Adaptive optimization techniques for accelerating battery charging protocols

Abstract

Methods and systems for implementing optimization techniques for accelerated battery charging protocols are disclosed. A state estimator model, configured as a synthetic data generator, generates a synthetic dataset from input data. The synthetic dataset includes one or more states of the battery. A surrogate model generates a feature set including singular parameters representing the internal properties of the battery by applying time series analysis to the synthetic dataset. A reduced-order ML model determines a minimum charging time based on the optimal control parameters and a subset of features derived from the feature set, while keeping the optimal control parameters within acceptable ranges. One or more charging protocols are generated based on the minimum charging time and the optimal control parameters. In some embodiments, an optimal charging protocol is further selected from one or more charging protocols.

Description

Cross-references

[0001] This application claims the benefit of U.S. Serial No. 63 / 606,837, filed December 6, 2023, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This manual generally relates to charging protocols for rechargeable batteries (secondary batteries). Background Technology

[0003] Rechargeable batteries are the desired power source for a wide variety of applications. Rechargeable batteries are based on a wide range of battery technologies. For example, rechargeable batteries can be lithium-ion batteries (LIBs), lithium metal batteries, liquid electrolyte-based batteries, gel polymer electrolyte-based batteries, or all-solid-state batteries (ASSBs). In some embodiments, the secondary battery is without a negative electrode. The advantages of these rechargeable battery technologies stem in part from their relatively high energy density. However, as battery charging / discharging activity increases or as the battery ages, rechargeable batteries (such as LIBs) experience performance degradation, such as a reduction in the amount of usable energy and an increased tendency to fail due to physical degradation. Summary of the Invention

[0004] This disclosure describes optimization techniques for accelerated charging protocols used in rechargeable batteries such as LIBs. For example, new or existing charging protocols can be optimized to accelerate battery charging while minimizing cell degradation caused by accumulated charge / discharge cycles and iterative battery use. More specifically, the optimization techniques are adaptive, allowing the charging protocol to be iteratively optimized based on battery aging and chemistry.

[0005] The disclosed techniques can be implemented using a computational system configured to accelerate the battery charging process based on inference computed by a machine learning (ML) model. More specifically, the computational system is configured to determine and / or select a specific charging protocol uniquely optimized for the underlying battery chemistry of Li-ion batteries. Furthermore, the disclosed techniques are used in the design and optimization of fast charging algorithms to achieve rapid battery charging speeds while concurrently maintaining battery health, cycle life, and overall safety. The proposed method efficiently adapts (or optimizes) charging strategies based on a comprehensive state value that accurately reflects the battery's charging level and overall condition.

[0006] The system includes a state estimator model as part of a digital twin of a specific battery. The state estimator model is an exemplary physics-based model configured as a synthetic data generator for the battery and operable to generate a synthetic dataset including one or more states of the battery. The state estimator model generates the synthetic dataset by processing an input dataset. The input data may include data and information such as material properties, battery chemistry, current-voltage curves, and battery temperature measurements. In some embodiments, the input data is processed by a data processing algorithm or formatted into preprocessed data such that the preprocessed input data is in a format recognizable by the state estimator model or otherwise processable. In some embodiments, data processing may facilitate data generation and / or enhance its accuracy. In some implementations, the state estimator model is a pseudo-two-dimensional (P2D) model representing an electrochemical model that captures the kinetics, transport processes, and thermodynamics of a rechargeable battery such as a LIB. In some embodiments, the input data includes one or more measurable parameters.

[0007] The system also includes a surrogate ML model that generates a set of engineered features (or alternatively, characteristics) based on time series analysis applied to synthetic datasets or a combination of synthetic and non-synthetic data (such as real experimental data derived from direct physical testing of batteries). In some implementations, the engineered feature set includes a set of feature types, each with a value or range representing that feature type. The engineered feature set may include exotic parameters, which are unmeasurable quantities that can represent specific internal properties of the battery. For example, the feature set may include ranges of voltage, temperature, and electrode and interface stress values, as well as exotic parameters indicating the battery's core temperature and internal diffusion. In some implementations, the surrogate model generates the feature set based on corresponding weights for different feature types, where the weights are determined by the surrogate model. The surrogate model is an engineering approach used when the result of interest cannot be easily measured or computed, and therefore an approximate mathematical model of the result is used instead. In some embodiments, the proxy model of this disclosure is a transformer-based ML model.

[0008] To generate accurate battery performance predictions, surrogate models, such as converter-based ML models, collaborate with other data models of the computational system to capture sequential data and determine long-range dependencies between observed sets of input parameters. In some implementations, the surrogate model collaborates with a reduced-order ML model to identify a charging protocol with a minimum charging time to charge the battery, while keeping maximum temperature and stress during charging within acceptable ranges for the battery in a specific state and / or condition. The minimum charging time in this disclosure refers to an acceptable duration for accelerated (fast) charging and can have a range of values, provided it conforms to the criteria for fast charging.

[0009] Based on the disclosed technology, a computing system can generate a charging protocol (algorithm) for a battery based at least on minimum charging time and optimized control parameters (such as temperature, aging, or stress). Optimized control parameters include one or more types of parameters used for control optimization. Each parameter includes a parameter type and a value or range of values ​​corresponding to that parameter type. In some implementations, all optimized control parameters can be user-defined. In some implementations, some optimized control parameters can be user-defined. In some embodiments, optimized control parameters are generated and recommended by the system according to their importance in affecting the final result. For example, the system uses a reduced-order ML model to compute or otherwise determine the minimum charging time of the battery based on one or more optimized control parameters and a subset of features derived from a feature set. The reduced-order model interprets the relationship between the feature set and the optimized control parameters, and identifies a subset of features crucial to accelerating battery charging time while keeping the values ​​of one or more optimized control parameters within acceptable ranges (such as keeping internal battery temperature and stress to a minimum and within acceptable ranges). In some cases, the feature set is as follows: Figure 4A The input is shown. The optimization control parameters differ from the feature set and are used to optimize the charging protocol. In some implementations, the optimization control parameters are user-specified.

[0010] Other implementations of this and other aspects include corresponding systems, devices, and computer programs encoded on computer storage devices that are configured to perform the actions of the method. A system of one or more computers can be configured by means of software, firmware, hardware, or combinations thereof installed on the system, which in operation cause the system to perform actions. A computer program can be configured by means of instructions that, when executed by a data processing device, cause the device to perform actions.

[0011] The subject matter described in this specification can be implemented in specific embodiments to achieve one or more of the following advantages. The disclosed techniques can be used to enhance existing battery charging protocols and / or algorithms in a way that balances accelerated charging speed with battery health, cycle life, and overall battery safety. In addition to enhancement, optimization techniques also allow for adaptation to dynamic factors such as battery health status and environmental conditions that affect or degrade battery performance.

[0012] One or more concepts of the disclosed techniques are closely related to computer technology. For example, the subject matter of this specification enables the efficient generation and distribution of data among various computing nodes in an exemplary computing system for optimized charging of rechargeable batteries. The system uses machine learning logic, including iterative tuning of prediction algorithms and model weights, to improve its analytical methods for optimizing battery charging protocols. The disclosed techniques can be seamlessly integrated into existing battery management systems (BMS) and embedded technologies.

[0013] The system's optimization engine executes specific computational rules for efficiently analyzing and detecting patterns, relationships, and dependencies among different parameters and / or latent variables in datasets of real and synthetic values. The optimization engine's unique computational methodology allows for optimization of the fast-charging protocol based on one or more battery states, including but not limited to State of Health (SOH), State of Charge (SOC), State of Power (SOP), State of Energy (SOE), or any other state (SOX). As indicated above, the system utilizes a unique approach to efficiently adapt (or optimize) the charging strategy based on one or more battery states (including but not limited to states reflecting charging levels and overall condition). In some embodiments, each of the one or more states includes a state type and a value representing that state type. In some embodiments, the charging protocol is adapted to one or more battery states such that the battery can be charged in a minimum time while keeping one or more optimized control parameters within acceptable ranges.

[0014] Details of one or more implementations of the subject matter described herein are set forth in the accompanying drawings and the following description. Other potential features, aspects, and advantages of the subject matter will become apparent from the description, drawings, and claims. Attached Figure Description

[0015] Figure 1 This is an exemplary computing system used while charging a battery.

[0016] Figure 2A A block diagram illustrating exemplary resources for optimizing various aspects of battery charging is provided.

[0017] Figures 2B to 2D A graphical example of the input data used in processing with a multidimensional data model is shown.

[0018] Figure 3A A block diagram illustrating a machine learning (ML) model.

[0019] Figures 3B to 3D Examples and Figure 3A A graphical example of the observation and prediction data associated with the ML model.

[0020] Figure 4A This shows representative resources used to optimize charging protocols for batteries.

[0021] Figure 4B , Figure 4C and Figure 4D Examples of graphs generated using multi-objective optimization processing are shown.

[0022] Figure 5 This is an exemplary process for determining the charging protocol used to charge the battery.

[0023] Figure 6 It is used for optimization Figure 1 An exemplary processing of the parameters (aspects) of an ML model in a computational system.

[0024] The same reference numerals and names in the various figures indicate the same elements. Detailed Implementation

[0025] Figure 1 This is an exemplary computing system 100 used when charging battery 102. In general, system 100 can represent a complex and innovative system physical structure that includes different components for enhancing existing battery charging protocols.

[0026] For example, components of system 100 collaborate to optimize and / or accelerate the charging protocol for battery 102. System 100 is configured to optimize the battery charging protocol while ensuring safe operation and efficient charging of battery 102. System 100 is also configured to optimize the battery charging protocol, resulting in improved battery 102 lifespan. In some implementations, system 100 may generate an optimized charging protocol to apply specific voltage and / or current values ​​(including target charging duration values) that minimizes internal degradation of battery 102 while increasing the charging speed of the battery.

[0027] For example, system 100 can optimize the battery charging protocol so that when the optimized protocol is used (e.g., for repeated use), battery 102 ages at a slower rate compared to a battery charged using an unoptimized protocol. Therefore, system 100 can be configured and utilized to optimize the battery charging protocol to extend the lifespan of battery 102 and / or ensure the sustained performance of battery 102.

[0028] exist Figure 1 In this context, battery 102 is a rechargeable battery. More specifically, battery 102 is a rechargeable battery having a specific battery chemistry system. In some embodiments, the battery chemistry system is selected from a wide variety of chemical systems based on chemical composition. For example, battery 102 may be a lithium-ion (Li-ion) battery, a lithium-sulfur (Li-S) battery, a sodium-sulfur battery, a magnesium-ion battery, or other types of rechargeable batteries. Generally, battery 102 can be designed based on a wide variety of battery technologies. In some examples, battery 102 includes a liquid electrolyte, a gel polymer electrolyte, a semi-solid (quasi-solid) electrolyte, an inorganic solid electrolyte (such as a sulfide-based electrolyte), or a combination thereof. For example, battery 102 is a lithium-ion battery, a lithium metal battery, an all-solid-state battery (ASSB), and an all-solid-state lithium-ion battery. In some embodiments, battery 102 is a negative electrode-less (without a negative electrode) battery.

[0029] In some implementations, battery 102 includes one or more battery cells 103. For example, battery 102 may include n cells 103, where n is an integer greater than or equal to 1. The number n of cells 103 in battery 102 can vary based on the intended end application. Figure 1 As typically shown, each cell 103 includes a positive electrode, a negative electrode, and a separator 105. In some cases, at least one of the cells 103 includes an electrolyte between any electrode and the separator 105. In some examples, the battery 102 can utilize an electrolyte with... Figure 1 The examples shown represent alternative design approaches for different cell structures (e.g., negative electrode-less designs). In some implementations, battery 102 is a rechargeable battery for electronic vehicles (such as cars, buses, or airplanes) or for electronic devices (such as smartphones or laptops).

[0030] System 100 also includes client device 104, sensor system 106, charging protocol optimization system 108, and charging control module 110.

[0031] although Figure 1The client device 104 is depicted as a desktop computer or console, but it can be any known computing device / system, such as a desktop computer, laptop computer, tablet device, mobile device, smartphone, or any other related computing device that receives user input and can transmit, send, or otherwise provide data and input commands to another device in system 100. The client device 104 may optionally be coupled to computing device / system 104-1. For example, the client device 104 may be coupled to computing device 104-1 via a wired or wireless interface connection. In some implementations, the client device 104 is coupled to or communicates with a remote or non-local computing server 104-2 or a cloud-based computing asset 104-2. For example, device 104-2 may be a client-server that receives input from client device 104, provides content to client device 104, or performs both.

[0032] Sensor system 106 includes one or more groups of sensors adapted to acquire a range of data (e.g., sensor data specific to battery 102). For example, sensor system 106 may include voltage sensors, current sensors, and thermocouples or associated temperature sensors. In general, sensor system 106 may include a variety of sensors or sensor types useful for evaluating or modeling parameters (aspects) of battery 102 (which may include inferring or predicting boundary conditions critical to the safe operation of battery 102).

[0033] Sensor system 106 collects and / or generates various types of data to characterize the performance of battery 102. In some implementations, sensor system 106 is configured to generate and / or analyze extensive sensor data and associated battery chemistry information to determine the state of charge (SOC), state of health (SOH), and state of power (SOP) of battery 102. The range of sensor values ​​and associated battery chemistry information analyzed and / or generated by sensor system 106 can generally be characterized as battery performance metrics. Sensor system 106 is configured to generate and / or analyze battery performance metrics specific to the charging and discharging activities of battery 102.

[0034] The charging protocol optimization system 108 includes at least one model 122, such as a machine learning (ML) model, and an optimization engine 124. The charging protocol optimization system 108 can be implemented in hardware, software, or both. In some embodiments, model 122 employs one or more ML algorithms to generate robust surrogate models. The optimization engine 124 performs its protocol optimization functions based on one or more algorithms. For example, at least one algorithm used by the optimization engine 124 could be a nondominated sorting genetic algorithm II (NSGA-II).

[0035] Generally speaking, machine learning involves a class of algorithms in computer science that, when implemented, enable systems to exhibit the ability to detect and learn patterns, relational connections, and dependencies in data. Machine learning can sometimes be categorized under the broader category of artificial intelligence. For example, learning about relationships and patterns could correspond to Model 122 learning or interpreting charging protocol samples, where differences exist between samples and certain distinct underlying features exist for each step or task within the sample protocol. Learning about relationships and patterns could also include Model 122 learning or interpreting how the details of adjusting certain steps or tasks in the sample protocol affect, for example, the charging duration and battery performance of a particular battery chemistry.

[0036] In other examples, machine learning may include mapping input sets, such as large sets of battery performance data for varying chemical systems (e.g., different combinations of elements, compounds, and chemical properties). Machine learning may also include mapping (one or more) output sets, which in this battery example could be groupings of good battery performance data detected in varying battery chemistry systems. In this example battery context, supervised learning is a subcategory of ML algorithms that can be utilized by model 122 to determine the mapping. For example, supervised learning methods could be labeled by the availability of actual samples of a specific battery charging protocol, voltage or current values, and battery chemistry systems, along with corresponding correct annotations or labels for samples validated by humans, another ML model, or both.

[0037] The charging control module 110 is a controller configured to generate and / or provide control signaling for controlling the charging and discharging activities of the battery 102. The charging control module 110 includes protocol logic 120 for identifying, selecting, generating, and / or determining a specific charging protocol. Example charging protocols include multi-step constant current (MSCC) protocols, multi-step constant current constant voltage (SCCCV) protocols, intermittent charging protocols, and constant power (CP) protocols. Additional or related methods (or protocols) for charging the battery 102 are also within the scope of this specification.

[0038] System 100 can use charging control module 110 to determine one or more details of the charger used or to be used for charging battery 102. For example, protocol logic 120 of charging control module 110 can be configured to determine the type of charger to be used when charging battery 102. In some implementations, charging control module 110 determines charger details based on details of battery 102 (such as battery type or chemistry). System 100 can pass the charger details to charging protocol optimization system 108 and determine an optimized charging protocol based at least on the charger type or other details.

[0039] In some implementations, a specific charging protocol is implemented and / or enhanced based on a corresponding charging algorithm. For example, charging protocol model 110 can implement the MSCC protocol by iteratively traversing the computational steps of a typical charging algorithm. At least one step of the algorithm involves the charging control module 110 applying a constant current with a selected value to the battery 102 for a specific duration. Exemplary optimization schemes may include adjusting the duration of the constant current (CC) and / or other parameters to achieve a specific charging speed. In some implementations, after an initial (first) application of CC with a first value during a first duration, there may be at least a second application of CC with a second value during a second duration. The first and second values ​​may be the same or different durations.

[0040] The charging control module 110 collaborates with the sensor system 106 to generate a comprehensive set of (one or more) battery performance metrics. At least one set of battery performance metrics can be provided as input data to the charging protocol optimization system 108 to generate a corresponding protocol optimization output. In some implementations, system 100 utilizes discrete values ​​from the set of battery performance metrics to analyze and / or annotate certain data or parameters. System 100 can generate one or more training datasets based on the analysis and / or annotation of values ​​in the set of performance metrics. For example, during the training period of system 108, the training dataset is provided as input data to the machine learning model of the charging protocol optimization system 108. This is referenced below. Figures 4A to 4D and Figure 6 Let me describe it in more detail.

[0041] Figure 2A A block diagram illustrating resource 200 for optimizing a charging protocol for a battery is shown. Resource 200 includes a synthetic data generator 202 and an optimization engine 124. Each of the synthetic data generator 202 and the optimization engine 124 can be implemented in hardware, software, or both. A digital twin of a battery is a virtual representation of a physical battery system that integrates real-time sensor data, computational models, and analytics to simulate and predict battery performance over time under different operating conditions. In some implementations, the synthetic data generator 202 is part of a digital twin of battery 102 that includes real-time data from sensor system 106. In some cases, the digital twin includes a pseudo-two-dimensional (P2D) model 204. For example, the P2D model 204 can be configured to model the constant current charging and discharging of battery cells, such as LIB and lithium metal batteries, which include electrolytes with or without polymers. In other implementations, the digital twin includes single-particle models, Doyle-Fuller-Newman (DFN) models, equivalent circuit models, or combinations thereof.

[0042] A digital twin can include a 3D model. For example, a digital twin (or model 204) can include a physics-based model, such as a porous electrode model, for predicting the performance and lifetime of the LIB. Some physics-based models can range from microscopic 3D models that spatially resolve the microstructural properties of all phases in a porous electrode to reduced-order, computationally efficient models that do not resolve the microstructure. Generally, a digital twin of battery 102 can include any model suitable for modeling a battery chemistry system that includes electrochemical parameters (aspects) for evaluating the overall performance of the rechargeable battery. In some implementations, the digital twin of battery 102 includes the aforementioned real-time data from sensor system 106. For example, the digital twin of battery 102 can incorporate or integrate some (or all) of the real-time data from sensor system 106.

[0043] The simulation model of the synthetic data generator 202 is calibrated using at least a portion of the real-time data from the digital twin of battery 102 as input data and / or calibration data. For example, a combination of input data 206 and calibration data 208 can be used to calibrate the P2D model 204. The calibration data includes a set of actually observed experimental results. In some implementations, Figure 2B Several graphical examples illustrate calibration data (or alternatively, calibration input data), which includes one or more items selected from a group consisting of: a time-series voltage at a charging rate, a time-series voltage at a discharging rate, a normalized time-series voltage at a charging rate, a normalized time-series voltage at a discharging rate, and their derivatives (such as differences and differentials). In some implementations, the P2D model 204 is calibrated by fine-tuning the parameters (aspects) based on the calibration data 208. For example... Figure 2B As shown, calibration data may include a set of experimental data related to various charge / discharge voltage curves at different charge / discharge rates. For example, system 100 may perform calibration based on Bayesian optimization. System 100 may perform calibration processing to ensure that the consistency between the simulation result set and the observed experimental data meets a minimum accuracy threshold. The accuracy threshold can be set to achieve a high-fidelity representation of the battery behavior curves for different fast charging protocols. In some embodiments, the state estimator model before calibration (such as a P2D model) is referred to as a general state estimator model. In some embodiments, the state estimator model after calibration (such as a P2D model) is referred to as a battery-specific calibrated model or a battery-specific model that generates a synthetic dataset with better accuracy.

[0044] P2D model 204 is configured to generate synthetic dataset 210 based on input data 206. Synthetic dataset 210 may include multiple data points corresponding to the state of battery 102. Additionally, P2D model 204 may implement specific sampling methods to generate portions of synthetic dataset 210, simulation results of various fast charging protocols, or both. Based on the sampling method system, P2D model 204 can generate synthetic dataset 210. Since synthetic dataset 210 includes a large amount of data, it captures the complex relationship between charging parameters and battery response. For example, when adjusting parameters associated with the battery's charging current curve, the relationship between these parameters and the cell temperature is captured.

[0045] The intensity of the charging current determines the rate of electrochemical reactions in the negative and positive electrodes of battery 102. Higher charging currents generally lead to faster reactions and increased heat generation. Heat is generated via the Joule heating effect as the current flows through the electrolyte from one electrode to another, overcoming the internal resistance of battery 102. This heat generation can be related to the solid-state electrode, electrolyte salt concentration, electrolyte salt concentration gradient, and electrode / electrolyte interface resistance. In some implementations, the captured relationships reveal that higher currents result in larger lithium-ion concentration gradients in the electrolyte and electrode particles. These gradients can lead to localized overpotentials, which can cause additional heat generation.

[0046] like Figure 2A As indicated, the input data 206 may contain a comprehensive set of parameters. For example, the input data 206 may include material properties 214, pseudo-open circuit voltage (OCV) 216, Galvanostatic Intermittent Titration Technique (GITT) experimental results 218, and core temperature measurements.

[0047] Material properties 214 may include a set of general physical properties and spatial properties. A representative set of general physical properties includes one or more of the following selected from the group consisting of: diffusion coefficients of components in the electrolyte (such as lithium ions and solvents), diffusion coefficients of electrodes, ionic conductivity of the electrolyte, and electronic conductivity of electrodes; while an example set of spatial properties includes particle size distribution, microstructure porosity, microstructure tortuosity, surface area of ​​active materials, or combinations thereof.

[0048] Figure 2C A graphical example is shown, including OCV input data 216 of the voltage processed by P2D model 204. Figure 2C In this context, the voltage relates to the pseudo-open circuit potential obtained from the C / 50 discharge activity. Figure 2D A graphical example is shown, including GITT input data 218 of time-series potentials processed by P2D model 204.

[0049] P2D model 204 may have a first calibration with a first accuracy and fidelity in relation to its output. In some embodiments, P2D model 204 may have a first calibration based on a first measurement. In some embodiments, P2D model 204 may have a first calibration based on a first measurement, followed by a second calibration based on a second measurement, resulting in a second accuracy and fidelity. In some implementations, the first calibration is based on a voltage measurement. In some embodiments, the second calibration is based on a core temperature measurement of the battery. In some implementations, P2D model 204 incorporates at least a core temperature measurement to obtain a second metric of accuracy and fidelity exceeding the first metric. For example, P2D model 204 incorporates the voltage as a first measurement used in the first calibration and the core temperature as a second measurement used in the second calibration to improve the accuracy and fidelity of the digital twin corresponding to battery 102. This enables P2D model 204 to provide a comprehensive representation of real battery behavior during fast charging scenarios. In some embodiments, P2D model 204 is calibrated taking into account three or more measurements such as voltage, core temperature, and stress.

[0050] In some implementations, the logical arrangement of resource 200 represents a workflow for implementing and / or performing multi-stage fast charging protocol optimization at system 100. As described in detail below, this workflow couples a digital twin of battery 102, a surrogate model 230 including its corresponding ML algorithm, and a charging protocol optimization block (240) for implementing multi-objective optimization processing. By coupling these resources in the disclosed manner, the example workflow represents a novel and efficient method for addressing the technical challenges of effectively optimizing existing battery charging protocols (algorithms).

[0051] exist Figure 2A In this process, the output of the synthetic data generator 202 (such as its experimental results and synthetic dataset 210) is provided as input data to the optimization engine 124 to generate an optimized charging protocol 250 (or alternatively, an optimal charging protocol). The optimization engine 124 includes a proxy model 230 and a charging protocol optimization block (240) that operate together. In some implementations, the optimized charging protocol 250 is battery-specific. For example, the optimization engine 124 collaborates with the proxy model 230 to generate the optimized charging protocol 250 as output, wherein the optimized charging protocol 250 may be one or more charging protocols uniquely optimized for specific design details of the battery 102.

[0052] Agent model 230 is Figure 1Examples of model 122 included in the charging protocol optimization system 108. In some implementations, the proxy model 230 is a transformer-based machine learning model that includes a transformer architecture based on a self-attention mechanism. For example, the proxy model 230 may include a recurrent neural network (RNN) having one or more multi-head attention blocks, each using a scaled dot product attention mechanism.

[0053] Generally speaking, a recurrent neural network (RNN) is an artificial neural network that is instantiated in software and receives an input sequence and generates an output sequence from it. An RNN can use some or all of the network's internal states from previous time steps to compute the output at the current time step. This allows RNNs to exhibit dynamic temporal behavior. In other words, a RNN summarizes all the information it has received up to the current time step and is able to capture long-term dependencies in the data sequence.

[0054] Regarding the self-attention mechanism, this mechanism allows data at each time step to reference other data at other time steps within the same sequence length. A surrogate model 230, such as a transformer-based ML model, can utilize this method to compute the encoded representation of the sequence data, where model 230 can perform these computations automatically without recursion or convolution operations. In some implementations, multi-head attention blocks enable the surrogate model 230 to collectively focus on information from different subspaces (such as the real data subspace and the synthetic data subspace) at different time locations.

[0055] Figure 3A A block diagram illustrating a machine learning (ML) model. More specifically, Figure 3A Examples are shown in the reference above. Figure 2A Additional aspects of the agent model 230 described.

[0056] The surrogate model 230 is configured to process one or more datasets, including a synthetic dataset 210, a non-synthetic dataset, or both. The non-synthetic dataset comprises a collection of real experimental data values ​​302 derived from physical tests of the battery 102 (such as data values ​​generated from the aforementioned sensor system 106). The synthetic dataset 210 and the non-synthetic dataset 302 are used as training datasets for training the surrogate model 230.

[0057] In some examples, the training dataset may include historical battery performance data with predefined target attributes (e.g., annotated or labeled data with values), such as charging duration, voltage, current, constant current duration, constant voltage duration, battery temperature, or stress parameters. The predefined target attributes or labels are defined for use during model training. The underlying algorithm employed during model training uses these labels to reliably and accurately identify those attributes, determine the corresponding weights for those target attributes, and ultimately train the model to make predictions using the determined weights.

[0058] During the example training period, the surrogate model 230 is iteratively trained by feeding labeled (or annotated) data from the training dataset into one or more transformer-based ML algorithms 304. Generally, data labeling (or annotation) involves preparing the training dataset to include certain target attributes and labeling each input sample according to the corresponding attribute, enabling the machine learning model to learn what predictions it expects to make. For example, the input samples could be a set of performance data for battery type A, and one or more corresponding attributes or labels could be battery aging, voltage, current, and temperature for a specific charging duration. This process is one of the stages for preparing data for supervised machine learning.

[0059] The surrogate model 230 can apply supervised ML algorithm 304 to annotated inputs of one or more training datasets. In some implementations, the surrogate model 230 is trained to threshold accuracy based on a specific underlying cost function. A representative algorithm 304 of the surrogate model 230 can be configured such that each input data point in the input samples is “tagged” and / or encoded into an embedding, which is used or processed by the model for all (or some) of its ML computation.

[0060] In some implementations, the surrogate model 230 is a transformation-based ML model, and it uniquely adapts one or more ML algorithms 304 to the encoding of numerical data in the time series domain. Accordingly, the surrogate model 230 utilizes one or more algorithms 304 to generate predicted data consistent with the observed data. For example, Figure 3B A graphical example 310B illustrates the observed data and predicted data 310 corresponding to the voltage of battery 102. Similarly, Figure 3C A graphical example 320C illustrating the observed data and predicted data 320 corresponding to the temperature values ​​of battery 102; and Figure 3D A graphical example 330D illustrates the observed and predicted data 330 corresponding to the stress associated with the same cell 102.

[0061] The surrogate model 230 is configured to generate a feature set (engineered features) 306, including a set of feature types with corresponding characteristics, based on the following analyses: i) time series analysis applied to synthetic dataset 210; ii) time series analysis applied to real experimental data 302; or iii) both. Features including feature types and feature values ​​include one or more singular parameters representing one or more internal properties of battery 102. For example, a singular parameter may indicate one or more internal states of battery 102 that are not directly measurable (e.g., unmeasurable) by physical testing of battery 102, or are associated with such one or more internal states. The engineered feature set from the transformer-based ML model may include, but is not limited to, slope, mean, standard deviation, kurtosis, entropy, skewness, and their derivatives relevant to time series analysis.

[0062] In some implementations, the surrogate model 230 is configured to capture complex relationships between charging parameters (e.g., current, voltage, stress, temperature) and enable dynamic optimization using time-varying data. The surrogate model 230 can also be configured as a pre-trained feature generator used to generate or identify a feature set from a user-specified feature set for the corresponding battery chemistry system of battery 102.

[0063] The surrogate model 230 is specifically designed and trained to capture sequential data and long-range dependencies, making it well-suited for generating realistic battery performance predictions. As an example, the surrogate model 230 is a transformer-based ML model and can be trained to effectively replicate the behavior of the corresponding digital twin of battery 102. Such training can significantly enhance the model's predictive capabilities.

[0064] A P2D model or a transformer-based ML model can be configured as a synthetic data generator to generate synthetic dataset 210. More specifically, a trained transformer-based ML model can represent a digital twin of battery 102. Transformer-based ML models can offer advantages over P2D models, such as faster computation time. In some implementations, a trained transformer-based ML model can be used to accelerate synthetic dataset generation, for example, by 100 to 1000 times compared to P2D model 204. Furthermore, speed improvements or accelerated computation are also within the scope of this specification.

[0065] Figure 4A The resources used to implement the charging protocol optimization block (240) for generating the charging protocol are shown. More specifically, it implements a multi-objective optimization (MOO) process 402 to generate one or more charging protocols 408, followed by identifying the optimal charging protocol (optimized charging protocol) 250 for the battery 102.

[0066] The goal of the multi-objective optimization process 402 is to generate an optimal charging protocol 250 with the minimum charging time. For example, the optimal charging protocol 250 has a minimum charging time while keeping the maximum temperature and stress within acceptable limits for a fixed state of charge (SOC) range. The charging protocol can be configured as a multi-stage fast charging method. In some implementations, the charging protocol can be configured to employ a multi-stage constant current (MSCC) approach, where each stage involves charging the battery with a constant current (CC) for a certain duration. The overall configuration of the charging protocol can be particularly advantageous for maximizing battery cycle life.

[0067] In some implementations, the surrogate model 230 generates a feature set and determines the corresponding feature values ​​based on single-head training, multi-head training, or both (404). In some cases, the surrogate model is a transformer-based ML model and is trained as a single-head attention network (or model), a multi-head attention network, or both. Single-head training and / or multi-head training are performed to determine the corresponding weights for a specific feature type. Generally, single-head and multi-head refer to attention networks that form the architecture of the transformer in a transformer-based ML model. Input samples are processed according to the specific structure of the single-head or multi-head attention network. In some implementations, the single-head attention network (or model) uses the same output layer for each task, while in the multi-head attention network (or model), the output layer assigns different sets of output units (heads) for each task.

[0068] To achieve its multi-objective optimization processing 402, the charging protocol optimization system 108 includes a reduced-order ML model configured to generate a simplified representation of time-series data derived from time-series analysis applied to the synthetic dataset 210. The reduced-order ML model of system 100 can transform the complete charging curve data into parameterized versions for each cycle. The reduced-order ML model is also configured to facilitate efficient handling of the diverse charging scenarios and battery states of battery 102. In some implementations and as... Figure 4A As shown, engineered features, synthetic datasets, and non-synthetic datasets are used as inputs to the charging protocol optimization 240 to generate an optimal charging protocol 250 as output. The multi-objective optimization process 402 includes training the machine learning model 404 and subsequent multi-objective optimization (MOO) processing 406. Figure 4B This is, according to one embodiment of the present disclosure, as the output from training 404. Figure 4A Graphical plot of 404B in the document. Figure 4B This demonstrates the accuracy of the trained machine learning model in predicting the maximum temperature by comparing the predictions with the actual results. The comparison shows that the trained machine learning model exhibits high accuracy and consistency in its predictions. Figure 4CThis is, according to one embodiment of the present disclosure, the output from MOO process 406. Figure 4A Graphical plot of 406C in [the image / database]. Figure 4C In this context, Δf represents the difference or improvement in the objective function value between successive generations during the optimization process. Figure 4C The convergence of the optimization is demonstrated. After iterative improvements, for example, after 50 to 80 generations of optimization, the system can obtain a stable solution.

[0069] Figure 4D According to one embodiment of this disclosure Figure 4A Graphical plotting of 408D in the middle. Figure 4D It demonstrates one or more optimal protocols for projection onto a 3D space defined by normalized time, normalized temperature, and normalized stress.

[0070] Process 402 further includes: i) generating a subset of features based on multi-objective optimization (406) for each feature type; and ii) performing multi-objective optimization for each feature type using a simplified representation of the time series data. Process 402 includes an optimization box 406 for instructing iterative optimization of the multi-objective method based on a reduced-order ML model over time. Figure 4D This demonstrates a typical example of one or more optimal protocols defined by normalized time, normalized temperature, and normalized stress in a 3D space. Figure 4D In the example, a set of data points and associated features, as illustrated in this article, can be used to define the selection of the exemplary optimal protocol 408. Also, Figure 4A As shown, experimental verification is considered to select an optimal charging protocol 250 from one or more charging protocols 408. In some implementations, system 100 employs an ML algorithm to extract central and / or high-weight features from a target space 412, which is more specifically defined and much narrower than a general space. In some implementations, the extracted features are used to understand and / or profile the underlying relationships between charging parameters and battery response. In some implementations, the extracted features are used to accelerate the identification of feature subsets that are crucial for the design and optimization of fast charging protocols.

[0071] Figure 5 This is an exemplary process for determining and / or selecting a charging protocol for charging a rechargeable battery. In some implementations, at least sensor system 106 and charging protocol optimization system 108 are used to implement or perform process 500 at system 100. In some examples, the steps or actions of process 500 are implemented by programmed software instructions, firmware instructions, or both. Instructions of various types may be stored in non-transitory machine-readable storage and may be executed by one or more of the processors or other resources described herein.

[0072] In some embodiments, the rechargeable battery includes a negative electrode comprising a negative electrode active material layer. In some embodiments, the negative electrode active material layer comprises a negative electrode active material such as lithium metal or a lithium alloy. In some embodiments, the negative electrode active material comprises at least one selected from the group consisting of lithium, sodium, magnesium, aluminum, silicon, calcium, titanium, manganese, iron, cobalt, nickel, zinc, molybdenum, silver, indium, tin, and tungsten.

[0073] In some embodiments, the negative electrode further includes a negative electrode current collector. In some embodiments, the negative electrode active material layer is assembled into the battery prior to the first charge. In some embodiments, the negative electrode active material layer is formed after the first charge.

[0074] In some embodiments, the rechargeable battery includes a liquid electrolyte, a polymer electrolyte, or a fully charged electrolyte.

[0075] In some embodiments, the polymer electrolyte includes an electrolyte salt, a solvent, and a polymer.

[0076] In some embodiments, the polymer is polymerized in situ after the electrolyte salt, solvent, and polymer precursor (or alternatively, monomer) are mixed.

[0077] In some embodiments, the monomer comprises one or more polymerizable groups. In some embodiments, non-limiting specific polymerizable groups include vinyl groups (-CH=CH2), substituted vinyl groups (-CR1=CR2R3), and combinations thereof, wherein R1, R2, and R3 are independently hydrogen, halogen, -CN, -NO2, C, etc. 1-6 Alkyl, C 1-6 Halogenated alkyl, C1-6 hydroxyalkyl, C1-6 aminoalkyl, C 2-6 alkenyl, C 2-6 alkynyl group, C 6-14 Aryl or any combination thereof.

[0078] In some embodiments, the monomer comprises at least one selected from the group consisting of: 2,2,3,3-tetrafluorobutane-1,4-diacrylate, 2,2,3,3,4,4,5,5-octafluorohexane-1,6-dimethyldiacrylate, 2,2,3,3,4,4,5,5-octafluorohexane-1,6-dimethylbis(2-methacrylate), poly(ethylene glycol) diacrylate (Mn=500-5000), triethylene glycol dimethacrylate (TEGDMA), dicarboxylate dimethacrylate, tetraallylsilane (T... AS), 2,4,6,8-tetramethyl-2,4,6,8-tetravinylcyclotetrasiloxane, triethoxyvinylsilane, allyltriethoxysilane, pentaerythritol tetraacrylate (PETA), pentaerythritol tetramethacrylate (PETMA), tris[2-(acryloyloxy)ethyl]isocyanurate (TAEI), di(trimethylolpropane)tetraacrylate (Di-TMPTA), trimethylolpropane propoxy triacrylate, trimethylolpropane trimethacrylate, pentaerythritol triacrylate, and dipentaerythritol hexaacrylate.

[0079] In some embodiments, the polymer in the polymer electrolyte is a cross-linked polymer. In some embodiments, the precursor (or monomer) for the cross-linked polymer comprises at least two or more polymerizable groups. In some embodiments, the precursor (or monomer) for the cross-linked polymer comprises at least three or more polymerizable groups.

[0080] In some embodiments, the electrolyte salt is selected from the group consisting of: lithium perchlorate (LiClO4), lithium nitrate (LiNO3), lithium hexafluorophosphate (LiPF6), lithium fluoroborate (LiBF4), lithium hexafluoroarsenate (LiAsF6), lithium trifluoromethanesulfonate (LiCF3), lithium bis(perfluoroethanesulfonyl)imide (LiBETI), lithium bis(fluorosulfonyl)imide (LiFSI), lithium bis(trifluoromethanesulfonyl)imide (LiN(CF3SO2)2, LiTFSI), lithium bis(oxalate)borate (LiBOB), lithium difluoro(oxalate)borate (LiDFOB), lithium fluoroalkyl phosphate (Li[PF6]), and lithium fluoride phosphate (Li[PF6]). x (C y F 2y+1-z H z ) 6-x(1≤x≤5, 1≤y≤8, and 0≤z≤2y-1), lithium fluorophosphate (Li2PO3F), lithium difluorophosphate (LiDFP), lithium difluoro(bis(oxalate)phosphate) (LiC4PO8F2), lithium tetrafluorooxalate phosphate (LiC2PO4F4), lithium tri(trifluoromethanesulfonyl)methylated ((LiC(CF3SO2)3), LiF, LiCl, LiBr, LiI, Li2SO4, Li3PO4, Li2CO3, lithium acetate, lithium trifluoromethylacetate, lithium oxalate and mixtures thereof.)

[0081] In some embodiments, the process includes: 1) Obtain input data; 2) Using a state estimator model configured as a synthetic data generator, generate one or more synthetic datasets based on the input data; 3) Utilize proxy models (such as transformer-based ML models) to generate feature sets based on time series analysis applied to synthetic datasets when operating in conjunction with an optimization engine; 4) Using a reduced-order ML model, the minimum charging time of the battery is determined based on one or more optimal control parameters and a subset of features derived from the feature set. The reduced-order ML identifies a subset of features for accelerated charging while maintaining optimal control parameters with values ​​within acceptable limits. 5) Generate one or more charging protocols for the battery, based at least on minimum charging time and optimized control parameters.

[0082] In some embodiments, the input data is preprocessed by a data processing algorithm. For example... Figure 5As shown, process 500 includes the following steps: obtaining input data (502) using system 100. In some embodiments, the input data includes one or more sets of sensing data from sensor system 106 embedded or integrated at battery 102. In some implementations, the process includes the step of preprocessing the input data into preprocessed input data. In some implementations, the process further includes the step of generating one or more synthetic datasets (504) based on the input data and / or the preprocessed input data using a state estimator model (such as a P2D model, an equivalent circuit model (ECM), and a converter-based ML model). For example, a state estimator model configured for a synthetic data generator specific to battery 102 can generate a synthetic dataset based on the input data and / or the preprocessed input data. In some embodiments, the state estimator model is a physics-based model. In some embodiments, process 500 includes the step of generating a synthetic dataset based on a portion or all of the preprocessed input data. In some embodiments, process 500 uses a P2D model to generate a synthetic dataset based on some (or all) of the input data 506 and / or the preprocessed data. The synthetic dataset includes one or more states of battery 102.

[0083] System 100 generates a feature set based on time series analysis (508). For example, a transformer-based ML model generates the feature set based on time series analysis of discrete data applied to a synthetic dataset. The feature set includes one or more singular parameters representing one or more internal properties of battery 102. The singular parameter can represent an unmeasurable quantity specific to battery 102 (including one or more of its cells) without disassembling the battery. For example, the singular parameter can represent unmeasurable quantities such as internal stress of battery 102, core temperature of battery 102, individual resistance of one or more electrodes of battery 102, electrolyte concentration at a specific cell of battery 102, or combinations thereof. Additional singular parameters representing internal properties of battery 102 or unmeasurable quantities specific to battery 102 may include the individual resistance of one or more electrodes of the battery, lithium concentration in solid active material, interfacial kinetic exchange current density, temperature gradient in a cell (e.g., large-scale cell), voltage gradient in a cell (e.g., large-scale cell), overpotential, or combinations thereof.

[0084] System 100 uses a reduced-order ML model to determine the minimum charging time for the battery (510). In some implementations, the minimum charging time is determined for a specific battery 102 based on optimized control parameters having values ​​within acceptable ranges and a subset of features derived from a feature set. Exemplary optimized control parameters may include aging, temperature, stress, voltage, or some other battery parameter. System 100 is configured to dynamically change or adjust the minimum charging time based on specific battery parameters or optimized control parameters. For example, system 100 may dynamically adjust the minimum charging time based on aging, battery state, or both. The minimum charging time may correspond to the desired charging duration provided as input to client device 104. In some implementations, the desired charging duration is one of various optimized control parameters that may be provided or received as input to protocol optimization system 108 (e.g., user input). System 100 generates one or more charging protocols for the battery based at least on the minimum charging time and optimized control parameters (512). In some embodiments, system 100 also considers experimental validation or analysis of accuracy or fidelity to select the optimal charging protocol from one or more charging protocols.

[0085] Figure 6 It is used for optimization Figure 1 An exemplary processing 600 of the parameters (aspects) of the ML model in the computing system 100. In some implementations, processing 600 is implemented or performed at the charging protocol optimization system 108 using input data including real-time data from the sensor system 106. In some examples, the steps or actions of processing 600 are implemented by programmed software instructions, firmware instructions, or both. Instructions of various types may be stored in non-transitory machine-readable storage and may be executed by one or more of the processors or other resources described herein.

[0086] like Figure 6 As shown, process 600 includes the following steps: training against a multi-head optimization technique associated with each feature type (602). Process 600 also includes the following steps: applying a supervised ML algorithm based on a training dataset including annotated inputs (604). Process 600 also includes the following steps: determining appropriate weights for the model for each feature type (606). In some implementations, process 600 includes the following steps: iteratively tuning one or more of its ML models to improve prediction accuracy for a specific battery chemistry system (608). For example, system 100 may iteratively tune the model based on the operations performed to perform at least steps 604 and 606 in process 600.

[0087] In some implementations, one or more steps of processing 500 and processing 600, or other processes described herein, are performed at one or more hardware integrated circuits as part of a larger computational operation to generate machine learning (ML) outputs, including outputs for neural network layers implementing one or more ML models. For example, the outputs may be part of computations for ML tasks or inference workloads to generate outputs for feature engineering processes used to optimize charging protocols for rechargeable batteries such as LIBs. As indicated above, one or more integrated circuits may include dedicated processors, such as neural network processors or hardware ML accelerators, configured to accelerate computations for generating different types of data processing outputs.

[0088] In some implementations, fast charging protocols include: after multi-step constant current (MSCC) charging, reducing the battery from a first threshold state of charge (SOC). a Charge to the second threshold (SOC) b Fast charging refers to the ability to charge a battery from its state of charge (SOC) in a duration of no more than 20 minutes, 15 minutes, or 10 minutes. a Charge to SOC b The charging protocol.

[0089] In some embodiments, multi-step constant current (MSCC) depends on the battery's initial state of charge (SOC). i This is part of a battery charging protocol. In some embodiments, a method of charging a battery includes: 1) If SOC i The first threshold of SOC (SOC) a If a low constant current (I) is used, then... a Charge the battery until the State of Charge (SOC) is reached. a Then, the battery is charged using a multi-step constant current (MSCC) charging protocol that includes multiple charging steps. 2) If SOC i The second threshold (SOC) is equal to or greater than the SOC threshold. b If a fixed constant current (I) is used, then... b To charge the battery, and 3) If SOC i Equal to or higher than SOC a But smaller than SOC b The battery is then charged according to the MSCC charging protocol, which includes multiple charging steps.

[0090] In some embodiments, the MSCC charging protocols in 1) and 3) above may be the same or different.

[0091] In some embodiments, the MSCC charging protocol includes at least three steps, at least four steps, or at least five steps. Batteries charged via the MSCC charging protocol exhibit improved cycle life. Batteries charged via the MSCC charging protocol exhibit at least 10% longer cycle life than batteries charged via a charging protocol with a single constant current.

[0092] In some embodiments, the MSCC charging protocol sequentially includes a first charging step, a second charging step, and a third charging step, wherein a first constant current (I1), a second constant current (I2), and a third constant current (I3) are used in the first charging step, the second charging step, and the third charging step, respectively, and wherein the constant current in the first step, the second step, and the third step represents I1. <i2>The mode of I3 or I1>I2<I3.

[0093] In some implementations, the MSCC charging protocol sequentially includes a first charging step, a second charging step, a third charging step, and a fourth charging step, where a first constant current (I1), a second constant current (I2), a third constant current (I3), and a fourth constant current (I4) are respectively used in the first charging step, the second charging step, the third charging step, and the fourth charging step, and where the constant currents in the first to fourth steps exhibit I1 <i2>I3 < I4 or I1 > I2 <i3>I4 mode.

[0094] In some embodiments, SOC a and SOC b Each is defined independently by the user or the program.

[0095] In some embodiments, SOC a Values ​​within the following ranges: 5% to 75%, 5% to 70%, 5% to 65%, 5% to 60%, 5% to 55%, 5% to 55%, 5% to 50%, 5% to 45%, 5% to 40%, 5% to 35%, 5% to 30%, 5% to 25%, 5% to 20%, 5% to 15%, or 5% to 10%.

[0096] In some embodiments, SOC b Greater than SOC a In some embodiments, SOC b It has values ​​within the following ranges: from 70% to 95%, from 70% to 90%, from 70% to 85%, from 70% to 80%, or from 70% to 75%.

[0097] Embodiments of the subject matter and functional operation described in this specification may be implemented in digital electronic circuit systems, in tangibly embodied computer software or firmware, in computer hardware including the structures disclosed in this specification and their equivalents, or in a combination of one or more of these. Embodiments of the subject matter described in this specification may be implemented as one or more computer programs (i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier) for execution by a data processing device or for controlling the operation of a data processing device.

[0098] Alternatively or additionally, program instructions may be encoded on artificially generated propagation signals (e.g., machine-generated electrical, optical, or electromagnetic signals) that are generated to encode information for transmission to a suitable receiver device for execution by a data processing device. Computer storage media may be machine-readable storage devices, machine-readable storage substrates, random or serial access memory devices, or combinations thereof.

[0099] The term "computing system" encompasses all kinds of devices, apparatuses, and machines used for processing data, including, by way of example, programmable processors, computers, or multiple processors or computers. The device may include dedicated logic circuit systems such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits). In addition to hardware, the device may also include code that creates the execution environment for the computer program in question, such as code constituting processor firmware, protocol stacks, database management systems, operating systems, or combinations thereof.

[0100] Computer programs (which may also be referred to or described as programs, software, software applications, modules, software modules, scripts, or code) can be written in any form of programming language (including compiled or interpreted languages, or declarative or processing languages), and computer programs can be deployed in any form (including as standalone programs or as modules, components, subroutines, or other units suitable for use in a computing environment).

[0101] A computer program may, but does not necessarily, correspond to a file in a file system. A program may be stored as part of a file that holds other programs or data (e.g., stored in one or more scripts within a markup language document), in a single file dedicated to the program in question, or in multiple coordinating files (e.g., files storing one or more modules, subroutines, or code sections). A computer program may be deployed to execute on a single computer or on multiple computers located in one place or distributed across multiple locations and interconnected via a communication network.

[0102] The processing and logic flow described in this specification can be performed by one or more programmable computers executing one or more computer programs to function by manipulating input data and generating output. The processing and logic flow can also be performed by dedicated logic circuit systems (e.g., FPGAs (Field-Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), or GPGPUs (General-Purpose Graphics Processing Units)), and the device can also be implemented as such a dedicated logic circuit system. Computers suitable for executing computer programs include, for example, those based on general-purpose microprocessors or dedicated microprocessors or both, or any other type of central processing unit.

[0103] Typically, the central processing unit (CPU) receives instructions and data from read-only memory (ROM) or random access memory (RAM), or both. Some components of a computer are the CPU for making or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices (e.g., disks, magneto-optical disks, or optical disks) for storing data, or be operatively coupled to receive data from or transfer data to one or more mass storage devices (e.g., disks, magneto-optical disks, or optical disks), or both. However, a computer does not need to have such devices. Furthermore, a computer may be embedded in another device (e.g., a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few).

[0104] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, including by way of example: semiconductor memory devices, such as EPROM, EEPROM, and flash memory devices; magnetic disks, such as internal hard disks or removable disks; magneto-optical disks; and CD ROMs and DVD-ROMs. The processor and memory may be supplemented by or incorporated into a dedicated logic circuit system.

[0105] To provide interaction with the user, embodiments of the subject matter described in this specification can be implemented on a computer having a display device (such as client device 104, etc.) for displaying information to the user and a keyboard and pointing device (e.g., mouse or trackball) via which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback, such as visual, auditory, or tactile feedback; and input from the user can be received in any form (including acoustic, speech, or tactile input). Additionally, the computer can interact with the user by sending documents to and receiving documents from the device used by the user; for example, by sending web pages to a web browser on the user's client device in response to a request received from a web browser.

[0106] Embodiments of the subject matter described in this specification can be implemented in a computing system that includes back-end components (e.g., as a data server), or middleware components (e.g., an application server), or front-end components (e.g., a client computer with a graphical user interface or a web browser through which a user can interact with an implementation of the subject matter described in this specification), or any combination of one or more such back-end, middleware, or front-end components. The components of the system can be interconnected via digital data communication (e.g., a communication network) of any form or medium. Examples of communication networks include local area networks ("LANs") and wide area networks ("WANs") (e.g., the Internet).

[0107] A computing system may include clients and servers. Clients and servers are typically located far apart and interact via communication networks. The client-server relationship is established by computer programs running on the respective computers and having a client-server relationship with each other.

[0108] This disclosure will be better understood by referring to the following experimental details; however, those skilled in the art will readily understand that the specific experiments detailed are merely illustrative and are not intended to limit the disclosure as described herein, as many variations and modifications of these exemplary embodiments are possible without the need for excessive experimentation. All such variations and modifications are within the scope of the teachings of this disclosure. It will be understood that the foregoing description and the following examples, however detailed they may appear in the text, can be practiced in many ways, and this disclosure should be interpreted in accordance with the appended claims and their equivalents. Example 1

[0109] Input data, including material properties, pseudo-open-circuit voltage (OCV) with C / 50, galvanostatic intermittent titration (GITT) results, and temperature measurements, are collected and fed into a state estimator model, such as a P2D model configured as a synthetic data generator. The state estimator model is calibrated based on calibration data from an experimental dataset including the battery, thereby obtaining a battery-specific calibrated state estimator model. The battery is a 0.75 Ah pouch cell comprising Li metal as the negative electrode, a microporous membrane as the separator, NMC811 as the positive electrode, and a polymer electrolyte as the electrolyte.

[0110] The calibrated state estimator model is then used to generate a synthetic dataset based on the input data and / or preprocessed input data. This synthetic dataset includes one or more states of the battery. The synthetic dataset captures the complex relationship between charging parameters and battery response. The generation of the synthetic dataset can be performed using one or more sampling methods.

[0111] Transducer-based ML models are used as typical surrogate models for generating feature sets based on time-series analysis applied to synthetic datasets, where the feature set includes singular parameters representing one or more internal properties of the battery. In some implementations, one or more internal properties cannot be directly measured by physical testing of the battery, particularly during battery operation. In some implementations, one or more internal properties include, but are not limited to, at least one selected from the group consisting of: internal stress of the battery, core temperature of the battery, electrolyte salt concentration, solvent concentration in the electrolyte, individual resistance of one or more electrodes, lithium concentration in the solid electrode active material, interfacial kinetic exchange current density, temperature gradient in the cell, voltage gradient in the cell, overpotential at any electrode, and electrode utilization. The transducer-based ML model is adapted to one or more states of the battery, which have a range of values ​​consistent with states detected from one or more sensors. In some implementations, the transducer-based ML model is adapted by generating and / or identifying feature sets such that charging protocol parameters are highly consistent with the actual state and condition of the battery.

[0112] Then, a reduced-order ML model is performed using typical multi-objective optimization (MOO) processing to determine the minimum charging time of the battery based on a subset of features derived from one or more optimization control parameters and a feature set. The minimum charging time is determined such that the charging duration is within an acceptable range while keeping the values ​​of one or more optimization control parameters within acceptable ranges (such as keeping internal battery temperature and stress within acceptable ranges (i.e., not exceeding thresholds)). In some implementations, the reduced-order ML model identifies one or more optimization control parameters that affect battery degradation and lifespan. In some implementations, the multi-objective optimization (MOO) model determines the minimum charging time such that the charging duration is an acceptable minimum while one or more optimization control parameters, such as battery degradation, maximum temperature, and maximum stress, are also controlled to an acceptable level.

[0113] Based on the minimum charging time and / or one or more optimized control parameters, one or more charging protocols are then generated. Table 1 summarizes one of these charging protocols. Table 1 MSCC charging protocol for Example 1

[0114]

[0115] The batteries prepared as described above were cycled at 25°C under external pressure between 2.8V and 4.25V. During each cycle, the pouch cells were charged according to the MSCC charging protocol in Table 1, and at 25°C... o Discharge is performed at 1C at point C. In some embodiments, the external pressure is in the range of 0.5 MPa to 5.0 MPa.

[0116] In addition to charging the pouch cells with a single constant current of 2.8C, the pouch cells are similarly fabricated and cycled.

[0117] Figure 7A The voltage curve of a battery charged using a standard protocol is shown, with a single constant current of 2.8C and initial and final SOCs of 10% and 80%, respectively. Figure 7B The voltage curves of batteries charged using the exemplary MSCC charging protocol summarized in Table 1 are shown.

[0118] Cycle life is the number of cycles a battery can complete before reaching a threshold (such as 80% of its original capacity) or a threshold (such as 98.0% of its coulombic efficiency, whichever comes first). Cycle life is commonly used to measure the cycling performance of rechargeable batteries. Figure 8A , Figure 8B As shown in Table 2, the cells charged with the MSCC charging protocol exhibited a cycle life of over 90 cycles, which is approximately 260% longer than the cycle life of the comparative example with a cycle life of 25 cycles. Table 2 Cycle life of cells charged with MSCC and a single constant current.

[0119] All aspects

[0120] In a first aspect, this disclosure provides a computer-implemented method for generating a charging protocol for a battery. The method includes: 1) Using data processing algorithms, the multiple input data of the battery are processed into multiple pre-processed input data; 2) Provide multiple preprocessed input data to the state estimator model configured as a synthetic data generator; 3) Using the state estimator model, a synthetic dataset is generated based on the multiple preprocessed input data, the synthetic dataset including one or more states of the battery; 4) Using a proxy model that operates in conjunction with the optimization engine, a feature set is generated based on time series analysis applied to the synthetic dataset, the feature set including singular parameters representing one or more internal properties of the battery; 5) Using a reduced-order ML model, determine the minimum charging time of the battery based on one or more optimization control parameters and a subset of features derived from the feature set; and 6) Generate one or more charging protocols for the battery, based at least on the minimum charging time and the one or more optimized control parameters.

[0121] In the second aspect according to the first aspect, the method further includes: selecting an optimal charging protocol from the one or more charging protocols.

[0122] In the third aspect according to the first or second aspect, the state estimator model is calibrated using calibration data from a set of experimental data including the battery.

[0123] In the fourth aspect according to the first aspect, the state estimator model is a pseudo-two-dimensional model, i.e., a P2D model, an equivalent circuit model, i.e., an ECM, or a converter-based ML model.

[0124] In the fifth aspect according to the first aspect, when the proxy model operates with the optimization engine, it adapts the feature set and / or the feature subset in consideration of the one or more states, the one or more states having values ​​consistent with or within a range of values ​​consistent with the one or more sets of sensed data retrieved from one or more sensors.

[0125] In the sixth aspect according to the fifth aspect, the one or more states include one or more selected from the group consisting of a healthy state (SOH), a charging state (SOC), a power state (SOP), and an energy state (SOE).

[0126] In the seventh aspect according to the first aspect, the reduced-order ML model takes into account one or more states of the battery and one or more sets of sensing data retrieved from one or more sensors to adapt the minimum charging time and / or the optimized control parameters.

[0127] In the eighth aspect according to the first aspect, the surrogate model generates the feature set based on time series analysis applied to the synthetic dataset and time series analysis applied to non-synthetic data, the non-synthetic data including real experimental data derived from physical tests of the battery. In some embodiments, the surrogate model is pre-trained based on the feature set, such that a subset of features is adapted to the corresponding battery chemistry system of the battery.

[0128] In the ninth aspect according to the eighth aspect, one or more internal properties cannot be directly measured by physical testing of the battery.

[0129] In the tenth aspect according to the ninth aspect, the one or more internal properties include at least one selected from the group consisting of: internal stress of the battery; core temperature of the battery; electrolyte salt concentration; solvent concentration in the electrolyte; individual resistance of one or more electrodes; lithium concentration in the solid electrode active material; interfacial kinetic exchange current density; temperature gradient in the cell; voltage gradient in the cell; overpotential at any electrode; and electrode utilization.

[0130] In the eleventh aspect, the surrogate model is pre-trained using multiple annotated inputs as one or more training datasets, employing a supervised machine learning algorithm, and a single-head attention network or a multi-head attention network. In some embodiments, the surrogate model is a transformer-based machine learning (ML) model. In some embodiments, the surrogate model is a transformer-based machine learning (ML) model trained using a single-head attention network to generate or identify a feature in a feature set and determine the corresponding weight of that feature. In some embodiments, the surrogate model is a transformer-based ML model trained using a multi-head attention network to generate or identify two or more features in a feature set and determine the corresponding weights of each of the two or more features.

[0131] In the twelfth aspect according to the first aspect, the reduced-order ML model generates a subset of feature values ​​by generating a simplified representation of time series data derived from time series analysis applied to the synthetic dataset, and then using the simplified representation of the time series data to perform multi-objective optimization for each feature type.

[0132] In the thirteenth aspect according to the first aspect, the proxy model is part of the digital twin of the battery.

[0133] In the fourteenth aspect according to the thirteenth aspect, the plurality of input data includes a set of data in the digital twin of the battery, and the plurality of input data is updated once the set of data in the digital twin is updated.

[0134] In the fifteenth aspect according to the first aspect, at least one of the one or more charging protocols is a multi-step constant current charging protocol, i.e., an MSCC charging protocol. In some embodiments, the MSCC charging protocol sequentially includes a first charging step, a second charging step, and a third charging step, wherein a first constant current (I1), a second constant current (I2), and a third constant current (I3) are used in the first charging step, the second charging step, and the third charging step, respectively, and wherein the constant current in the first step, the second step, and the third step exhibits I1. <i2>The mode of I3 or I1>I2<I3. In some embodiments, the MSCC charging protocol sequentially includes a first charging step, a second charging step, a third charging step, and a fourth charging step, where a first constant current (I1), a second constant current (I2), a third constant current (I3), and a fourth constant current (I4) are used in the first charging step, the second charging step, the third charging step, and the fourth charging step, respectively, and where the constant currents in the first step to the fourth step exhibit I1 <i2>I3 < I4 or I1 > I2 <i3>I4 mode.

[0135] In a sixteenth aspect, this disclosure provides a method for charging a battery, comprising: 1) Using data processing algorithms, the multiple input data of the battery are processed into multiple pre-processed input data; 2) Provide multiple preprocessed input data to the state estimator model configured as a synthetic data generator; 3) Using the state estimator model, a synthetic dataset is generated based on the multiple preprocessed input data, the synthetic dataset including one or more states of the battery; 4) Using a proxy model that operates in conjunction with the optimization engine, a feature set is generated based on time series analysis applied to the synthetic dataset, the feature set including singular parameters representing one or more internal properties of the battery; 5) Using a reduced-order ML model, determine the minimum charging time of the battery based on one or more optimization control parameters and a subset of features derived from the feature set; 6) Generate one or more charging protocols for the battery, based at least on the minimum charging time and the one or more optimized control parameters; 7) Select the optimal charging protocol from the one or more charging protocols; and 8) Charge the battery according to the optimal charging protocol.

[0136] In its seventeenth aspect, this disclosure provides a system comprising: Processing device; and A non-transitory machine-readable storage medium storing instructions that can be executed by the processing device to perform operations, including: 1) Using data processing algorithms, the multiple input data of the battery are processed into multiple pre-processed input data; 2) Provide multiple preprocessed input data to the state estimator model configured as a synthetic data generator; 3) Using the state estimator model, a synthetic dataset is generated based on the multiple preprocessed input data, the synthetic dataset including one or more states of the battery; 4) Using a proxy model that operates in conjunction with the optimization engine, a feature set is generated based on time series analysis applied to the synthetic dataset, the feature set including singular parameters representing one or more internal properties of the battery; 5) Using a reduced-order ML model, determine the minimum charging time of the battery based on one or more optimization control parameters and a subset of features derived from the feature set; and 6) Generate one or more charging protocols for the battery, based at least on the minimum charging time and the one or more optimized control parameters.

[0137] In the eighteenth aspect according to the seventeenth aspect, the surrogate model generates the feature set based on time series analysis applied to the synthetic dataset and time series analysis applied to non-synthetic data, the non-synthetic data including real experimental data derived from physical tests of the battery.

[0138] In a nineteenth aspect, this disclosure provides a non-transitory machine-readable storage device for storing instructions, the instructions being executable by a processing means to cause operations including: 1) Using data processing algorithms, the multiple input data of the battery are processed into multiple pre-processed input data; 2) Provide multiple preprocessed input data to the state estimator model configured as a synthetic data generator; 3) Using the state estimator model, a synthetic dataset is generated based on the multiple preprocessed input data, the synthetic dataset including one or more states of the battery; 4) Using a proxy model that operates in conjunction with the optimization engine, a feature set is generated based on time series analysis applied to the synthetic dataset, the feature set including singular parameters representing one or more internal properties of the battery; 5) Using a reduced-order ML model, determine the minimum charging time of the battery based on one or more optimization control parameters and a subset of features derived from the feature set; 6) Generate one or more charging protocols for the battery, based at least on the minimum charging time and the one or more optimized control parameters; and 7) Select the optimal charging protocol from the one or more charging protocols.

[0139] In a twentieth aspect, this disclosure provides a vehicle that includes a non-transitory machine-readable storage device as described in the nineteenth aspect.

[0140] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope that may be claimed, but rather as descriptions of features specific to particular embodiments of a particular invention. Certain features described in the specification within the context of separate embodiments may also be implemented in combination in a single embodiment. On the other hand, various features described in the context of a single embodiment may also be implemented separately or in any suitable sub-combination in multiple embodiments. Furthermore, although features may be described above as functioning in certain combinations and even initially claimed in this way, in some cases, one or more features from the claimed combination may be removed from the combination, and the claimed combination may involve sub-combinations or variations of sub-combinations.

[0141] All transitional phrases (such as "including," "contains," "carries," "has," "includes," "involves," "holds," "composes of," etc.) should be understood as open-ended, meaning including but not limited to. Only the transitional phrases "composes of" and "consistently composed of" should be closed or semi-closed transitional phrases, respectively. Phrases such as "based on," "at least based on," "on the basis of," "considering," "taking into account," "taking into account," etc., should be understood as open-ended, meaning considering at least one factor, including but not limited to that at least one factor.

[0142] Although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order, or to perform all illustrated operations to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of the various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0143] This document describes various embodiments of the features of this disclosure. However, it should be understood that such embodiments are provided merely as examples, and many variations, modifications, and substitutions will occur to those skilled in the art without departing from the scope of this disclosure. It should also be understood that various alternatives to the specific embodiments described herein are also within the scope of this disclosure.

Claims

1. A computer-implemented method for generating a battery charging protocol, comprising: 1) Using data processing algorithms, the multiple input data of the battery are processed into multiple pre-processed input data; 2) Provide multiple preprocessed input data to the state estimator model configured as a synthetic data generator; 3) Using the state estimator model, a synthetic dataset is generated based on the multiple preprocessed input data, the synthetic dataset including one or more states of the battery; 4) Using a proxy model that operates in conjunction with the optimization engine, a feature set is generated based on time series analysis applied to the synthetic dataset, the feature set including singular parameters representing one or more internal properties of the battery; 5) Using a reduced-order ML model, determine the minimum charging time of the battery based on one or more optimization control parameters and a subset of features derived from the feature set; as well as 6) Generate one or more charging protocols for the battery, based at least on the minimum charging time and the one or more optimized control parameters.

2. The computer-implemented method according to claim 1 further includes: Select the optimal charging protocol from one or more charging protocols.

3. The computer-implemented method according to claim 1, wherein, The state estimator model is calibrated using calibration data from a set of experimental data including the battery.

4. The computer-implemented method according to claim 1, wherein, The state estimator model is a pseudo-two-dimensional model, i.e., a P2D model, or a transformer-based ML model.

5. The computer-implemented method according to claim 1, wherein, When the proxy model operates in conjunction with the optimization engine, it adapts the feature set and / or the feature subset by taking into account one or more states having values ​​consistent with or within a range of values ​​consistent with one or more sets of sensed data retrieved from one or more sensors.

6. The computer-implemented method according to claim 5, wherein, The one or more states include one or more selected from the group consisting of the state of health (SOH), the state of charge (SOC), the state of power (SOP), and the state of energy (SOE).

7. The computer-implemented method according to claim 1, wherein, The reduced-order ML model takes into account one or more states of the battery and one or more sets of sensing data retrieved from one or more sensors to adapt the minimum charging time and / or the optimized control parameters.

8. The computer-implemented method according to claim 1, wherein, The proxy model generates the feature set based on time series analysis applied to the synthetic dataset and time series analysis applied to non-synthetic data, the non-synthetic data including real experimental data derived from physical tests of the battery.

9. The computer-implemented method according to claim 1, wherein, One or more of the internal properties cannot be directly measured by physical testing of the battery.

10. The computer-implemented method according to claim 9, wherein, The one or more internal attributes include at least one selected from the group consisting of: The internal stress of the battery; The core temperature of the battery; The electrolyte salt concentration of the battery; The solvent concentration in the electrolyte of the battery; The individual resistance of one or more electrodes of the battery; Lithium concentration in solid electrode active materials; Interfacial dynamics exchange current density; Temperature gradient within the battery cell; Voltage gradient in the battery cell; Overpotential at any electrode; and Electrode utilization rate.

11. The computer-implemented method according to claim 11, wherein, The proxy model is pre-trained using multiple annotated inputs as one or more training datasets, and employing supervised machine learning algorithms and single-head or multi-head attention networks.

12. The computer-implemented method according to claim 1, wherein, The reduced-order ML model generates a subset of feature values ​​by generating a simplified representation of the time series data derived from time series analysis applied to the synthetic dataset, and then using the simplified representation of the time series data to perform multi-objective optimization for each feature type.

13. The computer-implemented method according to claim 1, wherein, The proxy model is part of the digital twin of the battery.

14. The computer-implemented method according to claim 13, wherein, The plurality of input data includes a set of data in the digital twin of the battery, and the plurality of input data is updated once the set of data in the digital twin is updated.

15. The computer-implemented method according to claim 1, wherein, At least one of the one or more charging protocols is a multi-step constant current charging protocol, i.e., the MSCC charging protocol.

16. A method for charging a battery, comprising: 1) Using data processing algorithms, the multiple input data of the battery are processed into multiple pre-processed input data; 2) Provide multiple preprocessed input data to the state estimator model configured as a synthetic data generator; 3) Using the state estimator model, a synthetic dataset is generated based on the multiple preprocessed input data, the synthetic dataset including one or more states of the battery; 4) Using a proxy model that operates in conjunction with the optimization engine, a feature set is generated based on time series analysis applied to the synthetic dataset, the feature set including singular parameters representing one or more internal properties of the battery; 5) Using a reduced-order ML model, determine the minimum charging time of the battery based on one or more optimization control parameters and a subset of features derived from the feature set; 6) Generate one or more charging protocols for the battery, based at least on the minimum charging time and the one or more optimized control parameters; 7) Select the optimal charging protocol from the one or more charging protocols; as well as 8) Charge the battery according to the optimal charging protocol.

17. A system comprising: Processing device; as well as A non-transitory machine-readable storage medium storing instructions that can be executed by the processing device to perform operations, including: 1) Using data processing algorithms, the multiple input data of the battery are processed into multiple pre-processed input data; 2) Provide multiple preprocessed input data to the state estimator model configured as a synthetic data generator; 3) Using the state estimator model, a synthetic dataset is generated based on the multiple preprocessed input data, the synthetic dataset including one or more states of the battery; 4) Using a proxy model that operates in conjunction with the optimization engine, a feature set is generated based on time series analysis applied to the synthetic dataset, the feature set including singular parameters representing one or more internal properties of the battery; 5) Using a reduced-order ML model, determine the minimum charging time of the battery based on one or more optimization control parameters and a subset of features derived from the feature set; and 6) Generate one or more charging protocols for the battery, based at least on the minimum charging time and the one or more optimized control parameters.

18. The system according to claim 17, wherein, The surrogate model generates the feature set based on time series analysis applied to the synthetic dataset and time series analysis applied to non-synthetic data, the non-synthetic data including real experimental data values ​​derived from physical tests of the battery.

19. A non-transitory machine-readable storage device for storing instructions, the instructions being executable by a processing means to cause operations, the operations including: 1) Using data processing algorithms, the multiple input data of the battery are processed into multiple pre-processed input data; 2) Provide multiple preprocessed input data to the state estimator model configured as a synthetic data generator; 3) Using the state estimator model, a synthetic dataset is generated based on the multiple preprocessed input data, the synthetic dataset including one or more states of the battery; 4) Using a proxy model that operates in conjunction with the optimization engine, a feature set is generated based on time series analysis applied to the synthetic dataset, the feature set including singular parameters representing one or more internal properties of the battery; 5) Using a reduced-order ML model, determine the minimum charging time of the battery based on one or more optimization control parameters and a subset of features derived from the feature set; 6) Generate one or more charging protocols for the battery, based at least on the minimum charging time and the one or more optimized control parameters; as well as 7) Select the optimal charging protocol from the one or more charging protocols.

20. A vehicle comprising a non-transitory machine-readable storage device according to claim 19.

Images