All-electric aircraft lithium battery SOC estimation method based on multi-field coupling physical information neural network

Abstract

The invention discloses an all-electric aircraft lithium battery SOC estimation method and device based on a multi-field coupling physical information neural network, a medium and equipment, and the method comprises the steps: introducing a temperature field and a mechanical deformation mechanical field into a single-particle model of an all-electric aircraft lithium battery, and obtaining a lithium battery SPM model; constructing a neural network, and taking the calibrated multi-field coupled all-electric aircraft lithium battery SPM model as a physical constraint of the neural network to obtain a multi-field coupled physical information neural network; constructing a joint loss function by using the sum of the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model, and training the multi-field coupling physical information neural network based on the joint loss function to obtain a trained physical information neural network; the SOC of the all-electric aircraft lithium battery is estimated based on the trained physical information neural network, the predicted SOC is obtained, and accurate estimation and prediction of the SOC of the all-electric aircraft lithium battery under the complex working condition can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of aircraft electronic equipment and flight control systems, and more specifically to a method, device, medium and equipment for estimating the SOC of a lithium battery in an all-electric aircraft based on a multi-field coupled physical information neural network. Background Art

[0002] Estimating the state of charge (SOC) of an all-electric aircraft's battery system is an essential and crucial monitoring indicator for flight control. It plays a key role in monitoring battery status and performance and ensuring the completion of scheduled flight missions. Accurate SOC estimation directly impacts an all-electric aircraft's flight time, range, payload, and flight safety. In particular, during actual flight, the accuracy of SOC estimation significantly influences flight strategy, ultimately impacting the aircraft's flight performance, economy, and safety. Therefore, developing a high-precision SOC estimation method is crucial for improving the overall performance and operational reliability of all-electric aircraft.

[0003] Data-driven methods (such as those based on machine learning and deep learning) directly estimate SOC by training models using battery data such as voltage, current, and temperature. These methods rely on large-scale, high-quality experimental data and lack physical constraints, resulting in insufficient generalization capabilities under complex operating conditions. This is particularly true in the high and low temperature environments and extreme mechanical deformations of all-electric aircraft, making prediction accuracy difficult to guarantee.

[0004] SOC estimation methods based on electrochemical models demonstrate good physical consistency by describing the electrochemical reaction processes within the battery. However, these methods often ignore the effects of temperature fluctuations and mechanical deformation on the battery during actual operation and fail to fully reflect the battery's true state. This is particularly true in all-electric aircraft applications, where the accuracy of SOC estimation is significantly reduced due to the influence of multiple field coupling factors such as vibration, shock, and thermal environment changes. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, in a first aspect, the present invention proposes a method for estimating the SOC of an all-electric aircraft lithium battery based on a multi-field coupled physical information neural network, comprising: introducing a temperature field and a mechanical deformation mechanics field into a single particle model of the all-electric aircraft lithium battery to obtain a multi-field coupled SPM model of the all-electric aircraft lithium battery; constructing a neural network, using the calibrated multi-field coupled SPM model of the all-electric aircraft lithium battery as a physical constraint of the neural network, to obtain a multi-field coupled physical information neural network; constructing a joint loss function based on the sum of the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model; training the multi-field coupled physical information neural network based on the joint loss function to obtain a trained physical information neural network; and estimating the SOC of the all-electric aircraft lithium battery based on the trained physical information neural network to obtain a predicted SOC.

[0006] Optionally, before introducing the temperature field and the mechanical deformation mechanics field into the single particle model of the all-electric aircraft lithium battery to obtain the multi-field coupled SPM model of the all-electric aircraft lithium battery, the method further includes: simulating the temperature field of the all-electric aircraft lithium battery by dynamically adjusting the reaction rate constant; and simulating the mechanical deformation mechanics field of the all-electric aircraft lithium battery by adjusting the electrode thickness formula.

[0007] Optionally, before using the calibrated multi-field coupled all-electric aircraft lithium battery SPM model as a physical constraint of the neural network to obtain the multi-field coupled physical information neural network, the method further includes: optimizing the mechanical deformation influencing factor of the multi-field coupled all-electric aircraft lithium battery SPM model by a genetic algorithm to obtain the calibrated multi-field coupled all-electric aircraft lithium battery SPM model.

[0008] Optionally, the method of optimizing the mechanical deformation influencing factors of the multi-field coupled all-electric aircraft lithium battery SPM model by a genetic algorithm to obtain a calibrated multi-field coupled all-electric aircraft lithium battery SPM model includes: initializing the population size of the genetic algorithm and setting the parameter range of the genetic algorithm; constructing an objective function of the genetic algorithm based on minimizing the error between the voltage curve of experimental data and the voltage obtained by simulating the multi-field coupled all-electric aircraft lithium battery SPM model, and constructing the fitness of the genetic algorithm as the inverse of the objective function; based on the objective function and the fitness, utilizing the selection, crossover, and mutation operations of the genetic algorithm to iteratively optimize the deformation influencing factors of the positive and negative electrodes of the all-electric aircraft lithium battery to obtain a calibrated multi-field coupled all-electric aircraft lithium battery SPM model.

[0009] Optionally, the training of the multi-field coupled physical information neural network based on the joint loss function includes: inputting a voltage time series and a current time series into the multi-field coupled physical information neural network, and outputting a predicted SOC value of the all-electric aircraft lithium battery; calculating the negative electrode surface concentration of the all-electric aircraft lithium battery according to the predicted SOC value; inputting the negative electrode surface concentration into a calibrated multi-field coupled all-electric aircraft lithium battery SPM model to obtain a calculated voltage of the all-electric aircraft lithium battery; calculating the SOC mean square error based on the predicted SOC value, calculating the voltage mean square error based on the calculated value, and constructing a joint loss function based on the SOC mean square error and the voltage mean square error; and training the multi-field coupled physical information neural network with the joint loss function as the objective function, and using an Adam optimizer and a preset learning rate.

[0010] Optionally, the calculating of the SOC mean square error based on the predicted SOC value, the calculating of the voltage mean square error, and the constructing of a joint loss function based on the SOC mean square error and the voltage mean square error include: calculating the SOC mean square error of the neural network based on the predicted SOC value and the experimental SOC value; calculating the voltage mean square error of the calibrated multi-field coupled all-electric aircraft lithium battery SPM model based on the calculated voltage value and the experimental voltage value; and constructing a joint loss function based on the weighted SOC mean square error and the voltage mean square error.

[0011] Optionally, the neural network is a long short-term memory network, which includes: an input layer, multiple LSTM hidden layers and a fully connected output layer; wherein each hidden layer contains 50 to 100 LSTM units, and the output layer includes a normalization layer.

[0012] In order to achieve the above-mentioned objectives, the present application also provides an all-electric aircraft lithium battery SOC estimation device based on a multi-field coupled physical information neural network, including: a first model determination module, used to introduce a temperature field and a mechanical deformation mechanics field into a single particle model of an all-electric aircraft lithium battery to obtain a multi-field coupled SPM model of an all-electric aircraft lithium battery; a second model determination module, used to construct a neural network, using the calibrated multi-field coupled SPM model of an all-electric aircraft lithium battery as a physical constraint of the neural network to obtain a multi-field coupled physical information neural network; a joint loss function is constructed using the sum of the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model, and the multi-field coupled physical information neural network is trained based on the joint loss function to obtain a trained physical information neural network; a prediction module, used to estimate the SOC of the all-electric aircraft lithium battery based on the trained physical information neural network to obtain a predicted SOC.

[0013] In order to achieve the above-mentioned objectives, the present application also provides an electronic device, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the aforementioned all-electric aircraft lithium battery SOC estimation method based on a multi-field coupled physical information neural network.

[0014] To achieve the above-mentioned objectives, the present application also provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, code set, or instruction set, and the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the aforementioned all-electric aircraft lithium battery SOC estimation method based on a multi-field coupled physical information neural network.

[0015] The embodiments of the present invention provide a method, apparatus, medium, and device for estimating the SOC of an all-electric aircraft lithium battery based on a multi-field coupled physical information neural network. Compared with the prior art, the method has the following advantages: a temperature field and a mechanical deformation mechanics field are introduced into a single particle model of an all-electric aircraft lithium battery to obtain a multi-field coupled SPM model of the all-electric aircraft lithium battery; a neural network is constructed, and the calibrated multi-field coupled SPM model of the all-electric aircraft lithium battery is used as a physical constraint of the neural network to obtain a multi-field coupled physical information neural network; a joint loss function is constructed using the sum of the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model, and the multi-field coupled SPM model is trained based on the joint loss function. The physical information neural network is combined to obtain a trained physical information neural network; the SOC of the all-electric aircraft lithium battery is estimated based on the trained physical information neural network to obtain the predicted SOC. The multi-field coupled SPM model can accurately describe the actual charge and discharge characteristics of the all-electric aircraft lithium battery. The fusion of the neural network and the physical model improves the interpretability of the model. The introduction of physical constraints ensures that the output of the model conforms to the physical laws and has high credibility and accuracy. This application has both physical consistency and practicality. It is a novel online monitoring method for all-electric aircraft lithium batteries, which can provide a reliable theoretical basis and technical support for the design of all-electric aircraft lithium battery management systems under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0017] Figure 1 A flow chart of a method for estimating the SOC of a lithium battery in an all-electric aircraft based on a multi-field coupled physical information neural network according to an embodiment of the present invention;

[0018] Figure 2 A schematic diagram showing the comparison results of the terminal voltages of a multi-field coupled SPM model of an all-electric aircraft lithium battery SOC estimation method based on a multi-field coupled physical information neural network provided by an embodiment of the present invention;

[0019] Figure 3 A schematic diagram of SOC estimation results of a method for estimating SOC of lithium batteries in an all-electric aircraft based on a multi-field coupled physical information neural network according to an embodiment of the present invention;

[0020] Figure 4 Schematic diagram of SOC estimation error in the SOC estimation method for lithium batteries of all-electric aircraft based on a multi-field coupled physical information neural network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0022] This specification provides method operation steps as described in the embodiments or flowcharts, but more or fewer operation steps may be included based on routine or non-creative work. When implemented in an actual system or server product, the methods shown in the embodiments or figures may be executed sequentially or in parallel (for example, in a parallel processor or multi-threaded processing environment).

[0023] In the first prior art, an improved lithium-ion battery SOC estimation method is proposed. Using the first-order RC model as the basis, the open circuit voltage module of the model is optimized by introducing an error term reflecting the solid-phase diffusion process, thereby reducing the error between the equivalent circuit model and the mechanism model while keeping the computational complexity low. The particle swarm algorithm is used to decouple and identify the model parameters, which improves the identification accuracy and reduces the complexity. Subsequently, the SOC estimation algorithm is improved by combining the unscented Kalman filter (UKF) with the weighted sliding window technology. The high accuracy and robustness of the algorithm are verified under dynamic working conditions. The estimation error is significantly lower than that of the traditional method, and even if there is a large deviation in the initial SOC value, the estimation result can still converge quickly. However, all its tests are carried out at room temperature, and the influence of temperature changes on parameter identification and state estimation is not considered.

[0024] In the second prior art, a method for estimating the state of charge (SOC) of lithium batteries for all-electric aircraft based on an improved electrochemical model is proposed to solve the problem that the SOC of lithium batteries for all-electric aircraft is difficult to accurately estimate under wide temperature environments and complex working conditions. The study combines the finite difference method to reduce the order of the solid and liquid phase equations of the P2D model, and combines the equivalent circuit model with two RC network structures to characterize the polarization process inside the battery. The square root cubic Kalman filter (SRCKF) algorithm is introduced to deal with the uncertainty and noise interference caused by model simplification. Through experimental verification, this method achieves high-precision SOC estimation in a wide temperature range of -20°C to 60°C, with a maximum error of less than 1.6%, and shows strong adaptability and fast convergence ability under complex working conditions. However, the coupling effect in multi-physical fields has not yet been considered.

[0025] Existing methods generally have the following problems in the SOC estimation process: traditional electrochemical models do not introduce the coupling effects of temperature changes and mechanical deformation on battery performance, lack multi-field coupling considerations, and cannot truly reflect the actual characteristics of batteries under complex working conditions; although data-driven methods have good fitting capabilities, due to the lack of physical constraints, the prediction results are not reliable enough in practical applications; in addition, existing SOC estimation methods have poor applicability in all-electric aircraft applications, and it is difficult to meet the strict requirements of battery management systems for high precision and high reliability, affecting the flight performance and safety of all-electric aircraft.

[0026] Furthermore, under high-voltage conditions, all-electric aircraft lithium batteries inevitably deform during operation, which has a certain impact on the change in SOC. However, few current state-of-charge estimation methods take the influence of deformation into account.

[0027] To solve the above problems, the present invention proposes a SOC estimation method for lithium batteries of all-electric aircraft based on a multi-field coupled physical information neural network. By combining the multi-field coupled single particle model (SPM) with a data-driven neural network, the SOC estimation accuracy is optimized on the basis of physical constraints, and the influence of temperature and deformation under complex working conditions is adapted to improve the reliability and practicality of the battery management system of all-electric aircraft. Specifically, the purpose of the present invention is to improve the following aspects: The existing SOC estimation method has poor adaptability under different working conditions. The commonly used SOC estimation method is not accurate enough and cannot meet the requirements of the aviation environment. The existing SOC estimation method has a slow response speed when facing changes in working conditions, cannot quickly adapt to new operating states or changes in the external environment, and is difficult to meet the needs of dynamic adjustment. The existing SOC estimation method does not fully consider the complexity and variability of the environment in which the onboard battery is located, resulting in deviations between the estimation results and the actual situation.

[0028] To address the existing challenges of SOC estimation, such as poor versatility, insufficient accuracy, slow response, and a failure to consider the effects of complex and changing environments, this paper proposes a SOC estimation method for lithium batteries in all-electric aircraft based on a multi-field coupled physical information neural network. This method combines a multi-field coupled single particle model (SPM) with a data-driven neural network to ensure high accuracy and physical consistency in SOC estimation, making it suitable for battery management systems in all-electric aircraft operating under complex conditions. Specifically, the present invention incorporates the effects of temperature and mechanical deformation on the traditional SPM model, dynamically adjusting battery reaction rates and electrode thickness to simulate the actual operating state of batteries under complex conditions. The neural network uses time, voltage, and current as inputs to predict SOC values. By constructing a physical information neural network (PINN) framework, the neural network prediction results are integrated with the multi-field coupled SPM model, and a joint loss function is defined for optimization, achieving an organic integration of data-driven and physical models. The experimental results are highly reliable and can provide technical support for future battery management systems.

[0029] The following further illustrates the SOC estimation method for lithium batteries of all-electric aircraft based on a multi-field coupled physical information neural network with reference to the accompanying drawings and embodiments.

[0030] refer to Figure 1 , Figure 1 This is a flow chart of a method for fusing point cloud and image features based on bird's-eye view hybrid coding according to an embodiment of the present application. This method can be executed by a processor in an aircraft's battery management chip. A method for estimating the SOC of lithium batteries in an all-electric aircraft based on a multi-field coupled physical information neural network can include the following execution process:

[0031] S10. Introducing the temperature field and mechanical deformation mechanics field into the single particle model of the all-electric aircraft lithium battery to obtain a multi-field coupled SPM model of the all-electric aircraft lithium battery;

[0032] In one embodiment of the present application, before introducing the temperature field and the mechanical deformation mechanics field into the single particle model of the all-electric aircraft lithium battery to obtain a multi-field coupled SPM model of the all-electric aircraft lithium battery, the method further includes:

[0033] Simulating the temperature field of lithium batteries for all-electric aircraft by dynamically adjusting the reaction rate constant.

[0034] The mechanical deformation field of lithium batteries for all-electric aircraft is simulated through the electrode thickness adjustment formula.

[0035] It should be noted that the SPM modeling process for all-electric aircraft lithium batteries with multi-field coupling requires first establishing the basic framework of the single particle model (SPM). Secondly, SPM modeling of temperature field coupling and coupled modeling of mechanical deformation fields are required. Finally, parameter identification of deformation influencing factors is required.

[0036] The single particle model (SPM) is a classic approach for modeling lithium-ion batteries. Its core is to simulate the electrochemical behavior and terminal voltage response of batteries by describing the lithium ion concentration distribution and diffusion process within electrode particles. The following is a detailed description of the SPM framework:

[0037] Solid-phase diffusion equation inside particles. The single-particle model assumes that the electrode is composed of a large number of spherical particles. The lithium ions in the particles are transported by solid-phase diffusion, and their concentration changes are described by the following diffusion equation:

[0038]

[0039] Among them, c s (r,t) is the solid phase lithium ion concentration in the particle at radius r and time t, D s is the solid phase diffusion coefficient, which indicates the diffusion rate of lithium ions in the particles, r is the radial position of the particles, and t is the reaction time.

[0040] Initial conditions

[0041] c s (r,0)=c s,initial

[0042] That is, at t = 0, the lithium ion concentration in the particle is the initial concentration c s,initial .

[0043] Boundary conditions

[0044] Centrosymmetric boundary of the particle:

[0045]

[0046] This means that the gradient of lithium ion concentration at the center of the particle is zero, which is consistent with the spherical symmetry characteristics.

[0047] Interfacial flux between the particle surface and the liquid phase:

[0048]

[0049] This formula indicates that the lithium ion flux on the particle surface is related to the electrochemical reaction, where j is the surface reaction current density, n is the number of electron transfers in the reaction, F is the Faraday constant, and R is the particle radius.

[0050] The relationship between the lithium ion concentration on the particle surface and the reaction current, the lithium ion concentration on the particle surface c s,surf =c s (R,t) is the key parameter of the lithium ion deintercalation reaction, and the reaction current density j is expressed by the following formula:

[0051]

[0052] Among them, i0 is the exchange current density, which represents the reaction rate, α a and α c The transfer coefficients of the anode and cathode. η is the overpotential, calculated using the following formula:

[0053] η=φ s -φ e -U(c s,surf )

[0054] Among them, φ s is the solid phase potential, φ e is the liquid phase potential, U(c s,surf ) is the open circuit voltage on the particle surface.

[0055] Calculation of battery terminal voltage. The battery terminal voltage is determined by the surface electrochemical reaction of the positive and negative electrodes and the internal resistance of the battery. The formula is:

[0056]

[0057] Where V is the battery terminal voltage. and are the solid phase potentials of the positive and negative electrodes respectively. ohm is the ohmic voltage drop, calculated by the following formula:

[0058] Δφ ohm =IR ohm

[0059] Where I is the battery current, R ohm is the total internal resistance of the battery.

[0060] The single particle model assumes uniformity in the distribution of lithium ion concentration and potential. Through the layer-by-layer correlation of internal particle diffusion, surface concentration, open circuit voltage and terminal voltage, the SPM can fully describe the electrochemical reaction process of the lithium battery of an all-electric aircraft.

[0061] Temperature field coupled SPM modeling, temperature changes will affect the electrochemical reaction rate k r , which in turn affects the discharge characteristics of the battery. The reaction rate can be expressed as an exponential function of temperature:

[0062]

[0063] Where k0 is the pre-exponential factor, E a is the activation energy of the reaction, R q is the gas constant and T is the temperature.

[0064] The increase in temperature will increase the diffusion coefficient of lithium ions, thereby increasing the diffusion rate of ions. The diffusion coefficient of lithium ions D(T) can also be expressed by the Arrhenius equation:

[0065]

[0066] Where D0 is the diffusion coefficient constant, E d is the activation energy of diffusion.

[0067] Coupling modeling of mechanical deformation and mechanical field: When the battery is mechanically deformed, the electrode thickness will change. The dynamic correction formula is as follows:

[0068] L new =L original +k·Δd

[0069] L new is the electrode thickness after deformation adjustment, L original is the initial electrode thickness, k is the deformation influencing factor (determined by optimization), and Δd is the deformation.

[0070] In one embodiment of the present application, before using the calibrated multi-field coupled all-electric aircraft lithium battery SPM model as a physical constraint of the neural network to obtain the multi-field coupled physical information neural network, the method further includes:

[0071] The mechanical deformation influencing factors of the multi-field coupled all-electric aircraft lithium battery SPM model are optimized by genetic algorithm, and the calibrated multi-field coupled all-electric aircraft lithium battery SPM model is obtained.

[0072] It is worth noting that the genetic algorithm (GA) is used to optimize the parameter k that affects the thickness deformation of the positive and negative electrodes on the battery model. pos and k neg By minimizing the voltage error between the simulation results and the experimental data, the genetic algorithm automatically adjusts the parameter k pos and k neg , to achieve accurate calibration of the battery model.

[0073] Specifically, the mechanical deformation influencing factors of the multi-field coupled all-electric aircraft lithium battery SPM model are optimized by genetic algorithm to obtain a calibrated multi-field coupled all-electric aircraft lithium battery SPM model, including:

[0074] Initialize the genetic algorithm population and set the parameter range.

[0075] It is understandable that before initializing the population, it is necessary to clarify the possible value range of each optimization parameter, set its reasonable upper and lower limits in advance, and randomly generate values ​​of the positive and negative electrode thicknesses to ensure that the parameters of each individual are within a reasonable range.

[0076] The objective function of the genetic algorithm is constructed by minimizing the error between the voltage curve of the experimental data and the voltage obtained by simulating the multi-field coupled all-electric aircraft lithium battery SPM model, and the fitness of the genetic algorithm is constructed as the inverse of the objective function.

[0077] The objective function measures the prediction accuracy of the battery model. In this invention, the battery model is optimized by minimizing the error between the voltage curve of the experimental data and the voltage obtained by model simulation. The objective function is as follows:

[0078]

[0079] Where N is the total number of data points, V exp (t) is the experimentally measured battery voltage, V model (t) is the battery voltage calculated based on the battery model using the parameters.

[0080] The goal is to find a set of parameters k pos and k neg , so that the MSE is minimized, that is, the error between the model voltage and the experimental voltage is minimized.

[0081] The purpose of the selection operation is to select outstanding individuals from the current population to generate the next generation. Common selection strategies include roulette wheel selection and tournament selection. The probability of selection is usually proportional to the individual's fitness (objective function value). The individual's fitness can be defined as the inverse of the objective function, that is, the smaller the MSE, the higher the fitness:

[0082]

[0083] Based on the objective function and fitness, the selection, crossover and mutation operations of the genetic algorithm are used to iteratively optimize the influencing factors of the deformation of the positive and negative electrodes of the all-electric aircraft lithium battery, and a calibrated multi-field coupled SPM model of the all-electric aircraft lithium battery is obtained.

[0084] The crossover operation simulates genetic recombination, mixing the genes of two parent individuals to produce a progeny individual. The crossover operation produces new individuals in the population. The crossover probability is usually a pre-set value, usually between 60% and 90%. The present invention uses single-point crossover.

[0085] The mutation operation simulates gene mutation and explores new solution space by randomly adjusting certain genes of individuals, which can increase the diversity of the population and prevent the algorithm from falling into the local optimal solution. pos and kneg ,The mutation operation is a small range of random adjustments to the parameter values.

[0086] Through selection, crossover and mutation operations, a new offspring population is generated. Then, the population is updated, usually by retaining some excellent parent individuals and replacing those with poor performance. After multiple generations of evolution, the individuals in the population gradually approach the optimal solution, that is, the electrode thickness parameter k pos and k neg The optimal value of .

[0087] Termination condition: The iterative process continues until certain termination conditions are met. In this invention, the termination condition is when the objective function value (MSE) reaches a predetermined threshold. Experimental verification shows that the SPM model under multi-field coupling conditions can accurately describe the charging and discharging behavior of lithium batteries in all-electric aircraft. Figure 2 shown.

[0088] from Figure 2 The simulated terminal voltage output by the multi-field coupled SPM model at different deformation levels is close to the experimental value, accurately describing the battery's behavior. The root mean square percentage error (RMS) remains within 0.24%, demonstrating high accuracy. The following table shows the estimated error indicators at different deformation levels.

[0089] Table 1 Evaluation index of estimation error of multi-field coupled SPM model

[0090]

[0091] S20. Construct a neural network, using the calibrated multi-field coupled all-electric aircraft lithium battery SPM model as a physical constraint of the neural network to obtain a multi-field coupled physical information neural network. Construct a joint loss function based on the sum of the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model. Train the multi-field coupled physical information neural network based on the joint loss function to obtain a trained physical information neural network.

[0092] Specifically, in order to achieve an organic combination of data-driven methods and physical models, the present invention designs a physics-informed neural network (PINN) framework, which combines deep learning-based neural network prediction with multi-field coupled SPM (single particle model) to optimize the SOC estimation accuracy through physical constraints.

[0093] In the Physical Information Neural Network (PINN) framework, a neural network model is responsible for learning the nonlinear mapping relationship of SOC from experimental data. The LSTM network (Long Short-Term Memory Network) can process time series data and is suitable for modeling the dynamic relationship between battery voltage, current, and time. The input of the neural network model consists of voltage and current time series. These features are processed through a sliding window to form a fixed-length input sequence used to predict the current SOC value. The output is the battery SOC, which is limited to a range between 0 and 1 to reflect the battery's state of charge.

[0094] The LSTM network consists of an input layer, multiple LSTM hidden layers, and a fully connected output layer. The number of LSTM units in the hidden layer is typically set to 50 to 100 to capture the complex dynamic characteristics of battery operation. The output layer uses a sigmoid activation function to constrain network predictions to a reasonable range. Before training, input features are normalized to the range [0, 1] to ensure a consistent data distribution and facilitate network learning. To ensure the continuity of time series features, the input data is segmented using a sliding window method. The length of each window determines the time span of historical information captured by the model.

[0095] The neural network training objective is to minimize the mean squared error (MSE) between the predicted SOC value and the true value. Using the Adam optimizer, combined with an appropriate learning rate, accelerates model convergence and improves training efficiency. During training, network weights are optimized using the training set, and the validation set is used to evaluate the model's generalization ability, ensuring robustness and accuracy across diverse datasets.

[0096] In one embodiment of the present application, training a multi-field coupled physical information neural network based on a joint loss function includes:

[0097] The voltage time series and current time series are input into the multi-field coupled physical information neural network, and the predicted SOC value of the all-electric aircraft lithium battery at the current moment is output.

[0098] Calculate the negative electrode surface concentration of the lithium battery of the all-electric aircraft based on the predicted SOC value.

[0099] The negative electrode surface concentration is input into the calibrated multi-field coupled SPM model of the all-electric aircraft lithium battery to obtain the calculated voltage of the all-electric aircraft lithium battery.

[0100] Specifically, based on the multi-field coupling SPM model, the influence of temperature field and deformation on battery performance is incorporated into the model, and the concentration is calculated through the predicted SOC value of the neural network as the input of the SPM model to further simulate the voltage response of the battery.

[0101] Based on the SPM model, the predicted SOC value of the neural network is calculated by the formula: n,surf :

[0102] c n,surf =SOC*(c n,max -c n,min )+c n,min

[0103] The calculated negative electrode surface concentration of the all-electric aircraft lithium battery is used as the input of the SPM model, and the battery voltage response is calculated in combination with the temperature and deformation.

[0104] The SOC mean square error is calculated based on the predicted SOC value, the voltage mean square error is calculated based on the predicted SOC value, and a joint loss function is constructed based on the SOC mean square error and the voltage mean square error.

[0105] The multi-field coupled physical information neural network is trained using the joint loss function as the objective function and the Adam optimizer and the preset learning rate.

[0106] In one embodiment of the present application, the SOC mean square error is calculated based on the predicted SOC value, the voltage mean square error is calculated based on the SOC mean square error, and a joint loss function is constructed based on the SOC mean square error and the voltage mean square error, including:

[0107] The predicted SOC value and the experimental SOC value are compared, and the SOC mean square error of the neural network is calculated.

[0108] Based on the calculated voltage values ​​and the experimental voltage values, the voltage mean square error of the calibrated multi-field coupled all-electric aircraft lithium battery SPM model is calculated.

[0109] A joint loss function is constructed based on the weighted SOC mean square error and voltage mean square error.

[0110] The joint loss function consists of two parts: the weighted sum of the mean square error (MSE) between the SOC value predicted by the neural network and the actual SOC value of the experiment, and the mean square error between the voltage calculated by the SPM model and the experimental voltage.

[0111] Loss data is the mean square error loss function of SOC, which is as follows:

[0112]

[0113] Loss physics is the mean square error loss function of voltage, which is as follows:

[0114]

[0115] The total loss function is:

[0116] Loss total =α·Loss data +β·Loss physics

[0117] Among them, α and β are weight coefficients used to balance the impact of SOC and voltage errors on optimization.

[0118] Based on the LSTM neural network training and combined with the joint loss function, the network weights are updated through the gradient descent algorithm to ensure that the predicted SOC value and voltage are consistent with both experimental data and physical models.

[0119] S30. Estimate the SOC of the lithium battery of the all-electric aircraft based on the trained physical information neural network to obtain a predicted SOC.

[0120] The experimental results show that the PINN framework can achieve high-precision SOC estimation under multi-field coupling conditions. Figure 3 shown.

[0121] In order to more intuitively present the superiority of the method proposed in this invention, the error evaluation index is presented in Table 2:

[0122] Table 2 SOC estimation error evaluation index

[0123]

[0124]

[0125] Depend on Figure 4 As can be seen from Table 2, the proposed method achieves lower error and more accurate estimation compared to the common EKF and LSTM methods. In summary, the experimental results demonstrate the accuracy and practicality of this algorithm. While the basic structure, principles, and methods of the present invention have been specifically illustrated through the examples above, the specific implementation of the present invention is not limited to these examples. Those skilled in the art will be able to make several simple deductions or substitutions without departing from the spirit of the present invention.

[0126] Based on the above method embodiments, the present application also provides an all-electric aircraft lithium battery SOC estimation device based on a multi-field coupled physical information neural network, which is used to solve the same technical problems as the method embodiments. The device may include a first model determination module, a second model determination module, and a prediction module. The first model determination module is configured to introduce a temperature field and a mechanical deformation mechanics field into a single-particle model of the all-electric aircraft lithium battery to obtain a multi-field coupled SPM model of the all-electric aircraft lithium battery. The second model determination module is configured to construct a neural network, using the calibrated multi-field coupled SPM model of the all-electric aircraft lithium battery as a physical constraint of the neural network to obtain a multi-field coupled physical information neural network. A joint loss function is constructed using the sum of the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model. The multi-field coupled physical information neural network is trained based on the joint loss function to obtain a trained physical information neural network. The prediction module is configured to estimate the SOC of the all-electric aircraft lithium battery based on the trained physical information neural network to obtain a predicted SOC.

[0127] In another embodiment of the present invention, a device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement the method for estimating SOC of a lithium battery for an all-electric aircraft based on a multi-field coupled physical information neural network in an embodiment of the present invention.

[0128] In another embodiment provided by the present invention, a computer-readable storage medium is further provided, wherein the storage medium stores at least one instruction, at least one program, code set, or instruction set, and the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the SOC estimation method for an all-electric aircraft lithium battery based on a multi-field coupled physical information neural network in an embodiment of the present invention.

[0129] In the above embodiments, they can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. The computer program product includes multiple computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates multiple available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0130] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0131] Each embodiment in this specification is described in a related manner. Similar portions between the various embodiments can be referenced to each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are described briefly because they are generally similar to the method embodiments. For related portions, reference can be made to the description of the method embodiments.

[0132] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A method for estimating the SOC of lithium batteries in all-electric aircraft based on a multi-field coupled physical information neural network, characterized in that: include: The temperature field and mechanical deformation mechanics field are introduced into the single particle model of the all-electric aircraft lithium battery to obtain the multi-field coupled SPM model of the all-electric aircraft lithium battery; A neural network is constructed, and a calibrated multi-field coupled SPM model of an all-electric aircraft lithium battery is used as a physical constraint of the neural network to obtain a multi-field coupled physical information neural network. A joint loss function is constructed using the sum of the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model. The multi-field coupled physical information neural network is trained based on the joint loss function to obtain a trained physical information neural network. The SOC of the lithium battery of the all-electric aircraft is estimated based on the trained physical information neural network to obtain the predicted SOC.

2. The method for estimating SOC of lithium batteries for all-electric aircraft based on a multi-field coupled physical information neural network according to claim 1, characterized in that: Before introducing the temperature field and the mechanical deformation mechanics field into the single particle model of the all-electric aircraft lithium battery to obtain a multi-field coupled SPM model of the all-electric aircraft lithium battery, the method further includes: Simulate the temperature field of lithium batteries for all-electric aircraft by dynamically adjusting the reaction rate constant; The mechanical deformation field of lithium batteries for all-electric aircraft is simulated through the electrode thickness adjustment formula.

3. The method for estimating SOC of lithium batteries for all-electric aircraft based on a multi-field coupled physical information neural network according to claim 1, characterized in that: Before using the calibrated multi-field coupled all-electric aircraft lithium battery SPM model as a physical constraint of the neural network to obtain the multi-field coupled physical information neural network, the method further includes: The mechanical deformation influencing factors of the multi-field coupled all-electric aircraft lithium battery SPM model are optimized by genetic algorithm, and the calibrated multi-field coupled all-electric aircraft lithium battery SPM model is obtained.

4. The method for estimating SOC of lithium batteries in all-electric aircraft based on multi-field coupled physical information neural network according to claim 1, characterized in that: The method of optimizing the mechanical deformation influencing factor of the multi-field coupled all-electric aircraft lithium battery SPM model by a genetic algorithm to obtain a calibrated multi-field coupled all-electric aircraft lithium battery SPM model comprises: Initialize the population size of the genetic algorithm and set the parameter range of the genetic algorithm; The objective function of the genetic algorithm is constructed by minimizing the error between the voltage curve of the experimental data and the voltage obtained by simulating the SPM model of the all-electric aircraft lithium battery through multi-field coupling, and the fitness of the genetic algorithm is constructed as the inverse of the objective function; Based on the objective function and fitness, the selection, crossover and mutation operations of the genetic algorithm are used to iteratively optimize the influencing factors of the deformation of the positive and negative electrodes of the all-electric aircraft lithium battery, and a calibrated multi-field coupled SPM model of the all-electric aircraft lithium battery is obtained.

5. The method for estimating SOC of lithium batteries for all-electric aircraft based on multi-field coupled physical information neural network according to claim 1, characterized in that: The multi-field coupled physical information neural network training based on the joint loss function includes: Input voltage time series and current time series into the multi-field coupled physical information neural network, and output the predicted SOC value of the lithium battery of the all-electric aircraft; Calculate the negative electrode surface concentration of the lithium battery of the all-electric aircraft based on the predicted SOC value; The negative electrode surface concentration is input into the calibrated multi-field coupled all-electric aircraft lithium battery SPM model to obtain the calculated voltage of the all-electric aircraft lithium battery; Calculate the SOC mean square error based on the predicted SOC value, calculate the voltage mean square error based on the SOC mean square error and construct a joint loss function based on the SOC mean square error and the voltage mean square error; The multi-field coupled physical information neural network is trained using the joint loss function as the objective function and the Adam optimizer and the preset learning rate.

6. The method for estimating SOC of lithium batteries for all-electric aircraft based on multi-field coupled physical information neural network according to claim 4, characterized in that: Obtaining an experimental SOC value and an experimental voltage, calculating an SOC mean square error based on the predicted SOC value, calculating a voltage mean square error, and constructing a joint loss function based on the SOC mean square error and the voltage mean square error, including: The predicted SOC value and the experimental SOC value are compared to calculate the SOC mean square error of the neural network; Based on the calculated voltage values ​​and the experimental voltage values, the voltage mean square error of the calibrated multi-field coupled all-electric aircraft lithium battery SPM model is calculated; A joint loss function is constructed based on the weighted SOC mean square error and voltage mean square error.

7. The method for estimating SOC of lithium batteries in all-electric aircraft based on multi-field coupled physical information neural network according to claim 1, characterized in that: The neural network is a long short-term memory network, and the long short-term memory network includes: Input layer, multiple LSTM hidden layers, and a fully connected output layer; Each hidden layer contains 50 to 100 LSTM units, and the output layer includes a normalization layer.

8. An all-electric aircraft lithium battery SOC estimation device based on a multi-field coupled physical information neural network, characterized in that: include: The first model determination module is used to introduce the temperature field and mechanical deformation mechanics field into the single particle model of the all-electric aircraft lithium battery to obtain a multi-field coupled SPM model of the all-electric aircraft lithium battery; The second model determination module is used to construct a neural network, using the calibrated multi-field coupled all-electric aircraft lithium battery SPM model as a physical constraint of the neural network to obtain a multi-field coupled physical information neural network; a joint loss function is constructed using the sum of the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model, and the multi-field coupled physical information neural network is trained based on the joint loss function to obtain a trained physical information neural network; The prediction module is used to estimate the SOC of the lithium battery of the all-electric aircraft based on the trained physical information neural network to obtain the predicted SOC.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the method for estimating SOC of a lithium battery for an all-electric aircraft based on a multi-field coupled physical information neural network according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the method for estimating SOC of a lithium battery for an all-electric aircraft based on a multi-field coupled physical information neural network according to any one of claims 1 to 7.

Images