Power shortage prediction method and device and electronic equipment

Abstract

The embodiment of the invention discloses a power shortage prediction method and device and electronic equipment. The method comprises the following steps: acquiring current vehicle data of a vehicle; and based on the cabin domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, obtaining a current power shortage prediction result of the vehicle. Through the mode, the data of the plurality of domain controllers in the vehicle can be fused with the vehicle operation data, so that the system architecture of the current vehicle is better fitted, and the accuracy of power shortage prediction is improved. Moreover, the linear regression model is obtained by fusing decision tree models of a plurality of fields, so that the linear regression model can execute different power-lack risk prediction strategies on the data of each field, and can also perform overall power-lack risk assessment on the data of the plurality of fields, thereby improving the accuracy of power-lack prediction.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and more specifically, to a method, device and electronic device for predicting power failure. Background Art

[0002] Whether it is a new energy vehicle or a fuel vehicle, a low battery in the vehicle will have a serious impact on the normal use of the customer. When the battery is seriously low, the vehicle may not be able to start, so the prediction of battery low battery is very important.

[0003] In a related method, a single data source such as battery current, voltage, and temperature in a battery management system (BMS) can be relied upon to predict power shortage through a rule-based algorithm. However, the related method still has the problem of low prediction accuracy. Summary of the invention

[0004] In view of the above problems, the present application proposes a power failure prediction method, device and electronic device to improve the above problems.

[0005] In a first aspect, the present application provides a method for predicting a low battery, the method comprising: obtaining current vehicle data of the vehicle, the vehicle data comprising cockpit domain data, intelligent driving domain data, whole vehicle domain data, and vehicle operation domain data; based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, obtaining a prediction result of the current low battery of the vehicle, wherein the linear regression model is obtained by fusing decision tree models of multiple fields, the decision tree models of the multiple fields comprise decision tree models trained using data from any field of first historical vehicle data, respectively, the first historical vehicle data comprising first cockpit domain data, first intelligent driving domain data, first whole vehicle domain data, and first vehicle operation domain data.

[0006] In a second aspect, the present application provides a low-battery prediction device, the device comprising: a vehicle data acquisition unit, used to acquire current vehicle data of the vehicle, the vehicle data including cockpit domain data, intelligent driving domain data, whole vehicle domain data, and vehicle operation domain data; a prediction result acquisition unit, used to obtain the current low-battery prediction result of the vehicle based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, wherein the linear regression model is obtained by fusing decision tree models of multiple fields, the decision tree models of multiple fields include decision tree models trained using data from any field of first historical vehicle data, respectively, the first historical vehicle data including first cockpit domain data, first intelligent driving domain data, first whole vehicle domain data, and first vehicle operation domain data.

[0007] In a third aspect, the present application provides an electronic device comprising one or more processors and a memory; one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the above-mentioned method.

[0008] In a fourth aspect, the present application provides a computer-readable storage medium, in which program code is stored, wherein the above method is executed when the program code is run.

[0009] The present application provides a method, device, electronic device and storage medium for predicting power failure. After obtaining the current vehicle data of the vehicle, the current power failure prediction result of the vehicle is obtained based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation field data and the pre-established linear regression model. Through the above method, the power failure prediction of the vehicle can be performed based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation field data and the linear regression model. The data of multiple domain controllers in the vehicle can be integrated with the vehicle operation data to make it more in line with the system architecture of the current vehicle, improve the interpretability of the power failure prediction, and increase the basis for power failure prediction through the integration of multiple field data, thereby improving the accuracy of power failure prediction. In addition, the linear regression model is obtained by integrating decision tree models based on multiple fields, and each decision tree model is obtained by training based on the data of any field in the first historical vehicle data, so that the linear regression model can not only perform different power failure risk prediction strategies for the data in each field to a certain extent, but also perform an overall power failure risk assessment on the data in multiple fields, thereby improving the accuracy of power failure prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 A flowchart of a power failure prediction method proposed in an embodiment of the present application;

[0012] Figure 2 A flow chart of a power failure prediction method proposed in another embodiment of the present application is shown;

[0013] Figure 3 Shows this application Figure 2 A flowchart of an implementation of S230;

[0014] Figure 4 A flow chart of a power failure prediction method proposed in another embodiment of the present application is shown;

[0015] Figure 5 A structural block diagram of a power failure prediction device proposed in an embodiment of the present application is shown;

[0016] Figure 6 A structural block diagram of an electronic device proposed in this application is shown. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] In an embodiment of the present application, the inventor proposes a method, device and electronic device for predicting power failure. After obtaining the current vehicle data of the vehicle, the current power failure prediction result of the vehicle is obtained based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and the pre-established linear regression model. Through the above method, the vehicle can be predicted to be power failure based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and the linear regression model. The data of multiple domain controllers in the vehicle can be integrated with the vehicle operation data to make it more in line with the system architecture of the current vehicle, improve the interpretability of the power failure prediction, and increase the basis for power failure prediction through the integration of multiple domain data, thereby improving the accuracy of power failure prediction. In addition, the linear regression model is obtained by integrating decision tree models based on multiple fields, and each decision tree model is obtained by training based on data from any field in the first historical vehicle data, so that the linear regression model can not only perform different power failure risk prediction strategies for the data in each field to a certain extent, but also perform an overall power failure risk assessment on the data in multiple fields, thereby improving the accuracy of power failure prediction.

[0019] See also Figure 1 , an embodiment of the present application provides a method for predicting power shortage, the method comprising:

[0020] S110: Acquire the current vehicle data of the vehicle, where the vehicle data includes cockpit domain data, intelligent driving domain data, whole vehicle domain data, and vehicle operation domain data.

[0021] Among them, cockpit domain data may include the temperature, voltage and current of the air conditioner, the voltage and current of the multimedia system, the voltage and current of the light, etc.; intelligent driving domain data may include the voltage and current of the automatic driving system, radar, camera, etc.; vehicle domain data may include the current, voltage and remaining power of the battery, the current, voltage and speed of the motor, the speed, acceleration and load of the vehicle, etc.; vehicle operation domain data may include the current mileage, cumulative mileage, number of charging times, charging power, charging time, etc. Exemplarily, cockpit domain data may be as shown in Table 1.

[0022] Table 1

[0023]

[0024] As a method, the vehicle can collect vehicle data through its own sensors and other data collection devices, and then report the collected vehicle data to the electronic device through TSP (Telematics Service Provider, remote communication service), so that the electronic device can obtain the current vehicle data of the vehicle.

[0025] Among them, the electronic device can be a cloud server, a computer, etc.

[0026] Optionally, in order to achieve real-time monitoring of the risk of low battery of the vehicle, the vehicle may periodically report vehicle data. For example, the vehicle data may be reported once every minute.

[0027] S120: Based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, obtain the current battery shortage prediction result of the vehicle, wherein the linear regression model is obtained by fusing decision tree models of multiple fields, and the decision tree models of the multiple fields include decision tree models trained using data from any field in the first historical vehicle data, respectively, and the first historical vehicle data includes first cockpit domain data, first intelligent driving domain data, first whole vehicle domain data, and first vehicle operation domain data.

[0028] As a method, the vehicle's current predicted low-battery risk value can be obtained based on cockpit domain data, intelligent driving domain data, vehicle domain data, vehicle operation domain data and a pre-established linear regression model; based on the predicted low-battery risk value and the preset low-battery threshold, the low-battery prediction result can be obtained.

[0029] Optionally, the vehicle's current predicted low-battery risk value may be obtained based on cockpit domain data, intelligent driving domain data, vehicle domain data, vehicle operation domain data, a second historical vehicle data, and a pre-established linear regression model.

[0030] There may be multiple pieces of second historical vehicle data, and each piece of second historical vehicle data may include second cockpit domain data, second intelligent driving domain data, second whole vehicle domain data, and second vehicle operation domain data.

[0031] Optionally, the features of the cockpit domain data, intelligent driving domain data, vehicle domain data, and vehicle operation domain data can be extracted respectively, and then a high-dimensional vector can be generated based on the extracted features; and multiple historical vehicle data can be extracted in the same way to obtain multiple high-dimensional historical vectors. Then, based on the high-dimensional vector, multiple high-dimensional historical vectors, and the linear regression model, the current predicted low-power risk value of the vehicle is obtained.

[0032] Among them, the linear regression model can be:

[0033]

[0034] Among them, β i It can represent the regression coefficient corresponding to the i-th high-dimensional history vector, k i (x,x i ) can represent the kernel function value corresponding to the high-dimensional vector and the i-th high-dimensional history vector, b can represent the intercept, β i and b are both known quantities.

[0035] Among them, the kernel function value can be obtained by the Gaussian kernel function (RBF kernel), and the Gaussian kernel function can be:

[0036]

[0037] Among them, σ can represent a known bandwidth parameter.

[0038] For example, when σ=10, the first high-dimensional history vector is x1=[60, 24, 12, 25], and the high-dimensional vector is x=[50, 20, 13, 25], k1(x, x1) can be:

[0039]

[0040] Through the above formula, the vehicle's current predicted power shortage risk value can be obtained, and then the predicted power shortage risk value can be compared with the preset power shortage threshold. If the predicted power shortage risk value is greater than or equal to the preset power shortage threshold, the power shortage prediction result is power shortage; if the predicted power shortage risk value is less than the preset power shortage threshold, the power shortage prediction result is no power shortage.

[0041] For example, when β=[0.1, 0.2, 0.3], b=0.05, k1(x, x1)=0.557, k2(x, x2)=0.951, k3(x, x3)=0.114, the predicted power shortage risk value is:

[0042]

[0043] When the preset power shortage threshold is 0.35, the power shortage prediction result is no power shortage.

[0044] Optionally, the preset low-battery threshold may be determined by setting a plurality of candidate low-battery thresholds after establishing a linear regression model, and then evaluating the linear regression model by using a ROC curve and an AUC method to obtain the preset low-battery threshold.

[0045] Optionally, for each candidate power-loss threshold, the TPR (true positive rate: the proportion of samples correctly predicted as positive) and FPR (false positive rate: the proportion of samples incorrectly predicted as positive) under the threshold can be calculated, and then the (FPR, TPR) points of all candidate power-loss thresholds are plotted into a curve to obtain the ROC curve. AUC can refer to the area under the ROC curve, and AUC reflects the overall classification ability of the model. This method can be used to understand the trade-off (sensitivity vs. specificity) of the linear regression model under the candidate power-loss threshold, so that the best candidate power-loss threshold can be selected as the preset power-loss threshold to balance false positives and false negatives.

[0046] Among them, the preset power-loss threshold may be a candidate power-loss threshold that minimizes the following loss function value.

[0047] L(θ)=c1.FPR+c2.(1-TPR)

[0048] Among them, θ can represent the candidate power-loss threshold, c1 and c2 can represent weights. In order to balance TPR and FPR, c1 and c2 can be 1.

[0049] Exemplarily, as shown in Table 2, when θ is 0.35, the balance between TPR and FPR is better, so the preset power-deficit threshold may be 0.35.

[0050] Table 2

[0051] θ TPR FPR 0.2 0.9 0.4 0.25 0.85 0.3 0.3 0.8 0.2 0.35 0.75 0.15 0.4 0.7 0.1

[0052] Optionally, in order to dynamically update the model involved in this application, the electronic device can store the reported vehicle data and the predicted power shortage risk value, power shortage prediction result, etc. corresponding to the vehicle data in a designated location to obtain the first historical vehicle data and the second historical vehicle data.

[0053] Optionally, when the power shortage prediction result is power shortage, the electronic device can send a power shortage reminder to the user or the maintenance point corresponding to the vehicle, so that the user can send the vehicle for inspection, maintenance, etc. in time.

[0054] The present embodiment provides a method for predicting power failure, after obtaining the current vehicle data of the vehicle, based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and the pre-established linear regression model, to obtain the current power failure prediction result of the vehicle. Through the above method, the power failure prediction of the vehicle can be performed based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and the linear regression model, and the data of multiple domain controllers in the vehicle can be integrated with the vehicle operation data, so that it is more in line with the system architecture of the current vehicle, and the interpretability of the power failure prediction is improved. In addition, by integrating data from multiple fields, the basis for power failure prediction is increased, thereby improving the accuracy of power failure prediction. In addition, the linear regression model is obtained by integrating decision tree models based on multiple fields, and each decision tree model is obtained by training based on data from any field in the first historical vehicle data, so that the linear regression model can not only execute different power failure risk prediction strategies for the data in each field to a certain extent, but also perform an overall power failure risk assessment on the data in multiple fields, thereby improving the accuracy of power failure prediction.

[0055] See also Figure 2 , an embodiment of the present application provides a method for predicting power shortage, the method comprising:

[0056] S210: Acquire second historical vehicle data.

[0057] As one approach, the second historical vehicle data may be acquired from a designated location.

[0058] S220: Based on the second historical vehicle data and the decision tree models in the multiple fields, obtain the power shortage risk scores output by the decision tree models in the multiple fields.

[0059] Among them, the decision tree models in multiple fields may include a cockpit domain model trained based on the first cockpit domain data, an intelligent driving domain model trained based on the first intelligent driving domain data, a vehicle domain model trained based on the first vehicle domain data, and a vehicle operation domain model trained based on the first vehicle operation domain data. The power-loss risk scores output by the decision tree models in multiple fields can represent the power-loss risk of the corresponding domain controller or vehicle operation field, that is, reflect the degree of power-loss loss in each domain controller or field of the vehicle.

[0060] As a method, a cockpit domain low-battery risk score can be obtained based on the second historical cockpit domain data and the cockpit domain model; a smart driving domain low-battery risk score can be obtained based on the second historical smart driving domain data and the smart driving domain model; a vehicle domain low-battery risk score can be obtained based on the second historical vehicle domain data and the vehicle domain model; and an operation domain low-battery risk score can be obtained based on the second historical vehicle operation domain data and the vehicle operation domain model.

[0061] It should be noted that since there can be multiple second historical vehicle data, there can be multiple second cockpit domain data, multiple second intelligent driving domain data, multiple second whole vehicle domain data, and multiple second vehicle operation domain data. For the decision tree model in each field, each input data can obtain an output score, which is equivalent to that the decision tree model in each field can output multiple scores. Based on the multiple scores corresponding to the decision tree model in each field, the power outage risk score of the decision tree model in each field can be obtained.

[0062] Optionally, the average value of multiple scores corresponding to the decision tree model in each field can be used as the power outage risk score of the corresponding decision tree model.

[0063] Optionally, the weighted sum of multiple scores corresponding to the decision tree model in each field can be used as the power outage risk score of the decision tree model, wherein the closer the data collection date is, the greater the corresponding weight is.

[0064] S230: Establishing the linear regression model based on the second historical vehicle data and the power-loss risk scores output by the decision tree models in the multiple fields.

[0065] The second historical vehicle data corresponds to a real power-loss risk value, which can be obtained by manual annotation or trained model annotation. When there are multiple second historical vehicle data, each second historical vehicle data can correspond to a real power-loss risk value.

[0066] As a method, a Gaussian kernel matrix of the ridge regression model to be trained can be constructed based on the second historical vehicle data and the power-loss risk scores output by the decision tree models in multiple fields; based on the kernel matrix and the actual power-loss risk value, the ridge regression model to be trained is trained to obtain the regression coefficient and intercept of the linear regression model; based on the regression coefficient and the intercept, a linear regression model is established.

[0067] Optional, such as Figure 3 As shown, based on the second historical vehicle data and the power-loss risk scores output by the decision tree models in the multiple fields, a Gaussian kernel matrix of the ridge regression model to be trained is constructed, including:

[0068] S231: Based on the multiple pieces of second historical vehicle data, the power-loss risk scores output by the decision tree models in the multiple fields, and a preset Gaussian kernel function, multiple fusion kernel function values ​​are obtained.

[0069] As a method, multiple cockpit domain kernel function values ​​can be obtained based on multiple second cockpit domain data and preset Gaussian kernel functions; multiple intelligent driving domain kernel function values ​​can be obtained based on multiple second intelligent driving domain data and preset Gaussian kernel functions; multiple vehicle domain kernel function values ​​can be obtained based on multiple second vehicle domain data and preset Gaussian kernel functions; multiple operation domain kernel function values ​​can be obtained based on multiple second vehicle operation domain data and preset Gaussian kernel functions; multiple fusion kernel function values ​​can be obtained based on multiple cockpit domain kernel function values, multiple intelligent driving domain kernel function values, multiple vehicle domain kernel function values, multiple operation domain kernel function values ​​and the low-power risk scores output by the decision tree models of multiple fields.

[0070] Optionally, the multiple second cockpit domain data, the multiple second intelligent driving domain data, the multiple second whole vehicle domain data, and the multiple second vehicle operation domain data can be preprocessed first, and then feature extraction can be performed on the preprocessed multiple second cockpit domain data, the multiple second intelligent driving domain data, the multiple second whole vehicle domain data, and the multiple second vehicle operation domain data, respectively, and then data standardization can be performed on the extracted features to obtain high-dimensional vectors corresponding to the multiple second cockpit domain data, the multiple second intelligent driving domain data, the multiple second whole vehicle domain data, and the multiple second vehicle operation domain data. Therefore, based on the high-dimensional vectors and preset Gaussian kernel functions corresponding to each of the multiple second cockpit domain data, multiple cockpit domain kernel function values ​​can be obtained; based on the high-dimensional vectors and preset Gaussian kernel functions corresponding to each of the multiple second intelligent driving domain data, multiple intelligent driving domain kernel function values ​​can be obtained; based on the high-dimensional vectors and preset Gaussian kernel functions corresponding to each of the multiple second vehicle domain data, multiple vehicle domain kernel function values ​​can be obtained; based on the high-dimensional vectors and preset Gaussian kernel functions corresponding to each of the multiple second vehicle operation domain data, multiple operation domain kernel function values ​​can be obtained.

[0071] Among them, preprocessing can include cleaning data, such as processing missing values ​​and outliers.

[0072] Data normalization can use z-score normalization to eliminate the impact between different dimensions. The normalization formula can be:

[0073]

[0074] Among them, x i,j It can represent the high-dimensional vector corresponding to the j-th data in the i-th field, x′ i,j It can represent the features corresponding to the jth data in the i-th field, μ iIt can represent the feature mean of the i-th field, ε i It can represent the characteristic standard deviation of the i-th field.

[0075] Optionally, since each data in each field may have multiple features, data standardization may be performed on each feature of each data to obtain a high-dimensional vector.

[0076] Exemplarily, the feature may be air conditioning power. There may be 100 pieces of second historical data, which may be 100 air conditioning powers. The average and standard deviation of the 100 air conditioning powers may be calculated first, and then each air conditioning power may be substituted into the above-mentioned standardized formula to obtain a vector corresponding to each air conditioning power.

[0077] Optionally, the preset Gaussian kernel function may be the same as the kernel function in step S120, and the expression for calculating the kernel function value in a single domain may be changed to:

[0078]

[0079] Among them, x i It can represent the high-dimensional vector corresponding to the i-th data, x j It can represent the high-dimensional vector corresponding to the j-th data, and σ can represent the bandwidth parameter.

[0080] Optionally, the fusion kernel function can be:

[0081]

[0082] Among them, K(x i ,x j ) can represent the fusion function value obtained based on the i-th second historical data and the j-th second historical data, α i It can represent the power outage risk score of the i-th area, k p (x i ,x j ) can represent the kernel function value obtained based on the i-th data and the j-th data in the p-th field.

[0083] Exemplarily, when the cockpit domain kernel function value obtained based on the second historical data of the first article and the second historical data of the second article is 0.024, the intelligent driving domain kernel function value is 0.03, and the vehicle domain kernel function value is 0.05, and the operation domain kernel function value is 0.02, the cockpit domain low battery risk score is 0.3, the intelligent driving domain low battery risk score is 0.2, the vehicle domain low battery risk score is 0.4, and the operation domain low battery risk score is 0.1, the fusion kernel function value can be 0.0312 (0.3×0.024+0.2×0.03+0.4×0.05+0.1×0.02).

[0084] S232: Based on the multiple fusion kernel function values, construct a Gaussian kernel matrix of the ridge regression model to be trained.

[0085] As a method, a Gaussian kernel matrix of the ridge regression model to be trained can be constructed based on multiple fused kernel function values.

[0086] For example, there are 3 pieces of second historical data in total, and the fusion kernel function corresponding to the first piece of second historical data and the second piece of second historical data may be 0.0312, then the Gaussian kernel matrix may be:

[0087]

[0088] Optionally, after obtaining the Gaussian kernel matrix, the ridge regression model can be trained. The objective function of the ridge regression model can be:

[0089]

[0090] Among them, y i It can represent the real power shortage risk value of the ith item, K(x i ,x j ) can represent the fusion kernel function value, β can represent the regression coefficient, It can be expressed as the sum of squares of prediction errors, It can represent a regularization term, which can prevent overfitting.

[0091] The matrix of regression coefficients and intercept is represented as:

[0092] Kβ+b=Y,β=(K T K+λI) -1 K T Y

[0093] Among them, K is the kernel matrix, β is the regression coefficient vector, b is the intercept, and Y is the actual power outage risk value.

[0094] In order to uniformly process β and b, the Gaussian kernel matrix and β can be expanded: -K' = [K, 1], where 1 is a column vector of all 1s; -β' = [β, b] T , and then the regression coefficient and intercept are obtained through the Gaussian kernel matrix and the true power shortage risk value.

[0095] It should be noted that λ and σ can be determined by presetting multiple candidate values ​​and then training the ridge regression model through cross-validation to determine a final value.

[0096] In the embodiment of the present application, a nonlinear regression model can be established by kernel techniques to predict the risk of power outage. At the same time, the idea of ​​ridge regression can be used to prevent overfitting, and the optimal β value can be found by optimizing the objective function. The regression coefficients obtained by the ridge regression model have the following advantages: (1) Solving the problem of multicollinearity: Ridge regression can effectively deal with the high correlation between features by adding the L2 regularization term, avoiding the problem of unstable estimation of ordinary least squares method under multicollinearity. (2) Preventing overfitting: The L2 regularization term can constrain the size of the coefficient and prevent overfitting caused by the model being too complex. (3) Numerical stability: When the number of features is greater than the number of samples, ridge regression can still obtain a stable solution. (4) Interpretability: The coefficients obtained by ridge regression still maintain the interpretability of the linear model, and the importance of each feature can be measured by the size of the coefficient.

[0097] S240: Acquire the current vehicle data of the vehicle, where the vehicle data includes cockpit domain data, intelligent driving domain data, whole vehicle domain data, and vehicle operation domain data.

[0098] S250: Based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, obtain the current battery shortage prediction result of the vehicle, wherein the linear regression model is obtained by fusing decision tree models of multiple fields, and the decision tree models of multiple fields include decision tree models trained using data from any field in the first historical vehicle data, respectively, and the first historical vehicle data includes first cockpit domain data, first intelligent driving domain data, first whole vehicle domain data, and first vehicle operation domain data.

[0099] In the embodiments of the present application, the use of a linear regression model to predict power failure results can have the following advantages: (1) Simple and intuitive: The linear model is simple in form, easy to understand and explain, and suitable for actual business applications. (2) High computational efficiency: When predicting, only the kernel function value and linear combination need to be calculated, and the calculation speed is fast. (3) Strong generalization ability: By mapping the kernel function to a high-dimensional space, nonlinear relationships can be captured while maintaining the simplicity of the linear model. (4) Flexibility: Different kernel functions can be easily replaced to adapt to different data distributions. (5) Scalability: For new vehicle data, only the kernel function value with the second historical data needs to be calculated, without the need to retrain the model. In general, the classification method proposed in the present application combines the stability of ridge regression and the nonlinear modeling capability of the kernel method, while maintaining the simplicity and efficiency of the prediction model. It is an option that balances complexity and practicality.

[0100] The present embodiment provides a method for predicting power failure. Through the above method, after obtaining the current vehicle data of the vehicle, the current power failure prediction result of the vehicle is obtained based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and the pre-established linear regression model. Through the above method, the vehicle can be predicted to be power failure based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and the linear regression model. The data of multiple domain controllers in the vehicle can be integrated with the vehicle operation data to make it more suitable for the system architecture of the current vehicle, improve the interpretability of the power failure prediction, and increase the basis for power failure prediction through the integration of multiple domain data, thereby improving the accuracy of power failure prediction. In addition, the linear regression model is obtained by integrating decision tree models based on multiple fields, and each decision tree model is obtained by training based on the data of any field in the first historical vehicle data, so that the linear regression model can not only execute different power failure risk prediction strategies for the data in each field to a certain extent, but also perform an overall power failure risk assessment on the data in multiple fields, thereby improving the accuracy of power failure prediction. Moreover, in this embodiment, the features of each field can be weighted and fused by using a fusion kernel function, thereby improving the accuracy of the kernel function value. Since the weights of the features of each field are derived from the corresponding decision tree models, the weights can reflect the degree of power outage risk in the corresponding field. Therefore, the kernel function value obtained can not only represent the feature-level fusion, but also the decision-level fusion of each decision tree model, making the kernel function value more interpretable and more accurate, thereby improving the accuracy of the linear regression model and the accuracy of the prediction results.

[0101] See also Figure 4 , an embodiment of the present application provides a method for predicting power shortage, the method comprising:

[0102] S310: Extracting features of the first cockpit domain data, the first intelligent driving domain data, the first whole vehicle domain data and the first vehicle operation domain data respectively to obtain cockpit domain features, intelligent driving domain features, whole vehicle domain features and vehicle operation domain features.

[0103] Among them, there may be multiple first cockpit domain data, each of which may include the temperature, voltage and current of the air conditioner, the voltage and current of the multimedia system, the voltage and current of the light, etc.; the first intelligent driving domain data may include the voltage and current of the automatic driving system, radar, camera, etc.; the first vehicle domain data may include the current, voltage and remaining power of the battery, the current, voltage and speed of the motor, the speed, acceleration and load of the vehicle, etc.; the first vehicle operation domain data may include the current mileage, cumulative mileage, number of charging times, charging power, charging time, etc. There may be multiple first cockpit domain data, first intelligent driving domain data, first vehicle domain data and first vehicle operation domain data respectively.

[0104] Among them, cockpit domain features may include air conditioning power, multimedia system power, lighting power, and cabin temperature; intelligent driving domain features may include automatic driving system power, radar power, and camera power; vehicle domain features may include battery current, battery voltage, remaining battery power, motor power, motor speed, vehicle speed, acceleration, and load; vehicle operation domain features may include current mileage, cumulative mileage, number of charges, charging power, and charging time.

[0105] As a method, part of the first cockpit domain data can be directly used as the corresponding features, and the remaining features can be obtained by: obtaining the air-conditioning power based on the air-conditioning voltage and current, obtaining the multimedia system power based on the voltage and current of the multimedia system, obtaining the light power based on the voltage and current of the lights, and obtaining the automatic driving system power, radar power, and camera power based on the voltage and current of the automatic driving system, radar, and camera.

[0106] It should be noted that the vehicle data, the first historical vehicle data, and the second historical vehicle data have the same data type, but may be collected at different times, quantities, etc. The first historical vehicle data and the second historical vehicle data may be completely the same, or may be collected at different times or quantities.

[0107] S320: Obtain the cockpit domain model based on the cockpit domain feature training.

[0108] Among them, each piece of first cockpit domain data may correspond to a labeled real score.

[0109] As a method, since there are multiple first historical vehicle data, there may also be multiple cockpit domain features. The multiple cockpit domain features may be subjected to data standardization processing to obtain multiple cockpit domain feature vectors. The initial cockpit domain model may be trained based on the multiple cockpit domain feature vectors and the true scores corresponding to the multiple cockpit domain feature vectors to obtain the cockpit domain model.

[0110] Exemplarily, the first cockpit domain data may be as shown in Table 2

[0111] S330: Obtain the intelligent driving domain model based on the intelligent driving domain feature training.

[0112] Among them, each piece of the first intelligent driving domain data may correspond to a labeled real score.

[0113] As a method, since there are multiple first historical vehicle data, there may also be multiple intelligent driving domain features. The multiple intelligent driving domain features may be subjected to data standardization processing to obtain multiple intelligent driving domain feature vectors. The initial intelligent driving domain model may be trained based on the multiple intelligent driving domain feature vectors and the true scores corresponding to the multiple intelligent driving domain feature vectors to obtain the intelligent driving domain model.

[0114] S340: Obtain the whole vehicle domain model based on the whole vehicle domain feature training.

[0115] Among them, each piece of first vehicle domain data may correspond to a labeled real score.

[0116] As a method, since there are multiple first historical vehicle data, there may also be multiple whole vehicle domain features. The multiple whole vehicle domain features may be subjected to data standardization processing to obtain multiple whole vehicle domain feature vectors. The initial whole vehicle domain model may be trained based on the multiple whole vehicle domain feature vectors and the true scores corresponding to the multiple whole vehicle domain feature vectors to obtain the whole vehicle domain model.

[0117] S350: Obtaining the vehicle operation domain model based on the vehicle operation domain feature training.

[0118] Among them, each piece of first vehicle operation field data may correspond to a labeled real score.

[0119] As a method, since there are multiple first historical vehicle data, there may also be multiple vehicle operation domain features. The multiple vehicle operation domain features may be subjected to data standardization processing to obtain multiple vehicle operation domain feature vectors. The initial vehicle operation domain model may be trained based on the multiple vehicle operation domain feature vectors and the true scores corresponding to the multiple vehicle operation domain feature vectors to obtain a vehicle operation domain model.

[0120] S360: Acquire the current vehicle data of the vehicle, where the vehicle data includes cockpit domain data, intelligent driving domain data, whole vehicle domain data, and vehicle operation domain data.

[0121] S370: Based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, obtain the current battery shortage prediction result of the vehicle, wherein the linear regression model is obtained by fusing decision tree models of multiple fields, and the decision tree models of multiple fields include decision tree models trained using data from any field in the first historical vehicle data, respectively, and the first historical vehicle data includes first cockpit domain data, first intelligent driving domain data, first whole vehicle domain data, and first vehicle operation domain data.

[0122] The present embodiment provides a method for predicting power failure. Through the above-mentioned method, the power failure prediction of the vehicle can be performed based on cockpit domain data, intelligent driving domain data, vehicle domain data, vehicle operation domain data and linear regression model. The data of multiple domain controllers in the vehicle can be integrated with the vehicle operation data to make it more suitable for the system architecture of the current vehicle, improve the interpretability of power failure prediction, and increase the basis for power failure prediction through the integration of multiple domain data, thereby improving the accuracy of power failure prediction. In addition, the linear regression model is obtained by integrating decision tree models based on multiple fields, and each decision tree model is trained based on data from any field in the first historical vehicle data, so that the linear regression model can not only perform different power failure risk prediction strategies for the data in each field to a certain extent, but also perform overall power failure risk assessment on the data in multiple fields, thereby improving the accuracy of power failure prediction. In addition, in the present embodiment, by training decision tree models in multiple fields, it is possible to evaluate the power failure risk of the vehicle from different fields, thereby improving the accuracy of power failure prediction.

[0123] See also Figure 5 The present application provides a power shortage prediction device 600, the device 600 comprising:

[0124] The vehicle data acquisition unit 610 is used to acquire the current vehicle data of the vehicle, wherein the vehicle data includes cockpit domain data, intelligent driving domain data, vehicle domain data, and vehicle operation domain data;

[0125] The prediction result acquisition unit 620 is used to obtain the current low-power prediction result of the vehicle based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, wherein the linear regression model is obtained by fusing decision tree models of multiple fields, and the decision tree models of the multiple fields include decision tree models trained using data from any field in the first historical vehicle data, and the first historical vehicle data includes first cockpit domain data, first intelligent driving domain data, first whole vehicle domain data, and first vehicle operation domain data.

[0126] As a method, the prediction result acquisition unit 620 is specifically used to obtain the current predicted low-battery risk value of the vehicle based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model; based on the predicted low-battery risk value and the preset low-battery threshold, obtain the low-battery prediction result.

[0127] The device 600 further includes:

[0128] The model generation unit 630 is used to obtain the second historical data; based on the second historical vehicle data and the decision tree models of the multiple fields, obtain the low-battery risk scores output by the decision tree models of the multiple fields; based on the second historical vehicle data and the low-battery risk scores output by the decision tree models of the multiple fields, establish the linear regression model.

[0129] Optionally, the second historical vehicle data includes second cockpit domain data, second intelligent driving domain data, second whole vehicle domain data, and second vehicle operation domain data, and the decision tree models of the multiple domains include a cockpit domain model trained based on the first cockpit domain data, an intelligent driving domain model trained based on the first intelligent driving domain data, a whole vehicle domain model trained based on the first whole vehicle domain data, and a vehicle operation domain model trained based on the first vehicle operation domain data, and the model generation unit 630 is specifically used to obtain a cockpit domain low-battery risk score based on the second historical cockpit domain data and the cockpit domain model; to obtain a smart driving domain low-battery risk score based on the second historical smart driving domain data and the smart driving domain model; to obtain a whole vehicle domain low-battery risk score based on the second historical whole vehicle domain data and the whole vehicle domain model; and to obtain an operation domain low-battery risk score based on the second historical vehicle operation domain data and the vehicle operation domain model.

[0130] Optionally, the second historical vehicle data corresponds to an actual low-battery risk value, and the model generation unit 630 is specifically used to construct a Gaussian kernel matrix of the ridge regression model to be trained based on the second historical vehicle data and the low-battery risk scores output by the decision tree models of the multiple fields; based on the kernel matrix and the actual low-battery risk value, the ridge regression model to be trained is trained to obtain the regression coefficient and intercept of the linear regression model; based on the regression coefficient and the intercept, the linear regression model is established.

[0131] Optionally, there are multiple second historical vehicle data, and the model generation unit 630 is specifically used to obtain multiple fused kernel function values ​​based on the multiple second historical vehicle data, the low-power risk scores output by the decision tree models of the multiple fields, and a preset Gaussian kernel function; based on the multiple fused kernel function values, construct a Gaussian kernel matrix of the ridge regression model to be trained.

[0132] Optionally, the multiple second historical vehicle data include multiple second cockpit domain data, multiple second intelligent driving domain data, multiple second whole vehicle domain data, and multiple second vehicle operation field data. The model generation unit 630 is specifically used to obtain multiple cockpit domain kernel function values ​​based on the multiple second cockpit domain data and a preset Gaussian kernel function; obtain multiple intelligent driving domain kernel function values ​​based on the multiple second intelligent driving domain data and the preset Gaussian kernel function; obtain multiple whole vehicle domain kernel function values ​​based on the multiple second whole vehicle domain data and the preset Gaussian kernel function; obtain multiple operation field kernel function values ​​based on the multiple second vehicle operation field data and the preset Gaussian kernel function; obtain the multiple fusion kernel function values ​​based on the multiple cockpit domain kernel function values, the multiple intelligent driving domain kernel function values, the multiple vehicle domain kernel function values, the multiple operation field kernel function values ​​and the low-power risk scores output by the decision tree models of the multiple fields.

[0133] As a method, the model generation unit 630 is specifically used to extract the respective features of the first cockpit domain data, the first intelligent driving domain data, the first whole vehicle domain data and the first vehicle operation domain data, so as to obtain cockpit domain features, intelligent driving domain features, whole vehicle domain features and vehicle operation domain features; the cockpit domain model is obtained based on the cockpit domain features training; the intelligent driving domain model is obtained based on the intelligent driving domain features training; the whole vehicle domain model is obtained based on the whole vehicle domain features training; the vehicle operation domain model is obtained based on the vehicle operation domain features training.

[0134] Optionally, the cabin domain characteristics include air conditioning power, multimedia system power, lighting power, and cabin temperature; the intelligent driving domain characteristics include automatic driving system power, radar power, and camera power; the vehicle domain characteristics include battery current, battery voltage, remaining battery power, motor power, motor speed, vehicle speed, acceleration, and load; the vehicle operation domain characteristics include current mileage, cumulative mileage, number of charges, charging power, and charging time.

[0135] Optionally, the cockpit domain data includes the temperature, voltage and current of the air conditioner, the voltage and current of the multimedia system, and the voltage and current of the lights; the intelligent driving domain data includes the voltage and current of the automatic driving system, radar, and camera; the vehicle domain data includes the current, voltage and remaining power of the battery, the current, voltage and speed of the motor, and the vehicle speed, acceleration and load; the vehicle operation domain data includes the current mileage, cumulative mileage, number of charging times, charging power, and charging time.

[0136] The following will be combined Figure 6 An electronic device provided by the present application is described.

[0137] See also Figure 6 Based on the above power failure prediction method and device, the embodiment of the present application also provides another electronic device 100 that can execute the above power failure prediction method. The electronic device 100 includes a processor 102, a memory 104, and a communication module 106. The memory 104 stores a program that can execute the content of the above embodiment, and the backup processor 102 can execute the program stored in the memory 104.

[0138] Among them, the processor 102 may include one or more processing cores. The processor 102 uses various interfaces and lines to connect various parts within the entire electronic device 100, and executes various functions and processes data of the electronic device 100 by running or executing instructions, programs, code sets or instruction sets stored in the memory 104, and calling data stored in the memory 104. Optionally, the processor 102 can be implemented in at least one hardware form of a network processor (Neural network Processing Unit, NPU), a digital signal processing (Digital Signal Processing, DSP), a field programmable gate array (Field-Programmable Gate Array, FPGA), and a programmable logic array (Programmable Logic Array, PLA). The processor 102 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processor (Graphics Processing Unit, GPU), a network processor (Neural network Processing Unit, NPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing display content; the NPU is responsible for processing multimedia data such as video and images; and the modem is used to process wireless communications. It is understandable that the above-mentioned modem may not be integrated into the processor 102, but may be implemented separately through a communication chip.

[0139] The memory 104 may include a random access memory (RAM), and may also include a read-only memory (RAM) and a double data rate synchronous dynamic random access memory (DDR). The memory 104 may be used to store instructions, programs, codes, code sets or instruction sets. The memory 104 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the following various method embodiments, etc. The data storage area may also store data (such as a phone book, audio and video data, chat record data) created by the electronic device 100 during use.

[0140] The communication module 106 may be used to implement information exchange between the vehicle 100 and other devices, for example, to transmit device control instructions, operation request instructions, and status information acquisition instructions, etc. When the other devices are different devices, the corresponding communication modules 106 may be different.

[0141] An embodiment of the present application provides a computer-readable storage medium in which a program code is stored, and the program code can be called by a processor to execute the method described in the above method embodiment.

[0142] The computer readable storage medium may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk, or a ROM. Optionally, the computer readable storage medium includes a non-transitory computer-readable storage medium. The computer readable storage medium has storage space for program codes that execute any of the method steps in the above method. These program codes can be read from or written to one or more computer program products. The program code can be compressed, for example, in an appropriate form.

[0143] In summary, the present application provides a method, device and electronic device for predicting power failure, after obtaining the current vehicle data of the vehicle, based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and the pre-established linear regression model, to obtain the current power failure prediction result of the vehicle. Through the above method, the power failure prediction of the vehicle can be performed based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and the linear regression model, and the data of multiple domain controllers in the vehicle can be integrated with the vehicle operation data, so that it is more in line with the system architecture of the current vehicle, and the interpretability of the power failure prediction is improved. In addition, by integrating data from multiple fields, the basis for power failure prediction is increased, thereby improving the accuracy of power failure prediction. In addition, the linear regression model is obtained by integrating decision tree models based on multiple fields, and each decision tree model is obtained by training data from any field in the first historical vehicle data, so that the linear regression model can not only execute different power failure risk prediction strategies for the data in each field to a certain extent, but also perform an overall power failure risk assessment on the data in multiple fields, thereby improving the accuracy of power failure prediction.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting power shortage, characterized in that: The method comprises: Obtain the current vehicle data of the vehicle, wherein the vehicle data includes cockpit domain data, intelligent driving domain data, vehicle domain data, and vehicle operation domain data; Based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, a current battery shortage prediction result of the vehicle is obtained, wherein the linear regression model is obtained by fusing decision tree models of multiple fields, and the decision tree models of multiple fields include decision tree models trained using data from any field in the first historical vehicle data, respectively, and the first historical vehicle data includes first cockpit domain data, first intelligent driving domain data, first whole vehicle domain data, and first vehicle operation domain data.

2. The method according to claim 1, characterized in that: Before obtaining the current vehicle data of the vehicle, the method further includes: Obtaining second historical data; Based on the second historical vehicle data and the decision tree models in the multiple fields, obtaining the power shortage risk scores output by the decision tree models in the multiple fields respectively; The linear regression model is established based on the second historical vehicle data and the power-loss risk scores output by the decision tree models in the multiple fields.

3. The method according to claim 2, characterized in that The second historical vehicle data includes second cockpit domain data, second intelligent driving domain data, second whole vehicle domain data, and second vehicle operation domain data. The decision tree models in the multiple fields include a cockpit domain model trained based on the first cockpit domain data, an intelligent driving domain model trained based on the first intelligent driving domain data, a whole vehicle domain model trained based on the first whole vehicle domain data, and a vehicle operation domain model trained based on the first vehicle operation domain data. The decision tree models in the multiple fields based on the second historical vehicle data and the multiple fields obtain respective power shortage risk scores, including: Based on the second historical cockpit domain data and the cockpit domain model, a cockpit domain power failure risk score is obtained; Based on the second historical intelligent driving domain data and the intelligent driving domain model, obtaining an intelligent driving domain power shortage risk score; Based on the second historical vehicle domain data and the vehicle domain model, a vehicle domain power shortage risk score is obtained; Based on the second historical vehicle operation domain data and the vehicle operation domain model, an operation domain power shortage risk score is obtained.

4. The method according to claim 2, characterized in that: The second historical vehicle data corresponds to a real power-loss risk value, and the linear regression model is established based on the second historical vehicle data and the power-loss risk scores output by the decision tree models in the multiple fields, including: Based on the second historical vehicle data and the power-loss risk scores output by the decision tree models in the multiple fields, a Gaussian kernel matrix of the ridge regression model to be trained is constructed; Based on the kernel matrix and the actual power failure risk value, the ridge regression model to be trained is trained to obtain the regression coefficient and intercept of the linear regression model; The linear regression model is established based on the regression coefficient and the intercept.

5. The method according to claim 4, characterized in that There are multiple pieces of the second historical vehicle data, and the Gaussian kernel matrix of the ridge regression model to be trained is constructed based on the second historical vehicle data and the power-loss risk scores output by the decision tree models in the multiple fields, including: Based on the plurality of second historical vehicle data, the power-loss risk scores outputted by the decision tree models in the plurality of fields, and a preset Gaussian kernel function, a plurality of fusion kernel function values ​​are obtained; Based on the multiple fused kernel function values, a Gaussian kernel matrix of the ridge regression model to be trained is constructed.

6. The method according to claim 5, characterized in that The plurality of second historical vehicle data includes a plurality of second cockpit domain data, a plurality of second intelligent driving domain data, a plurality of second whole vehicle domain data, and a plurality of second vehicle operation domain data. The plurality of second historical vehicle data, the power-loss risk scores output by the decision tree models of the plurality of domains, and the preset Gaussian kernel function are used to obtain a plurality of fused kernel function values, including: Based on the plurality of pieces of second cockpit domain data and a preset Gaussian kernel function, obtaining a plurality of cockpit domain kernel function values; Based on the plurality of second intelligent driving domain data and the preset Gaussian kernel function, obtaining a plurality of intelligent driving domain kernel function values; Based on the plurality of second whole vehicle domain data and the preset Gaussian kernel function, obtaining a plurality of whole vehicle domain kernel function values; Based on the plurality of second vehicle operation domain data and the preset Gaussian kernel function, obtaining a plurality of operation domain kernel function values; The multiple fusion kernel function values ​​are obtained based on the multiple cockpit domain kernel function values, the multiple intelligent driving domain kernel function values, the multiple vehicle domain kernel function values, the multiple operation domain kernel function values ​​and the low-battery risk scores output by the decision tree models of the multiple domains.

7. The method according to claim 1, characterized in that Before obtaining the current vehicle data of the vehicle, the method further includes: Respectively extracting features of the first cockpit domain data, the first intelligent driving domain data, the first whole vehicle domain data, and the first vehicle operation domain data to obtain cockpit domain features, intelligent driving domain features, whole vehicle domain features, and vehicle operation domain features; Obtaining the cockpit domain model based on the cockpit domain feature training; Obtaining the intelligent driving domain model based on the intelligent driving domain feature training; Obtaining the whole vehicle domain model based on the whole vehicle domain feature training; The vehicle operation domain model is obtained based on the vehicle operation domain feature training.

8. The method according to claim 7, characterized in that The cockpit domain features include air conditioning power, multimedia system power, lighting power, and cabin temperature; The intelligent driving domain characteristics include autonomous driving system power, radar power, and camera power; The vehicle domain characteristics include battery current, battery voltage, battery remaining capacity, motor power, motor speed, vehicle speed, acceleration and load; The vehicle operation field characteristics include current mileage, cumulative mileage, number of charging times, charging power, and charging time.

9. The method according to claim 1, characterized in that: The method of obtaining a current power shortage prediction result of the vehicle based on the cockpit domain data, the intelligent driving domain data, the vehicle domain data, the vehicle operation domain data and a pre-established linear regression model includes: Based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, a current predicted power shortage risk value of the vehicle is obtained; Based on the predicted power shortage risk value and the preset power shortage threshold, the power shortage prediction result is obtained.

10. The method according to any one of claims 1 to 9, characterized in that: The cockpit domain data includes the temperature, voltage and current of the air conditioner, the voltage and current of the multimedia system, and the voltage and current of the lighting; Intelligent driving domain data includes voltage and current of the autonomous driving system, radar, and camera; The vehicle domain data includes the battery current, voltage and remaining power, the motor current, voltage and speed, and the vehicle speed, acceleration and load; Vehicle operation data includes current mileage, cumulative mileage, number of charges, charging power, and charging time.

11. A power shortage prediction device, characterized in that: The device comprises: A vehicle data acquisition unit, used to acquire the current vehicle data of the vehicle, wherein the vehicle data includes cockpit domain data, intelligent driving domain data, whole vehicle domain data, and vehicle operation domain data; A prediction result acquisition unit is used to obtain the current low-power prediction result of the vehicle based on the cockpit domain data, the intelligent driving domain data, the whole vehicle domain data, the vehicle operation domain data and a pre-established linear regression model, wherein the linear regression model is obtained by fusing decision tree models of multiple fields, and the decision tree models of the multiple fields include decision tree models trained using data from any field in the first historical vehicle data, respectively, and the first historical vehicle data includes first cockpit domain data, first intelligent driving domain data, first whole vehicle domain data, and first vehicle operation domain data.

12. An electronic device, characterized in that: comprising one or more processors and memory; One or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the method according to any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, wherein when the program code is run, the method according to any one of claims 1 to 10 is executed.

Images