A new energy truck carbon data tampering detection and tracing method and system

Abstract

The application relates to the technical field of data analysis, and discloses a new energy truck carbon data tampering detection and tracing method and system. The method comprises the following steps: continuously acquiring vehicle operation data and carbon data, and arranging operation data sequences and carbon data sequences; performing boundary prediction on each carbon data in the carbon data sequences based on the operation data sequences, and identifying data abnormal sections; establishing a three-dimensional energy consumption tensor of the data abnormal sections, performing slice inspection on the three-dimensional energy consumption tensor, and identifying tampered data points; positioning a data tampering section based on the tampered data points, fitting a reference sequence section for the data tampering section based on the carbon data sequences; comparing the data tampering section with the reference sequence section, and identifying a data tampering mode. The application realizes high-precision identification, accurate positioning and interpretable tracing of carbon data abnormalities of new energy trucks, thereby improving the credibility and applicability of carbon data.

Description

Technical Field

[0001] This application relates to the technical field of data analysis, specifically to a method and system for detecting and tracing carbon data tampering in new energy freight vehicles. Background Technology

[0002] Under current technological conditions, carbon emission data for new energy trucks is typically not obtained through direct measurement, but rather through carbon emission equivalent data calculated based on vehicle powertrain operating parameters using preset models or algorithms. For example, carbon emission levels per unit time or per unit mileage are estimated based on parameters such as motor power, battery state of charge, mileage, or time, combined with fixed or semi-fixed emission factors. This indirect accounting method based on operational data has become the mainstream path for generating carbon data for new energy vehicles and is widely used in scenarios such as internal energy consumption management, regulatory statistics, and carbon emission accounting within enterprises. However, there are still shortcomings in the technology for regulating and verifying carbon data for new energy trucks. Existing carbon data management systems lack independent and systematic verification mechanisms for the inherent rationality of carbon data itself. When carbon data undergoes abnormal changes during collection, transmission, storage, or processing, existing systems often struggle to identify in a timely and accurate manner whether these changes are reasonable fluctuations or caused by human or systematic intervention. Current methods for analyzing carbon data anomalies mostly remain at the level of statistical threshold judgment or simple rule comparison. While these methods are effective in dealing with extreme anomalies, they often fall short when faced with distributed, small-amplitude, intermittent, or structural anomalies. This is especially true given the complex and variable operating conditions of new energy freight vehicles; relying on a single dimension or threshold is insufficient to accurately distinguish between normal operating condition changes and data anomalies caused by non-natural factors. Furthermore, when carbon data anomalies are detected, existing systems typically only provide a conclusion that the anomaly exists, failing to pinpoint its specific distribution characteristics across time or operating conditions, and even less to effectively analyze the underlying causes.

[0003] The information disclosed in this background section is intended only to enhance the understanding of the overall background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0004] The technical problem to be solved by this application is to overcome the defects of the prior art and provide a method and system for detecting and tracing carbon data tampering of new energy trucks, so as to achieve high-precision identification, accurate positioning and interpretable tracing of carbon data anomalies of new energy trucks, and improve the credibility and applicability of carbon data.

[0005] To solve the above-mentioned technical problems, this application provides the following technical solution:

[0006] On the one hand, this application provides a method for detecting and tracing carbon data tampering in new energy freight vehicles, including the following steps:

[0007] Continuously acquire vehicle operation data and carbon data, and organize the operation data sequence and carbon data sequence;

[0008] Based on the running data sequence, boundary prediction is performed for each carbon data in the carbon data sequence, and abnormal data segments are identified.

[0009] A three-dimensional energy consumption tensor is established for the abnormal data segment, and the three-dimensional energy consumption tensor is sliced ​​for inspection to identify tampered data points;

[0010] Based on the tampered data points, the tampered data segment is located, and a reference sequence segment is fitted to the tampered data segment based on the carbon data sequence;

[0011] Compare the data tampered section with the reference sequence section and identify the data tampering method.

[0012] As a preferred embodiment of the carbon data tampering detection and traceability method for new energy freight vehicles described in this application, the vehicle operation data includes power consumption data, driving status data, and operating condition data; wherein, the power consumption data includes at least the vehicle's traction power; the driving status data includes at least the vehicle speed; the operating condition data includes at least the vehicle load; and the carbon data is carbon emission equivalent.

[0013] The operational data sequence includes a time series of each vehicle operational data item; the carbon data sequence is a time series of carbon data.

[0014] The method for boundary prediction of carbon data at any given time in a carbon data sequence is as follows:

[0015] Extract each vehicle operation data point at the corresponding time based on the operation data sequence, and encode each vehicle operation data point at the corresponding time into a feature vector at the corresponding time.

[0016] The feature vector at the corresponding time point is input into the trained generative adversarial network, which performs boundary prediction and outputs the carbon emission sample set at the corresponding time point; the carbon emission sample set contains at least M carbon emission samples, where M is a positive integer; any carbon emission sample is a reference value for carbon data;

[0017] A reference interval for carbon data is generated based on the carbon emission sample set at the corresponding time point, which serves as the boundary prediction result.

[0018] As a preferred embodiment of the method for detecting and tracing carbon data tampering in new energy freight vehicles described in this application, the identification of abnormal data segments specifically includes:

[0019] If the carbon data at any point in the carbon data sequence is within the corresponding reference interval, then the carbon data at that point is normal; otherwise, the carbon data at that point is marked as abnormal carbon data.

[0020] The carbon data sequence is divided into continuous subsequences, and the time period covered by each subsequence is marked as a detection period. If the number of abnormal carbon data in any detection period is greater than the preset abnormal threshold, the corresponding detection period is a data abnormal segment.

[0021] As a preferred embodiment of the carbon data tampering detection and tracing method for new energy freight vehicles described in this application, the method for establishing the three-dimensional energy consumption tensor for any abnormal data segment is as follows:

[0022] Each core operational data point at each time point in the abnormal data segment is extracted based on the operational data sequence and the carbon data sequence; the core operational data includes traction power, vehicle speed, and vehicle load.

[0023] Each core operational data item is discretized to obtain the discrete interval of each core operational data item corresponding to the abnormal data segment. The discrete interval of each core operational data item is then numbered and recorded as the discrete value of the corresponding core operational data item.

[0024] The discrete values ​​of traction power, vehicle speed, and vehicle load are sorted from smallest to largest and used as indices for the three dimensions to establish a three-dimensional energy consumption tensor. Each element in the three-dimensional energy consumption tensor is then assigned a value.

[0025] The method for assigning values ​​to any element in the three-dimensional energy consumption tensor is as follows: determine the associated discrete interval of the element based on the element's index; the associated discrete interval includes the discrete intervals of traction power, vehicle speed, and vehicle load corresponding to the indices of the element in the three dimensions, respectively.

[0026] In the data anomaly segment, the moment when traction power, vehicle speed, and vehicle load all fall within the corresponding associated discrete interval is marked as the associated moment of the element; the mean of the carbon data at each associated moment is calculated as the element value of the element.

[0027] As a preferred embodiment of the method for detecting and tracing carbon data tampering in new energy freight vehicles described in this application, the method includes: performing a slice inspection on the three-dimensional energy consumption tensor to identify tampered data points, specifically including:

[0028] Power load slices, power speed slices, and speed load slices are extracted from the three-dimensional energy consumption tensor, respectively; wherein, the power load slice is a two-dimensional tensor obtained by fixing the index corresponding to the vehicle speed in the three-dimensional energy consumption tensor; the power speed slice is a two-dimensional tensor obtained by fixing the index corresponding to the vehicle load; and the speed load slice is a two-dimensional tensor obtained by fixing the index corresponding to the traction power.

[0029] Each power load slice, power velocity slice, and velocity load slice is inspected to identify anomalous elements in the three-dimensional energy consumption tensor; each associated time of each anomalous element is marked as a tampered data point.

[0030] The method for performing a slice check on any velocity load slice is as follows: For any specified target element, extract the neighboring elements of the target element on the velocity load slice as reference elements of the target element; if the absolute value of the difference between the mean value of each reference element and the element value of the target element is greater than a preset jump threshold, then the target element is an abnormal element in the three-dimensional energy consumption tensor.

[0031] As a preferred embodiment of the carbon data tampering detection and tracing method for new energy freight vehicles described in this application, the method for performing slice inspection on any power load slice is as follows:

[0032] Calculate the first local gradient of each element along the direction from smallest to largest index corresponding to traction power, and calculate the second local gradient of each element along the direction from smallest to largest index corresponding to vehicle load.

[0033] For any given target element, extract the first local gradients of a total of R elements before and after the target element in the corresponding index direction, and use them as reference values ​​for the first local gradient of the target element; if the absolute value of the difference between the mean of the R reference values ​​and the first local gradient of the target element is greater than a preset first gradient difference threshold, then the target element is an anomalous element in the three-dimensional energy consumption tensor; R is a positive integer.

[0034] For any given target element, extract the second local gradients of a total of N elements before and after the target element in the corresponding index direction, and use them as reference values ​​for the second local gradient of the target element; if the absolute value of the difference between the mean of the N reference values ​​and the second local gradient of the target element is greater than the preset second gradient difference threshold, then the target element is an anomalous element in the three-dimensional energy consumption tensor; N is a positive integer.

[0035] As a preferred embodiment of the carbon data tampering detection and tracing method for new energy freight vehicles described in this application, the method for performing slice inspection on any power-speed slice is as follows:

[0036] Calculate the first local slope of each element along the direction from smallest to largest index corresponding to traction power, and calculate the second local slope of each element along the direction from smallest to largest index corresponding to vehicle speed.

[0037] For any given target element, extract the first local slope of a total of P elements before and after the target element in the corresponding index direction, and use them as the first reference slope of the target element; if the signs of the P first reference slopes are opposite to the first local slope of the target element, then the target element is an anomalous element in the three-dimensional energy consumption tensor; P is a positive integer;

[0038] For any given target element, extract the second local slopes of a total of Q elements before and after the target element in the corresponding index direction, and use them as the second reference slopes of the target element; if the signs of the Q second reference slopes are opposite to the second local slopes of the target element, then the target element is an anomalous element in the three-dimensional energy consumption tensor; Q is a positive integer.

[0039] As a preferred embodiment of the method for detecting and tracing carbon data tampering in new energy freight vehicles described in this application, the method specifically includes: locating the tampered data segment based on the tampered data point.

[0040] Mark each tampered data point on the time axis; based on the tampered data points, divide the time axis into candidate tampered segments, specifically including: setting a gap threshold; if the time difference between any two tampered data points is less than the gap threshold, then divide the corresponding two tampered data points and the time period between them into a candidate tampered segment;

[0041] Calculate the coverage length and anomaly density of each candidate tampered segment; the coverage length of any candidate tampered segment is the duration of the time covered by the candidate tampered segment; the anomaly density of any candidate tampered segment is the ratio of the number of tampered data points in the candidate tampered segment to the total number of time points contained in the candidate tampered segment.

[0042] If the coverage length of any candidate tampered segment is greater than a preset length threshold and the abnormal density is greater than a preset abnormal threshold, then the corresponding candidate tampered segment is a data tampered segment.

[0043] As a preferred embodiment of the carbon data tampering detection and tracing method for new energy freight vehicles described in this application, the method includes: fitting a reference sequence segment based on the carbon data sequence as any data tampering segment, specifically including:

[0044] The sequence segment corresponding to the data tampering section in the carbon data sequence is marked as the tampered sequence segment; the sequence segments corresponding to the m consecutive time steps before the tampered sequence segment and the sequence segments corresponding to the n consecutive time steps after the tampered sequence segment are extracted from the carbon data sequence and used as the reference sequence segment of the tampered sequence segment; m and n are both positive integers;

[0045] Based on the carbon data at each time point in the reference sequence segment of the tampered sequence segment, a change curve of the carbon data is fitted; the carbon data corresponding to each time point in the tampered sequence segment is extracted from the change curve and a first reference sequence is formed; the mean of the carbon data samples corresponding to different times in the tampered sequence segment is calculated respectively and a second reference sequence is formed.

[0046] The first reference sequence and the second reference sequence are fused point by point with weights to obtain the reference sequence segment of the data tampering segment; wherein, the fusion weight at any time is calculated by the dispersion of the corresponding carbon data sample set;

[0047] Compare the data tampered section with the reference sequence section and identify the data tampering method; specifically including:

[0048] The tampered sequence segment corresponding to the data tampered segment and the reference sequence segment are input into the trained recognition model; the recognition model identifies and outputs the data tampering method of the data tampered segment.

[0049] Secondly, this application provides a system for detecting and tracing carbon data tampering in new energy freight vehicles, including a data acquisition module, an anomaly detection module, a modeling module, an anomaly identification module, a data reconstruction module, and a tampering tracing module; wherein:

[0050] The data acquisition module is used to continuously acquire vehicle operation data and carbon data, and to organize the operation data sequence and carbon data sequence;

[0051] The anomaly detection module performs boundary prediction for each carbon data in the carbon data sequence based on the running data sequence and identifies abnormal data segments;

[0052] The modeling module is used to establish the three-dimensional energy consumption tensor of the data anomaly segment;

[0053] The anomaly detection module is used to perform slice inspection on the three-dimensional energy consumption tensor, identify tampered data points, and locate the data tampering segment based on the tampered data points;

[0054] The data reconstruction module fits a reference sequence segment to the data tampering segment based on the carbon data sequence;

[0055] The tampering tracing module is used to compare the data tampering segment with the reference sequence segment and identify the data tampering method.

[0056] Compared with the prior art, the beneficial effects achieved by this application are as follows:

[0057] This application starts from the inherent constraint relationship between vehicle operating conditions and carbon emission results, and systematically describes the reasonable range and variation boundary of carbon data. This enables a more robust judgment on whether carbon data conforms to objective operating laws under complex operating conditions, and significantly reduces the risk of misjudgment and omission.

[0058] To address the problem that existing technologies struggle to detect small-scale, intermittent, or distributed anomalies, this application employs joint analysis of carbon data across time and operational conditions. This approach enables the identification of anomalous behaviors that are difficult to detect using single-point detection or simple statistical methods. It is particularly suitable for addressing anomalous data patterns that circumvent traditional detection rules, thereby improving the overall coverage of carbon data anomaly detection.

[0059] This application can not only determine whether carbon data is abnormal, but also pinpoint the abnormality to a specific time period and the corresponding tampering method, giving the abnormal results clear spatial and temporal orientation and providing a clear basis for subsequent verification, tracing and processing. Attached Figure Description

[0060] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0061] Figure 1 A flowchart of a method for detecting and tracing carbon data tampering in new energy freight vehicles provided in this application;

[0062] Figure 2 This is a schematic diagram of the structure of a new energy truck carbon data tampering detection and traceability system provided in this application. Detailed Implementation

[0063] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0064] Example 1

[0065] This embodiment introduces a method for detecting and tracing carbon data tampering in new energy freight vehicles, referring to... Figure 1 The method includes the following steps:

[0066] Continuously acquire vehicle operation data and carbon data, and organize the operation data sequence and carbon data sequence;

[0067] The vehicle operation data includes power consumption data, driving status data, and operating condition data; wherein, the power consumption data includes at least the vehicle's traction power; optionally, the power consumption data also includes the battery's state of charge, motor current, and motor voltage; the driving status data includes at least the vehicle speed; optionally, the driving status data also includes the vehicle acceleration; the operating condition data includes at least the vehicle load; in this embodiment, the vehicle load is equivalently represented by the motor's output torque or wheel-end traction force, used to describe the actual driving resistance level borne by the vehicle's power system; optionally, the operating condition data also includes road gradient.

[0068] The carbon data refers to carbon emission equivalents. These carbon emission equivalents are the equivalent carbon emissions of a vehicle. Vehicle terminals or interconnected platforms can convert powertrain data, such as real-time motor power, remaining battery power, and motor torque, into equivalent carbon emission values ​​for the vehicle using models or formulas, such as the mass of carbon dioxide emitted per meter or second of travel. In application scenarios such as real-time energy consumption monitoring and carbon accounting platforms for new energy trucks, carbon emission accounting platforms for regulatory or carbon trading purposes, and driving behavior analysis and anomaly detection systems, carbon data is susceptible to tampering.

[0069] The operational data sequence includes a time series of each vehicle operational data item; the carbon data sequence is a time series of carbon data; the operational data sequence and the carbon data sequence are aligned by timestamps so that the carbon data at any moment in the carbon data sequence corresponds to a set of vehicle operational data with the same timestamp.

[0070] Based on the running data sequence, boundary prediction is performed for each carbon data in the carbon data sequence, and abnormal data segments are identified.

[0071] The method for boundary prediction of carbon data at any given time in a carbon data sequence is as follows:

[0072] Extract each vehicle operation data point at the corresponding time based on the operation data sequence, and encode each vehicle operation data point at the corresponding time into a feature vector at the corresponding time.

[0073] The feature vector at the corresponding time point is input into the trained generative adversarial network, which performs boundary prediction and outputs the carbon emission sample set at the corresponding time point; the carbon emission sample set contains at least M carbon emission samples, where M is a positive integer; any carbon emission sample is a reference value for carbon data;

[0074] A reference interval for carbon data is generated based on the carbon emission sample set at the corresponding time point, which serves as the boundary prediction result.

[0075] Optionally, a method for generating a reference interval for carbon data based on a carbon emission sample set is as follows: Sort the carbon emission samples in the carbon emission sample set by size, use the maximum value among the carbon emission samples as the upper limit of the reference interval, and use the minimum value among the carbon emission samples as the lower limit of the reference interval. Alternatively, calculate the mean and variance of the carbon emission samples, use the mean as the midpoint, and set a symmetrical interval based on the variance as the reference interval for the carbon data.

[0076] In this embodiment, the generative adversarial network (GAN) includes a generator and a discriminator. The generator's input is a feature vector encoded from each vehicle operation data point at any given time, and its output is a carbon emission sample. The discriminator determines whether the generated carbon emission sample conforms to the distribution characteristics of historical normal operation data. Optionally, the generator's input also includes random noise to introduce distribution randomness. The generator and discriminator establish their ability to characterize the carbon data distribution through adversarial training. This GAN learns the distribution characteristics of carbon data under different operating conditions for new energy trucks during normal operation and generates a carbon emission sample set for the corresponding operating conditions. The generated samples are used to determine the reasonable range of carbon data values ​​under those operating conditions. During training, the generator continuously adjusts its output so that the carbon emission samples it generates under given conditions can be discriminated by the discriminator; the discriminator continuously updates its discrimination ability to distinguish between real samples and samples generated by the generator. Through the above adversarial training process, the GAN learns the distribution characteristics of carbon data under different operating conditions. After training, the network parameters are saved. During actual operation, only the generator is used to generate multiple carbon emission samples under given conditions, forming a carbon emission sample set for the corresponding time period. Optionally, the generator adopts a regression or generative neural network structure, including but not limited to multi-layer fully connected networks, recurrent neural networks based on time-series modeling, and time-series convolutional networks; the discriminator's network structure can adopt a multi-layer fully connected network. In a specific implementable example, the generator adopts a conditional generative neural network structure, whose input consists of a vehicle operation data feature vector and a random perturbation vector. The vehicle operation data feature vector can be set to 12 dimensions to represent continuous variables such as traction power, vehicle speed, vehicle load, SOC, motor current, motor voltage, and energy recovery status. The random perturbation vector can be set to 16 dimensions and follow a Gaussian distribution with zero mean and unit variance. The two are concatenated at the input end and then fed into the generator. The generator's main network employs a four-layer fully connected structure. The first hidden layer has 128 neurons, the second has 64, and the third has 32. Each hidden layer uses the LeakyReLU activation function with a leakage coefficient of 0.2 to enhance the representation of nonlinear operating conditions and avoid the gradient vanishing problem. The output layer has only one neuron and outputs a single carbon emission sample value. A linear activation function is used in the output layer to meet the requirements of continuous regression output. The discriminator uses a multi-layer fully connected network structure consistent with the generator. Its input is a 13-dimensional input vector formed by concatenating the feature vector of the same 12-dimensional vehicle operation data with the corresponding 1-dimensional carbon data sample. The discriminator has three hidden layers with 128, 64, and 32 neurons respectively. Each hidden layer also uses the LeakyReLU activation function. The output layer has only one neuron and uses the Sigmoid activation function to output the probability value that the sample is real carbon data.During training, both the generator and discriminator use the Adam optimizer for parameter updates, with a learning rate of 0.0001, first-order and second-order moment estimation parameters of 0.5 and 0.9 respectively, and a batch size of 256. The discriminator and generator alternately update the weight parameters at a 1:1 ratio. By minimizing the adversarial loss function, the generator gradually learns the true distribution characteristics of carbon data under given operating conditions, thereby providing stable and statistically consistent generated samples for the subsequent construction of a carbon data reference interval.

[0077] The identified abnormal data segments specifically include:

[0078] If the carbon data at any point in the carbon data sequence is within the corresponding reference interval, then the carbon data at that point is normal; otherwise, the carbon data at that point is marked as abnormal carbon data.

[0079] The carbon data sequence is divided into continuous subsequences, and the time period covered by each subsequence is marked as a detection period. If the number of abnormal carbon data points in any detection period exceeds a preset abnormality threshold, the corresponding detection period is considered an abnormal data segment. Those skilled in the art can set the length of the detection period and the specific size of the abnormality threshold based on actual needs. For example, with a detection period length of 10 minutes, a sampling period of 1 second, and an abnormality threshold of 30, if the number of abnormal carbon data points exceeds 30 within 10 minutes, that time period is judged as an abnormal data segment. By dividing the detection period, distributed tampering can be captured, preventing intermittent tampering methods from circumventing continuous detection methods.

[0080] A three-dimensional energy consumption tensor is established for the abnormal data segment, and the three-dimensional energy consumption tensor is sliced ​​for inspection to identify tampered data points;

[0081] The method for constructing the three-dimensional energy consumption tensor for any data anomaly segment is as follows:

[0082] Each core operational data point at each time point in the abnormal data segment is extracted based on the operational data sequence and the carbon data sequence; the core operational data includes traction power, vehicle speed, and vehicle load.

[0083] Each core operational data item is discretized to obtain the discrete interval of each core operational data item corresponding to the abnormal data segment. The discrete interval of each core operational data item is then numbered and recorded as the discrete value of the corresponding core operational data item.

[0084] For example, vehicle speed can be discretized as follows: [0,5) is divided into the first discrete interval (unit: km / h), numbered 1, meaning the discrete value of this interval is 1; [5,10) is divided into the second discrete interval, numbered 2, meaning the discrete value of this interval is 2; if the vehicle speed at any time is greater than or equal to 5 and less than 10 km / h, then the value is located in the discrete interval numbered 2, meaning the discrete value of the vehicle speed at that time is 2.

[0085] The discrete values ​​of traction power, vehicle speed, and vehicle load are sorted from smallest to largest and used as indices for the three dimensions to establish a three-dimensional energy consumption tensor. Each element in the three-dimensional energy consumption tensor is then assigned a value.

[0086] The method for assigning values ​​to any element in the three-dimensional energy consumption tensor is as follows: determine the associated discrete interval of the element based on the element's index; the associated discrete interval includes the discrete intervals of traction power, vehicle speed, and vehicle load corresponding to the indices of the element in the three dimensions, respectively.

[0087] In the data anomaly segment, the moment when traction power, vehicle speed, and vehicle load are all located in the corresponding associated discrete interval is marked as the associated moment of the element;

[0088] Calculate the mean of the carbon data at each associated time point, and use it as the element value.

[0089] In this embodiment, the three-dimensional energy consumption tensor is used to characterize the structural characteristics of carbon data under different combinations of power, load, and vehicle speed. For example, the element with index (1,2,3) in the three-dimensional energy consumption tensor has associated discrete intervals including the first discrete interval of traction power, the second discrete interval of vehicle speed, and the third discrete interval of vehicle load. If the traction power at a certain moment is located within its first discrete interval, that is, the discrete value of the traction power at that moment is 1, the discrete value of the vehicle speed at that moment is 2, and the discrete value of the vehicle load at that moment is 3, then that moment is an associated moment of the element with index (1,2,3).

[0090] The three-dimensional energy consumption tensor is sliced ​​for inspection to identify tampered data points, specifically including:

[0091] Power load slices, power speed slices, and speed load slices are extracted from the three-dimensional energy consumption tensor. The power load slice is a two-dimensional tensor obtained by fixing the index corresponding to vehicle speed within the three-dimensional energy consumption tensor. The power speed slice is a two-dimensional tensor obtained by fixing the index corresponding to vehicle load. The speed load slice is a two-dimensional tensor obtained by fixing the index corresponding to traction power. For example, fixing the index corresponding to vehicle speed to 1 yields a power load slice where the discrete value of vehicle speed at the associated time for all elements in this power load slice is 1, meaning the vehicle speed is less than 5 km / h. Fixing the index corresponding to vehicle speed to 2 yields another power load slice where the discrete value of vehicle speed at the associated time for all elements in this power load slice is 2, meaning the vehicle speed is greater than or equal to 5 km / h and less than 10 km / h.

[0092] Each power load slice, power velocity slice, and velocity load slice is inspected to identify anomalous elements in the three-dimensional energy consumption tensor; each associated time of each anomalous element is marked as a tampered data point.

[0093] The method for performing a slice check on any power load slice is as follows:

[0094] Calculate the first local gradient of each element along the direction from smallest to largest index corresponding to traction power, and calculate the second local gradient of each element along the direction from smallest to largest index corresponding to vehicle load.

[0095] For any given target element, extract the first local gradients of a total of R elements before and after the target element in the corresponding index direction, and use them as reference values ​​for the first local gradient of the target element; if the absolute value of the difference between the mean of the R reference values ​​and the first local gradient of the target element is greater than a preset first gradient difference threshold, then the target element is an anomalous element in the three-dimensional energy consumption tensor; R is a positive integer.

[0096] For any given target element, extract the second local gradients of a total of N elements before and after the target element in the corresponding index direction, and use them as reference values ​​for the second local gradient of the target element; if the absolute value of the difference between the mean of the N reference values ​​and the second local gradient of the target element is greater than the preset second gradient difference threshold, then the target element is an anomalous element in the three-dimensional energy consumption tensor; N is a positive integer.

[0097] Those skilled in the art can set specific values ​​for the first gradient difference threshold and the second gradient difference threshold based on actual needs. For example, the first gradient difference threshold is 0.08 gCO2 / s, and the second gradient difference threshold is 0.12 gCO2 / s. In this embodiment, the first local gradient or the second local gradient of any element is the difference between its element value and the element value of its adjacent preceding element in the corresponding index direction, i.e., the difference in their carbon data. In a power load slice at a fixed vehicle speed, carbon emissions are driven by both power and load, and under the same vehicle speed conditions, they should show a continuous and smooth trend. If the local gradient of a certain element relative to its adjacent elements changes abruptly, it indicates that the carbon data structure under this operating condition combination has been destroyed and is at risk of being tampered with.

[0098] The method for performing a slice check on any power-velocity slice is as follows:

[0099] Calculate the first local slope of each element along the direction from smallest to largest index corresponding to traction power, and calculate the second local slope of each element along the direction from smallest to largest index corresponding to vehicle speed.

[0100] For any given target element, extract the first local slope of a total of P elements before and after the target element in the corresponding index direction, and use them as the first reference slope of the target element; if the signs of the P first reference slopes are opposite to the first local slope of the target element, then the target element is an anomalous element in the three-dimensional energy consumption tensor; P is a positive integer;

[0101] For any given target element, extract the second local slopes of a total of Q elements before and after the target element in the corresponding index direction, and use them as the second reference slopes of the target element; if the signs of the Q second reference slopes are opposite to the second local slopes of the target element, then the target element is an anomalous element in the three-dimensional energy consumption tensor; Q is a positive integer.

[0102] In this embodiment, the first or second local slope of any element is the difference between its element value and the element value of its adjacent element in the corresponding index direction, divided by its element value. In a speed-power slice under a fixed load, carbon emissions typically exhibit a stable, monotonic trend with changes in vehicle speed or power. By observing the local slope of an element, it can be determined whether this monotonic relationship has been broken, thereby identifying anomalous carbon data structures under specific load conditions.

[0103] The method for performing a slice check on any speed load slice is as follows:

[0104] For any given target element, the neighboring elements of the target element are extracted on the velocity load slice as reference elements of the target element; if the absolute value of the difference between the mean of the element values ​​of each reference element and the element value of the target element is greater than a preset jump threshold, then the target element is an anomalous element in the three-dimensional energy consumption tensor.

[0105] Those skilled in the art can set the specific value of the jump threshold and the size of the neighborhood extracted in the speed-load slice based on actual needs. For example, elements in the 8-neighborhood or 24-neighborhood of the target element can be extracted as its neighborhood elements, with a jump threshold of 0.30 gCO2 / s, or 25% of the average value of the neighborhood elements. In the speed-load slice with fixed power, the traction power, as the main energy consumption constraint, has been fixed, and the carbon data should form a smooth numerical distribution in the speed and load dimensions. If the value of a certain element deviates significantly from the overall level of its neighborhood, i.e., a local abnormal depression or protrusion occurs, it indicates that there is a risk of abnormal tampering with the carbon data under this operating condition.

[0106] Based on the tampered data points, the tampered data segment is located, and a reference sequence segment is fitted to the tampered data segment based on the carbon data sequence;

[0107] Locating the data tampering segment based on the tampered data points specifically includes:

[0108] Mark each tampered data point on the time axis; based on the tampered data points, divide the time axis into candidate tampered segments, specifically including: setting a gap threshold; if the time difference between any two tampered data points is less than the gap threshold, then divide the corresponding two tampered data points and the time period between them into a candidate tampered segment;

[0109] Calculate the coverage length and anomaly density of each candidate tampered segment; the coverage length of any candidate tampered segment is the duration of the time covered by the candidate tampered segment; the anomaly density of any candidate tampered segment is the ratio of the number of tampered data points in the candidate tampered segment to the total number of time points contained in the candidate tampered segment.

[0110] If the coverage length of any candidate tampered segment exceeds a preset length threshold and the anomaly density exceeds a preset anomaly threshold, then the corresponding candidate tampered segment is considered a data tampered segment. Those skilled in the art can set specific values ​​for the gap threshold, length threshold, and anomaly threshold based on actual needs. For example, if the sampling period is 1 second, then the gap threshold can be 120 seconds, the length threshold can be 600 seconds, and the anomaly threshold can be 0.2.

[0111] Based on carbon data sequences, a reference sequence segment is fitted for any data tampering segment, specifically including:

[0112] The sequence segment corresponding to the data tampering section in the carbon data sequence is marked as the tampered sequence segment;

[0113] Extract the sequence segments corresponding to m consecutive moments before the tampered sequence segment and the sequence segments corresponding to n consecutive moments after the tampered sequence segment from the carbon data sequence, and use them as reference sequence segments for the tampered sequence segment; m and n are both positive integers;

[0114] Based on the carbon data at each moment in the reference sequence segment of the tampered sequence segment, a change curve of the carbon data is fitted; the carbon data corresponding to each moment in the tampered sequence segment is extracted from the change curve and a first reference sequence is formed; optionally, the change curve of the carbon data is fitted by a nonlinear regression model such as gradient boosting tree or multilayer perceptron, thereby characterizing the change law of carbon data over time.

[0115] The mean values ​​of carbon data samples corresponding to different times in the tampered sequence segment are calculated respectively, and a second reference sequence is formed. In this embodiment, the elements in the first reference sequence and the second reference sequence are one-to-one with the tampered sequence segment in time. For any carbon data in the tampered sequence segment, there is a fitted value in the first reference sequence that is the same as its time, and there is a mean value of carbon data samples in the second reference sequence that is the same as its time.

[0116] The first reference sequence and the second reference sequence are fused point by point with weights to obtain the reference sequence segment of the data tampering segment; wherein, the fusion weight at any time is calculated by the dispersion of the corresponding carbon data sample set.

[0117] In this embodiment, the elements in the reference sequence segment and the tampered sequence segment correspond one-to-one; the element value of any element in the reference sequence segment is the weighted sum of the elements at the corresponding positions in the first reference sequence and the second reference sequence, and the sum of their fusion weights is 1; optionally, the dispersion of the carbon data sample set is the standard deviation of the carbon data samples, and the fusion weight of the element at the corresponding position in the second reference sequence is assigned based on the dispersion at each time point, and the larger the dispersion, the smaller the fusion weight, that is, the fusion weight of the element at the corresponding position in the first reference sequence is relatively larger.

[0118] Compare the data tampered section with the reference sequence section and identify the data tampering method; specifically including:

[0119] The tampered sequence segment corresponding to the data tampered segment and the reference sequence segment are input into the trained recognition model; the recognition model identifies and outputs the data tampering method of the data tampered segment.

[0120] By comparing the reported tampered sequence segments with reference sequence segments inferred from reliable data, data tampering methods can be identified. For example, if the tampered sequence segment is smoother than the reference sequence segment, it indicates interpolation smoothing tampering; if the tampered sequence segment shows a fixed delay increase or decrease compared to the reference sequence segment, it indicates backend rewriting; if the reference sequence segment is smooth while the tampered sequence segment has jagged edges, it indicates batch noise injection tampering. Using classification models, such as decision trees, random forests, and lightweight neural networks, the mapping relationship between the differences between the tampered and reference sequence segments and the tampering methods can be learned. These judgment rules can be implicitly embedded into the model's decision boundary, thereby achieving fast and accurate identification of data tampering methods.

[0121] Example 2

[0122] This embodiment is the second embodiment of this application; it is based on the same inventive concept as Embodiment 1, and refers to... Figure 2 This embodiment introduces a carbon data tampering detection and traceability system for new energy freight vehicles, including a data acquisition module, an anomaly detection module, a modeling module, an anomaly identification module, a data reconstruction module, and a tampering traceability module; wherein:

[0123] The data acquisition module is used to continuously acquire vehicle operation data and carbon data, and organize the operation data sequence and carbon data sequence. This module aligns the operation data sequence and carbon data sequence based on timestamps to form a one-to-one time series data pair, providing a basis for subsequent consistency analysis.

[0124] The anomaly detection module performs boundary prediction for each carbon data in the carbon data sequence based on the operational data sequence and identifies abnormal data segments; based on the vehicle operating conditions, the module performs reasonable boundary prediction for the carbon data and identifies concentrated abnormal segments in the time dimension.

[0125] The modeling module is used to establish the three-dimensional energy consumption tensor of the data anomaly segment; this module can construct a three-dimensional carbon data distribution structure to characterize the carbon emission structure characteristics under different operating conditions.

[0126] The anomaly detection module is used to perform slice inspection on the three-dimensional energy consumption tensor, identify tampered data points, and locate the data tampering segment based on the tampered data points;

[0127] The data reconstruction module fits a reference sequence segment to the data tampering section based on the carbon data sequence. This module identifies abnormal operating points that disrupt the continuity of the carbon data structure through multi-directional slicing analysis of the three-dimensional energy consumption tensor, and generates a reliable carbon data reference sequence segment for the tampered section.

[0128] The data tampering tracing module is used to compare the tampered data segment with the reference sequence segment and identify the data tampering method. This module compares the differences between the tampered sequence segment and the reference sequence segment in terms of smoothness, latency, and fluctuation structure, and outputs the corresponding data tampering method category, such as interpolation smoothing, backend rewriting, and noise injection; it can provide an interpretable chain of tampering evidence for regulatory, auditing, or carbon trading scenarios.

[0129] The specific functions of each module described above are implemented with reference to the relevant content in the method for detecting and tracing carbon data tampering of new energy freight vehicles described in Example 1, and will not be repeated here.

[0130] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0131] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of protection of this application, and these forms are all within the protection scope of this application.

Claims

1. A method for detecting and tracing carbon data tampering in new energy freight vehicles, characterized in that: Includes the following steps: Continuously acquire vehicle operation data and carbon data, and organize the operation data sequence and carbon data sequence; Based on the running data sequence, boundary prediction is performed for each carbon data in the carbon data sequence, and abnormal data segments are identified. A three-dimensional energy consumption tensor is established for the abnormal data segment, and the three-dimensional energy consumption tensor is sliced ​​for inspection to identify tampered data points; The method for constructing the three-dimensional energy consumption tensor for any data anomaly segment is as follows: Each core operational data point at each time point in the abnormal data segment is extracted based on the operational data sequence and the carbon data sequence; the core operational data includes traction power, vehicle speed, and vehicle load. Each core operational data item is discretized to obtain the discrete interval of each core operational data item corresponding to the abnormal data segment. The discrete interval of each core operational data item is then numbered and recorded as the discrete value of the corresponding core operational data item. The discrete values ​​of traction power, vehicle speed, and vehicle load are sorted from smallest to largest and used as indices for the three dimensions to establish a three-dimensional energy consumption tensor. Each element in the three-dimensional energy consumption tensor is then assigned a value. The method for assigning values ​​to any element in the three-dimensional energy consumption tensor is as follows: determine the associated discrete interval of the element based on the element's index; the associated discrete interval includes the discrete intervals of traction power, vehicle speed, and vehicle load corresponding to the indices of the element in the three dimensions, respectively. In the data anomaly segment, the moment when traction power, vehicle speed, and vehicle load all fall within the corresponding associated discrete interval is marked as the associated moment of the element; the mean of the carbon data at each associated moment is calculated as the element value of the element. The three-dimensional energy consumption tensor is sliced ​​for inspection to identify tampered data points, specifically including: Power load slices, power speed slices, and speed load slices are extracted from the three-dimensional energy consumption tensor, respectively; wherein, the power load slice is a two-dimensional tensor obtained by fixing the index corresponding to the vehicle speed in the three-dimensional energy consumption tensor; the power speed slice is a two-dimensional tensor obtained by fixing the index corresponding to the vehicle load; and the speed load slice is a two-dimensional tensor obtained by fixing the index corresponding to the traction power. Each power load slice, power velocity slice, and velocity load slice is inspected to identify anomalous elements in the three-dimensional energy consumption tensor; each associated time of each anomalous element is marked as a tampered data point. The method for performing a slice check on any velocity load slice is as follows: For any specified target element, extract the neighboring elements of the target element on the velocity load slice as reference elements of the target element; if the absolute value of the difference between the mean value of each reference element and the element value of the target element is greater than a preset jump threshold, then the target element is an abnormal element in the three-dimensional energy consumption tensor. Based on the tampered data points, the tampered data segment is located, and a reference sequence segment is fitted to the tampered data segment based on the carbon data sequence; Compare the data tampered section with the reference sequence section and identify the data tampering method.

2. The method for detecting and tracing carbon data tampering in new energy freight vehicles as described in claim 1, characterized in that: The vehicle operation data includes power consumption data, driving status data, and operating condition data; wherein, the power consumption data includes at least the vehicle's traction power; the driving status data includes at least the vehicle speed; the operating condition data includes at least the vehicle load; and the carbon data is carbon emission equivalent. The operational data sequence includes a time series of each vehicle operational data item; the carbon data sequence is a time series of carbon data. The method for boundary prediction of carbon data at any given time in a carbon data sequence is as follows: Extract each vehicle operation data point at the corresponding time based on the operation data sequence, and encode each vehicle operation data point at the corresponding time into a feature vector at the corresponding time. The feature vector at the corresponding time point is input into the trained generative adversarial network, which performs boundary prediction and outputs the carbon emission sample set at the corresponding time point; the carbon emission sample set contains at least M carbon emission samples, where M is a positive integer; any carbon emission sample is a reference value for carbon data; A reference interval for carbon data is generated based on the carbon emission sample set at the corresponding time point, which serves as the boundary prediction result.

3. The method for detecting and tracing carbon data tampering in new energy freight vehicles as described in claim 2, characterized in that: The identified abnormal data segments specifically include: If the carbon data at any point in the carbon data sequence is within the corresponding reference interval, then the carbon data at that point is normal; otherwise, the carbon data at that point is marked as abnormal carbon data. The carbon data sequence is divided into continuous subsequences, and the time period covered by each subsequence is marked as a detection period. If the number of abnormal carbon data in any detection period is greater than the preset abnormal threshold, the corresponding detection period is a data abnormal segment.

4. The method for detecting and tracing carbon data tampering in new energy freight vehicles as described in claim 3, characterized in that: The method for performing a slice check on any power load slice is as follows: Calculate the first local gradient of each element along the direction from smallest to largest index corresponding to traction power, and calculate the second local gradient of each element along the direction from smallest to largest index corresponding to vehicle load. For any given target element, extract the first local gradients of a total of R elements before and after the target element in the corresponding index direction, and use them as reference values ​​for the first local gradient of the target element; if the absolute value of the difference between the mean of the R reference values ​​and the first local gradient of the target element is greater than a preset first gradient difference threshold, then the target element is an anomalous element in the three-dimensional energy consumption tensor; R is a positive integer. For any given target element, extract the second local gradients of a total of N elements before and after the target element in the corresponding index direction, and use them as reference values ​​for the second local gradient of the target element; if the absolute value of the difference between the mean of the N reference values ​​and the second local gradient of the target element is greater than the preset second gradient difference threshold, then the target element is an anomalous element in the three-dimensional energy consumption tensor; N is a positive integer.

5. The method for detecting and tracing carbon data tampering in new energy freight vehicles as described in claim 4, characterized in that: The method for performing a slice check on any power-velocity slice is as follows: Calculate the first local slope of each element along the direction from smallest to largest index corresponding to traction power, and calculate the second local slope of each element along the direction from smallest to largest index corresponding to vehicle speed. For any given target element, extract the first local slope of a total of P elements before and after the target element in the corresponding index direction, and use them as the first reference slope of the target element; if the signs of the P first reference slopes are opposite to the first local slope of the target element, then the target element is an anomalous element in the three-dimensional energy consumption tensor; P is a positive integer; For any given target element, extract the second local slopes of a total of Q elements before and after the target element in the corresponding index direction, and use them as the second reference slopes of the target element; if the signs of the Q second reference slopes are opposite to the second local slopes of the target element, then the target element is an anomalous element in the three-dimensional energy consumption tensor; Q is a positive integer.

6. The method for detecting and tracing carbon data tampering in new energy freight vehicles as described in claim 5, characterized in that: Locating the data tampering segment based on the tampered data points specifically includes: Mark each tampered data point on the time axis; based on the tampered data points, divide the time axis into candidate tampered segments, specifically including: setting a gap threshold; if the time difference between any two tampered data points is less than the gap threshold, then divide the corresponding two tampered data points and the time period between them into a candidate tampered segment; Calculate the coverage length and anomaly density of each candidate tampered segment; the coverage length of any candidate tampered segment is the duration of the time covered by the candidate tampered segment; the anomaly density of any candidate tampered segment is the ratio of the number of tampered data points in the candidate tampered segment to the total number of time points contained in the candidate tampered segment. If the coverage length of any candidate tampered segment is greater than a preset length threshold and the abnormal density is greater than a preset abnormal threshold, then the corresponding candidate tampered segment is a data tampered segment.

7. The method for detecting and tracing carbon data tampering in new energy freight vehicles as described in claim 6, characterized in that: Based on carbon data sequences, a reference sequence segment is fitted for any data tampering segment, specifically including: The sequence segment corresponding to the data tampering section in the carbon data sequence is marked as the tampered sequence segment; the sequence segments corresponding to the m consecutive time steps before the tampered sequence segment and the sequence segments corresponding to the n consecutive time steps after the tampered sequence segment are extracted from the carbon data sequence and used as the reference sequence segment of the tampered sequence segment; m and n are both positive integers; Based on the carbon data at each time point in the reference sequence segment of the tampered sequence segment, a change curve of the carbon data is fitted; the carbon data corresponding to each time point in the tampered sequence segment is extracted from the change curve and a first reference sequence is formed; the mean of the carbon data samples corresponding to different times in the tampered sequence segment is calculated respectively and a second reference sequence is formed. The first reference sequence and the second reference sequence are fused point by point with weights to obtain the reference sequence segment of the data tampering segment; wherein, the fusion weight at any time is calculated by the dispersion of the corresponding carbon data sample set; Compare the data tampered section with the reference sequence section and identify the data tampering method; specifically including: The tampered sequence segment corresponding to the data tampered segment and the reference sequence segment are input into the trained recognition model; the recognition model identifies and outputs the data tampering method of the data tampered segment.

8. A system for detecting and tracing carbon data tampering in new energy freight vehicles, used to implement the method for detecting and tracing carbon data tampering in new energy freight vehicles as described in any one of claims 1-7, characterized in that: It includes a data acquisition module, an anomaly detection module, a modeling module, an anomaly identification module, a data reconstruction module, and a tampering tracing module; among which: The data acquisition module is used to continuously acquire vehicle operation data and carbon data, and to organize the operation data sequence and carbon data sequence; The anomaly detection module performs boundary prediction for each carbon data in the carbon data sequence based on the running data sequence and identifies abnormal data segments; The modeling module is used to establish the three-dimensional energy consumption tensor of the data anomaly segment; The anomaly detection module is used to perform slice inspection on the three-dimensional energy consumption tensor, identify tampered data points, and locate the data tampering segment based on the tampered data points; The data reconstruction module fits a reference sequence segment to the data tampering segment based on the carbon data sequence; The tampering tracing module is used to compare the data tampering segment with the reference sequence segment and identify the data tampering method.

Images