System for estimating power consumption of vehicle and method for estimating power consumption of vehicle

Abstract

The invention provides a system for estimating the power consumption of a vehicle, which appropriately determines the necessity of learning and performs learning. A system for estimating the power consumption of a vehicle is provided with: a first storage unit for storing a prediction error of the power consumption of the vehicle; a second storage unit that stores an input / output pair of a model that estimates the power consumption of the vehicle; and a learning unit that, when the estimation accuracy of the model for estimating the power consumption of the vehicle is poor, uses the average of the prediction errors of the power consumption of the vehicle in the first storage unit and the standard deviation to determine whether or not an improvement in the estimation accuracy can be expected by learning using the second storage unit. And learning is performed only when it is expected that the estimation accuracy is improved.

Description

Technical Field

[0001] This invention relates to a system and method for estimating the power consumption of a vehicle. Background Technology

[0002] Patent document 1 describes a device that obtains appropriate fuel economy or electrical efficiency in accordance with the vehicle's driving conditions and the driver's requirements.

[0003] Patent Document 1: Japanese Patent Application Publication No. 2017-9460 Summary of the Invention

[0004] However, the device in Patent Document 1 is configured to learn on a per-elapsed time basis, thus learning occurs even when it is not necessary. Therefore, the object of the present invention is to provide a vehicle power consumption estimation system that appropriately determines the necessity of learning and performs learning accordingly.

[0005] The vehicle power consumption estimation system of the present invention includes:

[0006] The first storage unit stores the prediction error of the power consumption of the vehicles.

[0007] The second storage unit stores the input-output pairs of the model for estimating the vehicle's power consumption; and

[0008] The learning unit, when the estimation accuracy of the model for estimating the vehicle's power consumption is poor, uses the average and standard deviation of the prediction error of the vehicle's power consumption in the first storage unit to determine whether learning through the second storage unit can be expected to improve the estimation accuracy, and only performs learning if the improvement in estimation accuracy can be expected.

[0009] The above structure provides a system for estimating the power consumption of a vehicle by appropriately determining the necessity of learning and then learning accordingly. This learning utilizes machine learning.

[0010] The vehicle power consumption estimation system of the present invention is characterized in that,

[0011] When the mean is small and the standard deviation is small, it is determined that no improvement in the estimation accuracy can be expected.

[0012] When the mean is large and the standard deviation is small, it is determined that an improvement in the estimation accuracy can be expected.

[0013] When the mean is small and the standard deviation is large, it is determined that no improvement in the estimation accuracy can be expected.

[0014] The above structure is an example of a determination method using the average and standard deviation of prediction errors.

[0015] The vehicle power consumption estimation system of the present invention is characterized in that,

[0016] The first storage unit and the second storage unit store data for each driver or each vehicle.

[0017] The above structure allows for consideration of deviations in power consumption based on the driver or vehicle.

[0018] The vehicle power consumption estimation system of the present invention is characterized in that,

[0019] The average and standard deviation of the power consumption prediction error are calculated by removing outlier data from the vehicle's power consumption.

[0020] The above structure provides a system for estimating the power consumption of a vehicle that more accurately and appropriately determines the necessity of learning and performs learning.

[0021] The method for estimating the power consumption of a vehicle according to the present invention includes:

[0022] The first storage unit stores the prediction error of the power consumption of the vehicles; and

[0023] The second storage unit stores the input-output pairs of the model that estimates the vehicle's power consumption.

[0024] If the estimation accuracy of the model for estimating the vehicle's power consumption is poor, the method uses the average and standard deviation of the prediction error of the vehicle's power consumption in the first storage unit to determine whether learning through the second storage unit can be expected to improve the estimation accuracy, and only performs learning if the improvement in estimation accuracy can be expected.

[0025] The above structure provides a method for estimating the power consumption of a vehicle that appropriately determines the necessity of learning and performs learning.

[0026] Invention Effects

[0027] This invention provides a system for estimating the power consumption of a vehicle that appropriately determines the necessity of learning and performs learning. Attached Figure Description

[0028] Figure 1A This is a first flowchart of a method for estimating the power consumption of a vehicle according to the implementation method.

[0029] Figure 1B This is the second flowchart of the method for estimating the power consumption of a vehicle according to the implementation method.

[0030] Figure 1C This is the third flowchart of the method for estimating the power consumption of a vehicle involved in the implementation method.

[0031] Figure 2 The flowchart (A) and the graph (B) showing the prediction error are for the case where the vehicle power consumption estimation method according to the implementation method has not undergone the first learning.

[0032] Figure 3 The diagram shows (A) a flowchart of the learning process of the vehicle power consumption estimation method involved in the implementation method and (B) a graph representing the prediction error.

[0033] Figure 4 The flowchart (A) and the graph (B) showing the prediction error are for the case where the second learning is not performed in the method for estimating the power consumption of the vehicle involved in the implementation method.

[0034] Figure 5A This is a diagram illustrating an example of driver learning data involved in the implementation method.

[0035] Figure 5B This is a diagram illustrating an example of learning by removing abnormal data in an implementation method.

[0036] Figure 6 This is a block diagram illustrating the structure of the power consumption estimation system for the vehicle involved in the implementation method. Detailed Implementation

[0037] Implementation

[0038] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, this is not intended to limit the invention to the following embodiments. Furthermore, not all structures described in the embodiments are necessary means to solve the problem. For clarity, the following description and drawings have been appropriately omitted and simplified. In the drawings, the same elements are labeled with the same symbols, and repeated descriptions are omitted as necessary.

[0039] (Explanation of the method for estimating the vehicle's power consumption in the implementation method)

[0040] Figure 1A This is a first flowchart of a method for estimating the power consumption of a vehicle according to the implementation method. Figure 1B This is the second flowchart of the method for estimating the power consumption of a vehicle according to the implementation method. Figure 1C This is the third flowchart of the method for estimating the power consumption of a vehicle according to the implementation method. (Reference) Figures 1A to 1C The method for estimating the power consumption of the vehicle according to the embodiment will be explained. The method for estimating the power consumption of the vehicle according to the embodiment uses machine learning. The vehicle described is an electric vehicle that runs on electricity, but it can also be a vehicle that runs on gasoline or diesel.

[0041] like Figure 1AAs shown, firstly, the input information of the model and the inherent weights of each vehicle are obtained (step S101). The model uses artificial intelligence. The model's machine learning uses deep learning such as neural networks. Therefore, the input information of the model and the inherent weights of each vehicle are obtained. However, the model only needs to be able to learn, so it is not limited to deep learning. The input of the model includes road information and vehicle information for the driving area.

[0042] Next, the predicted value of power consumption is calculated (step S102). The model output is power consumption. However, hybrid or electric vehicles consume electricity, while gasoline or diesel vehicles consume fuel. The method for estimating the vehicle's power consumption calculates predicted values ​​for fuel economy or energy efficiency. Thus, the model can be pre-learned using driving data from multiple vehicles. Assuming that when learning with weights of initial values, the prediction result is returned for average driving behavior, the prediction error is smaller under average driving behavior, and it can be determined that no further learning is needed.

[0043] Next, driving begins (step S103). Driving is performed while acquiring driving data. Next, driving ends (step S104). Next, the driving data is segmented (step S105). The driving data can be segmented as needed.

[0044] Next, driving data is input into the driver determination unit (step S106). The driver determination unit can be used to divide the stored data into queues based on the different drivers. The queues are either storage units or storage sections. Driver identification can be performed using in-vehicle cameras, etc. Even with the same vehicle, different drivers may have different driving tendencies, thus enabling a more accurate determination of the necessity for learning.

[0045] Next, driving data is input into the abnormal data determiner (step S107). Abnormal data can be deleted using the abnormal data determiner. If the driving route differs significantly from the suggested route, the abnormal data can be attributed to reasons such as accidents, traffic congestion, or prolonged stops during driving. By evaluating without using abnormal data, the necessity of learning can be accurately determined.

[0046] like Figure 1B Next, it is determined whether the data is abnormal (step S108). If the data is determined to be abnormal (if "yes" in step S108), the data is deleted (step S109). If the data is not determined to be abnormal (if "no" in step S108), the prediction error is calculated by comparing the measured value and the predicted value of power consumption for each data point (step S110).

[0047] After step S110, the calculated prediction error is appended to queue 1, which stores the prediction error per vehicle (step S111). The prediction error is appended to queue 1, which serves as the first storage unit. The capacity of the first storage unit is limited.

[0048] Next, the information input to the model and the measured power consumption are added to queue 2, which stores the model input / output pairs for each vehicle (step S112). The information input to the model and the measured power consumption as output are added to queue 2, which serves as the second storage unit. The capacity of the second storage unit is limited. Because the capacities of the first and second storage units are limited, the system size can be reduced.

[0049] However, efficient learning is essential within the system. Therefore, it is necessary to select and prioritize the data used for learning.

[0050] like Figure 1C As shown, next, it is determined whether queue 1 is full (step S113). If queue 1 is not full ("No" in step S113), the process ends.

[0051] If queue 1 is full ("Yes" in step S113), the average and standard deviation of the prediction errors stored in queue 1 are calculated (step S114). Next, it is determined whether learning is needed based on the average and standard deviation (step S115). When evaluating the prediction error, thresholds for the average and standard deviation as benchmarks are preset. If the error exceeds the threshold for the average but is below the threshold for the standard deviation, it is determined that learning is needed. If learning is not needed ("No" in step S115), the process ends.

[0052] If learning is required ("Yes" in step S115), the model input-output pairs stored in queue 2 are used for learning (step S116). Next, the weights are updated and saved (step S117). The weights of the neural network are updated. Next, queues 1 and 2 are cleared (step S118). Finally, the process ends.

[0053] The above structure provides a method for estimating the power consumption of a vehicle that appropriately determines the necessity of learning and performs learning.

[0054] (Explanation of the case where the first learning was not performed in the vehicle power consumption estimation method according to the implementation method)

[0055] Figure 2 These are (A) a flowchart and (B) a graph showing the prediction error in the case where the vehicle power consumption estimation method according to the implementation method has not undergone the first learning step. (See reference) Figure 2 The method for estimating the power consumption of the vehicle involved in the implementation method is explained in the case where the first learning is not performed.

[0056] like Figure 2 As shown in (B), if the average error and standard deviation of the power consumption prediction are small, it is determined that no learning is needed. Figure 2 As shown in (A), the process ends via the "No" path in step 115. This is equivalent to displaying the average driving behavior after learning and conforming to each vehicle, using a pre-learned model.

[0057] (Explanation of the learning process in the vehicle power consumption estimation method involved in the implementation method)

[0058] Figure 3 These are (A) a flowchart illustrating the learning process of the vehicle's power consumption estimation method according to the implementation method, and (B) a graph representing the prediction error. (Reference) Figure 3 The learning process for estimating the power consumption of vehicles involved in the implementation method will be explained.

[0059] like Figure 3 As shown in (B), when the average prediction error of power consumption is large and the standard deviation is small, it is determined that improvement can be expected through learning. Figure 3 As shown in (A), the path is "Yes" in step 115. This is equivalent to situations such as driving at extremely high speeds or other driving behaviors that easily lead to increased power consumption, or situations where the battery deteriorates over time.

[0060] (Explanation of the case where the second learning was not performed in the vehicle power consumption estimation method involved in the implementation method)

[0061] Figure 4 These are (A) a flowchart and (B) a graph showing the prediction error for the method of estimating the vehicle's power consumption according to the implementation method without the second learning step. (See reference) Figure 4 The second learning process is not performed in the method for estimating the power consumption of the vehicle involved in the implementation method.

[0062] like Figure 4 As shown in (B), when the average prediction error of power consumption is small and the standard deviation is large, since it cannot be expected to improve through learning, it is determined that learning is not necessary. Figure 4 As shown in (A), the process ends via the "No" path in step 115. This corresponds to a situation where factors not considered in the model have an impact, making it difficult to predict using the model. For example, if learning is set to occur when the prediction error exceeds a threshold, then learning would occur frequently in that case, but in this invention, it can be determined that improvement cannot be expected through learning.

[0063] (Explanation of the learning method for the driver determination device according to the implementation method)

[0064] Figure 5A This is a diagram illustrating an example of driver-specific learning data involved in the implementation method. (Reference) Figure 5A The learning method using the driver decision-maker is explained.

[0065] Assume that driving is assigned to each driver using a driver determiner. In the case where the vehicle is driven by driver A and driver B, the prediction error and input-output pairs are stored in different queues, for example, driver A in queues 1A and 2A, and driver B in queues 1B and 2B. Figure 5A As shown, even when the standard deviation of the prediction error increases due to different driving tendencies, and it is determined that no learning is needed, the necessity of learning can be more appropriately determined by dividing the queue.

[0066] (Explanation of the learning method using the anomaly data determiner described in the implementation method)

[0067] Figure 5B This diagram illustrates an example of learning to remove outlier data involved in the implementation method. (See reference) Figure 5B The learning method of the abnormal data determiner involved in the implementation method will be explained.

[0068] If the driver is traveling on a route significantly different from the suggested route, the abnormal data could be attributed to accidents, traffic congestion, or prolonged stops during the journey. Figure 5B As shown, even when the standard deviation of the prediction error increases due to the presence of outliers, and learning is deemed unnecessary, the necessity of learning can be determined more accurately by evaluating without using outliers.

[0069] (Description of the vehicle power consumption estimation system involved in the implementation)

[0070] Figure 6 This is a block diagram illustrating the structure of a power consumption estimation system for a vehicle according to the implementation method. (Reference) Figure 6 The system for estimating the power consumption of vehicles involved in the implementation method is described.

[0071] The vehicle power consumption estimation system 600 according to the embodiment includes a first storage unit 601, a second storage unit 602 and a learning unit 603.

[0072] The first storage unit 601 stores the prediction error of the vehicle's power consumption. The second storage unit 602 stores the input-output pairs of the model for estimating the vehicle's power consumption. The first storage unit 601 and the second storage unit 602 store data per driver or per vehicle.

[0073] If the model for estimating the vehicle's power consumption has poor estimation accuracy, the learning unit 603 uses the average and standard deviation of the prediction error of the vehicle's power consumption from the first storage unit 601 to determine whether the estimation accuracy can be improved by learning from the second storage unit 602, and only performs machine learning on the vehicle's power consumption if the estimation accuracy can be improved.

[0074] The vehicle power consumption estimation system 600 described in this embodiment is composed of an information processing device. The information processing device includes a processor for executing programs and a memory for storing programs. The information processing device can be a single device or multiple devices. The information processing device can be a cloud server that provides some or all of the distributed processing capabilities.

[0075] Furthermore, the present invention is not limited to the above-described embodiments, and appropriate modifications can be made without departing from its spirit.

[0076] Symbol Explanation

[0077] 600 - Vehicle power consumption estimation system, 601 - First storage unit, 602 - Second storage unit, 603 - Learning unit.

Claims

1. A system for estimating the power consumption of a vehicle, characterized in that, have: The first storage unit stores the prediction error of the power consumption of the vehicles. The second storage unit stores the input-output pairs of the model for estimating the vehicle's power consumption; and The learning unit, when the estimation accuracy of the model for estimating the vehicle's power consumption is poor, uses the average and standard deviation of the prediction error of the vehicle's power consumption in the first storage unit to determine whether learning through the second storage unit can be expected to improve the estimation accuracy, and only performs learning if the improvement in estimation accuracy can be expected.

2. The vehicle power consumption estimation system according to claim 1, characterized in that, When the mean is small and the standard deviation is small, it is determined that no improvement in the estimation accuracy can be expected. When the mean is large and the standard deviation is small, it is determined that an improvement in the estimation accuracy can be expected. When the mean is small and the standard deviation is large, it is determined that no improvement in the estimation accuracy can be expected.

3. The vehicle power consumption estimation system according to claim 1, characterized in that, The first storage unit and the second storage unit store data for each driver or each vehicle.

4. The vehicle power consumption estimation system according to claim 1, characterized in that, The average and standard deviation of the power consumption prediction error are calculated by removing outlier data from the vehicle's power consumption.

5. A method for estimating the power consumption of a vehicle, characterized in that, have: The first storage unit stores the prediction error of the power consumption of the vehicles; and The second storage unit stores the input-output pairs of the model that estimates the vehicle's power consumption. If the estimation accuracy of the model for estimating the vehicle's power consumption is poor, the method uses the average and standard deviation of the prediction error of the vehicle's power consumption in the first storage unit to determine whether learning through the second storage unit can be expected to improve the estimation accuracy, and only performs learning if the improvement in estimation accuracy can be expected.

Images