A charging method suitable for battery pack replacement

Abstract

The present invention discloses a charging method suitable for a battery swap pack, which belongs to the technical field of battery charging. The method adopts an artificial intelligence model to learn the data relationship between the battery pack state characteristics of the battery swap pack and the charge state corresponding to the battery pack state characteristics manually input, determine the charge state prediction rule, and identify the real-time battery pack state characteristics of the battery swap pack during the charging process based on the charge state prediction rule to determine the real-time charge state. Finally, according to different real-time charge states, different charging strategies are adopted to quickly charge the battery swap pack, which can achieve fast and safe charging and help improve charging efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery charging, and in particular relates to a charging method suitable for a battery replacement battery pack. Background Art

[0002] New energy vehicles refer to vehicles that use non-traditional fuels or advanced vehicle power technologies and have energy-saving and environmentally friendly characteristics. They usually include electric vehicles (EVs), plug-in hybrid electric vehicles (PHEVs) and fuel cell vehicles (FCEVs), which can effectively reduce greenhouse gas emissions and improve energy efficiency. They are an important force in promoting green transformation in the transportation sector. With the rapid development of new energy vehicles, the battery swap model has gradually attracted attention as an effective endurance solution. During the charging and discharging process of the battery swap pack, due to the influence of battery performance, charging equipment and other factors, there are problems such as low charging efficiency and great safety hazards. The existing charging methods cannot meet the efficient and safe charging requirements of the battery swap pack. Summary of the invention

[0003] The present invention provides a charging method suitable for a battery-swap battery pack, so as to solve the technical problems of low charging efficiency and great safety hazards of the battery-swap battery pack in the prior art.

[0004] A charging method for a battery pack for battery replacement, comprising:

[0005] Obtaining a battery pack status characteristic of a battery replacement battery pack and a manually input state of charge corresponding to the battery pack status characteristic; wherein the state of charge corresponding to the battery pack status characteristic is manually measured and input data;

[0006] An artificial intelligence model is used to learn the data relationship between the battery pack state characteristics of the battery swapping battery pack and the state of charge corresponding to the battery pack state characteristics input manually, and determine the state of charge prediction rule;

[0007] Collect the real-time battery pack status characteristics of the battery swapping battery pack during the charging process, and use the real-time battery pack status characteristics as input data for the state of charge prediction rule to determine the real-time state of charge of the battery swapping battery pack during the charging process;

[0008] Based on the real-time charge state of the battery swap pack during the charging process, the battery swap pack is controlled to charge using different charging strategies.

[0009] In a possible implementation, obtaining a battery pack state characteristic of a battery replacement battery pack and a manually input state of charge corresponding to the battery pack state characteristic includes:

[0010] For any battery pack, obtain the corresponding current parameter, voltage parameter and temperature parameter to obtain the battery pack status characteristics of the battery pack; and at the same time obtain the battery pack code corresponding to the battery pack;

[0011] The battery pack code corresponding to the battery replacement battery pack and the battery pack status characteristics of the battery replacement battery pack are displayed to the staff through a visual interface to obtain the charge state corresponding to the battery pack status characteristics manually input by the staff.

[0012] In a possible implementation, an artificial intelligence model is used to learn the data relationship between the battery pack state characteristics of the battery replacement battery pack and the state of charge corresponding to the battery pack state characteristics manually input, and determine the state of charge prediction rule, including:

[0013] Build a CNN-LSTM model to obtain an artificial intelligence model;

[0014] The battery pack state characteristics of the battery replacement battery pack are used as the input of the artificial intelligence model, and the state of charge corresponding to the battery pack state characteristics is used as the expected output of the artificial intelligence model. The artificial intelligence model is trained using an adaptive artificial intelligence training algorithm to learn the data relationship and obtain the state of charge prediction rule.

[0015] In a possible implementation, the battery pack state characteristics of the battery replacement battery pack are used as the input of the artificial intelligence model, and the state of charge corresponding to the battery pack state characteristics is used as the expected output of the artificial intelligence model. The artificial intelligence model is trained using an adaptive artificial intelligence training algorithm to learn the data relationship and obtain the state of charge prediction rule, including:

[0016] For artificial intelligence models, an adaptive optimization strategy is used to generate hyperparameters multiple times to obtain multiple model parameter vectors;

[0017] After applying the model parameter vector to the artificial intelligence model, the battery pack state characteristics of the battery replacement battery pack are used as the input of the artificial intelligence model, and the state of charge corresponding to the battery pack state characteristics is used as the expected output of the artificial intelligence model to determine the loss function value corresponding to each model parameter vector;

[0018] According to the loss function value corresponding to each model parameter vector, the optimal model parameter vector is determined, and the other model parameter vectors except the optimal model parameter vector are divided into two parts according to the loss function value, one part is used as the first target model parameter vector, and the other part is used as the second target model parameter vector; wherein the loss function value of any first target model parameter vector is less than the loss function values ​​of all second target model parameter vectors;

[0019] For the optimal model parameter vector, the optimal direction search strategy is used to search for the global optimal position, and the optimal model parameter vector after the optimal position search is obtained;

[0020] For the first target model parameter vector, an adaptive collaborative search strategy is used to perform a local optimal area search to obtain the first target model parameter vector after the local optimal area search;

[0021] For the second target model parameter vector, according to the optimal model parameter vector after the optimal position search and the first target model parameter vector after the local optimal area search, an adaptive multi-strategy guiding strategy is used to conduct a guided search on the second target model parameter vector to obtain the second target model parameter vector after the guided search;

[0022] The optimal model parameter vector after the optimal position search, the first target model parameter vector after the local optimal area search, and the second target model parameter vector after the guided search are fused into a population to obtain a fused population;

[0023] For the fusion population, an adaptive mutation strategy is used to perform mutation search on each model parameter vector to obtain the fusion population after mutation search;

[0024] Determine whether the current number of training times is greater than the preset maximum number of training times. If so, redetermine the optimal model parameter vector based on the fusion population after the mutation search, and use the re-obtained optimal model parameter vector as the final parameter of the artificial intelligence model. The final parameters of the artificial intelligence model and the artificial intelligence model together form the state of charge prediction rule. Otherwise, return to the step of obtaining the loss function value corresponding to each model parameter vector.

[0025] In a possible implementation, for an artificial intelligence model, an adaptive optimization strategy is used to generate hyperparameters multiple times to obtain multiple model parameter vectors, including:

[0026] For the artificial intelligence model, the initialization space vector is:

[0027]

[0028] in, Indicates i The space vector j The hyperparameters of dimensions, i =1,2,…,Q, where Q represents the total number of model parameter vectors that need to be generated. j =1,2,…,D, where D represents the total dimension of the model parameter vector, represents pi, Express satisfaction The smallest prime number of ;

[0029] Based on the space vector, the first space position of the model parameter vector is determined in the solution space as:

[0030]

[0031] in, Indicates i The first spatial position of the model parameter vector j The hyperparameters of dimensions, Indicates j The lower threshold of the dimension, Indicates j The upper threshold of the dimension, mod represents the remainder function;

[0032] Based on the first spatial position of the model parameter vector, the second spatial position of the model parameter vector is determined in the solution space as:

[0033]

[0034] in, Indicates i The second spatial position of the model parameter vector j The hyperparameters of dimensions, Represents the first random number between [0,1];

[0035] The loss function value of each first spatial position and the second spatial position is obtained, and Q spatial positions are taken in order of the loss function value from small to large to obtain multiple model parameter vectors.

[0036] In a possible implementation, for the optimal model parameter vector, an optimal direction search strategy is used to perform a global optimal position search to obtain the optimal model parameter vector after the optimal position search, including:

[0037] For the optimal model parameter vector, get the optimal model parameter vector in front t The historical optimal value during the training process is: ;

[0038] The historical optimal value corresponding to the optimal model parameter vector , determine the first update item as: ;in, represents the second random number generated according to the normal distribution N(0,1), Represents the historical best value Middle j The hyperparameters of dimensions, j =1,2,…,D, where D represents the total dimension of the model parameter vector, represents the optimal model parameter vector j Hyperparameters of dimensions;

[0039] The global optimal position search is performed according to the first update item, and the optimal model parameter vector after the optimal position search is obtained as follows: ;in, represents the optimal model parameter vector after the optimal position search. j dimensional hyperparameters.

[0040] In a possible implementation, for the first target model parameter vector, an adaptive collaborative search strategy is used to perform a local optimal region search to obtain the first target model parameter vector after the local optimal region search, including:

[0041] Based on the current number of training times, the adaptive inertia weight is determined as: ;in, t Indicates the current number of training times. Indicates t The adaptive inertia weight during the training process, represents a third random number uniformly distributed in [0,1], Indicates the preset maximum value of the inertia weight, represents the preset minimum value of the inertia weight, and T represents the preset maximum number of training times;

[0042] For the first target model parameter vector, randomly match a different other first target model parameter vector to obtain a collaboration vector corresponding to the first target model parameter vector;

[0043] According to the adaptive inertia weight and the cooperation vector, a local optimal region search is performed on the first target model parameter vector, and the first target model parameter vector after the local optimal region search is obtained as follows:

[0044]

[0045] in, Indicates n The first target model parameter vector after searching the local optimal region, Indicates n first target model parameter vectors, n=1,2,…,M1, M1 represents the total number of first target model parameter vectors, Indicates n The collaboration vector corresponding to the first target model parameter vector, represents the second random number generated according to the normal distribution N(0,1), represents the collaborative control factor, and , represents the first search step, and exp represents the exponential function with the natural constant e as the base.

[0046] In a possible implementation, for the second target model parameter vector, according to the optimal model parameter vector after the optimal position search and the first target model parameter vector after the local optimal area search, an adaptive multi-strategy guiding strategy is used to perform a guided search on the second target model parameter vector to obtain the second target model parameter vector after the guided search, including:

[0047] Based on the current number of training times, the adaptive decision factor is determined as: ;

[0048] Generate a random fourth number between (0,1) ;

[0049] When the fourth random number Less than 0.5 When , the second target model parameter vector is guided to search according to the optimal model parameter vector after the optimal position search, and the second target model parameter vector after the guided search is obtained as: ;in, Indicates t During the training process m The second target model parameter vector, represents the optimal model parameter vector after the optimal position search, Indicates m The second target model parameter vector after guided search, m=1,2,…,M2, M2 represents the total number of second target model parameter vectors, represents the second random number generated according to the normal distribution N(0,1), represents the guided search control factor, and ; represents the second search step length;

[0050] When the fourth random number Greater than 0.5 and less than When , a guided search is performed based on the first target model parameter vector after the local optimal region search to obtain the second target model parameter vector after the guided search: ;in, Represents the first target model parameter vector after a random local optimal region search;

[0051] When the fourth random number Greater than When , the outlier position vector is determined as: ; and perform guided search according to the outlier position vector, and obtain the second target model parameter vector after guided search as: ;in, represents the outlier position vector, Represents the vector composed of the minimum values ​​of each dimension in all model parameter vectors, Represents a vector consisting of the maximum values ​​of each dimension in all model parameter vectors.

[0052] In a possible implementation, for the fusion population, an adaptive mutation strategy is used to perform mutation search on each model parameter vector to obtain a fusion population after the mutation search, including:

[0053] Based on the current number of training times, the probability of mutation decision is determined as: ;in, v represents the probability control factor, e represents a natural constant, represents the full gamma function;

[0054] For any model parameter vector in the fusion population, a fifth random number is randomly generated between (0,1) for the model parameter vector ;

[0055] Determine the fifth random number Is it less than the mutation decision probability? If so, perform mutation search on the model parameter vector to obtain the mutation search value. Otherwise, do not process the model parameter vector and put the model parameter vector into the fusion population after the mutation search.

[0056] The mutation search is:

[0057]

[0058]

[0059]

[0060] in, Indicates the first k A model parameter vector, Represents the model parameter vector The corresponding mutation search value, represents the coefficient of variation, Represents the upper bound vector composed of the upper limit thresholds of all dimensions, Represents the lower bound vector composed of the lower bound thresholds of all dimensions, represents the scaling factor, means [-2.5 , 2.5 ], e represents a natural constant, Indicates the upper limit of the scaling factor, represents the adjustment parameter of the scaling factor;

[0061] Determine the mutation search value Is the loss function value less than the model parameter vector If the loss function value is , then accept the mutation and change the mutation search value Put it into the fusion population after the mutation search, otherwise reject the mutation and change the model parameter vector Put it into the fusion population after mutation search;

[0062] According to the above method, after traversing all model parameter vectors in the fusion population for mutation search, the fusion population after mutation search is obtained.

[0063] In a possible implementation, based on the real-time state of charge of the battery pack during the charging process, the battery pack is controlled to be charged using different charging strategies, including:

[0064] Determine a target charging strategy based on the real-time state of charge of the battery pack during the charging process; wherein the target charging strategy includes a constant current charging strategy, a staggered charging strategy or a trickle charging strategy;

[0065] Based on the target charging strategy, the battery-swap battery pack is controlled to charge.

[0066] The present invention provides a charging method suitable for a battery swap pack. The method adopts an artificial intelligence model to learn the data relationship between the battery pack state characteristics of the battery swap battery pack and the charge state corresponding to the battery pack state characteristics manually input, determine the charge state prediction rule, and identify the real-time battery pack state characteristics of the battery swap battery pack during the charging process based on the charge state prediction rule to determine the real-time charge state. Finally, different charging strategies are adopted for different real-time charge states to quickly charge the battery swap pack, which can achieve fast and safe charging and help improve charging efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0068] Figure 1 A flow chart of a method for charging a battery pack suitable for use in an embodiment of the present invention.

[0069] Figure 2 A flowchart of obtaining a state of charge prediction rule provided by an embodiment of the present invention.

[0070] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0071] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0072] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0073] like Figure 1 As shown, an embodiment of the present invention provides a charging method for a battery pack for battery replacement, comprising:

[0074] S101, obtaining a battery pack status characteristic of a battery replacement battery pack and a manually input state of charge (SOC) corresponding to the battery pack status characteristic; wherein the state of charge corresponding to the battery pack status characteristic is manually measured and input data;

[0075] The battery pack for battery swapping is generally equipped with current sensors, voltage sensors and temperature sensors, so the data transmitted by these sensors can be collected. The state of charge corresponding to the battery pack status characteristics cannot generally be obtained directly, but can be measured by staff using professional instruments.

[0076] S102, using an artificial intelligence model to learn the data relationship between the battery pack state characteristics of the battery replacement battery pack and the state of charge corresponding to the battery pack state characteristics manually input, and determine a state of charge prediction rule;

[0077] By learning the data relationship between the battery pack status characteristics of the swap battery pack and the manually input charge state corresponding to the battery pack status characteristics, the charge state can be predicted based on the data measured by the sensor, so that different charging strategies can be adopted under different charge states, ultimately achieving safe and fast charging.

[0078] S103, collecting real-time battery pack status characteristics of the battery swapping battery pack during the charging process, and using the real-time battery pack status characteristics as input data for the state of charge prediction rule to determine the real-time state of charge of the battery swapping battery pack during the charging process;

[0079] The real-time battery pack status characteristics can be collected by sensors, and the state of charge prediction rules can identify the real-time battery pack status characteristics, thereby determining the real-time state of charge corresponding to the real-time battery pack status characteristics.

[0080] S104. Based on the real-time charge state of the battery swap battery pack during the charging process, control the battery swap battery pack to charge using different charging strategies.

[0081] The present invention provides a charging method suitable for a battery swap pack. The method adopts an artificial intelligence model to learn the data relationship between the battery pack state characteristics of the battery swap battery pack and the charge state corresponding to the battery pack state characteristics manually input, determine the charge state prediction rule, and identify the real-time battery pack state characteristics of the battery swap battery pack during the charging process based on the charge state prediction rule to determine the real-time charge state. Finally, different charging strategies are adopted for different real-time charge states to quickly charge the battery swap pack, which can achieve fast and safe charging and help improve charging efficiency.

[0082] In a possible implementation, obtaining a battery pack state characteristic of a battery replacement battery pack and a manually input state of charge corresponding to the battery pack state characteristic includes:

[0083] For any battery pack, obtain the corresponding current parameter, voltage parameter and temperature parameter to obtain the battery pack status characteristics of the battery pack; and at the same time obtain the battery pack code corresponding to the battery pack;

[0084] The battery pack code corresponding to the battery replacement battery pack and the battery pack status characteristics of the battery replacement battery pack are displayed to the staff through a visual interface to obtain the charge state corresponding to the battery pack status characteristics manually input by the staff.

[0085] Different models of battery packs may have different operating parameters, so the battery pack model parameters can also be collected as features. The battery pack coding helps staff find the corresponding battery pack, which can realize manual collection of the state of charge.

[0086] Optionally, the current parameters, voltage parameters and temperature parameters can also be detected for over-limit. If any parameter exceeds the limit, it can be determined whether there is any abnormality in the charging process of the battery swap pack.

[0087] In a possible implementation, an artificial intelligence model is used to learn the data relationship between the battery pack state characteristics of the battery replacement battery pack and the state of charge corresponding to the battery pack state characteristics manually input, and determine the state of charge prediction rule, including:

[0088] Construct a CNN-LSTM (convolutional neural network-long short-term memory neural network) model to obtain an artificial intelligence model;

[0089] It is worth mentioning that since the data to be identified is relatively simple, only CNN or LSTM can be used for identification, or a classification neural network such as the BP neural network model can be used to build an artificial intelligence model.

[0090] The battery pack state characteristics of the battery replacement battery pack are used as the input of the artificial intelligence model, and the state of charge corresponding to the battery pack state characteristics is used as the expected output of the artificial intelligence model. The artificial intelligence model is trained using an adaptive artificial intelligence training algorithm to learn the data relationship and obtain the state of charge prediction rule.

[0091] Training the artificial intelligence model mainly involves training the weights and connection thresholds. However, the existing technologies (such as gradient descent algorithm, particle swarm algorithm) often have technical problems such as slow convergence speed, poor convergence accuracy and easy falling into local optimality when training the artificial intelligence model. The embodiment of the present invention provides an adaptive artificial intelligence training algorithm to solve the problems existing in the existing technology, improve the prediction accuracy of the charge state, and ultimately achieve high-precision charging process control.

[0092] like Figure 2 As shown, the battery pack state characteristics of the battery pack are used as the input of the artificial intelligence model, and the state of charge corresponding to the battery pack state characteristics is used as the expected output of the artificial intelligence model. The artificial intelligence model is trained using an adaptive artificial intelligence training algorithm to learn the data relationship and obtain the state of charge prediction rules, including:

[0093] S201. For the artificial intelligence model, an adaptive optimization strategy is used to generate hyperparameters multiple times to obtain multiple model parameter vectors;

[0094] Among them, each model parameter vector may include all hyperparameters or part of the hyperparameters of the artificial intelligence model (such as only part of the weights).

[0095] S202, after applying the model parameter vector to the artificial intelligence model, using the battery pack state characteristics of the battery replacement battery pack as the input of the artificial intelligence model, using the state of charge corresponding to the battery pack state characteristics as the expected output of the artificial intelligence model, and determining the loss function value corresponding to each model parameter vector;

[0096] The loss function value may be a loss function value obtained by a root mean square loss function or a cross entropy loss function.

[0097] S203, determining the optimal model parameter vector according to the loss function value corresponding to each model parameter vector, and dividing the other model parameter vectors except the optimal model parameter vector into two parts according to the loss function value, one part as the first target model parameter vector, and the other part as the second target model parameter vector; wherein the loss function value of any first target model parameter vector is less than the loss function values ​​of all second target model parameter vectors;

[0098] For example, the ratio of the number of the second target model parameter vectors to the number of the first target model parameter vectors may be 8: 2. If there is a decimal, the last model parameter vector is randomly assigned as the first target model parameter vector or the second target model parameter vector.

[0099] S204, for the optimal model parameter vector, using the optimal direction search strategy to perform a global optimal position search, to obtain the optimal model parameter vector after the optimal position search;

[0100] S205, for the first target model parameter vector, using an adaptive collaborative search strategy to perform a local optimal region search to obtain the first target model parameter vector after the local optimal region search;

[0101] S206, for the second target model parameter vector, according to the optimal model parameter vector after the optimal position search and the first target model parameter vector after the local optimal area search, an adaptive multi-strategy guiding strategy is used to perform a guided search on the second target model parameter vector to obtain the second target model parameter vector after the guided search;

[0102] S207, fusing the optimal model parameter vector after the optimal position search, the first target model parameter vector after the local optimal area search, and the second target model parameter vector after the guided search into a population to obtain a fused population;

[0103] S208, for the fusion population, using an adaptive mutation strategy to perform mutation search on each model parameter vector to obtain a fusion population after the mutation search;

[0104] S209. Determine whether the current number of training times is greater than the preset maximum number of training times. If so, redetermine the optimal model parameter vector based on the fusion population after the mutation search, and use the re-obtained optimal model parameter vector as the final parameter of the artificial intelligence model. The final parameters of the artificial intelligence model and the artificial intelligence model together form a state of charge prediction rule. Otherwise, return to the step of obtaining the loss function value corresponding to each model parameter vector.

[0105] In a possible implementation, for an artificial intelligence model, an adaptive optimization strategy is used to generate hyperparameters multiple times to obtain multiple model parameter vectors, including:

[0106] For the artificial intelligence model, the initialization space vector is:

[0107]

[0108] in, Indicates i The space vector j The hyperparameters of dimensions, i =1,2,…,Q, where Q represents the total number of model parameter vectors that need to be generated. j =1,2,…,D, where D represents the total dimension of the model parameter vector, represents pi, Express satisfaction The smallest prime number of ;

[0109] Based on the space vector, the first space position of the model parameter vector is determined in the solution space as:

[0110]

[0111] in, Indicates i The first spatial position of the model parameter vector j The hyperparameters of dimensions, Indicates j The lower threshold of the dimension, Indicates j The upper threshold of the dimension, mod represents the remainder function;

[0112] Based on the first spatial position of the model parameter vector, the second spatial position of the model parameter vector is determined in the solution space as:

[0113]

[0114] in, Indicates i The second spatial position of the model parameter vector j The hyperparameters of dimensions, Represents the first random number between [0,1];

[0115] The loss function value of each first spatial position and the second spatial position is obtained, and Q spatial positions are taken in order of the loss function value from small to large to obtain multiple model parameter vectors.

[0116] In the prior art, a random initialization strategy is often used for initialization, which results in a poor initial position leading to a slow convergence speed. Therefore, an embodiment of the present invention provides an adaptive optimization strategy for initialization, which can not only make the model parameter vector more evenly distributed in the solution space, but also be more suitable for solving hyperparameters, which can greatly improve the convergence speed of the algorithm.

[0117] In a possible implementation, for the optimal model parameter vector, an optimal direction search strategy is used to perform a global optimal position search to obtain the optimal model parameter vector after the optimal position search, including:

[0118] For the optimal model parameter vector, get the optimal model parameter vector in front t The historical optimal value during the training process is: ;

[0119] The historical optimal value corresponding to the optimal model parameter vector , determine the first update item as: ;in, represents the second random number generated according to the normal distribution N(0,1), Represents the historical best value Middle j The hyperparameters of dimensions, j =1,2,…,D, where D represents the total dimension of the model parameter vector, represents the optimal model parameter vector j Hyperparameters of dimensions;

[0120] The global optimal position search is performed according to the first update item, and the optimal model parameter vector after the optimal position search is obtained as follows: ;in, represents the optimal model parameter vector after the optimal position search. j dimensional hyperparameters.

[0121] The optimal direction search strategy provided by the embodiment of the present invention can enable the optimal model parameter vector to always be searched near the optimal position, which can effectively ensure the convergence speed of the algorithm while improving the convergence accuracy.

[0122] In a possible implementation, for the first target model parameter vector, an adaptive collaborative search strategy is used to perform a local optimal region search to obtain the first target model parameter vector after the local optimal region search, including:

[0123] Based on the current number of training times, the adaptive inertia weight is determined as: ;in, t Indicates the current number of training times. Indicatest The adaptive inertia weight during the training process, represents a third random number uniformly distributed in [0,1], Indicates the preset maximum value of the inertia weight, represents the preset minimum value of the inertia weight, and T represents the preset maximum number of training times;

[0124] For the first target model parameter vector, randomly match a different other first target model parameter vector to obtain a collaboration vector corresponding to the first target model parameter vector;

[0125] According to the adaptive inertia weight and the cooperation vector, a local optimal region search is performed on the first target model parameter vector, and the first target model parameter vector after the local optimal region search is obtained as follows:

[0126]

[0127] in, Indicates n The first target model parameter vector after searching the local optimal region, Indicates n first target model parameter vectors, n=1,2,…,M1, M1 represents the total number of first target model parameter vectors, Indicates n The collaboration vector corresponding to the first target model parameter vector, represents the second random number generated according to the normal distribution N(0,1), represents the collaborative control factor, and , represents the first search step, and exp represents the exponential function with the natural constant e as the base.

[0128] The adaptive collaborative search strategy provided by the embodiment of the present invention can perform more searches in the neighborhood of the first target model parameter vector, thereby improving the convergence of the algorithm within the range, thereby improving the algorithm's search ability in the local area. At the same time, an adaptive inertia weight is set, which can effectively prevent the algorithm from falling into a local optimum in the middle of the algorithm.

[0129] In a possible implementation, for the second target model parameter vector, according to the optimal model parameter vector after the optimal position search and the first target model parameter vector after the local optimal area search, an adaptive multi-strategy guiding strategy is used to perform a guided search on the second target model parameter vector to obtain the second target model parameter vector after the guided search, including:

[0130] Based on the current number of training times, the adaptive decision factor is determined as: ;

[0131] Generate a random fourth number between (0,1) ;

[0132] When the fourth random number Less than 0.5 When , the second target model parameter vector is guided to search according to the optimal model parameter vector after the optimal position search, and the second target model parameter vector after the guided search is obtained as: ;in, Indicates t During the training process m The second target model parameter vector, represents the optimal model parameter vector after the optimal position search, Indicates m The second target model parameter vector after guided search, m=1,2,…,M2, M2 represents the total number of second target model parameter vectors, represents the second random number generated according to the normal distribution N(0,1), represents the guided search control factor, and ; represents the second search step length;

[0133] By guiding the search through the optimal model parameter vector, the second target model parameter vector with a relatively poor position can be quickly searched, which can effectively improve the convergence speed of the algorithm.

[0134] When the fourth random number Greater than 0.5 and less than When , a guided search is performed based on the first target model parameter vector after the local optimal region search to obtain the second target model parameter vector after the guided search: ;in, Represents the first target model parameter vector after a random local optimal region search;

[0135] By guiding the search through the first target model parameter vector, the second target model parameter vector can be used to search unfamiliar areas, which helps to find a better solution, thereby improving the ability of the global search.

[0136] When the fourth random number Greater than When , the outlier position vector is determined as: ; and perform guided search according to the outlier position vector, and obtain the second target model parameter vector after guided search as: ;in, represents the outlier position vector, Represents the vector composed of the minimum values ​​of each dimension in all model parameter vectors, Represents a vector consisting of the maximum values ​​of each dimension in all model parameter vectors.

[0137] By guiding the search of the second target model parameter vector through the outlier position vector, the search of unfamiliar areas can be further increased and the ergodicity of the solution space can be increased.

[0138] In a possible implementation, for the fusion population, an adaptive mutation strategy is used to perform mutation search on each model parameter vector to obtain a fusion population after the mutation search, including:

[0139] Based on the current number of training times, the probability of mutation decision is determined as: ;in, v Represents the probability control factor, which can be set to 7; e represents a natural constant, represents the full gamma function;

[0140] For any model parameter vector in the fusion population, a fifth random number is randomly generated between (0,1) for the model parameter vector ;

[0141] Determine the fifth random number Is it less than the mutation decision probability? If so, perform mutation search on the model parameter vector to obtain the mutation search value. Otherwise, do not process the model parameter vector and put the model parameter vector into the fusion population after the mutation search.

[0142] The mutation search is:

[0143]

[0144]

[0145]

[0146] in, Indicates the first k A model parameter vector, Represents the model parameter vector The corresponding mutation search value, represents the coefficient of variation, Represents the upper bound vector composed of the upper limit thresholds of all dimensions, Represents the lower bound vector composed of the lower bound thresholds of all dimensions, represents the scaling factor, means [-2.5 , 2.5 ], e represents a natural constant, Indicates the upper limit of the scaling factor, represents the adjustment parameter of the scaling factor;

[0147] Determine the mutation search value Is the loss function value less than the model parameter vector If the loss function value is , then accept the mutation and change the mutation search value Put it into the fusion population after the mutation search, otherwise reject the mutation and change the model parameter vector Put it into the fusion population after mutation search;

[0148] According to the above method, after traversing all model parameter vectors in the fusion population for mutation search, the fusion population after mutation search is obtained.

[0149] In order to further prevent the algorithm from falling into the local optimum, an embodiment of the present invention provides an adaptive mutation strategy, which can provide a larger mutation probability in the early stage of the algorithm, improve the possibility of the algorithm searching for the global optimal position, and there is also a certain probability of mutation in the later stage of the algorithm, so as to always maintain the ability to jump out of the local optimum.

[0150] Optionally, each time the model parameter vector changes, an out-of-bounds process may be performed on the model parameter vector to ensure that the parameters are always valid.

[0151] In a possible implementation, based on the real-time state of charge of the battery pack during the charging process, the battery pack is controlled to be charged using different charging strategies, including:

[0152] Determine a target charging strategy based on the real-time state of charge of the battery pack during the charging process; wherein the target charging strategy includes a constant current charging strategy, a staggered charging strategy or a trickle charging strategy;

[0153] Based on the target charging strategy, the battery-swap battery pack is controlled to charge.

[0154] Optionally, determining a target charging strategy based on the real-time state of charge of the battery pack during charging may include:

[0155] During the initial charging process, before the real-time state of charge reaches the first threshold, a constant current charging strategy can be used to achieve pre-charging; then when the real-time state of charge exceeds the first threshold, an interleaved charging strategy can be used for fast charging; finally, when the real-time state of charge reaches the second threshold, a trickle charging strategy can be used for trickle charging, thereby ensuring the charging safety of the swap battery pack.

[0156] The constant current charging strategy may refer to: charging with a constant current less than a first preset current threshold;

[0157] The staggered charging strategy may refer to: charging by staggering positive pulses, negative pulses and pulse gaps; for example: the staggered time of positive pulses, negative pulses and pulse gaps may be 100S:8S:2S.

[0158] The trickle charging strategy may refer to: using a current less than a second preset current threshold and a voltage less than a preset voltage threshold for trickle charging. The second preset current threshold is less than the first preset current threshold, for example: the first preset current threshold may be set to 2A, and the second preset current threshold may be set to 0.5A.

[0159] The present invention uses artificial intelligence technology to intelligently select different charging strategies, which can not only effectively ensure the charging efficiency of the rechargeable battery pack, but also avoid the generation of high temperature in the battery swap battery pack during the charging process, weakening the battery polarization, and can effectively improve the service life of the battery swap battery pack, meeting the needs of the battery swap battery pack for fast and safe charging.

[0160] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and interpreted scripting language JavaScript, etc.

[0161] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0162] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0164] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0165] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for charging a battery pack suitable for battery replacement, characterized in that: include: Obtaining a battery pack status characteristic of a battery replacement battery pack and a manually input state of charge corresponding to the battery pack status characteristic; wherein the state of charge corresponding to the battery pack status characteristic is manually measured and input data; An artificial intelligence model is used to learn the data relationship between the battery pack state characteristics of the battery swapping battery pack and the state of charge corresponding to the battery pack state characteristics input manually, and determine the state of charge prediction rule; Collect the real-time battery pack status characteristics of the battery swapping battery pack during the charging process, and use the real-time battery pack status characteristics as input data for the state of charge prediction rule to determine the real-time state of charge of the battery swapping battery pack during the charging process; Based on the real-time charge state of the battery pack during the charging process, the battery pack is controlled to be charged using different charging strategies; The artificial intelligence model is used to learn the data relationship between the battery pack status characteristics of the battery pack and the state of charge corresponding to the battery pack status characteristics input manually, and determine the state of charge prediction rules, including: Build a CNN-LSTM model to obtain an artificial intelligence model; The battery pack state characteristics of the battery replacement battery pack are used as the input of the artificial intelligence model, and the state of charge corresponding to the battery pack state characteristics is used as the expected output of the artificial intelligence model. The artificial intelligence model is trained using an adaptive artificial intelligence training algorithm to learn the data relationship and obtain a state of charge prediction rule; The battery pack state characteristics of the battery replacement battery pack are used as the input of the artificial intelligence model, and the state of charge corresponding to the battery pack state characteristics is used as the expected output of the artificial intelligence model. The artificial intelligence model is trained using an adaptive artificial intelligence training algorithm to learn the data relationship and obtain the state of charge prediction rules, including: For artificial intelligence models, an adaptive optimization strategy is used to generate hyperparameters multiple times to obtain multiple model parameter vectors; After applying the model parameter vector to the artificial intelligence model, the battery pack state characteristics of the battery replacement battery pack are used as the input of the artificial intelligence model, and the state of charge corresponding to the battery pack state characteristics is used as the expected output of the artificial intelligence model to determine the loss function value corresponding to each model parameter vector; According to the loss function value corresponding to each model parameter vector, the optimal model parameter vector is determined, and the other model parameter vectors except the optimal model parameter vector are divided into two parts according to the loss function value, one part is used as the first target model parameter vector, and the other part is used as the second target model parameter vector; wherein the loss function value of any first target model parameter vector is less than the loss function values ​​of all second target model parameter vectors; For the optimal model parameter vector, the optimal direction search strategy is used to search for the global optimal position, and the optimal model parameter vector after the optimal position search is obtained; For the first target model parameter vector, an adaptive collaborative search strategy is used to perform a local optimal area search to obtain the first target model parameter vector after the local optimal area search; For the second target model parameter vector, according to the optimal model parameter vector after the optimal position search and the first target model parameter vector after the local optimal area search, an adaptive multi-strategy guiding strategy is used to conduct a guided search on the second target model parameter vector to obtain the second target model parameter vector after the guided search; The optimal model parameter vector after the optimal position search, the first target model parameter vector after the local optimal area search, and the second target model parameter vector after the guided search are fused into a population to obtain a fused population; For the fusion population, an adaptive mutation strategy is used to perform mutation search on each model parameter vector to obtain the fusion population after mutation search; Determine whether the current number of training times is greater than the preset maximum number of training times. If so, redetermine the optimal model parameter vector based on the fusion population after the mutation search, and use the re-obtained optimal model parameter vector as the final parameter of the artificial intelligence model. The final parameters of the artificial intelligence model and the artificial intelligence model together form the state of charge prediction rule. Otherwise, return to the step of obtaining the loss function value corresponding to each model parameter vector.

2. The method for charging a battery pack suitable for battery replacement according to claim 1, characterized in that: Obtaining the battery pack status characteristics of the battery replacement battery pack and the manually input state of charge corresponding to the battery pack status characteristics, including: For any battery pack, obtain the corresponding current parameter, voltage parameter and temperature parameter to obtain the battery pack status characteristics of the battery pack; and at the same time obtain the battery pack code corresponding to the battery pack; The battery pack code corresponding to the battery replacement battery pack and the battery pack status characteristics of the battery replacement battery pack are displayed to the staff through a visual interface to obtain the charge state corresponding to the battery pack status characteristics manually input by the staff.

3. The method for charging a battery pack suitable for battery replacement according to claim 1, characterized in that: For artificial intelligence models, an adaptive optimization strategy is used to generate hyperparameters multiple times to obtain multiple model parameter vectors, including: For the artificial intelligence model, the initialization space vector is: in, Indicates i The space vector j The hyperparameters of dimensions, i =1,2,…,Q, where Q represents the total number of model parameter vectors that need to be generated. j =1,2,…,D, where D represents the total dimension of the model parameter vector, represents pi, Express satisfaction The smallest prime number of ; Based on the space vector, the first space position of the model parameter vector is determined in the solution space as: in, Indicates i The first spatial position of the model parameter vector j The hyperparameters of dimensions, Indicates j The lower threshold of the dimension, Indicates j The upper threshold of the dimension, mod represents the remainder function; Based on the first spatial position of the model parameter vector, the second spatial position of the model parameter vector is determined in the solution space as: in, Indicates i The second spatial position of the model parameter vector j The hyperparameters of dimensions, Represents the first random number between [0,1]; The loss function value of each first spatial position and the second spatial position is obtained, and Q spatial positions are taken in order of the loss function value from small to large to obtain multiple model parameter vectors.

4. The method for charging a battery pack for battery replacement according to claim 1, characterized in that: For the optimal model parameter vector, the optimal direction search strategy is used to search for the global optimal position, and the optimal model parameter vector after the optimal position search is obtained, including: For the optimal model parameter vector, get the optimal model parameter vector in front t The historical optimal value during the training process is: ; The historical optimal value corresponding to the optimal model parameter vector , determine the first update item as: ;in, represents the second random number generated according to the normal distribution N(0,1), Represents the historical best value Middle j The hyperparameters of dimensions, j =1,2,…,D, where D represents the total dimension of the model parameter vector, represents the optimal model parameter vector j Hyperparameters of dimensions; The global optimal position search is performed according to the first update item, and the optimal model parameter vector after the optimal position search is obtained as follows: ;in, represents the optimal model parameter vector after the optimal position search. j dimensional hyperparameters.

5. The method for charging a battery pack suitable for battery replacement according to claim 4, characterized in that: For the first target model parameter vector, an adaptive collaborative search strategy is used to search for a local optimal region, and the first target model parameter vector after the local optimal region search is obtained, including: Based on the current number of training times, the adaptive inertia weight is determined as: ;in, t Indicates the current number of training times. Indicates t The adaptive inertia weight during the training process, represents a third random number uniformly distributed in [0,1], Indicates the preset maximum value of the inertia weight, represents the preset minimum value of the inertia weight, and T represents the preset maximum number of training times; For the first target model parameter vector, randomly match a different other first target model parameter vector to obtain a collaboration vector corresponding to the first target model parameter vector; According to the adaptive inertia weight and the cooperation vector, a local optimal region search is performed on the first target model parameter vector, and the first target model parameter vector after the local optimal region search is obtained as follows: in, Indicates n The first target model parameter vector after searching the local optimal region, Indicates n first target model parameter vectors, n=1,2,…,M1, M1 represents the total number of first target model parameter vectors, Indicates n The collaboration vector corresponding to the first target model parameter vector, represents the second random number generated according to the normal distribution N(0,1), represents the collaborative control factor, and , represents the first search step, and exp represents the exponential function with the natural constant e as the base.

6. The method for charging a battery pack suitable for battery replacement according to claim 5, characterized in that: For the second target model parameter vector, according to the optimal model parameter vector after the optimal position search and the first target model parameter vector after the local optimal area search, an adaptive multi-strategy guiding strategy is used to guide the search of the second target model parameter vector to obtain the second target model parameter vector after the guided search, including: Based on the current number of training times, the adaptive decision factor is determined as: ; Generate a random fourth number between (0,1) ; When the fourth random number Less than 0.5 When , the second target model parameter vector is guided to search according to the optimal model parameter vector after the optimal position search, and the second target model parameter vector after the guided search is obtained as: ;in, Indicates t During the training m The second target model parameter vector, represents the optimal model parameter vector after the optimal position search, Indicates m The second target model parameter vector after guided search, m=1,2,…,M2, M2 represents the total number of second target model parameter vectors, represents the second random number generated according to the normal distribution N(0,1), represents the guided search control factor, and ; represents the second search step length; When the fourth random number Greater than 0.5 and less than When , a guided search is performed based on the first target model parameter vector after the local optimal region search to obtain the second target model parameter vector after the guided search: ;in, Represents the first target model parameter vector after a random local optimal region search; When the fourth random number Greater than When , the outlier position vector is determined as: ; and perform guided search according to the outlier position vector, and obtain the second target model parameter vector after guided search as: ;in, represents the outlier position vector, Represents the vector composed of the minimum values ​​of each dimension in all model parameter vectors, Represents a vector consisting of the maximum values ​​of each dimension in all model parameter vectors.

7. The method for charging a battery pack for battery replacement according to claim 6, characterized in that: For the fusion population, an adaptive mutation strategy is used to perform mutation search on each model parameter vector to obtain the fusion population after mutation search, including: Based on the current number of training times, the probability of mutation decision is determined as: ;in, v represents the probability control factor, e represents a natural constant, represents the full gamma function; For any model parameter vector in the fusion population, a fifth random number is randomly generated between (0,1) for the model parameter vector ; Determine the fifth random number Is it less than the mutation decision probability? If so, perform mutation search on the model parameter vector to obtain the mutation search value. Otherwise, do not process the model parameter vector and put the model parameter vector into the fusion population after the mutation search. The mutation search is: in, Indicates the first k A model parameter vector, Represents the model parameter vector The corresponding mutation search value, represents the coefficient of variation, Represents the upper bound vector composed of the upper limit thresholds of all dimensions, Represents the lower bound vector composed of the lower bound thresholds of all dimensions, represents the scaling factor, means [-2.5 , 2.5 ], e represents a natural constant, Indicates the upper limit of the scaling factor, represents the adjustment parameter of the scaling factor; Determine the mutation search value Is the loss function value less than the model parameter vector If the loss function value is , then accept the mutation and change the mutation search value Put it into the fusion population after the mutation search, otherwise reject the mutation and change the model parameter vector Put it into the fusion population after mutation search; According to the above method, after traversing all model parameter vectors in the fusion population for mutation search, the fusion population after mutation search is obtained.

8. The method for charging a battery pack suitable for battery replacement according to any one of claims 1 to 7, characterized in that: Based on the real-time state of charge of the battery pack during the charging process, the battery pack is controlled to be charged using different charging strategies, including: Determine a target charging strategy based on the real-time state of charge of the battery pack during the charging process; wherein the target charging strategy includes a constant current charging strategy, a staggered charging strategy or a trickle charging strategy; Based on the target charging strategy, the battery-swap battery pack is controlled to charge.

Images