Battery state prediction method, electronic device, storage medium and program product

Abstract

The present application provides a battery status prediction method, electronic device, storage medium and program product. The battery status prediction method includes: obtaining first battery data of a past time period; determining second battery data of a future time period, the second battery data is different from the first battery data type; determining a battery data time series sequence according to the first battery data of the past time period and the second battery data of the future time period; and inputting the battery data time series sequence into a time series prediction model to obtain the first battery data of the future time period. This solution can more accurately obtain the battery data that needs to be predicted in the future time period, thereby helping to more accurately predict the battery status in the future time period.

Description

Technical Field

[0001] The present application relates to the technical field of battery data processing, and in particular, to a battery state prediction method, electronic device, storage medium and computer program product. Background Art

[0002] For energy storage batteries, the battery performance will be lost during the charging and discharging process. This characteristic can be reflected by battery status indicators, such as SOH (State of Health) and DOD (Depth of Discharge). These battery status indicators are generally calculated by measuring the relevant battery data of the battery during the cyclic charging and discharging process.

[0003] In the related art, a deep learning model can be used to predict battery data at future times based on battery data at past times to determine the battery state change. For example, the battery voltage and temperature recorded over a period of time can be input into a deep learning model, and then the deep model can be used to predict the battery voltage and temperature to obtain the voltage and temperature at future times.

[0004] This method only uses battery data from the past to predict the same type of battery data in the future. The accuracy of the prediction results is easily affected by the performance of the deep learning model itself, thereby reducing the accuracy. Summary of the invention

[0005] The present application has been made in view of the above-mentioned problems.

[0006] According to a first aspect of the present application, a battery state prediction method is provided. The method includes: obtaining first battery data of a past period; determining second battery data of a future period, the second battery data being different from the first battery data type; determining a battery data time series sequence based on the first battery data of the past period and the second battery data of the future period; and inputting the battery data time series sequence into a time series prediction model to obtain first battery data of the future period.

[0007] According to another aspect of the present application, an electronic device is provided, including: a processor and a memory, wherein the memory stores computer program instructions, and the computer program instructions are used to execute the above-mentioned battery status prediction method when the processor is running.

[0008] According to another aspect of the present application, a storage medium is provided, on which program instructions are stored, and the program instructions are used to execute the above-mentioned battery status prediction method when running.

[0009] According to another aspect of the present application, a computer program product is provided, including computer program instructions, wherein the computer program instructions are used to execute the above-mentioned battery status prediction method when running.

[0010] In the above technical solution, the first battery data of the past period and the second battery data of the future period are obtained, the second battery data is different from the first battery data type, and the first battery data of the future period is predicted based on the first battery data of the past period and the second battery data of the future period. By considering the impact of other types of battery data in the future period on the battery data to be predicted, the battery data to be predicted in the future period is predicted, and the battery data to be predicted in the future period can be obtained more accurately, which helps to more accurately predict the battery status in the future period.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0013] Figure 1 A schematic block diagram showing an example electronic device for implementing the method and device for blurring the background of an image according to an embodiment of the present application;

[0014] Figure 2 A schematic flow chart of a battery state prediction method according to an embodiment of the present application is shown;

[0015] Figure 3 A schematic diagram of a time series prediction model according to an embodiment of the present application is shown;

[0016] Figure 4 A schematic flow chart of inputting a battery data time series sequence into a time series prediction model to obtain first battery data in a future period according to an embodiment of the present application is shown;

[0017] Figure 5 A schematic flow chart of determining a battery data time series sequence according to first battery data of a past period and second battery data of a future period according to one embodiment of the present application is shown;

[0018] Figure 6 A schematic diagram of an attention calculation process according to an embodiment of the present application is shown;

[0019] Figure 7 A schematic flow chart of determining a battery data time series sequence according to first battery data of a past period and second battery data of a future period according to one embodiment of the present application is shown;

[0020] Figure 8 A schematic flow chart of obtaining a battery data timing sequence according to each third subsequence and the corresponding second subsequence according to an embodiment of the present application is shown;

[0021] Fig. 9 A schematic diagram of a time series prediction model according to an embodiment of the present application is shown;

[0022] Fig.10 A schematic diagram showing a method of determining a battery data time series sequence according to first battery data in a past period and second battery data in a future period according to an embodiment of the present application is shown;

[0023] Fig.11 A schematic flow chart of training a time series prediction model according to an embodiment of the present application is shown;

[0024] Fig.12 A schematic flow chart of training a time series prediction model according to another embodiment of the present application is shown;

[0025] Fig.13 A schematic flow chart of training a time series prediction model according to another embodiment of the present application is shown;

[0026] Fig.14 A schematic block diagram of a battery state prediction device according to an embodiment of the present application is shown;

[0027] Fig.15 A schematic block diagram of an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present application more obvious, the example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application described in the present application, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present application.

[0029] In recent years, research on computer vision, deep learning, machine learning, image processing, image recognition and other technologies based on artificial intelligence has made important progress. Artificial Intelligence (AI) is an emerging science and technology that studies and develops theories, methods, technologies and application systems for simulating and extending human intelligence. Artificial intelligence is a comprehensive discipline involving many types of technologies such as chips, big data, cloud computing, the Internet of Things, distributed storage, deep learning, machine learning, neural networks, etc. Computer vision, as an important branch of artificial intelligence, specifically allows machines to recognize the world. Computer vision technology usually includes face recognition, unlocking, fingerprint recognition and anti-counterfeiting verification, biometric recognition, face detection, pedestrian detection, target detection, pedestrian recognition, image processing, image recognition, image semantic understanding, image retrieval, text recognition, video processing, video content recognition, 3D reconstruction, virtual reality, augmented reality, simultaneous localization and mapping (SLAM), computational photography, robot navigation and positioning and other technologies. With the research and advancement of artificial intelligence technology, this technology has been applied in many fields, such as security control, urban management, traffic management, building management, park management, facial access, facial attendance, logistics management, warehouse management, robots, intelligent marketing, computational photography, mobile phone imaging, cloud services, smart homes, wearable devices, unmanned driving, automatic driving, smart medical care, facial payment, facial unlocking, fingerprint unlocking, identity verification, smart screens, smart TVs, cameras, mobile Internet, live streaming, beauty, makeup, medical beauty, and smart temperature measurement.

[0030] As mentioned above, the charging and discharging process of rechargeable batteries will cause loss of battery performance. This characteristic can be reflected by battery status indicators, such as SOH (State of Health). SOH can be calculated by measuring the voltage and temperature of the battery during the cyclic charging and discharging process. Although a deep learning model can be used to predict battery data such as battery voltage and temperature to determine the battery status at a future moment, in the actual use of the battery, battery data such as battery voltage and temperature at a future moment are affected by other types of battery data. For example, the voltage at a future moment will be affected by the future artificially controlled current. At this time, if only the battery data such as the voltage at the past moment is predicted, the accurate voltage at the future moment may not be obtained, thereby affecting the accuracy of the battery status prediction result. Therefore, the present application provides a battery status prediction method, which uses a time series prediction model to predict the battery data that needs to be predicted at a future moment according to the battery data of a different type from the battery data that needs to be predicted at a future moment, so as to obtain accurate battery data that needs to be predicted at a future moment.

[0031] First, refer to Figure 1An example electronic device 100 for implementing the method and apparatus for blurring the background of an image according to an embodiment of the present application is described.

[0032] like Figure 1 As shown, the electronic device 100 includes one or more processors 102 and one or more storage devices 104. Optionally, the electronic device 100 may also include an input device 106 and an output device 108, and these components are interconnected via a bus system 110 and / or other forms of connection mechanisms (not shown). It should be noted that Figure 1 The components and structures of the electronic device 100 shown are merely exemplary and non-limiting. The electronic device may also have other components and structures as required.

[0033] The processor 102 can be implemented in at least one hardware form of a digital signal processor (DSP), a field programmable gate array (FPGA), a programmable logic array (PLA), and a microprocessor. The processor 102 can be a central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), or one or a combination of other forms of processing units with data processing capabilities and / or instruction execution capabilities, and can control other components in the electronic device 100 to perform desired functions.

[0034] The storage device 104 may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 102 may run the program instructions to implement the functions (implemented by the processor) in the embodiments of the present application described below. Various applications and various data, such as various data used and / or generated by the application, may also be stored in the computer-readable storage medium.

[0035] The input device 106 may be a device used by a user to input instructions, and may include one or more of a keyboard, a mouse, a microphone, a touch screen, and the like.

[0036] The output device 108 may output various information to the outside (eg, a user), and may include one or more of a display, a speaker, etc. Optionally, the input device 106 and the output device 108 may be integrated together and implemented using the same interactive device (eg, a touch screen).

[0037] Exemplarily, an example electronic device for implementing the battery status prediction method and apparatus according to an embodiment of the present application may be implemented on a device such as a personal computer or a remote server.

[0038] Figure 2 FIG. 2 shows a schematic flow chart of a battery state prediction method according to an embodiment of the present application. Figure 2 As shown, the battery status prediction method may include steps S210 to S240.

[0039] In step S210 , first battery data of a past period is acquired.

[0040] The first battery data is the battery data that needs to be predicted. For example, the first battery data may be the temperature, voltage, battery capacity, etc. of the battery. It is understandable that the first battery data may be related to the battery status indicator that needs to be determined. For example, when it is necessary to determine the battery status of the battery at a future moment, the first battery data may be voltage, temperature, etc., so as to determine the battery status after predicting the first battery data at a future moment. The first battery data of the past time period is the first battery data from the current moment to a period of time. According to the time interval of data sampling, the first battery data of the past time period may include the first battery data of multiple time points, and the first battery data of each time point has its corresponding data value. Therefore, the first battery data of the past time period may reflect the change of the first battery data in the past time period. The past time period may be a preset time period determined in advance, or it may be determined when executing the battery status prediction method. It is understandable that the more first battery data in the past time period, the more it can reflect the change of the first battery data over time, and thus the more conducive to the accuracy of the first battery data predicted for the future time period.

[0041] Exemplarily, the past period is a period of battery charging. During the battery charging process, relevant data detection can be performed on the battery to obtain the first battery data of the past period, such as the voltage and temperature of the battery.

[0042] In step S220 , second battery data for a future period is determined, where the second battery data is different from the first battery data type.

[0043] The second battery data of the future time period is the second battery data within a period of time after the current moment. The future time period is the time period in which the first battery data to be predicted is located. For example, when the first battery data is the battery voltage, the second battery data can be battery data other than the battery voltage, such as the battery temperature and current. According to the time interval of data sampling, the second battery data of the future time period may include the second battery data of multiple time points, and the second battery data of each time point has its corresponding data value. Therefore, the second battery data of the future time period can reflect the change of the second battery data in the future time period. The future time period can be a preset time period determined in advance, or it can be determined when executing the battery state prediction method. For example, when it is necessary to predict the first battery data within 10 minutes after the current moment, the second battery data of the future time period can be the second battery data within 10 minutes after the current moment. It can be understood that the more second battery data in the future time period, the more it can reflect the change of the second battery data over time, and thus the more it is conducive to improving the accuracy of the predicted first battery data of the future time period.

[0044] Exemplarily, the future period is also a period for charging the battery. It is understandable that for the charging process of the rechargeable battery, the second battery data in the entire charging process, such as the current of the battery during the charging process, can be preset according to the charging rule.

[0045] In step S230 , a battery data time series sequence is determined based on the first battery data of the past period and the second battery data of the future period.

[0046] The battery data time series sequence includes at least the first battery data of the past time period and the second battery data of the future time period. By way of example, the battery data time series sequence may also include the second battery data of the past time period. In the aforementioned example, the battery may be tested, such as current testing, in the past time period relative to the current moment to obtain the second battery data of the past time period. In this way, the complex relationship between the first battery data and the second battery data of the predicted battery may be determined more accurately to more accurately predict the first battery data of the future time period.

[0047] For example, the past time period and the future time period may be continuous. For example, when the current time is 5:00, the past time period is 4:50-5:00, and the future time period is 5:00-5:10, the time period of the past time period and the time period of the future time period are continuous. In this way, the connection between the first battery data of the past time period and the second battery data of the future time period can be determined more accurately.

[0048] In step S240, the battery data time series sequence is input into a time series prediction model to obtain first battery data of a future period.

[0049] The time series prediction model can be any transformer-type deep learning model for predicting future time data, such as PatchTST, or a deep learning model improved based on part of PatchTST. The time series prediction model can analyze the relationship between the first battery data of the past time period and the second battery data of the future time period in the battery data time series sequence to predict the first battery data of the future time period.

[0050] Before inputting the battery data time series sequence into the time series prediction model, the battery data time series sequence can also be processed, such as segmenting the battery data time series sequence, filtering the data, etc., to reduce the calculation amount of the time series prediction model and improve the prediction efficiency of the time series prediction model.

[0051] Exemplarily, the first battery data includes battery voltage and / or battery temperature, and the second battery data includes battery current.

[0052] It can be understood that the second battery data can be battery data that can be controlled and recorded in a future period, and has an impact on the change of the first battery data in the future period. In this way, when determining the battery data time series, the impact of the second battery data in the future period on the first battery data in the future period is taken into account, so that the first battery data in the future period can be predicted more accurately. For example, when it is necessary to predict the battery state (SOH) in the future period during the battery energy storage process, it is necessary to determine the voltage and / or temperature in the future period. The voltage and temperature in the future period will be affected by the artificially controlled current in the future period. At this time, the current in the future period can be determined in advance, and the battery data time series can be determined in combination with the voltage and temperature in the past period and the current in the future period. The battery data time series sequence can then be input into the time series prediction model to obtain the first battery data in the future period. The battery state (SOH) in the future period can be determined based on the first battery data in the future period.

[0053] In the above technical solution, the first battery data of the past period and the second battery data of the future period are obtained, the second battery data is different from the first battery data type, and the first battery data of the future period is predicted based on the first battery data of the past period and the second battery data of the future period. By considering the impact of other types of battery data in the future period on the battery data to be predicted, the battery data to be predicted in the future period is predicted, and the battery data to be predicted in the future period can be obtained more accurately, which helps to more accurately predict the battery status in the future period.

[0054] Exemplarily, the time series prediction model includes a vector extraction module and a processing module, wherein the vector extraction module and the processing module are connected in series in sequence, and the processing module includes a decoder. Figure 3FIG. 1 shows a schematic diagram of a time series prediction model according to an embodiment of the present application. Figure 3 As shown, the timing model may include a vector extraction module and a processing module, and the two are connected in series in sequence, wherein the processing module may include n decoders, and n is greater than or equal to 1. In the timing prediction model, compared with the encoder, the decoder has a flexible output generation mechanism, the application of the attention mechanism, and efficient parallel processing capabilities, which can improve the accuracy and efficiency of predicting the first battery data in the future period. Exemplarily, the decoder can be a decoder of the transformer architecture.

[0055] For example, Figure 3 As shown, the last decoder in the processing module may be connected to a fully connected layer, and the fully connected layer may be used to determine the first battery data of the future time period according to the features corresponding to the first battery data of the future time period output by the last decoder.

[0056] For example, Figure 3 As shown, each decoder may include a multi-head attention module, two normalization layers, and a convolutional layer. The multi-head attention module, one of the two normalization layers, the convolutional layer, and the other of the two normalization layers are sequentially connected in series.

[0057] Figure 4 FIG. 2 shows a schematic flow chart of inputting the battery data time series sequence into the time series prediction model to obtain the first battery data of the future period according to the above step S240 of one embodiment of the present application. Figure 4 As shown, the above step S240 may include steps S410 to S420.

[0058] In step S410, the battery data time series sequence is input into a vector extraction module to obtain a correlation vector corresponding to the first battery data of the past time period and the second battery data of the future time period, wherein the correlation vector includes a query vector (Query, Q), a key vector (Key, K) and a value vector (Value, V).

[0059] The vector extraction module can position encode and map the input battery data time series sequence to obtain a query vector, a key vector, and a value vector corresponding to the input battery data time series sequence. The vector module can convert the battery data time series sequence into a representation more suitable for processing by the processing module, and the position encoding is used to determine the time order information in the battery data time series sequence. For example, the battery data time series sequence data can be mapped to the input potential space of the encoder in the processing module through a trainable linear parameter matrix and a position encoding matrix.

[0060] The query vector is used to represent the contextual information of the current moment or position. It can tell the time series model which parts of the input battery data time series sequence should be focused on. In the process of prediction by the above time series prediction model, the association with the information at the current time point will be found from the historical data (such as the first battery data in the past period) according to the query vector. The key vector is a vector corresponding to each position or time step in the sequence, which helps determine the relative importance of other positions. In time series prediction, the key vector contains the feature representation of each time point in the battery data time series sequence. The above time series prediction model can determine which time points have higher correlation between battery data by comparing the query vector and the key vector. The value vector contains the actual data or information. If a time point is determined to be highly correlated with the query vector, its value vector will be used to construct the final output. In the process of prediction by the above time series prediction model, the value vector carries the detailed features of the corresponding time point, which will be used to generate the prediction results. Among them, the query vector, key vector and value vector can be represented in matrix form.

[0061] In step S420, the portion of the query vector corresponding to the second battery data of the future period, the key vector and the value vector are input into the processing module to obtain the first battery data of the future period.

[0062] The decoder in the processing module can perform attention mechanism calculation according to the query vector and the key vector to obtain an attention weight matrix. The attention weight matrix calculated in this way only contains the weights of the second battery data in the future period and the first battery data in the past period.

[0063] The decoder may also perform matrix multiplication calculation on the attention weight matrix and the value vector to obtain the features of the first battery data information containing the future period, wherein the features of the first battery data information containing the future period may be represented in a matrix form.

[0064] For example, Figure 3 As shown, when the processing module includes multiple decoders, the multiple decoders are connected in sequence, and the part of the query vector corresponding to the second battery data of the future time period, the key vector and the value vector can be input into the first encoder of the processing module to obtain the first feature, that is, the feature containing the first battery data information of the future time period. For each encoder connected in sequence after the first encoder, a new first feature will be output according to the entirety of the query vector, the key vector and the value vector corresponding to the first feature output by the previous encoder.

[0065] Exemplarily, when the processing module includes multiple decoders, a linear layer may be connected between each decoder to output the corresponding query vector q, key vector k and value vector v according to the first feature output by the previous decoder.

[0066] For example, Figure 3 Each decoder of the processing module shown can determine the feature attention (q, k, v) corresponding to the first battery data of the future period according to the following formula 1:

[0067]

[0068] Among them, k represents the key vector, T represents the current time, d represents the feature dimension of the model, and v represents the value vector. represents the attention weight matrix after the attention mechanism is calculated based on the query vector and the key vector. When the time series prediction model is as follows Figure 3 As shown, for the first decoder, q represents the portion of the query vector corresponding only to the second battery data of the future period, and for other decoders except the first decoder, q represents the entire query vector.

[0069] In the above technical solution, the vector extraction module of the time series prediction model is used to extract the query vector, key vector and value vector corresponding to the first battery data of the past period and the second battery data of the future period, and the processing module of the time series prediction model is used to determine the first battery data of the future period according to the part of the query vector corresponding to the second battery data of the future period, the key vector and the value vector. In this way, the processing module of the time series prediction model can focus on the correlation between the second battery data of the future period and the first battery data of the future period, and can predict more accurate first battery data of the future period.

[0070] Exemplarily, the past period includes multiple first moments, and the future period includes multiple second moments. It is understandable that when actually collecting data in a certain period, the moment at which the data collected in the period is located will be affected by the sampling interval when collecting the data, and for the data in the same period, the smaller the sampling interval, the more moments at which the data collected in the period is located, and the more data volume collected in the period. Therefore, the number of first moments in the past period and the number of multiple second moments in the future period can be determined according to the actual sampling interval. Figure 5 FIG. 2 shows a schematic flow chart of determining a battery data timing sequence according to the first battery data of a past period and the second battery data of a future period in step S230 according to an embodiment of the present application. Figure 5 As shown, the above step S230 may include steps S510 to S520.

[0071] In step S510, the first battery data of the past time period and the second battery data of the future time period are concatenated to obtain a first data sequence including the first battery data of the past time period and the second battery data of the future time period, wherein each first data sequence includes a single type of first battery data and a single type of second battery data.

[0072] In the first data sequence, the data at each moment is arranged in order from front to back according to the order of the moments. For example, when the first battery data is voltage and the second battery data is current, when the voltage of the past period can be {V T-k …V T}, the second battery data in the future period can be {A T+1 …A T+n}. Tk to T correspond to the first moment of the past period, and T+1 to T+n correspond to the second moment of the future period. T-k …V T} and {A T+1 …A T+n} are concatenated to obtain the first data sequence {V T-k ·…V T ,A T+1 …A T+n Each first data sequence includes a single type of first battery data and a single type of second battery data, which can avoid the influence of other irrelevant data on the information in the first data sequence. For example, the battery temperature in the future period will also be affected by the current in the future period. If the temperature of the past period is also spliced ​​into the first data sequence, the temperature of the past period may not affect the voltage in the future period. However, the presence of temperature data of the past period in the first data sequence may cause the time series prediction model to associate the temperature of the past period with the voltage of the past period, thereby affecting the prediction of the voltage of the future period by the time series prediction model.

[0073] In step S520, the first data sequence is segmented according to the first time window to obtain a battery data timing sequence including a plurality of first subsequences, wherein each of the plurality of first subsequences includes a plurality of first battery data at a first moment or a plurality of second battery data at a second moment.

[0074] The time corresponding to the first time window may include multiple moments, and may be a preset time window, or may be configured when the time series prediction model performs prediction. Splitting the first data sequence according to the first time window can reduce the complexity of the attention mechanism calculation of the time series prediction model. For example, the first data sequence is {V T-k …V T ,A T+1 …A T+n}, each first time window can contain 3 moments, and the battery data time series after segmentation can be:

[0075] {patch(V T-k ,V T-k+1 ,V T-k+2 )…patch(V T-2 ,V T-1 ,V T ),

[0076] patch(A T+1 ,A T+2 ,A T+3 )…patch(A T+n-2 ,A T+n-1 ,A T+n )}.

[0077] Among them, patch(V T-k ,V T-k+1 ,V T-k+2 ), patch(V T-2 ,V T-1 ,V T ), patch(A T+1 ,A T+2 ,A T+3 ), patch(A T+n-2 ,A T+n-1 ,A T+n ) are the first subsequences in the battery data timing sequence.

[0078] In the above technical solution, the first battery data of the past period and the second battery data of the future period are spliced ​​to obtain a first data sequence, and the first data sequence is segmented according to the first time window to obtain a battery data time series sequence containing multiple first subsequences. This can avoid the influence of irrelevant battery data on the time series prediction model predicting the first battery in the future period, and can also reduce the complexity of the attention mechanism calculation of the time series prediction model, so as to improve the prediction efficiency and accuracy of the time series prediction model.

[0079] For example, the above Figure 3 The timing prediction model shown may also include a first submodule, and the above step S520 may also be completed by the first submodule. For example, when the voltage and temperature of the future time period need to be predicted during the battery charging process, the spliced ​​first data sequence {V T-k …V T ,A T+1 …A T+n}, the first data sequence {TT-k …T T ,A T+1 …A T+n} is input into the time series prediction model. The first submodule of the time series prediction model can respectively convert {V T-k …V T ,A T+1 …A T+n} and {T T-k …T T ,A T+1 …A T+n} are normalized and segmented to obtain a battery time series containing the temperature of the past period and the current of the future period, and a battery time series containing the voltage of the past period and the current of the future period. Then, the time series prediction model can positionally encode the first subsequences in the two battery time series and map out the query vector, key vector and value vector corresponding to each battery time series. Figure 6 FIG. 1 is a schematic diagram showing an attention calculation process according to an embodiment of the present application. Figure 6 As shown, for Figure 3 The first decoder of the processing module of the timing prediction model shown can determine the features of the first battery data information containing the future time period outputted by it according to the part of the query vector q corresponding to each battery timing sequence that only corresponds to the second battery data of the future time period, the key vector k, the value vector v and the above formula 1. The features corresponding to the currents of N future time periods in the q matrix can be retained as a new q matrix, so that q∈R N*d . Use the new q matrix obtained and k∈R n*d Perform matrix multiplication to obtain the attention weight matrix. t+1,1 Represents the relationship weight between the feature at time T = t + 1 and the feature at T = 1. The attention weight matrix and v∈R n*d Continue to perform matrix multiplication to output the feature x containing the first battery data information of the future period out ∈R N *d .

[0080] The other decoders after the first decoder can be sequentially decoded according to the features x output by the previous decoder. out ∈R N*d The corresponding query vector, key vector and value vector output a new feature x out ∈R N*d When the first data sequence input contains the voltage of the past period, x out is a feature containing voltage information for a future period. When the first input data sequence contains the temperature for a past period, x out is a feature containing temperature information for future periods. Figure 3 The last decoder output feature x of the temporal model shown is out After that, the fully connected layer will convert x out Mapped to the first battery data of the final predicted future period and output. out When x is a feature containing voltage information for future periods, the fully connected layer will convert x out Mapped to the voltage of the future period; when x out When x is a feature containing temperature information for future periods, the fully connected layer will convert x out Mapped as temperature for future time periods.

[0081] Exemplarily, the above-mentioned battery status prediction method may further include step S221: acquiring second battery data of a past period. Figure 7 FIG. 2 shows a schematic flow chart of determining a battery data timing sequence according to the first battery data of a past period and the second battery data of a future period in step S230 according to an embodiment of the present application. Figure 7 As shown, the above step S230 may include steps S710 to S730.

[0082] In step S710, for each type of first battery data, the first battery data of the type in the past period is segmented according to the second time window to obtain a second data sequence including a plurality of second subsequences.

[0083] The second time window can be the same size as the first time window, or it can be different. The time corresponding to the second time window can include multiple moments, and can be a preset time window, or it can be configured when the time series prediction model performs prediction. It can be understood that the amount of data of the first battery data in each second subsequence is affected by the data sampling interval and the size of the second time window. The smaller the data sampling interval, the larger the amount of data of the first battery data in each second subsequence, and the larger the second time window, the larger the amount of data of the first battery data in each second subsequence.

[0084] For each type of first battery data, a second data sequence including multiple second subsequences can be obtained respectively, so that the time series prediction model can predict multiple types of first battery data at the same time, and each type of first battery data will not interfere with each other.

[0085] In step S720, a third subsequence corresponding to each of the multiple second subsequences is determined based on the third time window, the second battery data of the past time period, and the second battery data of the future time period, wherein each third subsequence includes the second battery data of the third time window, and the length of the third time window is the same as the length of the future time period.

[0086] The third time window may be the same size as the first time window or the second time window, or may be different. The time corresponding to the second time window may include multiple moments, and may be a preset time window, or may be configured when the time series prediction model performs prediction.

[0087] Each third subsequence may contain the second battery data at the same time point. For example, when the sequence {A T-1 ,A T ,A T+1}, if the third time window contains 2 time points, the sequence can be divided into the third subsequence {A T-1 ,A T} and the third subsequence {A T ,A T+1}. At this time, two third subsequences with the same partial data can be obtained. That is, when there are second battery data of X time points in the sequence containing the second battery data of the past period and the second battery data of the future period, and each third time window contains Y time points, the sequence can be divided into X-Y+1 third subsequences.

[0088] Each of the multiple second subsequences corresponds to a third subsequence, so multiple third subsequences can be obtained. When the number of time points contained in the third time window is the same as the number of time points corresponding to the future time period, the duration of the third time window can also be regarded as the same as the duration of the future time period. For example, when the duration of the third time window is 10s and the future time period is 10s after the current moment, the duration of the third time window is the same as the duration of the future time period. If the duration of the third time window is 10s, the sampling interval is 5s, and the future time period is 14s after the current moment, each third subsequence may only have the second battery data of 2 time points. Although the duration of the third time window is different from the duration of the future time period, due to the sampling interval, the predicted first battery data of the future time period is still the first battery data of 2 time points. In this case, the duration of the third time window can also be regarded as the same as the duration of the future time period.

[0089] In step S730, a battery data timing sequence is obtained according to each third subsequence and the corresponding second subsequence.

[0090] Each third subsequence and the corresponding second subsequence can be spliced ​​to directly obtain a battery data time series sequence. The spliced ​​result can also be further processed to obtain a sequence as the battery data time series sequence to improve the efficiency of the time series prediction model in predicting the first battery data in the future period.

[0091] In the above technical solution, the second battery data of the past time period is obtained, and the first battery data of the same type of the past time period is divided according to the second time window to obtain a second data sequence containing multiple second subsequences; according to the third time window, the second battery data of the past time period and the second battery data of the future time period, the third subsequence corresponding to each of the multiple second subsequences is determined; according to each third subsequence and the corresponding second subsequence, the battery data time series sequence is obtained. By linking the second battery data of the past time period, the first battery data of the past time period and the second battery data of the future time period to obtain the battery data time series sequence, it is beneficial for the time series prediction model to also pay attention to the correlation between the second battery data of the past time period and the first battery data of the past time period when making predictions, so that the first battery data of the future time period predicted by the time series prediction model is more accurate.

[0092] Figure 8 FIG. 4 is a schematic flow chart showing the step S730 of obtaining a battery data timing sequence according to each third subsequence and the corresponding second subsequence according to an embodiment of the present application. Figure 7 As shown, the above step S730 may include step S731 to step S732.

[0093] In step S731, each third subsequence is concatenated with the corresponding second subsequence to obtain multiple fourth subsequences, wherein the fourth subsequence includes the first battery data of the second time window and the second battery data of the third time window, and in each fourth subsequence, the time corresponding to the second time window is before the time corresponding to the third time window and is continuous with the time corresponding to the third time window.

[0094] It can be understood that the time points corresponding to the data in the second subsequence are continuous, and the time points corresponding to the data in the third subsequence are also continuous. Therefore, when the time corresponding to the second time window is before the time corresponding to the third time window and is continuous with the time corresponding to the third time window, the time points corresponding to the data in the fourth subsequence will be continuous time points. And the number of corresponding time points in each fourth subsequence is the sum of the number of time points in the second time window and the number of time points in the third time window.

[0095] In step S732, multiple fourth subsequences are concatenated to obtain a battery data timing sequence.

[0096] In each fourth sequence of the battery data timing sequence, the first battery data at a time point of a partial past period is included, and the second battery data in a third window after the time point of the partial past period is included. The first battery data of the partial past period and the second battery data in the third window after the time point of the partial past period can be associated.

[0097] In the above technical solution, each third subsequence is spliced ​​with the corresponding second subsequence to obtain multiple fourth subsequences, and multiple fourth subsequences are spliced ​​to obtain a battery data time series sequence. When the time series prediction model makes a prediction based on the input battery data time series sequence, it can pay attention to the correlation between the first battery data in the partial past time period in each fourth sequence and the second battery data in the time period after the partial past time period in the fourth sequence, so as to more accurately predict the first battery data in the future time period.

[0098] Fig. 9 FIG. 1 shows a schematic diagram of a time series prediction model according to an embodiment of the present application. Figure 3 As shown, the timing model may include a vector extraction module and a processing module, and the two are connected in series in sequence, wherein the processing module may include n encoders, and n is greater than or equal to 1. Exemplarily, the encoder may be an encoder of a transformer architecture. For example, the timing prediction model may be a PatchTST model.

[0099] For example, Fig. 9 As shown, when the processing module includes multiple encoders, the multiple encoders are connected in sequence, and the query vector, key vector and value vector can be input into the first encoder of the processing module to obtain the first feature, that is, the feature containing the first battery data information of the future time period. For each encoder connected in sequence after the first encoder, a new first feature will be output according to the query vector, key vector and value vector corresponding to the first feature output by the previous encoder.

[0100] For example, Fig. 9 As shown, the last encoder in the processing module may be connected to a fully connected layer, and the fully connected layer may be used to determine the first battery data of the future time period according to the features corresponding to the first battery data of the future time period output by the last encoder.

[0101] For example, Fig. 9 As shown, each encoder may include a multi-head attention module, two normalization layers, and a convolution layer. The multi-head attention module, one of the two normalization layers, the convolution layer, and the other of the two normalization layers are sequentially connected in series.

[0102] Exemplarily, when the processing module includes multiple encoders, a linear layer may be connected between each encoder to output the corresponding query vector q, key vector k and value vector v according to the first feature output by the previous encoder.

[0103] Fig.10FIG. 2 is a schematic diagram showing a method of determining a battery data timing sequence according to the first battery data of a past period and the second battery data of a future period in step S230 according to an embodiment of the present application. Fig. 9 The time series prediction model shown may also include a second submodule, and the above steps S710 to S730 may also be completed by the second submodule.

[0104] like Fig.10 As shown in the figure, during the battery charging process, when the voltage of the future period needs to be predicted, the voltage sequence of the past period {V T-k …V T}、Contains the current sequence of the past period and the future period {A T-k …A T , A T+1 …A T+n}, and input these two sequences into the time series prediction model. The second submodule of the time series prediction model can normalize and segment the input sequence according to the process in the above steps S710 to S730 to obtain the battery data time series sequence. For example, when the size of the second time window is 3 (i.e., it includes 3 time points), {V T-k …V T} is split into {patch(V T-k , V T-k+1 , V T-k+2 )…patch(V T-2 , V T-1 , V T )}, where each patch() sequence is a second subsequence. When the future period includes 3 time points, the size of the third time window is 3 (i.e., it includes 3 time points), and {A T-k …A T , A T+1 …A T+n} is split into {patch(A T-k , A T-k+1 , A T-k+2 ), patch(A T-k+1 , A T-k+2 , A T-k+3 ),…patch(A T-2 , A T-1 , A T )). Among them, for the second subsequence patch (V T-k , V T-k+1 , V T-k+2 ), the corresponding third subsequence is patch(A T-k+3 , A T - k+4 , A T-k+5The segmented sequences can be spliced ​​to obtain the battery data timing sequence: {patch(V T-k , V T-k+1 , V T-k+2 , A T-k+3 …A T-k+2+n )…patch(V T-2 , V T-1 , V T , A T+1 …A T+n )). At this time, the size of the time window corresponding to each fourth subsequence in the battery data time series is the sum of the second time window and the third time window in the fourth subsequence.

[0105] Then the vector extraction module of the time series prediction model can convert the battery data time series x in matrix form into in ∈R N *(pl+n) Mapped to x in ∈R N*d . Where pl is the size of the second time window, n is the length of the sequence to be predicted by the model, and d is the feature dimension of the model. in ∈R N*d After the position encoding is performed by the vector extraction module, it can be input into the processing module and calculated using multiple encoders of the processing module to obtain the feature x containing the voltage information of the future period. out ∈R N*d Finally, x out After expansion, it is sent to the fully connected layer, and x out The dimension is mapped from 1*Nd to 1*n, where n is the length of the sequence to be predicted, that is, the time point in the future period, to obtain the voltage in the future period.

[0106] Similarly, when it is necessary to predict the temperature in the future period, the voltage can be replaced with temperature and the same process can be performed.

[0107] Fig.11 FIG. 1 shows a schematic flow chart of a training time series prediction model according to an embodiment of the present application. Fig.11 As shown, the above-mentioned battery status prediction method may further include steps S810 to S840.

[0108] In step S810, first battery data of a first preset period, first battery data of a first target period, and second battery data of the first target period are acquired, wherein the first preset period is before the first target period.

[0109] The first battery data of the first preset period, the first battery data of the first target period, and the second battery data of the first target period may be existing real first battery data and second battery data.

[0110] In step S820 , a first training battery data timing sequence is determined according to the first battery data of the first preset period and the second battery data of the first target period.

[0111] The first battery data of the first preset time period can be used as the first battery data of the past time period of the above step S230, and the second battery data of the first target time period can be used as the second battery data of the future time period, and step S820 can be executed according to the process of the above step S230, which will not be described in detail here.

[0112] In step S830, the first training battery data time series sequence is input into the time series prediction model to obtain first predicted battery data for the first target period.

[0113] Step S830 is similar to the above step S240 and will not be described in detail here. It can be understood that in order to enable the time series prediction model to better predict the first battery data of the future time period, the duration corresponding to the first target time period can be the same as the duration corresponding to the future time period.

[0114] In step S840, parameters of the time series prediction model are adjusted according to the difference between the first battery data and the first predicted battery data in the first target period.

[0115] Exemplarily, when the first battery data is voltage and temperature, the loss value between the first battery data and the first predicted battery data in the first target period can be determined according to the following formula 2:

[0116]

[0117] Among them, V i represents the voltage of the first target period, represents the first predicted battery data corresponding to the voltage, T i represents the temperature of the first target period, The first predicted battery data corresponding to the temperature is represented.

[0118] In the above technical solution, the first battery data of the first target time period is predicted by a timing prediction model, and the parameters of the timing prediction model are adjusted according to the difference between the predicted first predicted battery data and the first battery data of the first target time period, so as to improve the performance of the timing prediction model.

[0119] Fig.12 FIG. 1 shows a schematic flow chart of a training time series prediction model according to an embodiment of the present application. Fig.12 As shown, the above-mentioned battery status prediction method may further include steps S910 to S930.

[0120] In step S910, first battery data of a second preset period, first battery data of a third preset period, first battery data of a second target period, second battery data of a third preset period, and second battery data of a second target period are acquired, wherein the third preset period is before the second target period.

[0121] The first battery data of the second preset period, the first battery data of the third preset period, the first battery data of the second target period, the second battery data of the third preset period and the second battery data of the second target period may be existing real first battery data and second battery data.

[0122] In step S920, the time series prediction model is pre-trained according to the first battery data of the second preset time period.

[0123] The first battery data of the second preset time period can be input into the timing prediction model, and the prediction result obtained by the timing model is compared with the first battery data of the second prediction time period to pre-train the timing prediction model, thereby optimizing the timing prediction model.

[0124] In step S930, the time series prediction model is adjusted according to the first battery data of the third preset period, the first battery data of the second target period, the second battery data of the third preset period, and the second battery data of the second target period.

[0125] The first battery data of the third preset time period, the first battery data of the second target time period, the second battery data of the third preset time period, and the second battery data of the second target time period can be input into the timing prediction model, and the prediction result obtained by the timing model can be compared with the first battery data of the second target time period to pre-train the timing prediction model, thereby optimizing the timing prediction model.

[0126] In the above technical scheme, the time series prediction model is pre-trained and adjusted according to the first battery data of the second preset time period, the first battery data of the third preset time period, the first battery data of the second target time period, the second battery data of the third preset time period and the second battery data of the second target time period, so as to improve the prediction performance of the time series prediction model.

[0127] Fig.13 FIG. 1 shows a schematic flow chart of a training time series prediction model according to an embodiment of the present application. Fig.13 As shown, the above step S920 may include steps S921 to S924, and the above step S930 may include steps S931 to S933.

[0128] In step S921, a portion of the data corresponding to the first moment in the first battery data of the second preset period is masked to obtain the masked first battery data of the second preset period.

[0129] The time series prediction model will not obtain the data information of the masked portion of the first battery data in the second preset time period.

[0130] In step S922, a second training battery data time series sequence is determined based on the masked first battery data of the second preset period.

[0131] Step S922 is similar to the above step S230 and will not be described in detail here. The time point corresponding to the masked first battery data still exists in the second training battery data time series sequence. In step S923, the second training battery data time series sequence is input into the time series prediction model to obtain the second predicted battery data for the second preset time period.

[0132] Step S922 is similar to the above step S240 and will not be described in detail here. It can be understood that in order to enable the time series prediction model to better predict the first battery data of the future time period, the duration corresponding to the first target time period can be the same as the duration corresponding to the past time period.

[0133] In step S924, the parameters of the time series prediction model are adjusted according to the difference between the first battery data and the second predicted battery data in the second preset period.

[0134] By masking the first battery data of the second preset time period and aligning the prediction using the time series prediction model, the adjusted time series prediction model can focus on the battery data information of the past time period.

[0135] Exemplarily, the loss value between the first battery data and the second predicted battery data for the second preset period may be determined according to the following formula 3:

[0136]

[0137] Among them, V i represents the voltage of the second preset period, The second predicted battery data corresponding to the voltage, T i represents the temperature of the second preset period, The second predicted battery data corresponding to the temperature can be adjusted according to the loss value obtained by formula 3.

[0138] In step S931 , a third training battery data timing sequence is determined according to the first battery data of the third preset period, the second battery data of the third preset period, and the second battery data of the second target period.

[0139] Here, the first battery data of the third preset period can be used as the first battery data of the past period, the second battery data of the third preset period can be used as the second battery data of the past period, and the second battery data of the second target period can be used as the second battery data of the future period, and the processing of the above steps S710 to S730 is performed. Step S931 is similar to the above steps S710 to S730, and will not be described in detail here.

[0140] In step S932, the third training battery data time series sequence is input into the time series prediction model to obtain third predicted battery data for the second target period.

[0141] Step S932 is similar to the above step S240 and will not be described in detail here. It can be understood that in order to enable the time series prediction model to better predict the first battery data of the future time period, the duration corresponding to the second target time period can be the same as the duration corresponding to the past time period.

[0142] In step S933 , the parameters of the time series prediction model are adjusted according to the difference between the second battery data and the third predicted battery data in the second target period.

[0143] Exemplarily, the loss value between the first battery data and the second predicted battery data for the second preset period may be determined according to the following formula 4:

[0144]

[0145] Among them, V i represents the voltage of the second target period, The third predicted battery data corresponding to the voltage, T i represents the temperature of the second target period, The third predicted battery data corresponding to the temperature can be represented. The parameters of the time series prediction model can be adjusted according to the loss value. The parameters of the time series prediction model can be adjusted according to the loss value obtained according to Formula 4.

[0146] In the above technical method, the time series prediction model is first used to predict the masked first battery data and adjust the parameters, and then the time series prediction model is used to predict the first battery data of the second target time period and adjust the parameters again, so that the adjusted time series prediction model can accurately predict the first battery data of the future time period.

[0147] Compared with the conventional time series prediction model, the method in the embodiment of the present application is as follows Figure 3 The time series prediction model (model 1) shown in Fig. 9 The time series prediction model (model 2) shown, combined with the above-mentioned battery state prediction method, can predict more accurate first battery data of the future time period.

[0148] Table 1

[0149]

[0150] Table 1 shows the results of the comparative experiment. When the input length and the prediction length are the same, the mean absolute error and mean square error of Model 1 and Model 2 are smaller than those of the conventional time series prediction model. The input length refers to the time length of the past period, and the prediction length refers to the time length of the future period.

[0151] According to another aspect of the present application, a battery status prediction device is provided. Fig.14 A schematic block diagram of a battery state prediction device 1000 according to an embodiment of the present application is shown.

[0152] like Fig.14 As shown, the battery state prediction device 1000 according to the embodiment of the present application includes a first acquisition module 1010, a second acquisition module 1020, a determination module 1030, and a prediction module 1040. Each module can respectively execute the above Figure 2 The following describes the steps of the method for blurring the background of an image. The following only describes the main functions of the components of the device 1000 for blurring the background of an image, and omits the details described above.

[0153] The first acquisition module 1010 is used to acquire the first battery data of the past period. Figure 2 The processor 202 in the electronic device shown executes program instructions stored in the storage device 204 to implement.

[0154] The second acquisition module 1020 is used to determine the second battery data in the future period, and the second battery data is different from the first battery data type. Figure 2 The processor 202 in the electronic device shown executes program instructions stored in the storage device 204 to implement.

[0155] The determination module 1030 is used to determine the battery data time series according to the first battery data in the past period and the second battery data in the future period. Figure 2 The processor 202 in the electronic device shown executes program instructions stored in the storage device 204 to implement.

[0156] The prediction module 1040 is used to input the battery data time series sequence into the time series prediction model to obtain the first battery data in the future period. Figure 2 The processor 202 in the electronic device shown executes program instructions stored in the storage device 204 to implement.

[0157] Fig.15A schematic block diagram of an electronic device 1100 according to an embodiment of the present application is shown. The electronic device 1100 includes a memory 1110 and a processor 1120 .

[0158] The memory 1110 stores computer program instructions for implementing corresponding steps in the battery state prediction method according to the embodiment of the present application.

[0159] The processor 1120 is used to run the computer program instructions stored in the memory 1110 to perform corresponding steps of the battery state prediction method according to the embodiment of the present application.

[0160] In one embodiment, the computer program instructions are used by the processor 1120 to perform the following steps when executed: obtaining first battery data for a past time period; determining second battery data for a future time period, the second battery data being different from the first battery data type; determining a battery data time series sequence based on the first battery data for the past time period and the second battery data for the future time period; and inputting the battery data time series sequence into a time series prediction model to obtain first battery data for the future time period.

[0161] In addition, according to another aspect of the present application, a storage medium is also provided, on which program instructions are stored, and when the program instructions are executed by a computer or a processor, the computer or the processor executes the corresponding steps of the above-mentioned battery state prediction method of the embodiment of the present application, and is used to implement the corresponding modules in the above-mentioned battery state prediction device according to the embodiment of the present application. The storage medium may include, for example, a storage component of a tablet computer, a hard disk of a personal computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0162] According to another aspect of the present application, a computer program product is provided, including computer program instructions, wherein the computer program instructions are used to execute the above-mentioned battery status prediction method when running.

[0163] A person skilled in the art can understand the specific implementation and beneficial effects of the above-mentioned battery status prediction device, electronic device, storage medium and computer program product by reading the above-mentioned detailed description of the battery status prediction method. For the sake of brevity, they will not be repeated here.

[0164] Although example embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above example embodiments are merely exemplary and are not intended to limit the scope of the present application to this. Those of ordinary skill in the art may make various changes and modifications therein without departing from the scope and spirit of the present application. All these changes and modifications are intended to be included within the scope of the present application as required by the appended claims.

[0165] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0166] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0167] In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0168] Similarly, it should be understood that in order to streamline the present application and help understand one or more of the various application aspects, in the description of the exemplary embodiments of the present application, the various features of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the method of the present application should not be interpreted as reflecting the following intention: the claimed application requires more features than the features clearly stated in each claim. More specifically, as reflected in the corresponding claims, the application point is that the corresponding technical problem can be solved with less than all the features of a single disclosed embodiment. Therefore, the claims following the specific embodiment are hereby explicitly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present application.

[0169] It will be understood by those skilled in the art that, except for mutually exclusive features, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this specification may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0170] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0171] The various component embodiments of the present application may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all of the functions of some modules in the battery state prediction device according to the embodiments of the present application. The present application may also be implemented as a device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present application may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0172] It should be noted that the above embodiments illustrate the present application rather than limit the present application, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets should not be constructed as a limitation to the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The present application may be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In a unit claim that lists several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names.

[0173] The above is only a specific implementation or description of a specific implementation of the present application, and the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. The protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A battery status prediction method, characterized in that: The method comprises: Acquire first battery data of a past period; determining second battery data for a future time period, the second battery data being different from the first battery data type; determining a battery data time series sequence according to the first battery data of the past time period and the second battery data of the future time period; and The battery data time series sequence is input into a time series prediction model to obtain the first battery data of the future time period.

2. The method according to claim 1, characterized in that The time series prediction model includes a vector extraction module and a processing module, wherein the vector extraction module and the processing module are connected in series in sequence, and the processing module includes a decoder.

3. The method according to claim 2, characterized in that The step of inputting the battery data time series sequence into a time series prediction model to obtain the first battery data of the future time period includes: Inputting the battery data time series sequence into the vector extraction module to obtain a correlation vector corresponding to the first battery data of the past period and the second battery data of the future period, wherein the correlation vector includes a query vector, a key vector and a value vector; The portion of the query vector corresponding to the second battery data of the future time period, the key vector and the value vector are input into the processing module to obtain the first battery data of the future time period.

4. The method according to any one of claims 1 to 3, characterized in that: The past period includes a plurality of first moments, and the future period includes a plurality of second moments. The determining of the battery data time series according to the first battery data of the past time period and the second battery data of the future time period comprises: concatenating the first battery data of the past time period and the second battery data of the future time period to obtain a first data sequence including the first battery data of the past time period and the second battery data of the future time period, wherein each first data sequence includes a single type of the first battery data and a single type of the second battery data; The first data sequence is segmented according to a first time window to obtain a battery data timing sequence including a plurality of first subsequences, wherein each of the plurality of first subsequences includes a plurality of first battery data at the first moment or a plurality of second battery data at the second moment.

5. The method according to claim 1, characterized in that The method further comprises: Acquire second battery data of a past period; The determining of the battery data time series according to the first battery data of the past time period and the second battery data of the future time period comprises: For each type of first battery data, segment the first battery data of the type in the past period according to the second time window to obtain a second data sequence including a plurality of second subsequences; Determine, according to the third time window, the second battery data of the past time period, and the second battery data of the future time period, a third subsequence corresponding to each of the plurality of second subsequences, wherein each of the third subsequences includes the second battery data of the third time window, and the duration of the third time window is the same as the duration of the future time period; The battery data timing sequence is obtained according to each third subsequence and the corresponding second subsequence.

6. The method according to claim 5, characterized in that The step of obtaining the battery data timing sequence according to each third subsequence and the corresponding second subsequence includes: splicing each of the third subsequences with the corresponding second subsequence respectively to obtain a plurality of fourth subsequences, wherein the fourth subsequences include the first battery data of the second time window and the second battery data of the third time window, and in each of the fourth subsequences, the time corresponding to the second time window is before the time corresponding to the third time window and is continuous with the time corresponding to the third time window; The plurality of fourth subsequences are concatenated to obtain the battery data timing sequence.

7. The method according to claim 1, characterized in that The method further comprises: Acquire first battery data of a first preset period, first battery data of a first target period, and second battery data of the first target period, wherein the first preset period is before the first target period; Determining the first training battery data timing sequence according to the first battery data of the first preset time period and the second battery data of the first target time period; Inputting the first training battery data time series sequence into the time series prediction model to obtain first predicted battery data for the first target period; According to the difference between the first battery data and the first predicted battery data in the first target period, the parameters of the time series prediction model are adjusted.

8. The method according to claim 1, characterized in that The method further comprises: Acquire the first battery data of the second preset period, the first battery data of the third preset period, the first battery data of the second target period, the second battery data of the third preset period, and the second battery data of the second target period, wherein the third preset period is before the second target period; Pre-training the time series prediction model according to the first battery data of the second preset time period; The time series prediction model is adjusted according to the first battery data of the third preset time period, the first battery data of the second target time period, the second battery data of the third preset time period and the second battery data of the second target time period.

9. The method according to claim 8, characterized in that The pre-training of the time series prediction model according to the first battery data of the second preset time period includes: Masking a portion of the data corresponding to the first moment in the first battery data of the second preset time period to obtain the masked first battery data of the second preset time period; Determine a second training battery data timing sequence according to the masked first battery data of the second preset time period; Inputting the second training battery data time series sequence into the time series prediction model to obtain second predicted battery data for the second preset time period; adjusting the parameters of the time series prediction model according to the difference between the first battery data and the second predicted battery data in the second preset time period; The adjusting the time series prediction model according to the first battery data of the third preset time period, the first battery data of the second target time period, the second battery data of the third preset time period, and the second battery data of the second target time period includes: Determine a third training battery data timing sequence according to the first battery data of the third preset period, the second battery data of the third preset period, and the second battery data of the second target period; Inputting the third training battery data time series sequence into the time series prediction model to obtain third predicted battery data for the second target period; According to the difference between the second battery data of the second target period and the third predicted battery data, the parameters of the time series prediction model are adjusted.

10. The method according to claim 1, characterized in that The first battery data includes a battery voltage and / or a battery temperature, and the second battery data includes a battery current.

Images