System and method for predicting endurance mileage of pure electric commercial vehicle and server

Abstract

The invention provides an endurance mileage prediction system and method for a pure electric commercial vehicle and a server, and relates to the technical field of electric vehicle intellectualization, the endurance mileage prediction system comprises a data acquisition layer, a data processing layer, a data fusion layer and a self-learning layer, the data acquisition layer is used for acquiring vehicle state information, external environment information and actual endurance information; the data processing layer is used for determining an energy consumption estimation value, a first endurance mileage prediction result and a second endurance mileage prediction result; the data fusion layer is used for performing fuzzy logic fusion processing on the vehicle state information and the external environment information, determining a weighted average weight coefficient, performing weighted average fusion processing based on the weight coefficient, and determining a target endurance prediction result at the current moment; and the self-learning layer is used for performing real-time analysis processing on the deviation value to determine a target weight coefficient, so that the data fusion layer performs endurance mileage prediction at the next moment based on the target weight coefficient. According to the method, the accuracy of predicting the endurance mileage of the pure electric commercial vehicle can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle intelligence, and in particular to a range prediction system, method and server for a pure electric commercial vehicle. Background Art

[0002] With the rapid development of new energy technologies and the increasing awareness of environmental protection, pure electric commercial vehicles have gradually become an important part of the global transportation field. However, the range limit of pure electric commercial vehicles is still a key obstacle to their large-scale application, especially in urban distribution, long-distance transportation and complex traffic environments. Users' anxiety about range will significantly affect the market acceptance of electric commercial vehicles. At present, the range can be calculated by average energy consumption and remaining power, but this solution fails to fully consider the influence of various complex factors, resulting in low accuracy of the prediction results. Related technologies have proposed that the loss data model can also be used to calculate the influencing factors during vehicle driving to obtain the total energy consumption of the vehicle, and further determine the remaining range based on the total energy consumption of the vehicle and the total battery power. However, since the loss data model cannot accurately express all factors that affect the range, there is still a large error in the prediction result. Summary of the invention

[0003] In view of this, an object of the present invention is to provide a range prediction system, method and server for a pure electric commercial vehicle, which can significantly improve the accuracy of the range prediction of the pure electric commercial vehicle.

[0004] In a first aspect, an embodiment of the present invention provides a cruising range prediction system for a pure electric commercial vehicle, the system comprising: a data acquisition layer, a data processing layer, a data fusion layer and a self-learning layer, the data processing layer is respectively connected to the data acquisition layer and the data fusion layer, the self-learning layer is respectively connected to the data fusion layer and the data acquisition layer; wherein the data acquisition layer is used to collect vehicle status information, external environment information and actual cruising range information through a preset sensor set, wherein the vehicle status information comprises: vehicle speed information, battery status information, vehicle acceleration and steering angle, throttle and brake pedal displacement, motor status and driving mode, and the external environment information comprises: ambient temperature information, road condition information; the data processing layer is used to perform energy consumption estimation processing on the vehicle status information and the external environment information, determine the energy consumption estimation value, and the first cruising range corresponding to the energy consumption estimation value The data fusion layer is used to perform fuzzy logic fusion processing on the vehicle status information and the external environment information, determine the weight coefficient of weighted average, and based on the weight coefficient, perform weighted average fusion processing on the first cruising range prediction result and the second cruising range prediction result to determine the target cruising range prediction result at the current moment; the self-learning layer is used to perform real-time analysis and processing on the deviation value between the energy consumption change in the target cruising range prediction result and the energy consumption change in the actual cruising range information, update the weight coefficient according to the deviation analysis result, and determine the target weight coefficient, so that the data fusion layer can perform cruising range prediction at the next moment based on the target weight coefficient.

[0005] In one embodiment, the data acquisition layer also includes: a self-diagnosis unit; wherein the self-diagnosis unit is used to perform abnormality detection processing on the collected vehicle status information and external environment information, and perform fault sensor switching and fault alarm when a sensor failure or abnormal data is detected.

[0006] In one embodiment, the data processing layer includes: an average energy consumption estimation module; wherein the average energy consumption estimation module is used to calculate the remaining battery power, driving resistance loss power, motor efficiency loss power, battery consumption power and energy recovery power according to vehicle status information and external environment information, so as to determine the energy consumption per unit time according to the driving resistance loss power, motor efficiency loss power, battery consumption power and energy recovery power, and determine the energy consumption per unit mileage according to the energy consumption per unit time.

[0007] In one embodiment, the data processing layer also includes: a deep learning prediction module, the deep learning prediction module includes: a random forest model; wherein the random forest model is used to analyze the driving situation and determine that the driver is a common vehicle driver, to perform long-term energy consumption change trend prediction processing on the cruising range, and determine a second cruising range prediction result.

[0008] In one embodiment, the deep learning prediction module also includes: a long short-term memory network model; wherein the long short-term memory network model is used to analyze the driving situation and determine that the driver is not an ordinary vehicle driver, to perform short-term energy consumption change trend prediction processing on the cruising range and determine a second cruising range prediction result.

[0009] In one embodiment, the data fusion layer includes: a fuzzy logic subsystem; wherein the fuzzy logic subsystem is used to fuzzify vehicle status information and external environment information, and use a preset reasoning engine to reason on the fuzzified information to defuzzify the reasoning result and obtain a weighted average weight coefficient.

[0010] In one embodiment, the data fusion layer also includes: a weighted average fusion unit; wherein the weighted average fusion unit is used to determine the target range prediction result at the current moment by summing the products of the first range prediction result and the second range prediction result and the corresponding weight coefficients based on the weight coefficient and the preset fusion model.

[0011] In one embodiment, the self-learning layer includes: a data feedback and storage module; wherein the data feedback and storage module is used to collect the deviation value of the target cruising range prediction result at each moment and the energy consumption change in the actual cruising range information, as well as the vehicle status information and external environment information corresponding to each moment, and construct a historical data set based on the deviation value, vehicle status information and external environment information at each moment.

[0012] In one embodiment, the self-learning layer also includes: a model optimization module; wherein the self-learning layer is used to update the weight coefficient based on the deviation value through the incremental learning model to determine the target weight coefficient.

[0013] In a second aspect, an embodiment of the present invention further provides a method for predicting the range of a pure electric commercial vehicle, which is applied to the range prediction system of a pure electric commercial vehicle according to any one of the first aspects, and the method comprises: collecting vehicle status information, external environment information and actual range information, wherein the vehicle status information comprises: vehicle speed information, battery status information, vehicle acceleration and steering angle, throttle and brake pedal displacement, motor status and driving mode, and the external environment information comprises: ambient temperature information, road condition information; performing energy consumption estimation processing on the vehicle status information and the external environment information, determining an energy consumption estimation value, and a first range prediction result corresponding to the energy consumption estimation value, and performing in-depth processing on the vehicle status information and the external environment information; The method further comprises the following steps: performing a fuzzy logic fusion process on the vehicle status information and the external environment information to determine a weight coefficient of weighted average, and based on the weight coefficient, performing a weighted average fusion process on the first cruising range prediction result and the second cruising range prediction result to determine a target cruising range prediction result at the current moment; performing real-time analysis and processing on the deviation value between the energy consumption change in the target cruising range prediction result and the energy consumption change in the actual cruising range information, updating the weight coefficient according to the deviation analysis result, and determining the target weight coefficient, so that the data fusion layer predicts the cruising range at the next moment based on the target weight coefficient.

[0014] In a third aspect, an embodiment of the present invention further provides a server, comprising a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement any one of the methods provided in the second aspect.

[0015] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement any one of the methods provided in the second aspect.

[0016] The embodiments of the present invention bring the following beneficial effects: The embodiment of the present invention provides a cruising range prediction system, method and server for a pure electric commercial vehicle. The system can use a data collection layer to collect vehicle state information, external environment information and actual cruising range information, and use a data processing layer to perform energy consumption estimation processing on the vehicle state information and external environment information to determine the energy consumption estimation value and the first cruising range prediction result corresponding to the energy consumption estimation value, and perform deep learning prediction processing on the vehicle state information and the external environment information to determine the second cruising range prediction result, and then perform fuzzy logic fusion processing on the vehicle state information and the external environment information through the data fusion layer to determine the weight coefficient of weighted average, and based on the weight coefficient, perform weighted average fusion processing on the first cruising range prediction result and the second cruising range prediction result to determine the target cruising range prediction result at the current moment, and finally perform real-time analysis and processing on the deviation value between the energy consumption change in the target cruising range prediction result and the energy consumption change in the actual cruising range information through the self-learning layer, update the weight coefficient according to the deviation analysis result, and determine the target weight coefficient, so that the data fusion layer performs cruising range prediction at the next moment based on the target weight coefficient. The embodiment of the present invention can significantly improve the accuracy of the cruising range prediction of the pure electric commercial vehicle.

[0017] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A schematic diagram of the structure of a range prediction system for a pure electric commercial vehicle provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of an average energy consumption estimation module provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a deep learning prediction module provided by an embodiment of the present invention; Figure 4A schematic diagram of the structure of a random forest model provided by an embodiment of the present invention; Figure 5 A structural diagram of a data fusion module provided by an embodiment of the present invention; Figure 6 A schematic diagram of a flow chart of a method for predicting the range of a pure electric commercial vehicle provided by an embodiment of the present invention; Figure 7 A schematic diagram of the structure of a server provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described in combination with the embodiments below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] At present, with the rapid development of new energy technologies and the increasing awareness of environmental protection, pure electric commercial vehicles have gradually become an important part of the global transportation field. As the number of electric commercial vehicles continues to increase, the driving range prediction function has become a key factor that users pay attention to when choosing commercial vehicles. Although most of the existing pure electric commercial vehicles are equipped with a driving range prediction function, because it only relies on basic factors such as charge state and average energy consumption for calculation, there is a large gap between the predicted result of the driving range and the actual usage, which affects the user experience.

[0023] Compared with traditional fuel commercial vehicles, pure electric commercial vehicles have significant advantages in reducing carbon emissions and environmental pollution. However, the range limitation is still a key obstacle to its large-scale application, especially in urban distribution, long-distance transportation and complex traffic environments. Users' anxiety about range has affected the market acceptance of electric commercial vehicles. Traditional range prediction methods usually rely on simple average energy consumption and remaining power calculations. Although this method can provide preliminary estimates, it fails to fully consider the impact of multiple complex factors, resulting in low accuracy of the prediction results.

[0024] Factors that affect the range of pure electric commercial vehicles include not only battery capacity and state of charge, but also driving behavior, road conditions, climate conditions, vehicle load, battery aging and other factors. For example, frequent starts and stops in urban distribution increase energy consumption, while energy consumption is relatively low when driving at high speeds. The impact of ambient temperature on battery performance is also very significant. The output capacity of the battery will drop significantly in low temperature environments, resulting in a reduction in available range. The degree of battery aging also affects the range, especially as the battery ages, even if the battery is fully charged, the actual range is far lower than that of a new battery.

[0025] The impact of the driver's driving style on energy consumption is also a factor that cannot be ignored; commercial vehicle drivers are usually faced with frequent driving behaviors such as starting, braking and sudden acceleration. Especially in complex urban traffic conditions, intense driving styles will greatly increase instantaneous energy consumption and reduce battery life. Road slope, traffic conditions, etc. also directly affect energy consumption; when going uphill, the vehicle needs more energy to overcome the influence of gravity, while when going downhill, the energy recovery system can partially recover kinetic energy, thereby reducing overall energy consumption.

[0026] Current technical solutions fail to fully integrate these influencing factors, resulting in inaccurate range predictions, which in turn exacerbates "range anxiety"; especially in areas with insufficient long-distance transportation and charging facilities, users have doubts about whether the vehicle can successfully complete the task. In addition, existing electric commercial vehicle systems generally lack intelligent energy consumption management and dynamic adjustment capabilities, and fail to make accurate range predictions based on real-time environmental and driving conditions.

[0027] Based on the above background, the industry urgently needs a more intelligent and accurate range prediction method, which should be able to comprehensively consider multiple factors such as the real-time status of the vehicle, historical driving data, driver habits, and environmental changes, and conduct in-depth data analysis through machine learning and deep learning algorithms to maximize the accuracy of range prediction, reduce user anxiety, and enhance user experience. This demand has spawned the research and development of a new generation of intelligent range prediction systems, which aims to combine data from different sources and use advanced machine learning technology to provide more accurate range estimates.

[0028] The existing technology proposes that the total energy consumption of the vehicle can be calculated by determining the influencing factors such as driving resistance loss, motor loss, battery consumption and energy recovery during vehicle driving, and the remaining cruising range can be further determined based on the total energy consumption of the vehicle and the total battery power. However, it ignores the factors in reality that affect the cruising range but cannot be accurately expressed through the energy loss mathematical model.

[0029] In addition, researchers and automakers are also exploring various possible solutions. For example, they collect driving data in real time through multiple sensors on the vehicle, and process and analyze the data through cloud computing and artificial intelligence algorithms to generate more accurate range predictions. Some studies focus on using neural networks to predict the discharge characteristics of batteries in order to better predict the remaining power. However, these studies still face many challenges in practical applications, such as the real-time nature of the data, the generalization ability of the model, and how to achieve efficient model reasoning with low computing resources.

[0030] Based on this, the present invention implements the system, method and server for predicting the range of pure electric commercial vehicles. First, the data acquisition module collects multi-dimensional data in real time, and the collected data is transmitted to the central processing unit and input into the average energy consumption estimation module and the deep learning prediction module at the same time. In the average energy consumption estimation module, the set mathematical model is used for calculation, and the result is output to the data fusion module. In the deep learning prediction module, the system makes an output decision according to the driving style, and finally outputs the selected result to the data fusion module. In the data fusion module, the system performs fuzzy logic reasoning according to the environmental state and the whole vehicle state, and imports the output weights of each module into the weighted average fusion algorithm. At the same time, the self-learning module performs deviation analysis and model optimization on the input data, and feeds back the optimized parameters to the deep learning module and the data fusion module. Finally, the data fusion module dynamically adjusts the weight value, and finally outputs the final range prediction value, which can significantly improve the accuracy of the range prediction of pure electric commercial vehicles.

[0031] To facilitate understanding of this embodiment, a method for predicting the range of a pure electric commercial vehicle disclosed in an embodiment of the present invention is first described in detail. The method is applied to a range prediction system for a pure electric commercial vehicle. The system includes: a data acquisition layer, a data processing layer, a data fusion layer, and a self-learning layer. To facilitate understanding of the method for predicting the range of a pure electric commercial vehicle, an embodiment of the present invention provides a schematic diagram of the structure of a range prediction system for a pure electric commercial vehicle, such as Figure 1 As shown, the data processing layer is connected to the data acquisition layer and the data fusion layer respectively, the self-learning layer is connected to the data fusion layer and the data acquisition layer respectively, the data acquisition layer is composed of a data acquisition module, the data processing layer is composed of an average energy consumption estimation module and a deep learning prediction module, the data fusion layer is composed of a data fusion module, and the self-learning layer is composed of a data feedback and storage module and a model optimization module.

[0032] The data acquisition layer is composed of data acquisition modules, which are used to collect vehicle status information, external environment information and actual endurance information through a preset sensor set, so as to provide accurate data support for subsequent data analysis, model prediction and decision-making. The data acquisition module comprehensively monitors the vehicle's operating status and dynamic changes in the external environment by deploying a variety of sensors at key parts of the vehicle to ensure that the system can obtain high-quality real-time data and make intelligent predictions and control decisions based on the data. Among them, the data acquisition module can collect vehicle status information and external environment information in real time. Vehicle status information includes: vehicle speed information, battery status information, vehicle acceleration and steering angle, throttle and brake pedal displacement, motor status and driving mode. External environment information includes: ambient temperature information, road condition information. Specifically, data collection can be achieved by deploying the following sensors at key parts of pure electric commercial vehicles: Vehicle speed sensor: Installed on the wheel bearing or gearbox output shaft, it monitors the vehicle's speed in real time. The sensor outputs a pulse signal, which is transmitted to the central processing unit (CPU) through the vehicle bus and displayed and calculated in km / h.

[0033] Battery status sensor: including voltage sensor, current sensor and temperature sensor. These sensors collect battery charging and discharging parameters in real time through the battery management system (BMS).

[0034] Ambient temperature sensor: It is placed on the bottom of the vehicle to avoid heat interference from vehicle emissions. The sensor uses an NTC thermistor with a measurement range of -30°C to 50°C.

[0035] Road condition sensor: LiDAR and inertial navigation unit (IMU) work together to obtain slope information. LiDAR is used to detect road surface features, and IMU records vehicle inclination data. The calculated results are filtered and used as model input.

[0036] Throttle and brake pedal sensor: detects the driver's throttle opening and brake force through a displacement potentiometer, with an opening range of 0%-100%.

[0037] Acceleration sensor: installed on the vehicle chassis, it records the acceleration and deceleration behavior of the vehicle in real time, and the data is used to identify driving style.

[0038] Steering angle sensor: integrated on the steering wheel shaft, records the changes in the vehicle's steering angle in real time.

[0039] In one implementation, the above vehicle status information and external environment information are specifically described in (1) to (7) below: (1) Vehicle speed information is used to monitor the vehicle’s current speed in real time, providing dynamic basic data for range prediction and energy efficiency analysis.

[0040] (2) Battery status information includes multiple key parameters such as voltage, current, and temperature, which comprehensively monitor the battery’s discharge status and health status.

[0041] (3) Ambient temperature information: Temperature changes in the external environment have a significant impact on battery performance and energy consumption, especially under extreme temperature conditions. Cold weather can reduce the battery's energy output capacity. Ambient temperature sensors can monitor external temperature changes in real time and assist in making range prediction adjustments.

[0042] (4) Road condition information: Road condition sensors and slope sensors are used to collect information about road slope, type, and other characteristics, providing dynamic data for energy efficiency optimization and range prediction.

[0043] (5) Vehicle acceleration and steering angle: The acceleration and deceleration of the vehicle are monitored by the acceleration sensor, and the driver's driving behavior is analyzed in combination with the steering angle data. This helps to evaluate the impact of driving style (such as sudden acceleration, sudden braking, etc.) on energy consumption, thereby providing a basis for optimizing driving behavior and extending battery life.

[0044] (6) The accelerator and brake pedal displacement is used to monitor the driver's control behavior, further analyze the impact of the driver's behavior on energy consumption, and provide data support for personalized energy efficiency optimization and driving recommendations.

[0045] (7) Motor status and driving mode are used to help determine the energy efficiency performance of the vehicle under different working conditions.

[0046] The collected vehicle and environmental information is transmitted in real time through an efficient and stable vehicle bus system and sent to the central processing unit (CPU) for unified processing and storage, ensuring the real-time, integrity and consistency of the data, and providing reliable support for subsequent data analysis, model prediction and decision-making.

[0047] In one embodiment, the data acquisition layer also includes: a self-diagnosis unit; wherein the self-diagnosis unit is used to perform abnormal detection and processing on the collected vehicle status information and external environment information, and when a sensor failure or abnormal data is detected, perform fault sensor switching and fault alarm. That is to say, in order to ensure the high quality of data and the stability of the system, the data acquisition module also adopts a redundant design and self-diagnosis mechanism, which can automatically perform sensor switching adjustment and alarm in the event of sensor failure or data abnormality, thereby minimizing the risk of data loss and erroneous data, and ensuring the high reliability and continuous operation of the system in harsh environments.

[0048] In addition, in order to achieve comprehensive perception of vehicle status and external environment, the data acquisition module includes the following types of sensors: vehicle speed sensor, battery status sensor, ambient temperature sensor, acceleration sensor, road condition sensor, motor status sensor, driving mode sensor, throttle and brake pedal displacement sensor. Through the collaborative work and data fusion of multiple types of sensors, comprehensive monitoring of vehicle status and environmental information is achieved, and a high-quality data foundation is provided for endurance prediction, energy efficiency analysis and intelligent control.

[0049] The data processing layer is used to perform energy consumption estimation processing on the vehicle status information and external environment information, determine the energy consumption estimation value, and the first cruising range prediction result corresponding to the energy consumption estimation value, and perform deep learning prediction processing on the vehicle status information and external environment information to determine the second cruising range prediction result, wherein the energy consumption estimation value includes energy consumption per unit time and energy consumption per unit mileage.

[0050] In one embodiment, see Figure 2 The structure diagram of an average energy consumption estimation module shown in the figure, the data processing layer includes: an average energy consumption estimation module; wherein the average energy consumption estimation module is used to calculate the remaining battery power, driving resistance loss power, motor efficiency loss power, battery consumption power and energy recovery power according to the vehicle status information and external environment information, so as to determine the energy consumption per unit time according to the driving resistance loss power, motor efficiency loss power, battery consumption power and energy recovery power, and determine the energy consumption per unit mileage according to the energy consumption per unit time, that is, the average energy consumption estimation module mainly estimates the energy consumption based on the current vehicle status and driving status, and the system can quickly calculate the energy consumption per unit time and per unit mileage through the battery current, battery voltage, vehicle speed, acceleration, road rolling resistance coefficient, road slope angle and motor speed measured by the sensor, and specifically includes the following calculation process: battery SOC value calculation, driving resistance loss power calculation, motor efficiency loss power calculation, battery consumption power calculation, energy recovery power calculation, total energy consumption per unit time calculation and energy consumption per unit mileage calculation.

[0051] In another embodiment, see Figure 3 A structural schematic diagram of a deep learning prediction module is shown, the data processing layer also includes: a deep learning prediction module, the deep learning prediction module includes: a long short-term memory network (Long Short-Term Memory, LSTM) and a random forest (Random Forest) model, which comprehensively utilizes these two algorithms to conduct in-depth analysis and modeling of the vehicle's historical driving data, driving behavior characteristics and environmental conditions to achieve accurate mileage prediction.

[0052] Before model training, data preprocessing is required. Data preprocessing aims to improve the quality and consistency of input data, thereby improving the generalization ability and prediction accuracy of the model. The specific preprocessing steps include the following aspects: data cleaning, cleaning the collected raw data, using statistical methods and the principle of three times the standard deviation to detect outliers in the data, identify and remove outliers, and ensure the reliability and consistency of the input data. After data cleaning, data normalization is performed to convert the data to the same scale range to improve training efficiency and accuracy, and avoid the order of magnitude differences of different features from having an adverse effect on model training. Through these preprocessing steps, the input data will have higher quality and consistency, ensuring that the deep learning model can be effectively trained and produce high-precision prediction results.

[0053] For further information, see Figure 4The schematic diagram of the structure of a random forest model is shown in the figure. The random forest model is used to analyze the driving situation and determine the long-term energy consumption trend of the cruising range when the driver is determined to be a common vehicle driver, and determine the second cruising range prediction result. Specifically, the random forest model training is an integrated learning method to improve the prediction ability and generalization performance of the model. By generating multiple independent decision trees and integrating their prediction results, it can effectively handle complex nonlinear relationships and is particularly suitable for capturing the impact of environmental conditions and driving behavior on cruising range. The training process of the random forest model includes generating multiple random feature subsets and constructing multiple decision trees based on these subsets. Each decision tree independently predicts the cruising range, and its final output is the weighted average of the prediction results of all decision trees. Specifically, let fᵢ(X) be the prediction result of the i-th decision tree, and the final output F(X) of the random forest is the weighted average of the outputs of all decision trees, that is:

[0054] Among them, N is the total number of decision trees. The random forest model can not only effectively reduce the overfitting risk of a single decision tree, but also improve the stability and robustness of the prediction by integrating the results of multiple trees.

[0055] In addition, the long short-term memory network model is used to analyze the driving situation and determine that the driver is an infrequent vehicle driver. The short-term energy consumption change trend of the cruising range is predicted and processed to determine the second cruising range prediction result. Specifically, the long short-term memory network model (LSTM) is an important module used by the deep learning prediction module to process time series data. It can effectively capture the time dependency in vehicle status and energy consumption changes. The LSTM network can retain long-term dependency information through its unique memory gating mechanism and selectively forget useless information during the training process, thereby effectively simulating the dynamic behavior characteristics of the vehicle on the timeline. During the model training process, the vehicle status data is input into the LSTM network as an input sequence. The LSTM model captures the trend of energy consumption over time by learning the relationship between these status data and the cruising range.

[0056] Since the LSTM network focuses on capturing the time series features in the vehicle status data, it is suitable for predicting short-term energy consumption trends, while the random forest can handle the complex nonlinear relationship between the input features, especially when analyzing the impact of environmental conditions and driving behavior on the cruising range, it has significant advantages and is suitable for predicting long-term energy consumption trends. In order to combine the advantages of the random forest model and the LSTM model to achieve accurate prediction of the cruising range. Output decisions are added to the deep learning prediction module of the system. Specifically, the system will judge whether the driver is a regular user of the vehicle through the driving conditions of the pure electric commercial vehicle. If so, it will choose to output the prediction results of the random forest. If not, it will choose to output the prediction results obtained by using the LSTM model to make short-term predictions on the vehicle's energy consumption trends. The prediction process can be adjusted dynamically, and the optimal cruising range prediction results are selected according to the real-time situation.

[0057] By integrating the advantages of the random forest model and the LSTM model, the deep learning prediction module of the system of the present invention can provide high-precision mileage estimation, has strong adaptability and robustness, and can provide stable and reliable mileage estimation in complex road environments and under different drivers.

[0058] The data fusion layer is used to perform fuzzy logic fusion processing on the vehicle status information and external environment information, determine the weight coefficient of weighted average, and based on the weight coefficient, perform weighted average fusion processing on the first cruising range prediction result and the second cruising range prediction result to determine the target cruising range prediction result at the current moment. The data fusion layer is composed of a data fusion module. The data fusion module generates a more accurate and reliable final cruising range prediction value by comprehensively analyzing the cruising range prediction results from different sources. In order to achieve this goal, a weighted average fusion unit and a fuzzy logic subsystem are used to effectively fuse the outputs from different sources and dynamically adjust the weights, so as to cope with complex and changeable driving environments and vehicle conditions.

[0059] In one embodiment, see Figure 5 The structural diagram of a data fusion module shown in FIG. 1 shows that the data fusion layer includes: a fuzzy logic subsystem; wherein the fuzzy logic subsystem is used to fuzzify the vehicle state information and the external environment information, and use a preset reasoning engine to reason about the fuzzified information to defuzzify the reasoning result and obtain the weight coefficient of the weighted average. In other words, the fuzzy logic subsystem can dynamically adjust the weight in the weighted average to cope with different driving environments and vehicle state changes. The core of fuzzy logic fusion is to transform the driving environment and vehicle state information into the basis for model weight adjustment by processing uncertainty and fuzziness. Specifically, the fuzzy logic system consists of three parts: fuzzification, reasoning engine and defuzzification.

[0060] Fuzzification: First, the actual collected environmental information and vehicle status are input into the fuzzy logic system, and these precise data are converted into membership in the fuzzy set.

[0061] Reasoning engine: Then the fuzzified input is inferred using the set fuzzy rule base, which contains a series of "if-then" rules to dynamically adjust the weights of different prediction models according to the driving environment and vehicle status.

[0062] Defuzzification: Finally, the fuzzy output obtained by the inference engine is converted into precise values, which are the adjusted weights w1 and w2.

[0063] In another embodiment, the data fusion layer further includes: a weighted average fusion unit; wherein the weighted average fusion unit is used to determine the sum of the products of the first cruising range prediction result and the second cruising range prediction result and the corresponding weight coefficients, respectively, as the target cruising range prediction result at the current moment, based on the weight coefficient and the preset fusion model. That is, the weighted average fusion is used to comprehensively process the cruising range prediction results from different models to obtain an overall cruising range prediction value. Specifically, the method performs weighted average on the results from the average energy consumption estimation model and the deep learning prediction model according to the weights, and the fusion formula is as follows:

[0064] Among them, Pfinal represents the final fused endurance prediction value, Pavg is the output value of the average energy consumption estimation module, PDL is the prediction value output by the deep learning prediction module, w1 and w2 are the weight coefficients of the two prediction models, respectively, satisfying w1+w2=1. These weight coefficients are not fixed, but are dynamically adjusted according to the current state of the vehicle and external environmental conditions. The specific adjustment values ​​are given by the fuzzy logic system.

[0065] By combining weighted average fusion and fuzzy logic systems, we can effectively cope with the changing external environment and unstable driving conditions during vehicle operation. For example, in the low temperature environment in winter, battery performance tends to decline. At this time, the fuzzy logic system will reduce its reliance on average energy consumption estimation and rely more on deep learning models to provide more targeted range predictions. In relatively stable highway driving scenarios, weighted average fusion will focus on the average energy consumption estimation model to improve the accuracy of range.

[0066] The self-learning layer is used to analyze and process the deviation between the energy consumption change in the target cruising range prediction result and the energy consumption change in the actual cruising range information in real time, update the weight coefficient according to the deviation analysis result, and determine the target weight coefficient, so that the data fusion layer can predict the cruising range at the next moment based on the target weight coefficient. In one embodiment, the self-learning layer includes: a data feedback and storage module; wherein the data feedback and storage module is used to collect the deviation between the target cruising range prediction result at each moment and the energy consumption change in the actual cruising range information, as well as the vehicle state information and external environment information corresponding to each moment, and construct a historical data set based on the deviation value, vehicle state information and external environment information at each moment, that is, the data feedback and storage module is responsible for real-time collection of the deviation data between the cruising range prediction value output by the data fusion module and the actual cruising range value of the vehicle during the actual driving process, as well as the corresponding environmental conditions and vehicle state information. These data are then stored in the vehicle storage and the cloud server to form a complete historical data set, which is convenient for the subsequent optimization process.

[0067] In another embodiment, the self-learning layer further includes: a model optimization module; wherein the self-learning layer is used to update the weight coefficient based on the deviation value through the incremental learning model to determine the target weight coefficient, that is, the model optimization module will update the weight parameters of the average energy consumption estimation module and the deep learning prediction module in the data fusion layer through online learning based on the collected data deviation through the incremental learning method to ensure that the system can quickly adapt to the latest environment and driving conditions. At the same time, through the reinforcement learning method, a reward mechanism is constructed between the model prediction and the actual performance to achieve the adaptive adjustment of the system and dynamically adjust the weight allocation strategy in the data fusion module.

[0068] See also Figure 6 The flowchart of a method for predicting the cruising range of a pure electric commercial vehicle is shown in FIG. 1 , and the method mainly includes the following steps S602 to S608: Step S602, collect vehicle status information, external environment information and actual endurance information, wherein the vehicle status information includes: vehicle speed information, battery status information, vehicle acceleration and steering angle, throttle and brake pedal displacement, motor status and driving mode, and the external environment information includes: ambient temperature information, road condition information. In one embodiment, data transmission and storage transmit data to the central processing unit (CPU) via the vehicle-mounted CAN bus, and the data is stored in the vehicle-mounted memory with a timestamp as the index. In order to ensure the reliability of data collection, a dual redundant design can be adopted: each group of key sensors is equipped with a backup sensor, and when the main sensor fails, it automatically switches to the backup sensor. In addition, the system uses a self-diagnosis algorithm to regularly detect whether the sensor output exceeds a preset range, and triggers an alarm if abnormal data is found.

[0069] Step S604: perform energy consumption estimation processing on the vehicle status information and the external environment information to determine an energy consumption estimation value and a first cruising range prediction result corresponding to the energy consumption estimation value, and perform deep learning prediction processing on the vehicle status information and the external environment information to determine a second cruising range prediction result, wherein the energy consumption estimation value includes energy consumption per unit time and energy consumption per unit mileage.

[0070] In one implementation, the energy consumption per unit time and the energy consumption per unit mileage may be calculated in steps based on real-time data. The specific steps are as follows: SOC value estimation:

[0071]

[0072] Among them, Iin is the bus voltage of the batteries connected in series; CN is the rated capacity of the battery, Ah; SOCInit is the initial SOC value of the battery; ηbat is the battery efficiency.

[0073] Calculation of power loss due to driving resistance:

[0074]

[0075] Among them, Ft is the vehicle's driving resistance, N; v is the vehicle's driving speed, m / s; m is the vehicle's mass, kg; f is the rolling resistance coefficient; α is the road ramp angle, (°); CD is the drag coefficient; A is the vehicle's frontal area; δ is the rotational inertia coefficient of the transmission system; a is the vehicle's acceleration, m / s2; PF is the vehicle's driving resistance loss power, W.

[0076] Motor efficiency loss power calculation:

[0077]

[0078] Where: Pm is the motor efficiency loss power, W; P is the motor theoretical output power, W; Pa is the motor actual output power, W; T is the vehicle required torque, Nm; n is the motor speed, r / min; η is the motor efficiency at the current torque and speed.

[0079] Battery power consumption:

[0080] Where: Pb is the power consumed by the battery, W; Uloss is the battery loss voltage, V.

[0081] Energy recovery power:

[0082] Where: Pg is the energy recovery power, W; I is the battery bus current, A; UB is the battery terminal voltage, V.

[0083] Calculation of energy consumption per unit time:

[0084] Where: Ecost is the energy consumption per unit time, J.

[0085] Calculation of energy consumption per unit mileage:

[0086] Where: Escost is the energy consumption per unit mileage, J.

[0087] In one implementation, the calculation results can be updated based on the data collected in real time by the data acquisition module and transmitted to the data fusion module, and deep learning prediction processing can be performed on the vehicle status information and the external environment information. When determining the second cruising range prediction result, if the deep learning prediction module is used, the data set collected by the data acquisition module is first divided into a training set and a test set. First, the training set data is randomly sampled to obtain i training set samples {D1(X), D2(X), D3(X), …, Di(X)} and i test set samples, a decision tree model is established, and the decision tree is randomly split to establish i decision trees. The test set data is brought into the trained decision tree model. Through model calculation, i results {C1(X), C2(X), C3(X), …, Ci(X)} can be obtained, and the average value of {Ci, i=1, 2, 3, …, i} can be used to obtain the predicted value of the cruising range.

[0088] In addition, the LSTM neural network model can also be used for deep learning prediction processing to determine the second cruising range prediction result. The LSTM neural network model is an improved variant based on the recurrent neural network (RNN). By introducing three "gate" structures (forget gate, input gate and output gate) to the RNN, the LSTM neural network can selectively retain the input sequence information, effectively solving the problems of gradient explosion and gradient disappearance in the long sequence training process of the RNN, making the LSTM neural network very effective in processing long sequence information. The steps of building a pure electric commercial vehicle cruising range prediction model based on the LSTM neural network are as follows: According to the data collected by the data acquisition module, the input and output of the LSTM neural network model at time t are defined as:

[0089] Among them, Xt is the input at time t; ht is the output at time t. After processing ht−1 and Xt information through the Sigmoid function, the forget gate outputs a vector between 0 and 1 to determine the degree of information retention. 0 means all information is discarded, and 1 means all information is retained. The formula is described as:

[0090] Among them, Wf and bf are the weight matrix and bias of the forgetting layer, σ is the Sigmoid nonlinear activation function, and ft is the forgetting gate. The input gate determines which important information is input to the current cell state, and the ht−1 and Xt information are respectively passed into the Sigmoid function and the tanh function. The Sigmoid function determines which information is updated, and a new candidate value vector gt is generated when the tanh function is passed. Finally, the output value of the Sigmoid function is multiplied by the output value of the tanh function, and the new cell information Ct is obtained by combining the forgetting gate to choose to forget part of the old cell information. The formula is described as follows:

[0091] Among them, Wi, bi, Wg, bg are the weight matrices and biases of the two layers, tanh is the nonlinear activation function, and it is the input gate.

[0092] The output gate is used to determine the value of the next hidden layer state. First, the Sigmoid function is used to obtain the state characteristics of the output cell, and then the tanh activation function is used to obtain the state value of the cell, and then the state value is multiplied by the output value of the Sigmoid function. The specific formula is described as follows:

[0093] Where Wo and bo are the weight matrix and bias respectively, and Ot is the output gate.

[0094] Implementation of data fusion module: In the data fusion module, the input variables in the fuzzy logic rule base are defined as follows: Ambient temperature: Fuzzy sets: low temperature (LT), moderate temperature (MD), high temperature (HT).

[0095] Domain: [−30℃, 10℃], [10℃, 25℃], [25℃, 50℃].

[0096] Road slope: Fuzzy sets: flat (FL), moderate slope (MD), steep slope (ST).

[0097] Domain: [0%, 5%], [5%, 10%], [10%, 20%].

[0098] Vehicle speed: Fuzzy sets: low speed (LS), medium speed (MS), high speed (HS).

[0099] Domain: [0km / h, 40km / h], [40km / h, 80 km / h], [80km / h, 120km / h].

[0100] Battery SOC (State of Charge): Fuzzy sets: low (LOW), medium (MED), high (HIGH).

[0101] Domain: [0%, 20%], [20%, 60%], [60%, 100%].

[0102] In the data fusion module, the output variables in the fuzzy logic rule base are defined as follows: Weight w1 (weight of average energy consumption estimation module output value): Fuzzy sets: low (LOW), medium (MED), high (HIGH).

[0103] Domain: 0.2, 0.5, 0.8.

[0104] Weight w2 (weight of output value of deep learning prediction module): Fuzzy sets: low (LOW), medium (MED), high (HIGH).

[0105] Domain: 0.2, 0.5, 0.8.

[0106] Step S606, fuzzy logic fusion processing is performed on the vehicle state information and the external environment information to determine the weight coefficient of weighted average, and based on the weight coefficient, weighted average fusion processing is performed on the first cruising range prediction result and the second cruising range prediction result to determine the target cruising range prediction result at the current moment. In one embodiment, the fuzzy logic system consists of three parts: fuzzification, reasoning engine and defuzzification. Among them, the reasoning engine uses a set fuzzy rule base to reason on the fuzzified input. The fuzzy logic rule base can be shown in Table 1 below: Table 1 Fuzzy logic rule base

[0107] Furthermore, the rules of the fuzzy logic rule base used by the inference engine are specifically explained as follows (1) to (10): (1) When the input variables after fuzzification are low ambient temperature (LT), flat road slope (FL), low vehicle speed (LS), and low battery SOC (LOW), due to the complexity of the environment and battery status, the weights are biased towards the deep learning model. Therefore, the output variable average energy consumption estimation module output value weight w1=LOW and the deep learning prediction module output value weight w2=HIGH are set.

[0108] (2) When the input variables after fuzzification are low ambient temperature (LT), steep road slope (ST), medium vehicle speed (MS), and medium battery SOC (MED), due to the complexity of the environment and battery status, the weights are biased towards the deep learning model. Therefore, the output variable average energy consumption estimation module output value weight w1=LOW, and the deep learning prediction module output value weight w2=HIGH are set.

[0109] (3) When the input variables after fuzzification are medium ambient temperature (MD), flat road slope (FL), medium vehicle speed (MS), and high battery SOC (HIGH), since the environment and battery status are not complicated, the weights tend to be in the average energy consumption model. Therefore, the output variable average energy consumption estimation module output value weight w1 = HIGH, and the deep learning prediction module output value weight w2 = LOW.

[0110] (4) When the input variables after fuzzification are medium ambient temperature (MD), medium road slope (MD), medium vehicle speed (MS), and medium battery SOC (MED), since the environment and battery status are medium, the weights are not biased. Therefore, the output variable average energy consumption estimation module output value weight w1 = MED, and the deep learning prediction module output value weight w2 = MED.

[0111] (5) When the input variables after fuzzification are high ambient temperature (HT), flat road slope (FL), high vehicle speed (HS), and high battery SOC (HIGH), since the environment and battery status are not complicated, the weights tend to be in the average energy consumption model. Therefore, the output variable average energy consumption estimation module output value weight w1 = HIGH, and the deep learning prediction module output value weight w2 = LOW.

[0112] (6) When the input variables after fuzzification are low ambient temperature (LT), steep road slope (ST), low vehicle speed (LS), and low battery SOC (LOW), due to the complexity of the environment and battery status, the weights are biased towards the deep learning model. Therefore, the output variable average energy consumption estimation module output value weight w1 = LOW, and the deep learning prediction module output value weight w2 = HIGH.

[0113] (7) When the input variables after fuzzification are medium ambient temperature (MD), medium road slope (MD), high vehicle speed (HS), and low battery SOC (LOW), due to the complexity of the environment and battery status, the weights are biased towards the deep learning model. Therefore, the output variable average energy consumption estimation module output value weight w1 = LOW, and the deep learning prediction module output value weight w2 = HIGH.

[0114] (8) When the input variables after fuzzification are high ambient temperature (HT), steep road slope (ST), high vehicle speed (HS), and medium battery SOC (MED), due to the complexity of the environment and battery status, the weights are biased towards the deep learning model. Therefore, the output variable average energy consumption estimation module output value weight w1 = LOW, and the deep learning prediction module output value weight w2 = HIGH.

[0115] (9) When the input variables after fuzzification are low ambient temperature (LT), medium road slope (MD), low vehicle speed (LS), and medium battery SOC (MED), since the environment and battery status are medium, the weights are not biased. Therefore, the output variable average energy consumption estimation module output value weight w1 = MED, and the deep learning prediction module output value weight w2 = MED.

[0116] (10) When the input variables after fuzzification are medium ambient temperature (MD), flat road slope (FL), low vehicle speed (LS), and high battery SOC (HIGH), since the environment and battery status are not complicated, the weights tend to be in the average energy consumption model. Therefore, the output variable average energy consumption estimation module output value weight w1 = HIGH, and the deep learning prediction module output value weight w2 = LOW.

[0117] Step S608, the deviation between the energy consumption change in the target cruising range prediction result and the energy consumption change in the actual cruising range information is analyzed and processed in real time, the weight coefficient is updated according to the deviation analysis result, and the target weight coefficient is determined, so that the data fusion layer predicts the cruising range at the next moment based on the target weight coefficient. In one embodiment, the data stream can be shared with the existing average energy consumption estimation module and deep learning prediction module by acquiring real-time operation data. After each trip, the system automatically evaluates the accuracy of the predicted value. When the deviation exceeds the set threshold, the self-learning module is triggered to optimize the model. Regarding the allocation of dynamic weights, the optimization results of the self-learning module are dynamically updated in real time through the fuzzy logic system to ensure that the fusion model can provide the best prediction under the new working conditions.

[0118] Furthermore, the operation process of the self-learning module is as follows: first, the data acquisition module obtains the vehicle operation data, and then passes the data to the deep learning prediction module, the average energy consumption estimation module and the self-learning module. The deep learning prediction module and the average energy consumption estimation module output preliminary prediction values ​​according to the existing algorithms. Then the self-learning module performs deviation analysis and model optimization on the input data, and feeds back the optimized parameters to the deep learning module and the data fusion module. Finally, the data fusion module dynamically adjusts the weight value and outputs the final endurance prediction value.

[0119] In summary, the present invention combines traditional average energy consumption estimation with historical data analysis based on deep learning to make more accurate predictions of cruising range in various complex environments and driving conditions. Specifically, the present invention uses a long short-term memory network (LSTM) to capture the energy consumption trend in the time series, combines the random forest algorithm for feature importance analysis, and performs weighted fusion of the results of different prediction models through fuzzy logic. Finally, the self-learning layer is used to improve the robustness and prediction accuracy of the system in diversified scenarios through continuous model optimization and data feedback mechanism.

[0120] By predicting the mileage through the above-mentioned fusion method, pure electric commercial vehicles can provide users with reliable mileage information under various driving conditions, reduce mileage anxiety, and further improve users' acceptance and satisfaction with pure electric commercial vehicles. This innovative solution can not only provide real-time mileage prediction when the vehicle is driving normally, but also continuously optimize the model through deep learning, so that its prediction accuracy gradually improves over time.

[0121] An embodiment of the present invention provides a server. Specifically, the server includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above-mentioned embodiments.

[0122] Figure 7 A structural diagram of a server provided in an embodiment of the present invention, the server 100 includes: a processor 70, a memory 71, a bus 72 and a communication interface 73, wherein the processor 70, the communication interface 73 and the memory 71 are connected via the bus 72; the processor 70 is used to execute an executable module stored in the memory 71, such as a computer program.

[0123] The memory 71 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 73 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0124] The bus 72 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0125] Among them, the memory 71 is used to store programs, and the processor 70 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 70 or implemented by the processor 70.

[0126] The processor 70 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 70. The above processor 70 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present invention can be directly embodied as a hardware decoding processor to execute, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 71, and the processor 70 reads the information in the memory 71 and completes the steps of the above method in combination with its hardware.

[0127] The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be referred to the previous method embodiments, which will not be repeated here.

[0128] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0129] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A range prediction system for a pure electric commercial vehicle, characterized in that: The system comprises: a data acquisition layer, a data processing layer, a data fusion layer and a self-learning layer, wherein the data processing layer is connected to the data acquisition layer and the data fusion layer respectively, and the self-learning layer is connected to the data fusion layer and the data acquisition layer respectively; wherein, The data collection layer is used to collect vehicle status information, external environment information and actual endurance information through a preset sensor set, wherein the vehicle status information includes: vehicle speed information, battery status information, vehicle acceleration and steering angle, throttle and brake pedal displacement, motor status and driving mode, and the external environment information includes: ambient temperature information and road condition information; The data processing layer is used to perform energy consumption estimation processing on the vehicle state information and the external environment information to determine an energy consumption estimation value and a first cruising range prediction result corresponding to the energy consumption estimation value, and perform deep learning prediction processing on the vehicle state information and the external environment information to determine a second cruising range prediction result, wherein the energy consumption estimation value includes energy consumption per unit time and energy consumption per unit mileage; The data fusion layer is used to perform fuzzy logic fusion processing on the vehicle state information and the external environment information, determine a weight coefficient of weighted average, and based on the weight coefficient, perform weighted average fusion processing on the first cruising range prediction result and the second cruising range prediction result to determine a target cruising range prediction result at a current moment; The self-learning layer is used to perform real-time analysis and processing on the deviation value between the energy consumption change in the target endurance prediction result and the energy consumption change in the actual endurance information, update the weight coefficient according to the deviation analysis result, and determine the target weight coefficient, so that the data fusion layer can make a range prediction at the next moment based on the target weight coefficient.

2. The range prediction system for pure electric commercial vehicles according to claim 1, characterized in that: The data collection layer also includes: a self-diagnosis unit; wherein, The self-diagnosis unit is used to perform abnormality detection processing on the collected vehicle status information and the external environment information, and to switch fault sensors and issue fault alarms when a sensor failure or abnormal data is detected.

3. The range prediction system for pure electric commercial vehicles according to claim 1, characterized in that: The data processing layer includes: an average energy consumption estimation module; wherein, The average energy consumption estimation module is used to calculate the remaining battery power, driving resistance loss power, motor efficiency loss power, battery consumption power and energy recovery power according to the vehicle status information and the external environment information, so as to determine the energy consumption per unit time according to the driving resistance loss power, the motor efficiency loss power, the battery consumption power and the energy recovery power, and determine the energy consumption per unit mileage according to the energy consumption per unit time.

4. The range prediction system for pure electric commercial vehicles according to claim 3, characterized in that: The data processing layer also includes: a deep learning prediction module, and the deep learning prediction module includes: a random forest model; wherein, The random forest model is used to analyze the driving situation and determine that the driver is a frequent vehicle driver, to predict the long-term energy consumption change trend of the cruising range and determine the second cruising range prediction result.

5. The range prediction system for pure electric commercial vehicles according to claim 4, characterized in that: The deep learning prediction module also includes: a long short-term memory network model; wherein, The long short-term memory network model is used to analyze the driving situation and determine that the driver is not an ordinary vehicle driver, to predict the short-term energy consumption change trend of the cruising range and determine the second cruising range prediction result.

6. The range prediction system for pure electric commercial vehicles according to claim 1, characterized in that: The data fusion layer includes: a fuzzy logic subsystem; wherein, The fuzzy logic subsystem is used to fuzzify the vehicle state information and the external environment information, and use a preset reasoning engine to reason on the fuzzified information to defuzzify the reasoning result and obtain the weight coefficient of the weighted average.

7. The range prediction system for a pure electric commercial vehicle according to claim 6, characterized in that: The data fusion layer also includes: a weighted average fusion unit; wherein, The weighted average fusion unit is used to determine the target cruising range prediction result at the current moment by summing the products of the first cruising range prediction result and the second cruising range prediction result with the corresponding weight coefficients based on the weight coefficient and the preset fusion model.

8. The range prediction system for pure electric commercial vehicles according to claim 1, characterized in that: The self-learning layer includes: a data feedback and storage module; wherein, The data feedback and storage module is used to collect the deviation value between the target endurance prediction result at each moment and the energy consumption change in the actual endurance information, as well as the vehicle status information and the external environment information corresponding to each moment, and construct a historical data set based on the deviation value, the vehicle status information and the external environment information at each moment.

9. The range prediction system for a pure electric commercial vehicle according to claim 8, characterized in that: The self-learning layer also includes: a model optimization module; wherein, The self-learning layer is used to update the weight coefficient based on the deviation value through an incremental learning model to determine the target weight coefficient.

10. A method for predicting the cruising range of a pure electric commercial vehicle, characterized in that: The method is applied to the range prediction system of a pure electric commercial vehicle according to any one of claims 1 to 9, and the method comprises: Collect vehicle status information, external environment information and actual endurance information, wherein the vehicle status information includes: vehicle speed information, battery status information, vehicle acceleration and steering angle, throttle and brake pedal displacement, motor status and driving mode, and the external environment information includes: ambient temperature information and road condition information; Performing energy consumption estimation processing on the vehicle state information and the external environment information to determine an energy consumption estimation value and a first cruising range prediction result corresponding to the energy consumption estimation value, and performing deep learning prediction processing on the vehicle state information and the external environment information to determine a second cruising range prediction result, wherein the energy consumption estimation value includes energy consumption per unit time and energy consumption per unit mileage; Performing fuzzy logic fusion processing on the vehicle state information and the external environment information to determine a weight coefficient of weighted average, and based on the weight coefficient, performing weighted average fusion processing on the first cruising range prediction result and the second cruising range prediction result to determine a target cruising range prediction result at the current moment; The deviation value between the energy consumption change in the target cruising range prediction result and the energy consumption change in the actual cruising range information is analyzed and processed in real time, the weight coefficient is updated according to the deviation analysis result, and the target weight coefficient is determined, so that the data fusion layer predicts the cruising range at the next moment based on the target weight coefficient.

11. A server, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the method of claim 10.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method of claim 10 .

Images