Method and device for predicting service life of energy storage battery

Abstract

The invention relates to the technical field of energy storage battery life prediction, and discloses an energy storage battery life prediction method and device, and the method comprises the steps: obtaining the historical operation information of an energy storage battery and the current state information of the energy storage battery, and carrying out the feature extraction of the historical operation information of the energy storage battery and the current state information of the energy storage battery, the method comprises the steps of obtaining an energy storage battery degradation evolution feature vector and an energy storage battery dynamic response feature vector, performing feature fusion on the energy storage battery degradation evolution feature vector and the energy storage battery dynamic response feature vector to obtain a fusion feature vector, and predicting the service life of the energy storage battery based on the fusion feature vector. And obtaining an energy storage battery life prediction result. According to the method, the robustness and the response capability of the energy storage battery life prediction result in a complex dynamic scene are improved, and the accuracy of the energy storage battery life prediction result is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage battery life prediction, and particularly relates to an energy storage battery life prediction method and device. BACKGROUND

[0002] With large-scale deployment of energy storage systems in photovoltaic, wind power, electric transportation and power frequency modulation scenarios, accurate prediction of the life status of electrochemical batteries as core energy units puts higher requirements on the reliability, safety and economy of system operation.

[0003] However, the related energy storage battery life prediction method cannot simultaneously consider long-term evolution modeling capability and response sensitivity to short-term disturbances, thereby resulting in low accuracy of energy storage battery life prediction. SUMMARY

[0004] Therefore, the present application provides an energy storage battery life prediction method and device to solve the problem that the related energy storage battery life prediction method cannot simultaneously consider long-term evolution modeling capability and response sensitivity to short-term disturbances, thereby resulting in low accuracy of energy storage battery life prediction.

[0005] In a first aspect, the present application provides an energy storage battery life prediction method, which comprises:

[0006] obtaining energy storage battery historical operation information and energy storage battery current state information, and performing feature extraction on the energy storage battery historical operation information and the energy storage battery current state information respectively to obtain an energy storage battery degradation evolution feature vector and an energy storage battery dynamic response feature vector;

[0007] performing feature fusion on the energy storage battery degradation evolution feature vector and the energy storage battery dynamic response feature vector to obtain a fusion feature vector;

[0008] predicting the life of the energy storage battery based on the fusion feature vector to obtain an energy storage battery life prediction result.

[0009] The energy storage battery life prediction method provided in the embodiment, by acquiring the historical operation information of the energy storage battery and the current state information of the energy storage battery, respectively extracting features from the historical operation information of the energy storage battery and the current state information of the energy storage battery, obtaining the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery, and through feature fusion of the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery, a fusion feature vector is obtained, so that the fusion feature vector contains both long-term evolution energy storage battery features and short-term disturbance response energy storage battery features, the energy storage battery life is predicted based on the fusion feature vector, the energy storage battery life prediction result is accurately obtained, the prediction result takes into account both long-term evolution modeling capability and short-term disturbance response sensitivity, the robustness and response capability of the energy storage battery life prediction result in a complex dynamic scenario are improved, and the accuracy of the energy storage battery life prediction result is further improved.

[0010] In an optional implementation, respectively extracting features from the historical operation information of the energy storage battery and the current state information of the energy storage battery, obtaining the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery, includes:

[0011] Extracting battery degradation features from the historical operation information of the energy storage battery to obtain the evolution feature vector of the degradation of the energy storage battery;

[0012] Extracting short-term features from the current state information of the energy storage battery to obtain the dynamic response feature vector of the energy storage battery.

[0013] The energy storage battery life prediction method provided in the embodiment, by extracting battery degradation features from the historical operation information of the energy storage battery to obtain the evolution feature vector of the degradation of the energy storage battery, the long-term degradation law of the battery can be captured, the aging cumulative effect is reflected, and the trend basis is provided for long-term prediction such as remaining service life, and by extracting short-term features from the current state information of the energy storage battery to obtain the dynamic response feature vector of the energy storage battery, the dynamic characteristics under real-time working conditions are captured, the influence of short-term load and environmental changes is reflected, and the sensitivity to sudden state changes is improved.

[0014] In an optional implementation, extracting battery degradation features from the historical operation information of the energy storage battery to obtain the evolution feature vector of the degradation of the energy storage battery includes:

[0015] Mapping the historical operation information of the energy storage battery to a high-dimensional feature space to obtain a historical time sequence feature vector;

[0016] Position encoding the historical time sequence feature vector to obtain an embedded feature vector;

[0017] Performing feature enhancement on the embedded feature vector to obtain an embedded feature vector after feature enhancement;

[0018] performing nonlinear transformation on the embedded feature vector after feature enhancement to obtain an updated embedded feature vector;

[0019] performing residual connection and normalization processing on the updated embedded feature vector;

[0020] performing multi-layer stacking on the embedded feature vector after residual connection and normalization processing to obtain an evolution feature vector of energy storage battery degradation.

[0021] The energy storage battery life prediction method provided in the embodiment maps the historical operation information of the energy storage battery to a high-dimensional feature space to obtain a historical time sequence feature vector, excavates the hidden nonlinear relationship in the data, enhances the expression ability of the feature, obtains an embedded feature vector by performing position coding on the historical time sequence feature vector, retains the time sequence information of the historical data, ensures that the feature vector can reflect the real time evolution logic, obtains an embedded feature vector after feature enhancement by performing feature enhancement on the embedded feature vector, improves the discriminability of the feature, weakens the interference of accidental fluctuations, and makes the feature more focused on the core law of battery degradation, obtains an updated embedded feature vector by performing nonlinear transformation on the embedded feature vector after feature enhancement, avoids that a certain type of feature is dominated by model learning due to excessively large factor value, ensures that the degradation features in each dimension are equally concerned, and obtains an evolution feature vector of energy storage battery degradation by performing multi-layer stacking on the embedded feature vector after residual connection and normalization processing, fully captures the complete evolution process of the battery from slight attenuation to significant aging, and provides more accurate feature support for energy storage battery life prediction.

[0022] In an optional implementation, short-term feature extraction is performed on the current state information of the energy storage battery to obtain an energy storage battery dynamic response feature vector, including:

[0023] constructing an initial state sequence based on the current state information of the energy storage battery;

[0024] extending the initial state sequence to a prediction target length to obtain an energy storage battery state sequence;

[0025] performing information extraction on the energy storage battery state sequence to obtain an energy storage battery dynamic response feature vector.

[0026] The energy storage battery life prediction method provided in the embodiment is based on current state information of the energy storage battery to construct an initial state sequence, captures the instant dynamic trend under the current working condition of the energy storage battery, expands the initial state sequence to a prediction target length to obtain an energy storage battery state sequence that adapts to the demand of matching the dynamic response characteristics to the time scale, provides sufficient time sequence dimension for extracting state trend changes, avoids incomplete dynamic feature capture due to too short sequence, and extracts information from the energy storage battery state sequence to obtain an energy storage battery dynamic response feature vector that accurately reflects the dynamic performance of the energy storage battery under the current working condition, thereby improving the sensitivity and response accuracy of the model to the current working condition change.

[0027] In an optional implementation, the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery are fused to obtain a fusion feature vector, including:

[0028] Similarity weights are calculated based on the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery.

[0029] The evolution feature vector of the degradation of the energy storage battery is weighted and summed based on the similarity weights to obtain a fusion information vector.

[0030] The fusion information vector and the dynamic response feature vector of the energy storage battery are spliced to obtain the fusion feature vector.

[0031] The energy storage battery life prediction method provided in the embodiment calculates similarity weights based on the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery, measures the correlation degree between the evolution feature of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery, avoids the interference of the irrelevant part of the long-term information on the fusion effect, weights and sums the evolution feature vector of the degradation of the energy storage battery based on the similarity weights to obtain a fusion information vector, so that the long-term information is more suitable for the current working condition, realizes the targeted screening and concentration of the long-term feature, splices the fusion information vector and the dynamic response feature vector of the energy storage battery to obtain the fusion feature vector, realizes the complementation of the long-term and short-term information, reflects the long-term degradation trend of the battery and the instant change of the current state, provides more comprehensive and higher correlation degree feature support for the energy storage battery life prediction, and avoids the one-sidedness of a single information source.

[0032] In an optional implementation, the energy storage battery life prediction is performed based on the fusion feature vector to obtain an energy storage battery life prediction result, including:

[0033] The fusion feature vector is subjected to feature enhancement to obtain a fusion feature vector after feature enhancement.

[0034] The fusion feature vector after feature enhancement is predicted step by step to obtain the energy storage battery life prediction result.

[0035] The energy storage battery life prediction method provided by the embodiment is characterized in that the fusion feature vector is enhanced to obtain the fusion feature vector after feature enhancement, so that the feature is focused on the core factor affecting the battery life, the subsequent prediction is reduced in interference, the effectiveness of the input feature is enhanced, the fusion feature vector after feature enhancement is predicted step by step to obtain the energy storage battery life prediction result, the prediction time resolution and accuracy are improved, and a more fine time node reference is provided for maintenance decision.

[0036] In a second aspect, the present application provides an energy storage battery life prediction device, which comprises:

[0037] The extraction module is configured to obtain the historical operation information and the current state information of the energy storage battery, and extract features from the historical operation information and the current state information of the energy storage battery respectively to obtain the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery.

[0038] The fusion module is configured to fuse the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery to obtain a fusion feature vector.

[0039] The prediction module is configured to predict the life of the energy storage battery based on the fusion feature vector to obtain an energy storage battery life prediction result.

[0040] In a third aspect, the present application provides a computer device, which comprises a memory and a processor, the memory and the processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the energy storage battery life prediction method of the first aspect or any of the corresponding embodiments thereof.

[0041] In a fourth aspect, the present application provides a computer readable storage medium, which stores computer instructions, and the computer instructions are used to make the computer execute the energy storage battery life prediction method of the first aspect or any of the corresponding embodiments thereof.

[0042] In a fifth aspect, the present application provides a computer program product, which comprises computer instructions, and the computer instructions are used to make the computer execute the energy storage battery life prediction method of the first aspect or any of the corresponding embodiments thereof. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the specific embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, the drawings described below are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0044] Figure 1 is a flowchart of a method for predicting the life of an energy storage battery according to an embodiment of the present application;

[0045] Figure 2 is a flowchart of another method for predicting the life of an energy storage battery according to an embodiment of the present application;

[0046] Figure 3 is a flowchart of still another method for predicting the life of an energy storage battery according to an embodiment of the present application;

[0047] Figure 4 is a flowchart of yet another method for predicting the life of an energy storage battery according to an embodiment of the present application;

[0048] Figure 5 is a structural block diagram of a device for predicting the life of an energy storage battery according to an embodiment of the present application;

[0049] Figure 6 is a hardware structure diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0051] With the large-scale deployment of energy storage systems in photovoltaic, wind power, electric transportation and power frequency modulation scenarios, the accurate prediction of the life status of electrochemical batteries as the core energy unit is required to ensure the reliability, safety and economy of system operation. However, the energy storage batteries exhibit significant time evolution characteristics and state disturbance complexity in actual operation. The degradation path is nonlinear and stage-based, and is easily affected by environmental conditions, charging and discharging strategies and other factors, which brings challenges to the life estimation method of energy storage batteries. In particular, in the multi-step prediction task, it is often difficult to maintain prediction accuracy while considering short-term response and long-term trend modeling capability. Especially in the multi-step prediction demand scenarios for 3-day, 7-day and other short-term life prediction cycles, related methods generally have problems such as response lag to dynamic changes and significant accuracy decline, which are difficult to meet the high-frequency update and real-time evaluation tasks under complex operating conditions.

[0052] As a key task in energy storage battery life modeling, multi-step prediction aims to continuously infer the SOH (State of Health) or RUL (Remaining Useful Life) at future time points based on existing observation information. It has been widely used in battery maintenance, scheduling optimization and safety evaluation scenarios. In recent years, with the development of deep learning technology, multi-step prediction methods based on RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) sequence models have emerged. In particular, the introduction of Encoder-Decoder structure and Transformer architecture has made the model show significant advantages in modeling long-term dependence and nonlinear degradation trend. The first aspect adopts multi-layer GRU to build a sequence prediction model, realizes rolling prediction of battery RUL, and improves the robustness of the model through feature selection. The second aspect proposes a long sequence SOH prediction model based on the Transformer structure, uses self-attention mechanism to enhance the long-term dependence modeling capability, and realizes excellent multi-step prediction accuracy under the condition of multi-data source input. In addition, by constructing a hybrid network that combines convolution extraction and LSTM memory mechanism, an end-to-end multi-step life prediction framework is built, which has strong cross-battery adaptability.

[0053] The attention mechanism and the multi-modal feature fusion strategy further improve the representation ability of the model for complex input states. However, the energy storage battery life prediction method still faces many challenges: first, the multi-step rolling prediction generally has the problem of error accumulation, and the early prediction error is easily enlarged with the time step, which seriously affects the long-term prediction accuracy; second, the sequence model has limited ability to model the "long-period weak trend" characteristics of the battery, and it is difficult to accurately capture the key turning points in the degradation process; third, most methods focus on learning from historical data, and the response to real-time state changes is insufficient, making it difficult to adapt to dynamic operating conditions; fourth, battery operation involves multiple heterogeneous variables, and related methods are still cumbersome in multi-modal information fusion, which affects the generalization ability and deployment efficiency of the model. In summary, the battery multi-step prediction method needs to break through the traditional single sequence driven paradigm and introduce a dual-effect mechanism modeling framework that combines historical trends and current states to simultaneously consider long-term evolution modeling ability and response sensitivity to short-term disturbances, thereby achieving higher accuracy and stronger robustness of life prediction.

[0054] To solve the above technical problems, the embodiment of the present application provides a kind of energy storage battery life prediction method, first, the historical operation information of energy storage battery is preprocessed and feature factor is extracted, as the basic feature representation of neural network input, then, improved encoder-decoder (Encoder-Decoder) structure is used to realize the collaborative modeling of historical trend and current state, i.e. in the encoder part, the historical operation information of energy storage battery is mapped to high-dimensional space by linear embedding, then the long-term dependence in time series is modeled by using multi-layer Transformer module, and the evolution feature vector of energy storage battery degradation is extracted;At the same time, the current state information of energy storage battery, such as SOC (State of Charge, State of Charge), temperature and internal resistance, etc. is input into GRU network to obtain the dynamic response feature vector of energy storage battery in short term, in order to effectively fuse two kinds of information, cross-branch attention mechanism is introduced, so that the current state can dynamically perceive the historical evolution trend, realize the effective fusion of multi-source information, second, the decoder part uses multi-step GRU structure to predict the feature representation fused by time step, and outputs the life prediction result of energy storage battery, in order to enhance the feature discriminability, CBAM (Convolutional Block Attention Module, Convolutional Block Attention Module) is added in the fusion stage, which strengthens the expression of key features from two dimensions of channel and time sequence, and the verification result of the experimental data of energy storage battery shows that the method is superior to related sequence model in multi-step prediction accuracy, response ability to sudden state change and the like.

[0055] Based on the synergistic fusion of the first effect time sequence modeling mechanism and the second effect current state feature mapping mechanism, multi-step life prediction is carried out by combining historical operation information of the energy storage battery and current state information of the energy storage battery, so as to improve the robustness and response ability of the model in a complex dynamic scene. Specifically, the periodic evolution trend is extracted by the historical mechanism, and the instantaneous response to the energy change of the battery is provided by the state mechanism, both of which jointly drive the prediction process, not only solving the problem of unstable multi-step prediction accuracy, but also greatly improving the ability to capture abnormal fluctuations. By introducing the double-effect mechanism, the model has stronger feature redundancy robustness and short-term prediction sensitivity, providing a higher value prediction basis for the actual battery management system (Battery Management System, BMS). In industrial applications, it can effectively guide the battery dynamic scheduling, capacity planning and retirement strategy optimization, and provide key technical support for the intelligent development of energy storage systems.

[0056] The embodiment of the present application provides a kind of energy storage battery life prediction method, it needs to be explained, the energy storage battery life prediction method provided in the embodiment of the present application, its execution subject can be the device of energy storage battery life prediction, the device of energy storage battery life prediction can be realized by software, hardware or software and hardware combination way becomes part or all of electronic equipment, wherein the electronic equipment can be server or terminal, wherein the server in the embodiment of the present application can be a server, can also be the server cluster of being made of multiple servers, the terminal in the embodiment of the present application can be smart phone, personal computer, tablet computer, wearable device and smart robot and other intelligent hardware equipment.The method embodiment described below, all with the execution subject being electronic equipment as an example to be explained.

[0057] According to the embodiment of the present application, an energy storage battery life prediction method embodiment is provided, it needs to be explained, the steps shown in the flowchart of the drawing can be executed in a computer system, such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from here.

[0058] In the present embodiment, an energy storage battery life prediction method is provided, which can be used in the above-mentioned electronic device, Figure 1 The flowchart of the energy storage battery life prediction method according to the embodiment of the present application is shown in FIG. Figure 1 As shown in the figure, the flow includes the following steps:

[0059] In step S101, the historical operation information of the energy storage battery and the current state information of the energy storage battery are obtained, and the feature extraction is performed on the historical operation information of the energy storage battery and the current state information of the energy storage battery respectively, to obtain the evolution feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery.

[0060] Specifically, the encoder is used to extract features of the historical operation information and the current state information of the energy storage battery respectively, to obtain an evolution feature vector of degradation of the energy storage battery and a dynamic response feature vector of the energy storage battery.

[0061] Further, in the life prediction task of the energy storage battery, the historical operation information of the energy storage battery is usually derived from periodic monitoring of an experimental bench or an actual operation system, covers multiple key operation parameters, forms a sequence feature in the time dimension, and can reflect the performance degradation trend of the battery in long-term operation. In order to fully mine the life information contained in the key operation parameters, the historical operation information of the energy storage battery must be preprocessed and feature factors extracted first, to provide high-quality input for subsequent neural network model training.

[0062] Further, the historical operation information of the energy storage battery is derived from a high-precision data acquisition platform, and has multiple original features such as Data Point (data point), Test Time (test time), Date Time (date time), Step Time (step time), Step Index (step index), etc. Through a scientific data extraction and processing process, seven core feature variables are constructed from these original measurement fields (i.e. energy storage battery operation data): Cycle (cycle), Disc Capacity (battery capacity), Heal Indicator (repair indicator), Resistance (resistance), CCCT (Constant Current Cycle Test, constant current cycle test), CVCT (Constant Voltage Cycle Test, constant voltage cycle test), and DISCT (Discharge Cycle Test, discharge cycle test).

[0063] Further, the current state information of the energy storage battery includes SOC, temperature, internal resistance, etc.

[0064] In step S102, the evolution feature vector of degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery are fused to obtain a fused feature vector.

[0065] Specifically, to effectively fuse the evolution feature vector of degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery, a cross-branch attention mechanism is introduced in the encoder, and the encoder is used to fuse the evolution feature vector of degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery to obtain a fused feature vector, so that the current state can dynamically perceive the historical evolution trend, and effectively fuse multiple sources of information.

[0066] Step S103, predicting the energy storage battery life based on the fusion feature vector to obtain an energy storage battery life prediction result.

[0067] Specifically, the decoder adopts a multi-step unidirectional GRU structure to perform time-step-by-time-step prediction on the fusion feature vector, and outputs the battery life indicators (i.e., the energy storage battery life prediction result) at multiple future time points.

[0068] The energy storage battery life prediction method provided in this embodiment, by obtaining the historical operation information of the energy storage battery and the current state information of the energy storage battery, respectively extracting features from the historical operation information of the energy storage battery and the current state information of the energy storage battery, obtaining the evolution feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery, and fusing the evolution feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery to obtain the fusion feature vector, so that the fusion feature vector contains both long-term evolution features of the energy storage battery and short-term disturbance response features of the energy storage battery, predicting the energy storage battery life based on the fusion feature vector, and accurately obtaining the energy storage battery life prediction result, the prediction result takes into account both long-term evolution modeling capability and short-term disturbance response sensitivity, improves the robustness and response capability of the energy storage battery life prediction result in complex dynamic scenarios, and further improves the accuracy of the energy storage battery life prediction result.

[0069] An energy storage battery life prediction method is provided in this embodiment, which can be used in the electronic device described above, Figure 2 is a flowchart of the energy storage battery life prediction method according to an embodiment of the present application, as Figure 2 shown, the flow includes the following steps:

[0070] Step S201, obtaining the historical operation information of the energy storage battery and the current state information of the energy storage battery, and respectively extracting features from the historical operation information of the energy storage battery and the current state information of the energy storage battery to obtain the evolution feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery.

[0071] Specifically, the above step S201 includes:

[0072] Step S2011, extracting battery degradation features from the historical operation information of the energy storage battery to obtain the evolution feature vector of the energy storage battery degradation.

[0073] Specifically, in the encoder part, the historical operation information of the energy storage battery is mapped to a high-dimensional space through linear embedding, and then a multi-layer Transformer module is used to model the long-term dependence in the time series to extract the evolution feature vector of the energy storage battery degradation.

[0074] In some optional embodiments, the above step S2011 includes:

[0075] Step a1, map the historical operation information of the energy storage battery to a high-dimensional feature space to obtain a historical time series feature vector.

[0076] Specifically, in the encoder stage, the historical operation information of the energy storage battery needs to be converted into a representation form that can be learned by the model. Let the feature vector of the energy storage battery at the t-th period time be The feature dimension is d = 7 core features are included, the historical sequence window has a length T, and an input sequence is constructed according to the feature vectors of the energy storage battery at multiple periods; wherein the expression of the input sequence X is:

[0077]

[0078] The feature dimension of the input sequence X is

[0079] Further, the input sequence X is dimensioned by a linear embedding layer and mapped to a high-dimensional latent space to obtain a historical time series feature vector; wherein the expression of the historical time series feature vector is:

[0080] Z0 = XW e +b e (2)

[0081] Wherein, Z0 is the historical time series feature vector, W e is the weight matrix, b e is the bias vector, the embedding dimension of the weight matrix W e is The embedding dimension of the bias vector is d model .

[0082] Step a2, position encoding is performed on the historical time series feature vector to obtain an embedded feature vector.

[0083] Specifically, in order to preserve the time order information of the historical time series feature vector, a position encoding PE is introduced to obtain an embedded feature vector; wherein the expression of the embedded feature vector is:

[0084]

[0085] Step a3, feature enhancement is performed on the embedded feature vector to obtain a feature-enhanced embedded feature vector.

[0086] Specifically, long-term dependency modeling is performed on the embedding feature vector (i.e., feature enhancement is performed on the embedding feature vector using a multi-layer Transformer Encoder structure), to obtain an embedding feature vector after feature enhancement; wherein each layer in the multi-layer Transformer Encoder structure comprises a multi-head self-attention (MHA) module and a feed forward network (FFN), which can prevent overfitting from occurring, and for the first layer, the i-th time step can be expressed as:

[0087]

[0088] wherein Q is a query vector, K is a key vector, V is a value vector, and W is a learnable parameter. i

[0089] Step a4, performing nonlinear transformation on the embedding feature vector after feature enhancement to obtain an updated embedding feature vector.

[0090] Specifically, a feed forward network is composed of two linear layers and a nonlinear activation, and the embedding feature vector after feature enhancement is nonlinearly transformed using the feed forward network, i.e., the features output by the self-attention are nonlinearly mapped, to enhance the ability of the model to fit complex relationships, and to independently process the features of each position in the channel dimension (feature dimension), to supplement the "local feature extraction" of the self-attention in addition to the "global correlation".

[0091] Step a5, performing residual connection and normalization processing on the updated embedding feature vector.

[0092] Step a6, performing multi-layer stacking on the embedding feature vector after residual connection and normalization processing to obtain an evolution feature vector of the energy storage battery degradation.

[0093] Step S2012, performing short-term feature extraction on the current state information of the energy storage battery to obtain a dynamic response feature vector of the energy storage battery.

[0094] Specifically, the current state information of the energy storage battery (such as SOC, temperature, internal resistance, etc.) is input into a GRU network to obtain a dynamic response feature vector of the energy storage battery.

[0095] In some optional embodiments, the above step S2012 comprises:

[0096] Step b1, constructing an initial state sequence based on the current state information of the energy storage battery.

[0097] Specifically, a state-aware branch is constructed to model the immediate operating state at the current time (i.e., the initial state sequence), and the initial state sequence ​The feature dimension of the initial state sequence is The extended features selected from the latest time point, such as the current battery temperature, the charging voltage, the energy efficiency of the previous period, and the like.

[0098] Step b2, extending the initial state sequence to a prediction target length to obtain a state sequence of the energy storage battery.

[0099] The expression of the state sequence S of the energy storage battery extended to the prediction target length (taking 4 steps of multi-step prediction as an example) is as follows:

[0100]

[0101] wherein, is the feature dimension of the state sequence S of the energy storage battery.

[0102] Step b3, information extraction is performed on the state sequence of the energy storage battery to obtain a dynamic response feature vector of the energy storage battery.

[0103] Specifically, the state sequence of the energy storage battery is input into a double-layer GRU network to extract the current mapping information (i.e., the dynamic response feature vector of the energy storage battery).

[0104] Step S202, feature fusion is performed on the evolution feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery to obtain a fusion feature vector. For details, refer to Figure 1 Step S102 of the embodiment shown in

[0105] Step S203, the life of the energy storage battery is predicted based on the fusion feature vector to obtain a prediction result of the life of the energy storage battery. For details, refer to Figure 1 Step S103 of the embodiment shown in

[0106] The energy storage battery life prediction method provided in this embodiment can capture the long-term degradation law of the battery by extracting the battery degradation features from the historical operation information of the energy storage battery to obtain the evolution feature vector of the energy storage battery degradation, reflect the aging cumulative effect, and provide a trend basis for the long-term prediction of the remaining service life, and can capture the dynamic characteristics under real-time working conditions by extracting short-term features from the current state information of the energy storage battery to obtain the dynamic response feature vector of the energy storage battery, reflect the influence of short-term load and environmental changes, and improve the sensitivity to sudden state changes.

[0107] In this embodiment, an energy storage battery life prediction method is provided, which can be used in the electronic device described above, Figure 3 is a flowchart of the energy storage battery life prediction method according to an embodiment of the present application, as shown in Figure 3 The flowchart includes the following steps:

[0108] In step S301, the historical operation information of the energy storage battery and the current state information of the energy storage battery are acquired, and feature extraction is performed on the historical operation information of the energy storage battery and the current state information of the energy storage battery respectively to obtain an evolution feature vector of degradation of the energy storage battery and a dynamic response feature vector of the energy storage battery. For details, please refer to Figure 2 In step S201 of the embodiment shown, no further elaboration is given here.

[0109] In step S302, the evolution feature vector of degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery are fused to obtain a fusion feature vector.

[0110] In the fusion, the dynamic response feature vector of the energy storage battery is taken as a query vector, and the evolution feature vector of degradation of the energy storage battery is taken as a key vector and a value vector to calculate the fusion feature vector.

[0111] Specifically, the above step S302 includes:

[0112] In step S3021, a similarity weight is calculated based on the evolution feature vector of degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery.

[0113] Specifically, the expression of the similarity weight is:

[0114]

[0115] wherein α i is the similarity weight, softmax is a normalization function, h i is the dynamic response feature vector of the energy storage battery, W Q and W K are learnable weight matrices, and H hist is the evolution feature vector of degradation of the energy storage battery.

[0116] In step S3022, the evolution feature vector of degradation of the energy storage battery is weighted and summed based on the similarity weight to obtain a fusion information vector.

[0117] Specifically, the expression of the fusion information vector is:

[0118]

[0119] wherein F is the fusion information vector, and W V is a learnable weight matrix.

[0120] In step S3023, the fusion information vector and the dynamic response feature vector of the energy storage battery are spliced to obtain the fusion feature vector.

[0121] Specifically, the expression of the fusion feature vector is:

[0122]

[0123] wherein f is a fusion feature vector, and Concat is a concatenation function.

[0124] Further, the design of the encoder realizes dynamic alignment and collaborative modeling of historical evolution and instantaneous state (i.e., feature fusion of the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery), provides a strong discriminative feature basis for the decoder in the next stage, and effectively improves the generalization ability and robustness of the model in the multi-step prediction scenario.

[0125] In step S303, the life of the energy storage battery is predicted based on the fusion feature vector to obtain an energy storage battery life prediction result. For details, please refer to Figure 2 The step S203 of the embodiment shown will not be repeated here.

[0126] The energy storage battery life prediction method provided in this embodiment calculates the similarity weight based on the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery, measures the correlation degree of the evolution feature of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery, avoids the interference of the irrelevant part of the long-term information on the fusion effect, and performs weighted summation on the evolution feature vector of the degradation of the energy storage battery based on the similarity weight to obtain a fusion information vector, so that the long-term information is more suitable for the current working condition, realizes the targeted screening and concentration of long-term features, realizes the complementation of long-term and short-term information by concatenating the fusion information vector and the dynamic response feature vector of the energy storage battery, reflects the long-term degradation trend of the battery and the instantaneous change of the current state, provides more comprehensive and higher correlation degree feature support for the life prediction of the energy storage battery, and avoids the one-sidedness of a single information source.

[0127] An energy storage battery life prediction method is provided in this embodiment, which can be used in the electronic device described above, Figure 4 is a flowchart of the energy storage battery life prediction method according to an embodiment of the present application, as shown in the figure, the flowchart includes the following steps: Figure 4

[0128] In step S401, the historical operation information of the energy storage battery and the current state information of the energy storage battery are obtained, and the historical operation information of the energy storage battery and the current state information of the energy storage battery are respectively extracted to obtain the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery. For details, please refer to Figure 3 The step S301 of the embodiment shown will not be repeated here.

[0129] In step S402, the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery are fused to obtain a fusion feature vector. For details, please refer to​Figure 3 At step S302 of the illustrated embodiment, no further elaboration is provided.

[0130] At step S403, the energy storage battery life is predicted based on the fusion feature vector, and an energy storage battery life prediction result is obtained.

[0131] Specifically, the above step S403 includes:

[0132] At step S4031, the fusion feature vector is enhanced in feature, and a fusion feature vector enhanced in feature is obtained.

[0133] Specifically, in order to enhance the model's ability to capture future time series, a convolutional block attention module (CBAM) is embedded before the input GRU to strengthen the model's attention ability to key features and time steps. In the fusion stage, a CBAM convolutional attention module is added to strengthen the expression of key features from the channel and time dimensions (i.e., the fusion feature vector is enhanced in feature), and a fusion feature vector enhanced in feature is obtained.

[0134] Further, the CBAM module is composed of a channel attention module and a time attention module in series. First, in the channel dimension, global pooling is performed on each feature channel, and importance modeling is performed through a shared multi-layer perceptron (i.e., the fusion feature vector is processed by pooling, and the fusion feature vector processed by pooling is fused in feature to obtain a channel attention weight vector). The expression of the channel attention weight is:

[0135]

[0136] wherein σ is an activation function, MLP represents a perceptron for processing the context feature sequence processed by pooling, AvgPool is an average pooling, MaxPool is a global maximum pooling, and M c (h t ) is a channel attention weight vector, is the dimension of the channel attention weight vector.

[0137] Further, the channel attention weight vector and the context feature sequence are multiplied to obtain a channel feature map; wherein the expression of the channel feature map F c is:

[0138]

[0139] Further, the channel feature map is processed by pooling, and the channel feature map processed by pooling is convolved to obtain a spatial attention weight vector; wherein the expression of the spatial attention weight vector is:

[0140]

[0141] wherein, Mt(Fc) is a spatial attention weight vector, ConvlD is a convolution processing, and the dimension of the spatial attention weight vector is

[0142] Further, the channel feature map and the spatial attention weight vector are multiplied to obtain a fusion feature vector after feature enhancement; wherein, the expression of the fusion feature vector F after feature enhancement is:

[0143]

[0144] In step S4032, the fusion feature vector after feature enhancement is time step by time step predicted to obtain the energy storage battery life prediction result.

[0145] Specifically, the decoder part adopts a multi-step unidirectional GRU structure to perform time step by time step prediction on the fusion feature vector after feature enhancement, and outputs the battery life index at multiple future time points (i.e. the energy storage battery life prediction result).

[0146] Further, the state representation after fusion is used as the initial input of the decoder, wherein K is the number of future prediction time steps, and df is the fusion feature dimension.

[0147] The energy storage battery life prediction method provided in the embodiment, by performing feature enhancement on the fusion feature vector, obtains the fusion feature vector after feature enhancement, so that the features are more focused on the core factors affecting the battery life, reducing interference for subsequent prediction, enhancing the effectiveness of the input features, and by performing time step by time step prediction on the fusion feature vector after feature enhancement, obtaining the energy storage battery life prediction result, improves the prediction time resolution and accuracy, and provides a more fine time node reference for maintenance decision.

[0148] In the embodiment, an energy storage battery life prediction device is also provided, which is used to implement the above embodiments and preferred embodiments, and will not be described again. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware, or a combination of software and hardware is also possible and contemplated.

[0149] The embodiment provides an energy storage battery life prediction device, as shown in the following formula: Figure 5 The device comprises:

[0150] The extraction module 501 is configured to acquire historical operation information of the energy storage battery and current state information of the energy storage battery, and perform feature extraction on the historical operation information of the energy storage battery and the current state information of the energy storage battery respectively to obtain an evolution feature vector of degradation of the energy storage battery and a dynamic response feature vector of the energy storage battery.

[0151] The fusion module 502 is configured to perform feature fusion on the evolution feature vector of degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery to obtain a fusion feature vector.

[0152] The prediction module 503 is configured to predict the service life of the energy storage battery based on the fusion feature vector to obtain a service life prediction result of the energy storage battery.

[0153] In some optional embodiments, the extraction module 501 comprises:

[0154] The first extraction unit is configured to perform battery degradation feature extraction on the historical operation information of the energy storage battery to obtain the evolution feature vector of degradation of the energy storage battery.

[0155] The second extraction unit is configured to perform short-term feature extraction on the current state information of the energy storage battery to obtain the dynamic response feature vector of the energy storage battery.

[0156] In some optional embodiments, the first extraction unit comprises:

[0157] The mapping subunit is configured to map the historical operation information of the energy storage battery to a high-dimensional feature space to obtain a historical time sequence feature vector.

[0158] The encoding subunit is configured to perform position encoding on the historical time sequence feature vector to obtain an embedded feature vector.

[0159] The enhancement subunit is configured to perform feature enhancement on the embedded feature vector to obtain an embedded feature vector after feature enhancement.

[0160] The transformation subunit is configured to perform nonlinear transformation on the embedded feature vector after feature enhancement to obtain an updated embedded feature vector.

[0161] The normalization subunit is configured to perform residual connection and normalization processing on the updated embedded feature vector.

[0162] The stacking subunit is configured to perform multi-layer stacking on the embedded feature vector after residual connection and normalization processing to obtain the evolution feature vector of degradation of the energy storage battery.

[0163] In some optional embodiments, the second extraction unit comprises:

[0164] The construction subunit is configured to construct an initial state sequence based on the current state information of the energy storage battery.

[0165] The extension sub-unit is configured to extend the initial state sequence to a prediction target length to obtain a state sequence of the energy storage battery.

[0166] The extraction sub-unit is configured to perform information extraction on the state sequence of the energy storage battery to obtain a dynamic response feature vector of the energy storage battery.

[0167] In some optional embodiments, the fusion module 502 includes:

[0168] The calculation unit is configured to calculate a similarity weight based on the evolution feature vector of the degradation of the energy storage battery and the dynamic response feature vector of the energy storage battery.

[0169] The summation unit is configured to perform weighted summation on the evolution feature vector of the degradation of the energy storage battery based on the similarity weight to obtain a fusion information vector.

[0170] The concatenation unit is configured to concatenate the fusion information vector and the dynamic response feature vector of the energy storage battery to obtain a fusion feature vector.

[0171] In some optional embodiments, the prediction unit 502 includes:

[0172] The enhancement unit is configured to perform feature enhancement on the fusion feature vector to obtain a feature-enhanced fusion feature vector.

[0173] The prediction unit is configured to perform time-step-by-time-step prediction on the feature-enhanced fusion feature vector to obtain a life prediction result of the energy storage battery.

[0174] Further function descriptions of the above-mentioned modules and units are the same as those of the corresponding embodiments, and will not be described here.

[0175] The energy storage battery life prediction apparatus in the embodiment is presented in the form of function units. The units herein refer to ASIC (Application Specific Integrated Circuit) circuits, processors and memories that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0176] The embodiment of the present application also provides a computer device having the above-mentioned Figure 5 energy storage battery life prediction apparatus.

[0177] Please refer to Figure 6 , Figure 6 is a structural schematic diagram of a computer device provided by an optional embodiment of the present application, as Figure 6As shown, the computer device includes one or more processors 10, memory 20, and interfaces 50 for external devices such as modems and network interfaces. The one or more processors 10 can be implemented as one or more central processing units (CPUs), one or more microprocessors, microcontrollers, digital signal processors, specialized processors or controller, or one or more processors of any equivalent known in the art. In some embodiments, the one or more processors 10 can be implemented as a combination of one or more of the above physical processors and / or one or more software or firmware modules. The software or firmware can reside in the memory 20 or in any memory available to the one or more processors 10. In some embodiments, the one or more processors 10 can be implemented as, include, operate, or otherwise be associated with one or more artificial intelligence (Al) models, systems, or devices. The one or more processors 10 can process instructions stored in the memory 20 to display a GUI on an external input / output device, such as a display device coupled to the interfaces 50. In some alternative embodiments, multiple processors and / or buses can be used, as appropriate, in conjunction with the memory 20. Also, multiple computer devices can be connected, with each device providing portions of the necessary operations (e.g., as a server array, a group of blade servers, or a multi-processor system). Figure 6 The processor 10 is taken as an example in the embodiments.

[0178] The processor 10 can be a central processing unit, a network processing unit, or a combination thereof. The processor 10 can further include a hardware chip. The hardware chip can be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device can be a complex programmable logic device, a field programmable logic device, a general array logic, or any combination thereof.

[0179] The memory 20 stores instructions executable by the at least one processor 10 to cause the at least one processor 10 to perform the methods shown in the above embodiments.

[0180] The memory 20 can include a program storage area and a data storage area. The program storage area can store an operating system, application programs required by at least one function, and the like. The data storage area can store data created according to the use of the computer device, and the like. In addition, the memory 20 can include a high-speed random access memory, and can further include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid state storage device. In some alternative embodiments, the memory 20 can optionally include a memory disposed remotely from the processor 10, which can be connected to the computer device through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0181] The memory 20 can include a volatile memory, such as a random access memory, and can also include a non-volatile memory, such as a flash memory, a hard disk, or a solid state disk. The memory 20 can further include a combination of the above-mentioned kinds of memories.

[0182] The computer device further includes input devices 30 and output devices 40. The processor 10, the memory 20, the input devices 30, and the output devices 40 can be connected through a bus or other means,Figure 6 The bus connection is taken as an example.

[0183] The input device 30 can receive inputted digital or character information, and generate key signal input related to user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touchpad, a pointing stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 can include a display device, an auxiliary lighting device (e.g., an LED), a tactile feedback device (e.g., a vibration motor), etc. The display device includes, but is not limited to, a liquid crystal display, a light-emitting diode, a display, and a plasma display. In some alternative embodiments, the display device can be a touch screen.

[0184] The embodiments of the present application also provide a computer readable storage medium, and the method according to the embodiments of the present application can be implemented in hardware, firmware, or recorded in a storage medium, or stored in a remote storage medium or a non-transitory machine readable storage medium and downloaded from a network and stored in a local storage medium, so that the method described herein can be processed by such software on a storage medium using a general purpose computer, a special purpose processor, or programmable or special purpose hardware. The storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid state disk, etc. Further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that the computer, the processor, the microprocessor controller, or the programmable hardware includes a storage component that can store or receive software or computer code, which, when accessed and executed by the computer, the processor, or the hardware, implements the method shown in the above embodiments.

[0185] Part of the present application can be applied as a computer program product, for example, computer program instructions, when executed by a computer, the operation of the computer can invoke or provide the method and / or technical solutions according to the present application. Those skilled in the art should understand that the form of computer program instructions in computer readable medium includes but is not limited to source file, executable file, installation package file, etc. Correspondingly, the way of computer program instructions executed by computer includes but is not limited to: the computer directly executes the instructions, or the computer executes the corresponding compiled program after compiling the instructions, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Here, the computer readable medium can be any available computer readable storage medium or communication medium accessible to the computer.

[0186] While embodiments of the application have been described in connection with the preferred embodiments of the various figures, those of ordinary skill in the art will appreciate that various modifications and changes can be made without departing from the spirit and scope of the application, and that such modifications and changes fall within the scope of the appended claims.

Claims

1. A method for predicting the lifespan of an energy storage battery, characterized in that, The method includes: The historical operation information and current status information of the energy storage battery are obtained. Feature extraction is performed on the historical operation information and current status information of the energy storage battery to obtain the evolution feature vector of energy storage battery degradation and the dynamic response feature vector of energy storage battery. The evolution feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery are fused to obtain a fused feature vector. The lifespan of the energy storage battery is predicted based on the fused feature vector, and the predicted lifespan of the energy storage battery is obtained.

2. The method according to claim 1, characterized in that, The step involves extracting features from the historical operating information and current state information of the energy storage battery to obtain the evolutionary feature vector of energy storage battery degradation and the dynamic response feature vector of the energy storage battery, including: Battery degradation features are extracted from the historical operating information of the energy storage battery to obtain the evolution feature vector of the energy storage battery degradation. Short-term features are extracted from the current state information of the energy storage battery to obtain the dynamic response feature vector of the energy storage battery.

3. The method according to claim 2, characterized in that, The step of extracting battery degradation features from the historical operating information of the energy storage battery to obtain the evolution feature vector of the energy storage battery degradation includes: The historical operation information of the energy storage battery is mapped to a high-dimensional feature space to obtain a historical time series feature vector; The historical time series feature vector is positionally encoded to obtain the embedded feature vector; The embedded feature vector is enhanced to obtain the enhanced embedded feature vector; A nonlinear transformation is performed on the enhanced embedded feature vector to obtain an updated embedded feature vector. The updated embedded feature vectors are then subjected to residual connection and normalization. The embedded feature vectors after residual connection and normalization are stacked in multiple layers to obtain the evolution feature vector of the energy storage battery degradation.

4. The method according to claim 2, characterized in that, The step of extracting short-term features from the current state information of the energy storage battery to obtain the dynamic response feature vector of the energy storage battery includes: An initial state sequence is constructed based on the current state information of the energy storage battery; The initial state sequence is extended to the predicted target length to obtain the energy storage battery state sequence; Information is extracted from the state sequence of the energy storage battery to obtain the dynamic response feature vector of the energy storage battery.

5. The method according to claim 1, characterized in that, The feature fusion of the evolutionary feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery to obtain a fused feature vector includes: Calculate similarity weights based on the evolutionary feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery; The evolutionary feature vector of the energy storage battery degradation is weighted and summed based on the similarity weight to obtain the fused information vector; The fused information vector and the dynamic response feature vector of the energy storage battery are concatenated to obtain the fused feature vector.

6. The method according to claim 1, characterized in that, The process of predicting the lifespan of an energy storage battery based on the fused feature vector, to obtain the predicted lifespan result, includes: The fused feature vector is then enhanced to obtain the enhanced fused feature vector. The enhanced fused feature vector is then used for time-step prediction to obtain the predicted lifespan of the energy storage battery.

7. A device for predicting the lifespan of an energy storage battery, characterized in that, The device includes: The extraction module is used to obtain historical operating information and current state information of the energy storage battery, and to extract features from the historical operating information and current state information of the energy storage battery to obtain the evolution feature vector of energy storage battery degradation and the dynamic response feature vector of energy storage battery. The fusion module is used to fuse the evolution feature vector of the energy storage battery degradation and the dynamic response feature vector of the energy storage battery to obtain a fused feature vector. The prediction module is used to predict the lifespan of the energy storage battery based on the fused feature vector, and obtain the prediction result of the energy storage battery lifespan.

8. A computer device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the energy storage battery life prediction method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to execute the energy storage battery life prediction method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes computer instructions for causing a computer to execute the energy storage battery life prediction method according to any one of claims 1 to 6.

Images