A life prediction method for pulse power supply system
By performing charging and discharging experiments of pulse power supplies and multi-neural network model training at different ambient temperatures, combined with the weight optimization of the IOWA operator, the problem of conventional IOWA operators ignoring the life change law, and improving the accuracy of pulse power supply life prediction and model reliability.
Patent Information
- Application Number
- CN202411322200.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-23
- Publication Date
- 2025-01-03
- Estimated Expiration
- 2044-09-23
AI Technical Summary
When predicting the remaining service life of the pulse power supply, the conventional IOWA operator only uses the magnitude relationship between the predicted value and the actual value, and ignores the change law of the pulse power supply life, resulting in poor prediction accuracy.
The charging and discharging experiment of pulsed power supply samples was carried out under different ambient temperature ranges. Through training of multiple neural network models and weight optimization of IOWA operators, combining the charging and discharging cycle times and parameter sequence, the prediction accuracy and consistency of each lifetime prediction value were calculated. to obtain more accurate life expectancy.
It improves the prediction accuracy of the remaining service life of the pulse power supply, enhances the reliability and comprehensiveness of the prediction model, and is suitable for practical application environments.
Smart Images
Figure CN118861604B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of life prediction using a neural network model, and more specifically, to a method for predicting the life of a pulse power supply system. Background Art
[0002] Pulse power systems are often used in demanding applications such as military, aerospace, medical equipment, etc. These systems require long-term stable performance, and life prediction helps evaluate whether the pulse power system can maintain stable performance during use. It can also help take appropriate measures to extend the life cycle of the equipment, thereby ensuring the reliability and effectiveness of the system.
[0003] In the related technology, for example, the Chinese patent document with the authorization announcement number CN113361692B discloses a combined prediction method for lithium battery life, including: based on the collected lithium battery capacity data, the constructed combined model is solved by an optimization algorithm to construct a new prediction method, the difference between each prediction is recorded in each prediction process, and the error influence is added when predicting the next cycle. The error of the previous prediction value is iterated to make the prediction algorithm more accurate after the combination, and the standard deviation is used as the error judgment to calculate the total error of the training set to obtain the optimal weight value. The above technical solution has greatly improved the accuracy of lithium battery life prediction.
[0004] In the related art, it is difficult to accurately predict the life of a battery by training a single neural network; multiple neural networks can be trained to predict the life of a pulse power supply. When the same data is input, multiple neural networks will output multiple prediction values. According to the error between the multiple prediction values and the true values, the contribution of different prediction values is quantified using the induced ordered weighted average operator (IOWA operator), and the prediction result of the remaining service life of the pulse power supply is obtained by combining multiple prediction values; while the conventional IOWA operator only uses the size relationship between the prediction value and the actual value, and ignores the change law of the remaining service life of the pulse power supply. Therefore, the prediction accuracy of the remaining service life of the pulse power supply predicted by the conventional IOWA operator is poor. Summary of the invention
[0005] In order to solve the technical problem that the conventional IOWA operator only utilizes the size relationship between the predicted value and the actual value and ignores the change law of the remaining service life of the pulse power supply, resulting in poor prediction accuracy of the remaining service life of the pulse power supply through the conventional IOWA operator, the present invention provides solutions in the following aspects.
[0006] The present invention provides a life prediction method for a pulse power supply system, comprising: performing a charge and discharge experiment on a pulse power supply sample under different ambient temperature ranges; calculating the actual life of the pulse power supply sample in each charge and discharge cycle according to the ratio of the number of each charge and discharge cycle in the charge and discharge experiment to the total number of all charge and discharge cycles; sampling at equal intervals on each parameter curve of each charge and discharge cycle to obtain each parameter sequence of each charge and discharge cycle; training multiple neural network models as life prediction models according to the parameter sequence of each charge and discharge cycle and the actual life; obtaining multiple life prediction values of the pulse power supply to be tested in the current charge and discharge cycle through the life prediction model according to each parameter sequence of the pulse power supply to be tested in the current charge and discharge cycle; and The target arrangement mode among different arrangements of the life prediction values is used, and the weighted summation of the life prediction values in the target arrangement mode is performed by the IOWA operator to obtain the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement mode; the prediction accuracy of each life prediction value is calculated according to the size relationship between each life prediction value and the life of the pulse power supply to be tested in the previous charge and discharge cycle of the current charge and discharge cycle and the difference between each life prediction value and the life estimation value; all life prediction values are sorted in order from large to small in terms of prediction accuracy to obtain a sequence sequence; the consistency of the sequence sequence and the target arrangement mode is calculated; and the life estimation value under the arrangement mode with the greatest consistency is used as the life of the pulse power supply to be tested in the current charge and discharge cycle.
[0007] The present invention combines the law that the remaining service life of the pulse power supply decreases with the increase of the number of charge and discharge cycles, calculates the prediction accuracy of each life prediction value according to the size relationship of the life of the pulse power supply to be tested in the previous charge and discharge cycle of the current charge and discharge cycle and the difference between each life prediction value and the life estimation value, thereby enhancing the prediction accuracy of the remaining service life of the pulse power supply;
[0008] Furthermore, the present invention assumes that the target arrangement is the result of sorting all the life prediction values in order of prediction accuracy from large to small, so as to obtain the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement, calculate the prediction accuracy of each life prediction value according to the difference between each life prediction value and the life estimation value, and sort the life prediction values in order of prediction accuracy from large to small to obtain a sequence sequence, and express the possibility that the target arrangement is the result of sorting all the life prediction values in order of prediction accuracy from large to small through the consistency between the sequence sequence and the target arrangement, thereby obtaining the life of the pulse power supply to be tested in the current charge and discharge cycle, thereby enhancing the prediction accuracy of the remaining service life of the pulse power supply.
[0009] Preferably, the charging and discharging experiment of the pulse power supply sample under different ambient temperature ranges includes: setting Different ambient temperature ranges, , Indicates the minimum ambient temperature, Indicates the maximum ambient temperature, represents the fourth preset value, Indicates rounding up; The ambient temperature range is ;No. Plant to The ambient temperature range is ,in, for An integer in the range; The ambient temperature range is ; During the charge and discharge experiment of the pulse power supply sample under any ambient temperature range, the ambient temperature at each moment is randomly adjusted, and the ambient temperature at each moment is required to be within the ambient temperature range.
[0010] When conducting charge and discharge experiments on pulse power supply samples, the present invention randomly adjusts the ambient temperature at each moment within the ambient temperature range. By conducting charge and discharge experiments under different temperature conditions, the working state of the pulse power supply in the actual application environment can be more realistically simulated. At the same time, the specific impact of temperature on the life of the pulse power supply system can be better identified and quantified, making the prediction model more comprehensive and reliable, which helps to improve the performance of the pulse power supply system in practical applications.
[0011] Preferably, the calculation of the actual life of the pulse power supply sample in each charge and discharge cycle includes: for the first charge and discharge cycles, calculate the Number of charge and discharge cycles The ratio of 1 to the total number of all charge and discharge cycles; the difference between 1 and the ratio is used as the pulse power sample at the first The actual life of the charge and discharge cycle.
[0012] Preferably, the parameter sequence includes a charging environment temperature sequence, a charging temperature sequence, a charging voltage sequence, a charging power sequence, a discharging environment temperature sequence, a discharging temperature sequence, a discharging current sequence, a discharging voltage sequence and a discharging power sequence; the input when training the first life prediction model is the charging environment temperature sequence and the charging temperature sequence of the charge and discharge cycle; the input when training the second life prediction model is the charging voltage sequence and the charging power sequence of the charge and discharge cycle; the input when training the third life prediction model is the discharging environment temperature sequence and the discharging temperature sequence of the charge and discharge cycle; the input when training the fourth life prediction model is the discharging current sequence and the discharging power sequence of the charge and discharge cycle; the input when training the fifth life prediction model is the discharging voltage sequence and the discharging power sequence of the charge and discharge cycle; the output of training all life prediction models is the actual life of the charge and discharge cycle.
[0013] Different neural networks can learn different features of parameter sequences. The present invention can more comprehensively capture the complex relationships in parameter sequences by training multiple life prediction models, thereby improving the overall prediction accuracy.
[0014] Preferably, obtaining the estimated life value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement includes: calculating the product of each target weight in the IOWA operator and each life prediction value in the target arrangement, and taking the sum of all products as the estimated life value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement.
[0015] Preferably, the prediction accuracy of the life prediction value satisfies the expression: ; In the formula, Indicates the first The prediction accuracy of the life prediction value is Indicates the first Life expectancy prediction value, Indicates the life of the pulse power supply to be tested in the previous charge and discharge cycle before the current charge and discharge cycle. , , represent the first preset value, the second preset value and the third preset value respectively, represents the natural exponential function, Indicates the estimated life of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement mode, Indicates taking the absolute value, Indicates taking the minimum value.
[0016] The present invention combines the law that the remaining service life of a pulse power supply decreases with the increase in the number of charge and discharge cycles, calculates the prediction accuracy of each life prediction value according to the size relationship of the life of the pulse power supply to be tested in the previous charge and discharge cycle of the current charge and discharge cycle and the difference between each life prediction value and the life estimation value, thereby enhancing the prediction accuracy of the remaining service life of the pulse power supply.
[0017] Preferably, the consistency between the order sequence and the target arrangement satisfies the expression: ; In the formula, Indicates the consistency between the order sequence and the target arrangement. represents the total number of all life prediction models, represents the first The target weights, Indicates the first Life expectancy prediction value, Indicates the first Life expectancy prediction value, Indicates taking the absolute value.
[0018] The consistency between the sequential sequence and the target arrangement is used to indicate the possibility that the target arrangement is the result of sorting all life prediction values in order from large to small according to the prediction accuracy, thereby obtaining the life of the pulse power supply to be tested in the current charge and discharge cycle, thereby enhancing the prediction accuracy of the remaining service life of the pulse power supply.
[0019] Preferably, the target weight is the optimal solution obtained by solving the weight optimization model of the IOWA operator, and constructing the weight optimization model of the IOWA operator includes: according to the parameter sequences of the charge and discharge cycle, obtaining multiple life prediction values of the charge and discharge cycle through multiple life prediction models, and recording each life prediction value of the charge and discharge cycle as the life reference value of the charge and discharge cycle; sorting the life reference values of the charge and discharge cycle in order from large to small according to the prediction accuracy of each life reference value of the charge and discharge cycle, and performing weighted summation of the multiple life reference values of the sorted charge and discharge cycle through the IOWA operator to obtain the life estimation value of the charge and discharge cycle, wherein all the weights in the IOWA operator are unknowns that need to be solved; and constructing the weight optimization model of the IOWA operator with the objective function of minimizing the error between the life estimation value and the actual life of the charge and discharge cycle of all pulse power supply samples.
[0020] Preferably, the prediction accuracy of each life reference value of the charge and discharge cycle satisfies the expression: ; In the formula, Indicates The first charge-discharge cycle The prediction accuracy of the lifetime reference value is Indicates The first charge-discharge cycle A life reference value, Indicates The actual life of charge and discharge cycles, , , represent the first preset value, the second preset value and the third preset value respectively, represents the natural exponential function, Indicates The actual life of charge and discharge cycles, Indicates taking the absolute value, Indicates taking the minimum value.
[0021] Preferably, the weight optimization model of the IOWA operator is: ; In the formula, represents the objective function, Represents the total number of all pulse power samples, Indicates The total number of all charge and discharge cycles of the pulse power supply samples, , Respectively represent The first pulse power supply sample Estimated and actual lifespan of charge and discharge cycles, Indicates The first pulse power supply sample The error between the estimated lifespan of a charge-discharge cycle and the actual lifespan, Indicates taking the absolute value, represents the total number of all life prediction models, represents the first weights, Indicates taking the minimum value.
[0022] The beneficial effects of the present invention are:
[0023] The present invention combines the law that the remaining service life of a pulse power supply decreases with the increase of the number of charge and discharge cycles, and calculates the prediction accuracy of each life prediction value according to the size relationship of the life of the pulse power supply to be tested in the previous charge and discharge cycle of the current charge and discharge cycle and the difference between each life prediction value and the life estimation value. At the same time, it is assumed that the target arrangement method is the result of sorting all the life prediction values in order from large to small according to the prediction accuracy, so as to obtain the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement method, and calculate the prediction accuracy of each life prediction value according to the difference between each life prediction value and the life estimation value, and sort the life prediction values in order from large to small according to the prediction accuracy to obtain a sequence sequence, and the consistency of the sequence sequence with the target arrangement method is used to indicate the possibility that the target arrangement method is the result of sorting all the life prediction values in order from large to small according to the prediction accuracy, thereby obtaining the life of the pulse power supply to be tested in the current charge and discharge cycle, thereby enhancing the prediction accuracy of the remaining service life of the pulse power supply. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] The above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below through the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:
[0025] Figure 1 It is a flow chart schematically showing a method for predicting the life of a pulse power supply system in the present invention. DETAILED DESCRIPTION
[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0027] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0028] The embodiment of the present invention discloses a pulse power system life prediction method, referring to Figure 1 , comprising steps S1 to S5:
[0029] S1. Carry out charge and discharge experiments on pulse power supply samples under different ambient temperature ranges, and calculate the actual life of the pulse power supply samples in each charge and discharge cycle.
[0030] It should be noted that excessively high or low ambient temperatures will accelerate the aging process of the pulse power supply. In order to more accurately predict the remaining service life of the pulse power supply in the cycle in which it is located, the present invention conducts charging and discharging experiments on pulse power supply samples under different ambient temperature ranges.
[0031] Specifically, set Different ambient temperature ranges, , Indicates the minimum ambient temperature, Indicates the maximum ambient temperature, represents the fourth preset value, Indicates rounding up; The ambient temperature range is ;No. Plant to The ambient temperature range is ,in, for An integer in the range; The ambient temperature range is .
[0032] Among them, the specific values of the minimum ambient temperature, the maximum ambient temperature and the fourth preset value can be set according to the actual application scenario and requirements. The value range of the minimum ambient temperature is [-35, -25], the value range of the maximum ambient temperature is [45, 60], and the value range of the fourth preset value is [10, 20]. The present invention sets the minimum ambient temperature to -30 degrees, the maximum ambient temperature to 50 degrees, and the fourth preset value to 10.
[0033] Furthermore, during the charging and discharging experiment of the pulse power supply sample under any ambient temperature range, the ambient temperature at each moment is randomly adjusted, and the ambient temperature at each moment is required to be within the ambient temperature range.
[0034] Furthermore, under each ambient temperature range, charge and discharge experiments were performed on the three pulse power supply samples, and the process of the charge and discharge experiments included multiple charge and discharge cycles.
[0035] It should be noted that when the ratio of the energy of the pulse power supply sample during a charge and discharge cycle to the energy of a new pulse power supply when it is fully charged is equal to a preset threshold, it means that the life of the pulse power supply sample has ended after the charge and discharge cycle. At this time, the charge and discharge experiment of the pulse power supply sample is stopped.
[0036] The specific value of the preset threshold is usually set according to the design specifications and application requirements of the pulse power supply. The value range of the preset threshold is [0.5, 0.9]. The present invention sets the preset threshold to 0.7.
[0037] Further, the actual life of the pulse power supply sample in each charge and discharge cycle is calculated, including: charge and discharge cycles, calculate the Number of charge and discharge cycles The ratio of 1 to the total number of all charge and discharge cycles; the difference between 1 and the ratio is used as the pulse power sample at the first The actual life of the charge and discharge cycle.
[0038] It should be noted that when conducting charging and discharging experiments on pulse power supply samples, the present invention randomly adjusts the ambient temperature at each moment within the ambient temperature range. By conducting charging and discharging experiments under different temperature conditions, it can more realistically simulate the working state of the pulse power supply in the actual application environment. At the same time, it can better identify and quantify the specific impact of temperature on the life of the pulse power supply system, making the prediction model more comprehensive and reliable, which helps to improve the performance of the pulse power supply system in actual applications.
[0039] S2. Collecting parameter curves of each charge-discharge cycle, sampling at equal intervals on the parameter curves of each charge-discharge cycle, and obtaining parameter sequences of each charge-discharge cycle.
[0040] Specifically, any pulse power supply sample is taken as an example for explanation below, and each charge and discharge cycle includes a charging process and a discharging process.
[0041] Among them, during the charging process of the charge and discharge cycle, the charging current is equal to the constant current. The charging capacity of the pulse power supply sample at each moment of the charge and discharge cycle is calculated according to the constant current through the ampere-hour integration method, and the charging capacity at all moments is fitted to obtain the charging capacity curve of the charge and discharge cycle; the charging temperature of the pulse power supply sample at each moment of the charge and discharge cycle is collected through a thermistor, and the charging temperature at all moments is fitted to obtain the charging temperature curve of the charge and discharge cycle; the charging voltage of the pulse power supply sample at each moment of the charge and discharge cycle is collected through a high-voltage differential probe, and the charging voltage at all moments is fitted to obtain the charging voltage curve of the charge and discharge cycle; the charging environment temperature at all moments of the charge and discharge cycle is fitted to obtain the charging environment temperature curve of the charge and discharge cycle.
[0042] Among them, the ampere-hour integration method is a method for calculating electric quantity, which estimates the electric quantity by calculating the integral of the charging current and the corresponding time or the integral of the discharging current and the corresponding time within a certain period of time. The ampere-hour integration method is a well-known technology and will not be described here in detail.
[0043] Furthermore, during the discharge process of the charge and discharge cycle, the discharge current of the pulse power supply sample at each moment of the charge and discharge cycle is collected by a Rogowski pulse current transformer, and the discharge current at all moments is fitted to obtain the discharge current curve of the charge and discharge cycle; the discharge capacity of the pulse power supply sample at each moment of the charge and discharge cycle is calculated according to the discharge current curve by the ampere-hour integration method, and the discharge capacity at all moments is fitted to obtain the discharge capacity curve of the charge and discharge cycle; the discharge temperature of the pulse power supply sample at each moment of the charge and discharge cycle is collected by a thermistor, and the discharge temperature at all moments is fitted to obtain the discharge temperature curve of the charge and discharge cycle; the discharge voltage of the pulse power supply sample at each moment of the charge and discharge cycle is collected by a high-voltage differential probe, and the discharge voltage at all moments is fitted to obtain the discharge voltage curve of the charge and discharge cycle; the discharge environment temperature at all moments of the charge and discharge cycle is fitted to obtain the discharge environment temperature curve of the charge and discharge cycle.
[0044] Therefore, the parameter curves of the charge and discharge cycle include a charge environment temperature curve, a charge temperature curve, a charge voltage curve, a charge capacity curve, a discharge environment temperature curve, a discharge temperature curve, a discharge current curve, a discharge voltage curve and a discharge capacity curve.
[0045] Among them, the thermistor determines the temperature by measuring the change of resistance, the Rogowski current transformer is a differential current transformer, and the high-voltage differential probe is a high-voltage measurement method that is both safe and accurate. The use of the high-voltage differential probe and the Rogowski pulse current transformer can effectively isolate the electrical connection between the high and low voltage systems to achieve safe measurement; the thermistor, the high-voltage differential probe and the Rogowski pulse current transformer are all installed inside the pulse power supply sample.
[0046] It should be noted that since the duration of each charge and discharge cycle is different, in order to ensure the same input to the neural network, it is necessary to sample each parameter curve of each charge and discharge cycle equidistantly and the total number of all parameter data obtained by sampling is the same, so as to ensure that the sequence composed of the collected parameter data can be used as the input of the neural network later.
[0047] Specifically, for any parameter curve of the charge-discharge cycle, when sampling is performed equidistantly on the parameter curve, the total number of all parameter data obtained by sampling is equal to the preset first number; the sequence composed of all the obtained parameter data is used as the parameter sequence of the charge-discharge cycle; wherein the specific value of the preset first number can be set according to the actual application scenario and requirements, and the value range of the preset first number is [100,500]. The present invention sets the preset first number to 300. Therefore, the parameter sequence of each charge-discharge cycle includes the charging environment temperature sequence, charging temperature sequence, charging voltage sequence, charging power sequence, discharging environment temperature sequence, discharging temperature sequence, discharging current sequence, discharging voltage sequence and discharging power sequence of the charge-discharge cycle.
[0048] It should be noted that predicting the remaining service life of a pulse power supply in its cycle requires comprehensive consideration of multiple factors, which have a direct impact on the health and life of the pulse power supply: the level of the charging voltage directly affects the charging efficiency and life of the pulse power supply, the size of the discharge current affects the heat generation and internal resistance of the pulse power supply, and the change in the discharge voltage reflects the discharge state and performance of the pulse power supply.
[0049] S3. Training multiple neural network models according to the parameter sequences of each charge and discharge cycle and the actual lifespan as life prediction models.
[0050] It should be noted that the charging environment temperature, charging temperature, charging voltage, charging capacity, discharging environment temperature, discharging temperature, discharging current, discharging voltage and discharging capacity are all important factors in the performance and life of the pulse power supply. By comprehensively considering these factors, an accurate model can be established to predict the remaining service life of the pulse power supply, which not only helps to optimize the use and maintenance of the pulse power supply, but also improves the reliability and safety of the pulse power supply system.
[0051] In this embodiment, the neural network model is a LSTM (Long short-term memory) neural network; in other embodiments, other neural networks that can be used to process data sequences can be selected, such as a recurrent neural network RNN or a BP neural network.
[0052] It should be noted that the LSTM neural network is mainly used to solve the problems of gradient disappearance and gradient explosion in the training process of long data sequences. It controls the transmission state through the gating state, remembers the information that needs to be remembered for a long time, and forgets unimportant information. Therefore, the LSTM neural network has better performance in long data sequences. Therefore, the present invention chooses to use the LSTM neural network to train the life prediction model of the pulse power supply.
[0053] In this example, the hidden layer size (hidden_size) of the LSTM neural network is set to , Indicates the preset first quantity, Indicates rounding up. The hidden layer size refers to the number of neurons in the hidden layer of the LSTM neural network. Set the number of LSTM layers (num_layers) of the LSTM neural network to 1.
[0054] Specifically, for the three pulse power supply samples subjected to charge and discharge experiments in each ambient temperature range, the parameter sequences and actual lifespans of all charge and discharge cycles of two of the pulse power supply samples are used as training sets for training the neural network, and the parameter sequences and actual lifespans of all charge and discharge cycles of the remaining one pulse power supply sample are used as a test set for testing the neural network.
[0055] It should be noted that a single neural network is easily affected by noise and outliers in the training set, resulting in overfitting; different neural networks can learn different characteristics of parameter sequences. The present invention can more comprehensively capture the complex relationships in parameter sequences by training multiple life prediction models, thereby improving the overall prediction accuracy.
[0056] Further, the multiple life prediction models include a first life prediction model, a second life prediction model, a third life prediction model, a fourth life prediction model and a fifth life prediction model.
[0057] The input for training the first life prediction model is the charging environment temperature sequence and charging temperature sequence of each charge and discharge cycle, and the output is the actual life of each charge and discharge cycle. Therefore, in the first life prediction model, the input vector is 2-dimensional and has a size of , Indicates the preset first quantity, the output vector is 1-dimensional and the size is The input of the second life prediction model is the charging voltage sequence and charging power sequence of each charge and discharge cycle, and the output is the actual life of each charge and discharge cycle. Therefore, in the second life prediction model, the input vector is 2-dimensional and has a size of , Indicates the preset first quantity, the output vector is 1-dimensional and the size is ; The input for training the third life prediction model is the discharge environment temperature sequence and discharge temperature sequence of each charge and discharge cycle, and the output is the actual life of each charge and discharge cycle. Therefore, in the third life prediction model, the input vector is 2-dimensional and has a size of , Indicates the preset first quantity, the output vector is 1-dimensional and the size is ; The input for training the fourth life prediction model is the discharge current sequence and discharge power sequence of each charge and discharge cycle, and the output is the actual life of each charge and discharge cycle. Therefore, in the fourth life prediction model, the input vector is 2-dimensional and has a size of , Indicates the preset first quantity, the output vector is 1-dimensional and the size is ; The input for training the fifth life prediction model is the discharge voltage sequence and discharge power sequence of each charge and discharge cycle, and the output is the actual life of each charge and discharge cycle. Therefore, in the fifth life prediction model, the input vector is 2-dimensional and has a size of , Indicates the preset first quantity, the output vector is 1-dimensional and the size is .
[0058] S4. Solve the weight optimization model of the IOWA operator to obtain the target weight in the IOWA operator.
[0059] It should be noted that in order to enable the IOWA operator to provide the optimal weighted average result in the decision-making problem of obtaining the estimated life value of the pulse power supply to be tested in the charge and discharge cycle, the present invention obtains the target weight in the IOWA operator by constructing and solving the weight optimization model of the IOWA operator, and meets the specific needs of the present invention by reasonably adjusting the weight.
[0060] Specifically, according to each parameter sequence of each charge and discharge cycle, multiple life prediction values of each charge and discharge cycle are obtained through multiple life prediction models, and each life prediction value of the charge and discharge cycle is recorded as the life reference value of the charge and discharge cycle. The total number of all life reference values of the charge and discharge cycle is equal to , Represents the total number of all life prediction models.
[0061] It should be noted that the prediction accuracy of the predicted value calculated by the conventional IOWA operator through the predicted value and the actual value only utilizes the relationship between the predicted value and the actual value, and does not combine the law of the remaining service life of the pulse power supply. Therefore, the prediction accuracy of the predicted value of each charge and discharge cycle of the pulse power supply sample calculated by the conventional IOWA operator is inaccurate.
[0062] It should be further explained that, considering that the remaining service life of the pulse power supply is constantly decreasing as the number of charge and discharge cycles increases, the present invention calculates the prediction accuracy of the life reference value based on the relationship between the life reference value of the charge and discharge cycle and the actual life of the previous charge and discharge cycle, as well as the difference between the life reference value of the charge and discharge cycle and the actual life of the charge and discharge cycle.
[0063] Further, according to the size relationship between the life reference value of the charge-discharge cycle and the actual life of the previous charge-discharge cycle and the difference between the life reference value of the charge-discharge cycle and the actual life of the charge-discharge cycle, the prediction accuracy of the life reference value is calculated, and the prediction accuracy of each life reference value of the charge-discharge cycle satisfies the expression:
[0064] ;
[0065] In the formula, Indicates The first charge-discharge cycle The prediction accuracy of the lifetime reference value is Indicates The first charge-discharge cycle A life reference value, Indicates The actual life of charge and discharge cycles, , , represent the first preset value, the second preset value and the third preset value respectively, represents the natural exponential function, Indicates The actual life of charge and discharge cycles, Indicates taking the absolute value, Indicates taking the minimum value.
[0066] Among them, the specific values of the first preset value, the second preset value and the third preset value can be set according to the actual application scenario and requirements. Among them, the value range of the first preset value is [0.01, 0.1], the value range of the second preset value is [20, 30], and the value range of the third preset value is [0.55, 0.9]. The present invention sets the first preset value to 0.05, the second preset value to 25, and the third preset value to 0.7.
[0067] It should be noted that when the life reference value of the charge and discharge cycle Greater than the actual life of the previous charge and discharge cycle , and the reference value of the life of the charge-discharge cycle is different from the actual life of the previous charge-discharge cycle Greater than or equal to the first preset value , indicating that the life reference value obtained at this time completely violates the characteristic that the remaining service life of the pulse power supply decreases with the increase of the number of charge and discharge cycles. At this time, the prediction accuracy of each life reference value of the charge and discharge cycle is ; When the life reference value of charge and discharge cycle Greater than the actual life of the previous charge and discharge cycle , and the reference value of the life of the charge-discharge cycle is different from the actual life of the previous charge-discharge cycle Less than the first preset value , indicating that the lifetime reference value obtained at this time violates the characteristic that the remaining lifetime of the pulse power supply decreases with the increase of the number of charge and discharge cycles, but the degree is small. At this time, the prediction accuracy of each lifetime reference value of the charge and discharge cycle is ,in, ,therefore, ,but ; When the life reference value of charge and discharge cycle Less than the actual life of the previous charge and discharge cycle , indicating that the life reference value obtained at this time fully meets the characteristics that the remaining service life of the pulse power supply decreases with the increase of the number of charge and discharge cycles. At this time, the prediction accuracy of each life reference value of the charge and discharge cycle is ,in, ,but .
[0068] Furthermore, the life reference values of the charge and discharge cycles are sorted in descending order according to the prediction accuracy of each life reference value of the charge and discharge cycle, and the weighted sum of multiple life reference values of each charge and discharge cycle is performed by the IOWA operator to obtain the life estimation value of each charge and discharge cycle. Then, the life estimation value of the charge and discharge cycle satisfies the expression:
[0069] ;
[0070] In the formula, Indicates Estimated lifetime of charge and discharge cycles, represents the total number of all life prediction models, represents the first weights, Indicates the sorted The first charge-discharge cycle A life reference value.
[0071] Furthermore, the objective function is to minimize the error between the estimated life and the actual life of all pulse power supply samples in charge and discharge cycles, and to construct a weighted optimization model of the IOWA operator, in which all the values in the IOWA operator are unknowns that need to be solved, specifically:
[0072] ;
[0073] In the formula, represents the objective function, Represents the total number of all pulse power samples, Indicates The total number of all charge and discharge cycles of the pulse power supply samples, Indicates The first pulse power supply sample Estimated lifetime of charge and discharge cycles, Indicates The first pulse power supply sample The actual life of charge and discharge cycles, Indicates The first pulse power supply sample The error between the estimated lifespan of a charge-discharge cycle and the actual lifespan, Indicates taking the absolute value, represents the total number of all life prediction models, represents the first weights, Indicates taking the minimum value.
[0074] Furthermore, the weight optimization model of the IOWA operator is solved, and the optimal solution obtained is used as the target weight in the IOWA operator.
[0075] S5. Calculate the consistency between the order sequence and each arrangement mode, and use the estimated lifespan of the pulse power supply to be tested in the current charge-discharge cycle under the arrangement mode with the greatest consistency as the lifespan of the pulse power supply to be tested in the current charge-discharge cycle.
[0076] Specifically, through the method of step S2, the parameter curves of the pulse power supply to be tested in the current charge and discharge cycle are collected, and sampling is performed at equal intervals in the current charge and discharge cycle to obtain the parameter sequences of the pulse power supply to be tested in the current charge and discharge cycle.
[0077] Further, according to the parameter sequences of the pulse power supply to be tested in the current charge and discharge cycle, multiple life prediction values of the pulse power supply to be tested in the current charge and discharge cycle are obtained through multiple life prediction models, and the total number of all life prediction values is equal to , Represents the total number of all life prediction models.
[0078] It should be noted that since the purpose of the present invention is to obtain the life of the pulse power supply to be tested in the current charge and discharge cycle, that is to say, there is no actual life of the pulse power supply to be tested in the current charge and discharge cycle, it is impossible to calculate the prediction accuracy of the predicted value based on the difference between the predicted value and the actual life, and further, it is impossible to obtain the life of the pulse power supply to be tested in the current charge and discharge cycle by weighted summing the sorted predicted values through the target weights in the IOWA operator.
[0079] It should be further explained that the present invention uses any arrangement of all life prediction values as the target arrangement, assuming that the target arrangement is the result of sorting all life prediction values in order of prediction accuracy from large to small, and then weighted summing the life prediction values in the target arrangement by the IOWA operator to obtain the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement, and according to the size relationship between each life prediction value and the life of the pulse power supply to be tested in the previous charge and discharge cycle of the current charge and discharge cycle and the difference between each life prediction value and the life estimation value, the prediction accuracy of each life prediction value is calculated, and all life prediction values are sorted in order of prediction accuracy from large to small to obtain a sequence sequence. When the consistency between the sequence sequence and the target arrangement is the largest, it means that the target arrangement is the result of sorting all life prediction values in order of prediction accuracy from large to small, and then it means that the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement is the life of the pulse power supply to be tested in the current charge and discharge cycle finally obtained.
[0080] Furthermore, all life expectancy prediction values are arranged, and there are Different arrangements, represents factorial, Represents the total number of all life prediction models; any one arrangement is used as the target arrangement, and the life prediction values in the target arrangement are weighted and summed by the IOWA operator to obtain the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement. The life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement satisfies the expression:
[0081] ;
[0082] In the formula, Indicates the estimated life of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement mode, represents the total number of all life prediction models, represents the first The target weights, Indicates the first Life expectancy prediction.
[0083] Further, according to the size relationship between each life prediction value and the life of the pulse power supply to be tested in the previous charge and discharge cycle of the current charge and discharge cycle and the difference between each life prediction value and the life estimation value, the prediction accuracy of each life prediction value is calculated, and the prediction accuracy of the life prediction value satisfies the expression:
[0084] ;
[0085] In the formula, Indicates the first The prediction accuracy of the life prediction value is Indicates the first Life expectancy prediction value, Indicates the life of the pulse power supply to be tested in the previous charge and discharge cycle before the current charge and discharge cycle. , , represent the first preset value, the second preset value and the third preset value respectively, represents the natural exponential function, Indicates the estimated life of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement mode, Indicates taking the absolute value, Indicates taking the minimum value.
[0086] It should be noted that the present invention combines the law that the remaining service life of the pulse power supply decreases with the increase of the number of charge and discharge cycles, calculates the prediction accuracy of each life prediction value according to the size relationship of the life of the pulse power supply to be tested in the previous charge and discharge cycle of the current charge and discharge cycle and the difference between each life prediction value and the life estimation value, thereby enhancing the prediction accuracy of the remaining service life of the pulse power supply.
[0087] Further, all life prediction values are sorted in descending order of prediction accuracy to obtain a sequence; the consistency between the sequence and the target arrangement is calculated, and the consistency between the sequence and the target arrangement satisfies the expression:
[0088] ;
[0089] In the formula, Indicates the consistency between the order sequence and the target arrangement. represents the total number of all life prediction models, represents the first The target weights, Indicates the first Life expectancy prediction value, Indicates the first Life expectancy prediction value, Indicates taking the absolute value.
[0090] It should be noted that, assuming that the target arrangement is the result of sorting all life prediction values in order from large to small according to prediction accuracy, the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement is obtained, and the prediction accuracy of each life prediction value is calculated according to the difference between each life prediction value and the life estimation value, and the life prediction values are sorted in order from large to small according to prediction accuracy to obtain a sequential sequence; the greater the consistency between the sequential sequence and the target arrangement, the more likely the target arrangement is to be the result of sorting all life prediction values in order from large to small according to prediction accuracy, and further, the more likely the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement is to be the finally obtained life of the pulse power supply to be tested in the current charge and discharge cycle.
[0091] Furthermore, the estimated lifespan of the pulse power supply to be tested in the current charge-discharge cycle under the arrangement with the greatest consistency is used as the lifespan of the pulse power supply to be tested in the current charge-discharge cycle.
[0092] Life prediction of the pulse power supply to be tested can help evaluate whether the pulse power supply to be tested can maintain stable performance during use, and can also help take appropriate measures to extend the service life of the pulse power supply to be tested and ensure the reliability and effectiveness of the pulse power supply to be tested.
[0093] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.
[0094] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, alternatives to the embodiments of the present invention described herein may be employed.
Claims
1. A pulse power system life prediction method, characterized in that: include: Carry out charge and discharge experiments on pulse power samples under different ambient temperature ranges; calculate the actual life of the pulse power sample in each charge and discharge cycle according to the ratio of the number of each charge and discharge cycle in the charge and discharge experiment to the total number of all charge and discharge cycles; equally sample on each parameter curve of each charge and discharge cycle to obtain each parameter sequence of each charge and discharge cycle; train multiple neural network models according to the parameter sequence of each charge and discharge cycle and the actual life to obtain multiple life prediction models; According to the parameter sequences of the pulse power supply to be tested in the current charge and discharge cycle, multiple life prediction values of the pulse power supply to be tested in the current charge and discharge cycle are obtained through multiple life prediction models; for the target arrangement in the different arrangements of all life prediction values, the product of each target weight in the IOWA operator and each life prediction value in the target arrangement is calculated, and the sum of all products is used as the life estimation value of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement; the target weight is the optimal solution obtained by solving the weight optimization model of the IOWA operator; the weight optimization model of the IOWA operator is composed of the objective function of minimizing the error between the life estimation value and the actual life of all pulse power supply samples in the charge and discharge cycle; according to the size relationship between each life prediction value and the life of the pulse power supply to be tested in the previous charge and discharge cycle of the current charge and discharge cycle and the difference between each life prediction value and the life estimation value, the prediction accuracy of each life prediction value is calculated; Sort all life prediction values in descending order of prediction accuracy to obtain a sequential sequence; Calculating the consistency of the order sequence with the target arrangement includes: ; In the formula, Indicates the consistency between the order sequence and the target arrangement. represents the total number of all life prediction models, represents the first The target weights, Indicates the first Life expectancy prediction value, Indicates the first Life expectancy prediction value, It means taking the absolute value; the estimated life value under the arrangement with the greatest consistency is taken as the life of the pulse power supply to be tested in the current charge and discharge cycle.
2. The pulse power system life prediction method according to claim 1, characterized in that: The charging and discharging experiments of the pulse power supply samples under different ambient temperature ranges include: set up Different ambient temperature ranges, , Indicates the minimum ambient temperature, Indicates the maximum ambient temperature, represents the fourth preset value, Indicates rounding up; The ambient temperature range is ;No. Plant to The ambient temperature range is ,in, for An integer in the range; The ambient temperature range is ; During the charging and discharging experiment of the pulse power supply sample under any ambient temperature range, the ambient temperature at each moment is randomly adjusted, and the ambient temperature at each moment is required to be within the ambient temperature range.
3. The pulse power system life prediction method according to claim 1, characterized in that: The calculation of the actual life of the pulse power supply sample in each charge and discharge cycle includes: For the charge and discharge experiments charge and discharge cycles, calculate the Number of charge and discharge cycles The ratio of 1 to the total number of all charge and discharge cycles; the difference between 1 and the ratio is used as the pulse power sample at the first The actual life of the charge and discharge cycle.
4. The pulse power system life prediction method according to claim 1, characterized in that: The parameter sequence includes a charging environment temperature sequence, a charging temperature sequence, a charging voltage sequence, a charging power sequence, a discharging environment temperature sequence, a discharging temperature sequence, a discharging current sequence, a discharging voltage sequence, and a discharging power sequence; the input when training the first life prediction model is the charging environment temperature sequence and the charging temperature sequence of the charge and discharge cycle; the input when training the second life prediction model is the charging voltage sequence and the charging power sequence of the charge and discharge cycle; the input when training the third life prediction model is the discharging environment temperature sequence and the discharging temperature sequence of the charge and discharge cycle; the input when training the fourth life prediction model is the discharging current sequence and the discharging power sequence of the charge and discharge cycle; The input for training the fifth life prediction model is the discharge voltage sequence and discharge power sequence of the charge and discharge cycle; the output for training all life prediction models is the actual life of the charge and discharge cycle.
5. The pulse power system life prediction method according to claim 1, characterized in that: The prediction accuracy of the life prediction value satisfies the expression: ； In the formula, Indicates the first The prediction accuracy of the life prediction value is Indicates the first Life expectancy prediction value, Indicates the life of the pulse power supply to be tested in the previous charge and discharge cycle before the current charge and discharge cycle. , , represent the first preset value, the second preset value and the third preset value respectively, represents the natural exponential function, Indicates the estimated life of the pulse power supply to be tested in the current charge and discharge cycle under the target arrangement mode, Indicates taking the absolute value, Indicates taking the minimum value.
6. The pulse power system life prediction method according to claim 1, characterized in that: Constructing the weight optimization model of the IOWA operator includes: According to each parameter sequence of the charge and discharge cycle, multiple life prediction values of the charge and discharge cycle are obtained through multiple life prediction models, and each life prediction value of the charge and discharge cycle is recorded as the life reference value of the charge and discharge cycle; according to the size relationship between the life reference value of the charge and discharge cycle and the actual life of the previous charge and discharge cycle and the difference between the life reference value of the charge and discharge cycle and the actual life of the charge and discharge cycle, the prediction accuracy of each life reference value of the charge and discharge cycle is calculated; according to the order of prediction accuracy of each life reference value of the charge and discharge cycle from large to small, the life reference values of the charge and discharge cycle are sorted, and the multiple life reference values of the sorted charge and discharge cycles are weighted and summed by the IOWA operator to obtain the life estimation value of the charge and discharge cycle, wherein all the values in the IOWA operator are unknowns that need to be solved; and the weight optimization model of the IOWA operator is constructed with the minimum error between the life estimation value and the actual life of the charge and discharge cycle of all pulse power supply samples as the objective function.
7. The pulse power system life prediction method according to claim 6, characterized in that: The prediction accuracy of each life reference value of the charge and discharge cycle satisfies the expression: ； In the formula, Indicates The first charge-discharge cycle The prediction accuracy of the lifetime reference value is Indicates The first charge-discharge cycle A life reference value, Indicates The actual life of charge and discharge cycles, , , represent the first preset value, the second preset value and the third preset value respectively, represents the natural exponential function, Indicates The actual life of charge and discharge cycles, Indicates taking the absolute value, Indicates taking the minimum value.
8. The pulse power system life prediction method according to claim 6, characterized in that: The weight optimization model of the IOWA operator is: ； In the formula, represents the objective function, Represents the total number of all pulse power samples, Indicates The total number of all charge and discharge cycles of the pulse power supply samples, , Respectively represent The first pulse power supply sample Estimated and actual lifespan of charge and discharge cycles, Indicates The first pulse power supply sample The error between the estimated lifespan of a charge-discharge cycle and the actual lifespan, Indicates taking the absolute value, represents the total number of all life prediction models, represents the first weights, Indicates taking the minimum value.