.
[0031] Thirdly, this application also provides an electrical power model training device. The device includes:
[0032] The first acquisition module is used to acquire a first training sample and a maximum number of iterations, wherein the first training sample includes various types of sample data of the target circuit;
[0033] An initialization module is used to select at least one type of sample data from the first training sample to initialize the learner, thereby obtaining multiple initialized learners. The types of sample data corresponding to the multiple initialized learners cover all types in the training sample.
[0034] The first computation module is used to calculate the negative gradient of each initialized learner;
[0035] The fitting module is used to fit the negative gradient of each initialized learner to obtain the corresponding decision tree, and to calculate the loss function of each leaf node in the decision tree to determine the target loss function with the minimum loss.
[0036] The recursive relationship acquisition module is used to obtain the recursive relationship of sample data at adjacent time points in each type based on the same type of sample data at adjacent time points in the training samples and the target loss function.
[0037] The first model acquisition module is used to perform weighted summation based on the recursive relationships corresponding to each type to obtain the power model of the target circuit.
[0038] Fourthly, this application also provides an electrical power transmission power acquisition device. The device includes:
[0039] The working data acquisition module is used to acquire the current working data of the target circuit. The working data includes the coil voltage, input power, output power, coil distance, receiving side frequency value, and actual impedance change of the target circuit.
[0040] The prediction module is used to input the working data into the power model to obtain the power transmission power of the target circuit, wherein the power model is obtained by the method described in the first aspect.
[0041] Fifthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method steps described in the first and second aspects.
[0042] Sixthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the method steps described in the first and second aspects.
[0043] In a seventh aspect, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the method steps described in the first and second aspects.
[0044] The aforementioned power model training method, transmission power acquisition method, and related equipment, in the aforementioned efficiency power model training method, involve acquiring a first training sample and a maximum number of iterations. The first training sample includes sample data of various variable parameters of the target circuit. At least one sample data is selected from the first training sample to initialize a learner, resulting in multiple initialized learners. The types of sample data corresponding to the multiple initialized learners cover all types in the training sample. The negative gradient of each initialized learner is calculated. A decision tree is obtained by fitting the negative gradient of each initialized learner, and the loss function of each leaf node in the decision tree is calculated to determine the target loss function with the minimum loss. The recursive relationship of sample data of the same type at adjacent time points in the first training sample, and the target loss function, is obtained. A weighted sum is performed based on the recursive relationship corresponding to each type to obtain the power model of the target circuit. In this manner, the present application initializes multiple learners with data of various variable parameters in the training sample, then calculates the negative gradient of each learner, and fits the negative gradient of each learner to obtain a decision tree. The function with the minimum loss among the loss functions of each leaf node in the decision book is taken as the target loss function. Then, the recursive relationship between adjacent data in a type is obtained using the target loss function. The weighted summation of this recursive relationship is used to obtain the trained power model. Thus, the pre-stored model in this application is obtained by training on the variable parameters in the circuit and can make predictions based on the actual situation of the circuit. Therefore, it can accurately predict the power transmission of the circuit. Attached Figure Description
[0045] Figure 1 This is an application circuit diagram of an efficient power model training method in one embodiment;
[0046] Figure 2 This is a simplified schematic diagram of the application circuit of the efficiency power model training method in one embodiment;
[0047] Figure 3 This is a flowchart illustrating the efficiency power model training method in one embodiment;
[0048] Figure 4 This is a flowchart illustrating the efficiency prediction method in one embodiment;
[0049] Figure 5 This is a structural block diagram of an efficiency power model training device in one embodiment;
[0050] Figure 6 This is a structural block diagram of an efficiency prediction device in one embodiment;
[0051] Figure 7This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0052] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0053] This application provides an efficient power model training method that can be applied to, for example... Figure 1 The circuit shown includes an input side and an output side, with a resonant circuit between them. The input side includes a voltage detection module, an MCU control unit (other types of processors can also be used in specific implementations), a power supply, and a first capacitor (with a capacitance value of C). p ) and the first inductor (whose inductance is L) p The output side includes a second inductor (whose inductance value is L), etc. s ), the second capacitor (whose capacitance value is C) s ) and the load (whose resistance is R) L The first inductor and the second inductor form a resonant circuit. Figure 1 The circuit diagram shown can also be simplified to be equivalent to Figure 2 The circuit diagram shown, i.e., the input side of the circuit, mainly includes a power supply (AC) and a first capacitor (whose capacitance value is C). p ), the first inductor (its inductance value is L) p ) and the equivalent resistance of the first inductor (whose resistance is R) p The output side of the circuit mainly includes a second capacitor (whose capacitance value is C). s ), the second inductor (whose inductance value is L) s The second inductor is equivalent to the second resistor (whose resistance is R). s ) and load resistor (its resistance value is R) L The power supply provides power to the entire circuit, supplying electrical energy to the load through a resonant circuit composed of the first and second inductors.
[0054] In one embodiment, such as Figure 3 As shown, an efficient power model training method is provided, which is then applied to... Figure 1 Taking the MCU in the example, the following steps are included:
[0055] Step 310: Obtain the first training sample and the maximum number of iterations. The first training sample includes sample data of various variable parameters of the target circuit.
[0056] Specifically, the MCU can obtain the first training samples and the maximum number of iterations via the network. These parameters can also be input into memory by the user and then read by the MCU. To ensure the accuracy of the training results, the first training samples contain sufficient sample data. These first training samples include various types of sample data. The first training samples may include variable parameters of the target circuit. For example, variable parameters may include one or more of the following: system coil voltage, input power, output power, coil distance, receiver frequency value, and actual impedance change. It is understood that the more variable parameters there are, the more accurate the training model will be.
[0057] Step 320: Select at least one sample data from the first training sample to initialize the learner, and obtain multiple initialized learners. The types of sample data corresponding to the multiple initialized learners cover all types in the training sample.
[0058] Suppose that the first training sample input is T = {(x1,y1),(x2,y2),(x3,y3),…(x n ,y n x1 is a 1*6 matrix, and the six features in the matrix can be: system coil voltage, input power, output power, coil distance, receiving side frequency value and actual impedance change.
[0059] The maximum number of iterations T and the loss function L output the strong learner f(x), where x n =(x n,1 ,x n,2 ,…,x n,k )) represents the features of the input sample (taken from a 1x6 matrix), k is the number of features, and y n It is the output value of the nth sample.
[0060] Input the sample data of the first training sample into the learner to initialize the learner:
[0061]
[0062] Wherein, x is selected from at least one x n y i Selected from y n Specifically, at least one feature can be selected from multiple features to initialize the learner. Based on the input features, a predicted value is output, and T iterations are performed to obtain the initialized learner that minimizes the learner's error (after multiple iterations, the error of the learner obtained in each iteration is calculated, and the learner with the smallest error is selected). Each time a different x is selected, a corresponding learner is obtained, and multiple initialized learners are obtained in this way.
[0063] Step 330: Calculate the negative gradient of each initialized learner;
[0064] For samples i = 1, 2, ..., n, each sample corresponds to an initialized learner. Calculate the negative gradient of each initialized learner:
[0065]
[0066] Step 340: Fit the negative gradient of each initialized learner to obtain the corresponding decision tree, and calculate the loss function of each leaf node in the decision tree to determine the target loss function with the minimum loss.
[0067] After obtaining the negative gradient of each initialized learner, a fitting process is performed based on the negative gradient of each initialized learner to obtain the corresponding decision tree. The leaf nodes of this decision tree are R... mj ,j=1,2,3,…,J, where J is the number of leaf nodes in the regression tree t.
[0068] For the leaf regions j = 1, 2, 3, ..., J, calculate the best-fit value that minimizes the loss function for all leaf nodes. This minimum loss function is defined as the target loss function. Specifically, it can be obtained using the following formula:
[0069]
[0070] Among them, C tj This represents the loss value.
[0071] Furthermore, after obtaining the decision tree, it is also necessary to ensure that each data point in the decision tree is discrete. Therefore, after obtaining the decision tree, the continuous features in the decision tree are discretized. The discretization of the continuous features in the decision tree can be done using existing techniques, which will not be elaborated here.
[0072] Step 350: Based on the same type of sample data at adjacent time points in the first training sample and the target loss function, obtain the recursive relationship of sample data at adjacent time points in each type;
[0073] After obtaining the target loss function, the predicted value at time t is related to the result at the previous time, which can be represented as a recursive form:
[0074]
[0075] Where I represents the target loss function, f t (x) represents the predicted value at time t, f t-1 (x) represents the predicted value at time t-1, R tj R is a variable parameter in the sample data. tjSelected from x n For example, R tj Selected from x n The system coil voltage.
[0076] Based on the above formula, the expression for the predicted value at time t is derived as follows:
[0077]
[0078] This expression can express the recursive relationship between adjacent data of each type.
[0079] Step 360: Perform a weighted summation based on the recursive relationships corresponding to each type to obtain the power model of the target circuit.
[0080] Then, a weighted sum is performed based on the recursive relationship between adjacent data of each type to obtain the power model of the target circuit. The weights of each type can be determined manually in advance.
[0081] In the above-described efficiency power model training method, a first training sample and a maximum number of iterations are obtained. The first training sample includes sample data of various variable parameters of the target circuit. At least one sample data is selected from the first training sample to initialize a learner, resulting in multiple initialized learners. The types of sample data corresponding to these multiple initialized learners cover all types in the training sample. The negative gradient of each initialized learner is calculated. A decision tree is obtained by fitting the negative gradient of each initialized learner, and the loss function of each leaf node in the decision tree is calculated to determine the target loss function with the minimum loss. The recursive relationship between sample data of the same type at adjacent time points in the first training sample and the target loss function is obtained. A weighted sum is performed based on the recursive relationship corresponding to each type to obtain the power model of the target circuit. In this way, this application initializes multiple learners with data of various variable parameters in the training sample, then calculates the negative gradient of each learner, and fits the negative gradient of each learner to obtain a decision tree. The function with the minimum loss among the loss functions of each leaf node in the decision book is taken as the target loss function. Then, the recursive relationship between adjacent data in a type is obtained using the target loss function. The weighted summation of this recursive relationship is used to obtain the trained power model. Thus, the pre-stored model in this application is obtained by training on the variable parameters in the circuit and can make predictions based on the actual situation of the circuit. Therefore, it can accurately predict the power transmission of the circuit.
[0082] In one embodiment, the method further includes:
[0083] Collect the current operating data of the target circuit and input the operating data into the power model to obtain the first power transmission power;
[0084] Determine the first deviation between the preset efficiency and the first power transmission power;
[0085] If the first deviation exceeds the preset deviation range, adjust the weights of each recursive relation to obtain the updated power model.
[0086] Specifically, in this embodiment, to ensure the accuracy of the power model, prediction is also performed using an actual circuit. First, the current operating data of the target circuit can be obtained. This data may include: the coil voltage, input power, output power, coil distance, receiving-side frequency value, and actual impedance change. The coil distance refers to the distance between the input-side inductor and the output-side inductor. Simultaneously, the actual power transmission power of the target circuit is obtained as a preset efficiency.
[0087] The current operating data of the target circuit is input into the trained power model to obtain the first power transmission power. The first power transmission power is compared with the preset efficiency corresponding to the target circuit to determine the deviation between the preset efficiency and the first power transmission power, which is taken as the first deviation.
[0088] If the first deviation does not exceed the preset deviation range, it indicates that the accuracy of the trained prediction template is high. If the first deviation exceeds the preset deviation range, it indicates that the accuracy of the trained prediction template is low. In this case, to further improve the accuracy of the prediction results, the power model can be tuned, that is, the weights of each recursive relation can be modified to obtain a new power model. Then, the working data of the target circuit and the preset efficiency are used for re-verification until the first deviation between the preset efficiency and the first power transmission power does not exceed the preset deviation range.
[0089] In one embodiment, if the first deviation exceeds a preset deviation range, after adjusting the weights of each recursive relation to obtain the updated power model, the method further includes:
[0090] The updated power model is input using the current operating data of the target circuit to obtain the second power transmission power.
[0091] Determine the second deviation between the preset efficiency and the second power transmission power;
[0092] If the second deviation exceeds the preset deviation range, a second training sample is obtained again, and the power model is updated according to the second training sample, wherein the number of sample data in the second training sample is greater than the number of sample data in the first training sample.
[0093] Specifically, after repeatedly adjusting the weights of each recursive relation to obtain multiple updated power models, if the deviation between the first power transmission power corresponding to the updated power model and the preset efficiency corresponding to the target circuit (defined as the second deviation) exceeds the preset deviation range, it indicates that the data in the first training sample is insufficient and cannot meet the training requirements. Therefore, a second training sample is obtained. The second training sample can be a new sample or a sample added to the first training sample, as long as the total number of the second training samples is greater than that of the first training samples.
[0094] After obtaining the second training sample, the power model is retrained using the second training sample. The specific process of retraining the power model using the second training sample is the same as in the above embodiment, and will not be repeated here.
[0095] In one embodiment, before determining the first deviation between the preset efficiency and the first power transmission power, the following steps are included:
[0096] Obtain the voltage, current, capacitance, and inductance values on the input side and the output side of the target circuit, and construct the node equations of the target circuit based on the capacitance and inductance values on the input side and the output side of the target circuit.
[0097] The angular frequency of the target circuit resonant point is obtained based on the nodal equation.
[0098] The preset efficiency is obtained based on the voltage and current on the input side and the voltage and current on the output side of the target circuit, as well as the angular frequency.
[0099] Specifically, in this embodiment, the actual power transmission power of the target circuit can also be calculated during the training process, serving as the preset efficiency of the target circuit. The specific calculation process may include:
[0100] Obtain the voltage, current, capacitance, and inductance values on the input side and the output side of the target circuit, and construct the nodal equations of the target circuit based on the capacitance and inductance values on the input and output sides.
[0101] The nodal equations of the target circuit include:
[0102]
[0103]
[0104]
[0105] Among them, Rp It is coil L p The circuit has a built-in resistor, jw is the impedance of the capacitor and inductor, and the AC voltage source is parameter U. p The inductance and capacitance parameters on the input side are L. p C p The current is I p The inductor and capacitor parameters on the output side are L. s C s The load is R L The load voltage is Us. The operating frequency of the primary coil is denoted as w, and the resulting loop current is denoted as I. p and I s Z 11 Z represents the resistance on the input side. 12 Z represents the resistance on the output side. M This represents mutual inductance resistance.
[0106] Then, based on the above nodal equations, the angular frequency of the target circuit's resonant point can be obtained. Specifically, by setting the input and output resistances of the target circuit to 0, the angular frequency w0 of the target circuit's resonant point can be obtained.
[0107]
[0108] The preset efficiency η is obtained based on the voltage and current on the input side and the voltage and current on the output side of the target circuit, as well as the angular frequency. The calculation formula is as follows:
[0109]
[0110]
[0111]
[0112] Where M is the mutual inductance between the input-side inductance and the output-side inductance.
[0113] In one embodiment, such as Figure 4 As shown, this application also provides a method for obtaining electrical power transmission power, the method comprising:
[0114] Step 410: Obtain the current operating data of the target circuit, including the coil voltage, input power, output power, coil distance, receiving frequency value, and actual impedance change of the target circuit.
[0115] Step 420: Input the working data into the power model to obtain the power transmission power of the target circuit.
[0116] Specifically, in this embodiment, the power model trained in any of the above embodiments is used, along with the current operating data of the target circuit. This operating data consists of variable parameters of the circuit, specifically including the coil voltage, input power, output power, coil distance, receiving frequency, and actual impedance change of the target circuit. This allows the current power transmission power of the target circuit to be obtained. The target circuit can be as follows: Figure 1 or Figure 2 As shown.
[0117] This embodiment can predict the power transmission power of the target circuit in real time based on the current operating data of the target circuit and the trained power model. Simultaneously, the collected operating data represents the variable parameters of the target circuit, allowing for prediction based on actual conditions, thus enabling accurate prediction of the circuit's power transmission power.
[0118] In one embodiment, the method further includes:
[0119] When a change in the load of the target circuit is detected, the operating frequency of the target circuit is adjusted so that the power transmission efficiency of the target circuit reaches a preset efficiency.
[0120] Specifically, the MCU can also detect changes in the load of the target circuit in real time or at regular intervals (through methods such as...). Figure 1 The S3 interface shown detects load changes. If the operating data of the target circuit changes, the MCU adjusts the operating frequency of the target circuit to restore its power transmission capacity. Specifically, for example... Figure 1 As shown, the MCU can be adjusted through the S1 and S2 interfaces to change the operating frequency of the target circuit.
[0121] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0122] Based on the same inventive concept, this application also provides an energy power model training device for implementing the energy power model training method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more energy power model training device embodiments provided below can be found in the limitations of the energy power model training method described above, and will not be repeated here.
[0123] In one embodiment, such as Figure 5 As shown, an electrical power model training device is provided, comprising:
[0124] The first acquisition module 510 is used to acquire a first training sample and a maximum number of iterations, wherein the first training sample includes various types of sample data of the target circuit;
[0125] The initialization module 520 is used to select at least one type of sample data from the first training sample to initialize the learner, thereby obtaining multiple initialized learners. The types of sample data corresponding to the multiple initialized learners cover all types in the training sample.
[0126] The first calculation module 530 is used to calculate the negative gradient of each initialized learner;
[0127] The fitting module 540 is used to fit based on the negative gradient of each initialized learner to obtain the corresponding decision tree, and to calculate the loss function of each leaf node in the decision tree to determine the target loss function with the minimum loss.
[0128] The recursive relationship acquisition module 550 is used to obtain the recursive relationship of sample data at adjacent times in each type based on the same type of sample data at adjacent times in the training samples and the target loss function.
[0129] The first model acquisition module 560 is used to perform weighted summation based on the recursive relationship corresponding to each type to obtain the power model of the target circuit.
[0130] In one embodiment, the power model training device further includes:
[0131] The first acquisition module (not shown in the figure) is used to acquire the current working data of the target circuit and input the working data into the power model to obtain the first power transmission power;
[0132] The first determining module (not shown) is used to determine the first deviation between the preset efficiency and the first power transmission power;
[0133] The second model acquisition module 560 is used to adjust the weights of each recursive relation to obtain an updated power model if the first deviation exceeds a preset deviation range.
[0134] In one embodiment, the power model training device further includes:
[0135] The second obtaining module (not shown) is used to input the updated power model with the current working data of the target circuit to obtain the second power transmission power;
[0136] The second determining module (not shown) is used to determine a second deviation between the preset efficiency and the second power transmission power;
[0137] The efficiency power model training device is further configured to, if the second deviation exceeds a preset deviation range, obtain a second training sample again and update the power model according to the second training sample, wherein the number of sample data in the second training sample is greater than the number of sample data in the first training sample.
[0138] In one embodiment, the power model training device further includes:
[0139] The third acquisition module (not shown in the figure) is used to acquire the voltage, current, capacitance and inductance values of the input side of the target circuit, as well as the voltage, current, capacitance and inductance values of the output side, and to construct the node equations of the target circuit based on the capacitance and inductance values of the input side and the output side of the target circuit.
[0140] An angular frequency acquisition module (not shown in the figure) is used to obtain the angular frequency of the target circuit resonant point according to the nodal equation.
[0141] An efficiency calculation module (not shown) is used to obtain the preset efficiency based on the voltage and current on the input side of the target circuit, the voltage and current on the output side, and the angular frequency.
[0142] In one embodiment, the fitting module 540 is further configured to discretize the continuous features in the decision tree.
[0143] Each module in the aforementioned power model training device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.
[0144] Based on the same inventive concept, this application also provides an efficiency prediction apparatus for implementing the efficiency prediction method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more efficiency prediction apparatus embodiments provided below can be found in the limitations of the efficiency prediction method described above, and will not be repeated here.
[0145] In one embodiment, such as Figure 6 As shown, an electrical power transmission and power acquisition device is provided, comprising:
[0146] The working data acquisition module 610 is used to acquire the current working data of the target circuit. The working data includes the coil voltage, input power, output power, coil distance, receiving side frequency value, and actual impedance change of the target circuit.
[0147] The prediction module 620 is used to input the working data into the power model to obtain the power transmission power of the target circuit, wherein the power model is obtained by the efficiency prediction method described in any of the above.
[0148] In one embodiment, the power transmission and acquisition device further includes:
[0149] An adjustment module (not shown) is used to adjust the operating frequency of the target circuit when a change in the load of the target circuit is detected, so that the power transmission power of the target circuit reaches the power transmission power.
[0150] Each module in the aforementioned power transmission and acquisition device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0151] In one embodiment, a computer device is provided, which may be a server, i.e., a computer device can be used to replace, for example, a server. Figure 1 The MCU described above. Its internal structure diagram can be seen as follows. Figure 7As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores operating data of the target circuit, first training samples, second training samples, etc. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements an electrical power model training method or an electrical power transmission power acquisition method.
[0152] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0153] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0154] Obtain the first training sample and the maximum number of iterations. The first training sample includes sample data of various variable parameters of the target circuit.
[0155] Select at least one sample data from the first training sample to initialize the learner, and obtain multiple initialized learners. The types of sample data corresponding to the multiple initialized learners cover all types in the training sample.
[0156] Calculate the negative gradient of each initialized learner;
[0157] The decision tree is obtained by fitting the negative gradient of each initialized learner, and the loss function of each leaf node in the decision tree is calculated to determine the target loss function with the minimum loss.
[0158] Based on the same type of sample data at adjacent time points in the first training sample and the target loss function, the recursive relationship of sample data at adjacent time points in each type is obtained;
[0159] The power model of the target circuit is obtained by weighted summation based on the recursive relationships corresponding to each type.
[0160] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0161] Collect the current operating data of the target circuit and input the operating data into the power model to obtain the first power transmission power;
[0162] Determine the first deviation between the preset efficiency and the first power transmission power;
[0163] If the first deviation exceeds the preset deviation range, adjust the weights of each recursive relation to obtain the updated power model.
[0164] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0165] The updated power model is input using the current operating data of the target circuit to obtain the second power transmission power.
[0166] Determine the second deviation between the preset efficiency and the second power transmission power;
[0167] If the second deviation exceeds the preset deviation range, a second training sample is obtained again, and the power model is updated according to the second training sample, wherein the number of sample data in the second training sample is greater than the number of sample data in the first training sample.
[0168] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0169] Obtain the voltage, current, capacitance, and inductance values on the input side and the output side of the target circuit, and construct the node equations of the target circuit based on the capacitance and inductance values on the input side and the output side of the target circuit.
[0170] The angular frequency of the target circuit resonant point is obtained based on the nodal equation.
[0171] The preset efficiency is obtained based on the voltage and current on the input side and the voltage and current on the output side of the target circuit, as well as the angular frequency.
[0172] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0173] Discretize the continuous features in the decision tree.
[0174] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0175] Acquire the current operating data of the target circuit, including the coil voltage, input power, output power, coil distance, receiving frequency value, and actual impedance change of the target circuit;
[0176] The working data is input into the power model to obtain the power transmission power of the target circuit. The power model is obtained by any of the methods described above.
[0177] In one embodiment, the processor, when executing a computer program, also performs the following steps:
[0178] When a change in the load of the target circuit is detected, the operating frequency of the target circuit is adjusted so that the power transmission power of the target circuit reaches the power transmission power.
[0179] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the efficiency prediction method and the efficiency power model training method as described above.
[0180] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the power model training method and the power transmission acquisition method as described in any one of the above embodiments.
[0181] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0182] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0183] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for training an electrical power model, characterized in that, The method includes: Obtain the first training sample and the maximum number of iterations. The first training sample includes sample data of various variable parameters of the target circuit. Select at least one sample data from the first training sample to initialize the learner, and obtain multiple initialized learners. The types of sample data corresponding to the multiple initialized learners cover all types in the training sample. Calculate the negative gradient of each initialized learner; The decision tree is obtained by fitting the negative gradient of each initialized learner, and the loss function of each leaf node in the decision tree is calculated to determine the target loss function with the minimum loss. Based on the same type of sample data at adjacent time points in the first training sample and the target loss function, the recursive relationship of sample data at adjacent time points in each type is obtained; The power model of the target circuit is obtained by weighted summation based on the recursive relationship corresponding to each type. Collect the current operating data of the target circuit and input the operating data into the power model to obtain the first power transmission power; Obtain the voltage, current, capacitance, and inductance values on the input side and the output side of the target circuit, and construct the node equations of the target circuit based on the capacitance and inductance values on the input side and the output side of the target circuit. The angular frequency of the target circuit resonant point is obtained based on the nodal equation. A preset efficiency is obtained based on the voltage and current on the input side, the voltage and current on the output side, and the angular frequency of the target circuit. Determine the first deviation between the preset efficiency and the first power transmission power; If the first deviation exceeds the preset deviation range, adjust the weights of each recursive relation to obtain the updated power model.
2. The method according to claim 1, characterized in that, If the first deviation exceeds a preset deviation range, the weights of each recursive relation are adjusted to obtain the updated power model, and the process further includes: The updated power model is input using the current operating data of the target circuit to obtain the second power transmission power. Determine the second deviation between the preset efficiency and the second power transmission power; If the second deviation exceeds the preset deviation range, a second training sample is obtained again, and the power model is updated according to the second training sample, wherein the number of sample data in the second training sample is greater than the number of sample data in the first training sample.
3. The method according to claim 1, characterized in that, The process involves fitting the negative gradient of each initialized learner to obtain a corresponding decision tree, calculating the loss function for each leaf node in the decision tree, and determining the target loss function that minimizes the loss. This includes: Discretize the continuous features in the decision tree.
4. A method for obtaining electrical energy transmission power, characterized in that, The method includes: Acquire the current operating data of the target circuit, including the coil voltage, input power, output power, coil distance, receiving frequency value, and actual impedance change of the target circuit; The working data is input into the power model to obtain the power transmission power of the target circuit, wherein the power model is obtained by the method described in any one of claims 1-3.
5. The method according to claim 4, characterized in that, The method further includes: When a change in the load of the target circuit is detected, the operating frequency of the target circuit is adjusted so that the power transmission efficiency of the target circuit reaches a preset efficiency.
6. An electrical power model training device, characterized in that, The device includes: The first acquisition module is used to acquire the first training sample and the maximum number of iterations, wherein the first training sample includes various types of sample data of the target circuit; An initialization module is used to select at least one type of sample data from the first training sample to initialize the learner, thereby obtaining multiple initialized learners. The types of sample data corresponding to the multiple initialized learners cover all types in the training sample. The first computation module is used to calculate the negative gradient of each initialized learner; The fitting module is used to fit the negative gradient of each initialized learner to obtain the corresponding decision tree, and to calculate the loss function of each leaf node in the decision tree to determine the target loss function with the minimum loss. The recursive relationship acquisition module is used to obtain the recursive relationship of sample data at adjacent time points in each type based on the same type of sample data at adjacent time points in the training samples and the target loss function. The first model acquisition module is used to perform weighted summation based on the recursive relationship corresponding to each type to obtain the power model of the target circuit. The first acquisition module is used to acquire the current working data of the target circuit and input the working data into the power model to obtain the first power transmission power. The third acquisition module is used to acquire the voltage, current, capacitance and inductance values of the input side of the target circuit, as well as the voltage, current, capacitance and inductance values of the output side, and to construct the node equations of the target circuit based on the capacitance and inductance values of the input side and the output side of the target circuit. An angular frequency acquisition module is used to obtain the angular frequency of the target circuit resonant point according to the nodal equation. An efficiency calculation module is used to obtain a preset efficiency based on the voltage and current on the input side of the target circuit, the voltage and current on the output side, and the angular frequency. The first determining module is used to determine a first deviation between the preset efficiency and the first power transmission power. The second model acquisition module is used to adjust the weights of each recursive relation to obtain an updated power model if the first deviation exceeds a preset deviation range.
7. A power acquisition device for electrical energy transmission, characterized in that, The device includes: The working data acquisition module is used to acquire the current working data of the target circuit. The working data includes the coil voltage, input power, output power, coil distance, receiving side frequency value, and actual impedance change of the target circuit. The prediction module is used to input the working data into the power model to obtain the power transmission power of the target circuit, wherein the power model is obtained by the method described in any one of claims 1-3.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.