Method for load detection and allocation in an electrical vehicle infrastructure, computer program product and corresponding device

Abstract

A method for load detection and allocation in an electrical vehicle infrastructure is described, wherein more than one of these consumers is protected by an electronic fuse, wherein a current consumption state of at least one consumer protected by the electronic fuse is changed via the electronic fuse, wherein a change in the load state is recorded, wherein signals relating to at least one change in a vehicle function are acquired via a data bus of the vehicle, wherein at least one environmental metadata is acquired, wherein the change in the load state of the electronic fuse, the change in the vehicle function and the at least one acquired environmental metadata are time-synchronized, wherein the change in the load state is correlated with the change in the vehicle function and stored in a data record.

Description

[0001] A method for load detection and allocation in an electrical vehicle infrastructure, a computer program product, and a device for load detection and allocation in an electrical vehicle infrastructure are described.

[0002] Methods for load detection and allocation in an electrical vehicle infrastructure, computer program products and corresponding devices of the type mentioned above are known in the prior art.

[0003] Modern motor vehicles are characterized by an increasingly complex electrical system architecture, in which a multitude of different electrical consumers are operated via zonally distributed circuits. Power is typically supplied via electronic fuses (so-called eFuses) that protect multiple electrical loads simultaneously. While these electronic fuses provide extensive measurement data such as current, voltage, and sometimes other data like temperature, the necessary resolution to precisely attribute the energy consumption of individual components or functions is currently lacking.

[0004] This is particularly problematic when multiple devices are simultaneously powered by the same fuse, as in such cases no clear correlation can be established between activity at the functional or component level and specific electrical behavior. This makes the dimensioning of corresponding electrical systems difficult.

[0005] Furthermore, the activation and load behavior of individual components is highly dependent on dynamic system states, such as the activated driving mode (e.g., ECO mode, activation of driver assistance systems or climate control systems) and external conditions like low ambient temperature or a low vehicle battery charge. These factors can lead to variable inrush currents and unpredictable activation patterns, which must also be considered in the analysis.

[0006] Due to this lack of granularity, it is currently only possible to a limited extent to analyze the real behavior of customer-relevant functions and control units, to optimize the power grid design in the vehicle using simulation, or to identify deviations in energy consumption at the component level.

[0007] Existing tools for analyzing such relationships are often based on manual evaluation, static rule sets, or an isolated consideration of CAN signals and are therefore unable to reliably link electrical signatures with the actual, context-dependent use of vehicle functions.

[0008] In the automotive industry, a variety of established tools are used for the analysis and development of ECU communication and for simulating internal vehicle processes. For example, tools like CANoe are routinely used to simulate CAN-based communication scenarios at the signal level and to emulate or validate ECU behavior. Furthermore, XCP-based data loggers, capable of providing high-resolution measurement data, are used for the detailed acquisition of system states and internal ECU parameters. However, the evaluation of this data is generally manual and involves considerable interpretation effort, especially when relationships between electrical measurements and the vehicle's functional states need to be established.Furthermore, common methods for simulating electrical loads in vehicle electrical systems are often based on synthetically generated load profiles that are not derived from real field measurements. This results in a discrepancy between model-based predictions and actual behavior during vehicle operation, which complicates the validation and optimization of energy distribution and protection.

[0009] From US Patent 2008 / 320349 A1, an eFuse device and a method for verifying data alignment are known. Alignment bars are provided in a series of locking units of a write scan chain, and a logic unit is coupled to the alignment bars. A sequence of data scanned into the series of locking units of the write scan chain preferably includes alignment data values. These alignment data values ​​are placed at positions within the data sequence that cause the data values ​​to be stored in the alignment bars when the data sequence is correctly scanned into the series of locking units. The logic unit receives data signals from the alignment bars and determines whether the correct pattern of data values ​​is stored in the alignment bars.If the correct pattern of data values ​​is present in the alignment bars, the data is aligned and a program release signal is sent to the bank of electronic fuses.

[0010] The task therefore arises to further develop methods for load detection and allocation in an electric vehicle infrastructure, computer program products and corresponding devices of the type mentioned above in such a way that a dynamic linking of the current vehicle behavior with functional identifiers is made possible.

[0011] The problem is solved by a method for load detection and allocation in an electric vehicle infrastructure according to claim 1, a computer program product according to dependent claim 8, and a device for load detection and allocation in an electric vehicle infrastructure according to dependent claim 9. Further embodiments and developments are the subject of the dependent claims.

[0012] A method for load detection and allocation in an electrical vehicle infrastructure is described, comprising a plurality of consumers, wherein more than one of these consumers is protected by an electronic fuse, wherein the electronic fuse is configured to detect a current load state of the electrical fuse, wherein a current consumption state of at least one consumer protected by the electronic fuse is changed via the electronic fuse, wherein a change in the load state is recorded, wherein signals relating to at least one change in a vehicle function are detected via a data bus of the vehicle, wherein at least one environmental metadata is also detected, wherein the change in the load state of the electronic fuse, the change in the vehicle function and the at least one detected environmental metadata are time-synchronized.where the change in the load state is correlated with the change in the vehicle function and stored in a data set.

[0013] Machine learning plays a crucial role in automotive development, particularly through the creation of digital twins and the validation of models. A fundamental requirement for accurate machine learning algorithms is high-quality training data. Such training data is generated using the presented method, which performs a real-time correlation of communication signals and telemetry measurement data. Furthermore, the method enables context-dependent and environment-adaptive modeling of electrical loads, significantly improving both fuse diagnostics and the preparation and execution of load simulations under realistic conditions. The described method offers advancements over existing approaches, including automation and contextualization through automated load tagging and context-aware analysis.The method enables the differentiation of loads for high-resolution simulations as well as load-dependent diagnostics, for safety optimization and energy modeling. Furthermore, the method allows for adaptation to real driving conditions, the structuring of data sets, and the identification, grouping, and classification of unique scenarios.

[0014] The method has several key features. First, synchronized recording takes place, simultaneously logging load states such as current waveforms in electronic fuses and functions that communicate via BUS / CAN signals. Furthermore, activated software components can be recorded, especially when certain functions are triggered by user actions, such as pressing a button.

[0015] Another feature is the mapping of fuse-load relationships, meaning the system enables the assignment of electronic fuses to their associated electrical loads. These relationships can be mapped based on data from development vehicles to accurately understand possible load distributions.

[0016] Furthermore, the process includes context-aware CAN matching. CAN signals are analyzed taking into account the current operating context, examining current electrical signatures to accurately identify and assign loads.

[0017] The method thus enables an assignment logic (tagging logic) that increases the accuracy of the analysis and allows for continuous adaptation and improvement, resulting in efficient monitoring and optimization of the electrical system in test vehicles under real operating conditions.

[0018] In a first further development, it is planned that the correlation of the change in the load state with the change in the vehicle function will be represented by a confidence algorithm.

[0019] Confidence assessment allows for a more reliable attribution of ambiguous states, such as a load increase that could be attributed to more than one change in the operating state of multiple electrical loads connected to the electronic fuse. This ensures that the recorded data can be reliably categorized and interpreted in many situations.

[0020] In a further, more advanced embodiment, it is envisaged that a temporal progression of the load state is recorded and evaluated.

[0021] In a further, more advanced embodiment, it is provided that the load condition includes electrical current, voltage and / or operating temperature.

[0022] In a further, more advanced version, it is planned that a machine learning algorithm will be used to detect outliers and / or ambiguous scenarios.

[0023] In a further, more advanced embodiment, the machine learning algorithm is designed to discard detected outliers.

[0024] In a further, more advanced version, it is planned that the machine learning algorithm will be trained using supervised learning.

[0025] A first independent subject matter relates to a computer program product comprising a computer-readable storage medium on which instructions are embedded which, when executed by at least one computing unit, cause that at least one computing unit to be equipped to execute the procedure of the aforementioned type.

[0026] The process can be executed on one or more computing units, so that certain process steps are executed on one computing unit and other process steps on at least one other computing unit, whereby calculated data can be transmitted between the computing units if necessary.

[0027] Another independent item relates to a device for load detection and allocation in an electrical motor vehicle infrastructure, comprising a telemetry logger connectable to at least one electronic fuse, at least one data bus signal logger, at least one time synchronization module and at least one storage medium comprising an allocation database.

[0028] In a first further embodiment, it is envisaged that a control system is provided which has at least one of the following modules: - Waveform function analysis module, - Confidence calculation module, - Reference signature library, - Table of environment and context modifiers, - Outlier detection and data validation, - Load profile generator module and / or - Export module.

[0029] Further advantages, features, and details will become apparent from the following description, in which – possibly with reference to the drawing – at least one embodiment is described in detail. Identical, similar, and / or functionally equivalent parts are identified by the same reference numerals.

[0030] They show schematically: Fig. 1 a process scheme in a device for load detection and assignment, as well as Fig. 2. A flowchart of the procedure.

[0031] Fig. Figure 1 shows a process diagram in a device for load detection and assignment.

[0032] The device 2 is connected to or installed on a motor vehicle 4, which has at least one electronic fuse and several electrical consumers protected by the electronic fuse, for example, lighting, driver assistance systems, electric heating elements, and the like. The motor vehicle 4 is a test vehicle. The electrical consumers are supplied with energy by a vehicle battery. The electrical consumers are controlled via a CAN bus system.

[0033] The device 2 includes a data logger 5, which initially connects to a data bus signal logger 6 that is connected to the CAN bus system of the motor vehicle 2. Signals transmitted via the motor vehicle's data bus are received by the data bus signal logger 6 and recorded with a timestamp.

[0034] Furthermore, the data logger 5 includes a telemetry logger 8, which is connected to at least one electronic fuse of the motor vehicle 4 and records the data output by the electronic fuse, such as current and voltage curves as well as temperatures in the electronic fuse, with timestamps.

[0035] An environmental data logger 10 is used to record relevant environmental data that is recorded via the motor vehicle 4, for example an outside temperature or a state of charge of the vehicle battery.

[0036] The various data streams of the logger 5 are synchronized via a time synchronization module 12 to ensure that causes and effects are correlated correctly in time.

[0037] A processor 14, running a machine learning algorithm used to detect outliers and unreliable data sets, can evaluate the combined data stream generated via the time synchronization module 12, as described in connection with Fig. 2 is described.

[0038] The time-synchronized data from the various loggers 6, 8, 10 are stored together with the evaluation data from the processor 14 in a raw data memory 16, from which they are forwarded to a central computer 18. Correlation and evaluation take place there.

[0039] Using an export module 20, the correlation data is then exported in a desired data format and stored on a storage medium 22 with an assignment database 20.

[0040] Fig. Figure 2 shows a flowchart of the procedure.

[0041] Process steps VS1 to VS5 are performed, each comprising several sub-steps. VS4 is a step that is not performed in some specific embodiments. VS1: Data acquisition in the test vehicle

[0042] The first step involves setting up comprehensive high-frequency data acquisition in vehicle 2. This includes installing data logger 5 in vehicle 4 and connecting it to the electronic fuse box and a CAN bus port. Data logger 5 supports protocols such as XCP or alternative protocols like CAN FD or a gateway.

[0043] Data logger 5 is configured to capture various relevant signals. This includes electronic fuse current and voltage telemetry from at least one electronic fuse via telemetry logger 8. As mentioned, the electrical power and voltages at the electronic fuse(s) are recorded using suitable protocols such as XCP, CAN FD, or a gateway.

[0044] Other signals captured by Data Logger 5 include decoded BUS signals indicating changes in vehicle functions. For example, signals such as "SeatHeat_Status," "FanSpeed_Level," and "ADAS_Active" are logged, representing the activation of specific functions by a vehicle user. In addition, Data Logger 5 captures internal activation logs of software components. When the vehicle user activates certain functions, such as an Advanced Driver Assistance System (ADAS), the associated software components in relevant control units, such as a camera controller or sensor controller, are activated. These activations and their precise timestamps are captured via the bus protocol by Data Bus Signal Logger 6, enabling analysis of load increases across the various control units.

[0045] In addition, environmental metadata signals are recorded using the environmental data logger 10. Data such as outside temperature (“OutsideTemp”), state of charge (SOC level) of a vehicle battery, and cabin humidity (“CabinHumidity”) are taken into account to analyze the environmental impact on current and voltage curves. Finally, vehicle metadata is also stored via the environmental data logger 10 to ensure that all data can be considered in the context of the specific vehicle 4. This metadata includes information about the model as well as the hardware and software versions of the control units.

[0046] All recorded data points are stored in high-resolution file formats. This ensures precise subsequent analysis of the vehicle data with regard to load distribution and changes during testing. VS2: Time synchronization of electronic backup data and bus signals

[0047] In this step, the time synchronization module 12 ensures that all acquired data are precisely synchronized. The data from the CAN bus and at least one electronic fuse are synchronized by synchronizing their timestamps using control unit (ECU) timestamps or alternative reference points such as GPS positions or mileage readings. This makes it possible to identify the exact moments when specific electrical load changes and bus signals coincide.

[0048] The accuracy of the time synchronization is validated by checking known vehicle events. For example, switching on the ignition lock could be used as a reference point to verify that all systems are correctly synchronized and that the timing of the data is valid.

[0049] All recorded timestamps are normalized to a global reference. This normalization allows all events in the millisecond range to be precisely mapped and compared. This is crucial for the accurate analysis of time-critical data and ensures that all components in the vehicle system are precisely coordinated. VS3: Data cleansing, processing and mapping

[0050] This step involves several processes for the detailed analysis and classification of vehicle data. First, event detection takes place on a communication channel of the electronic fuse. Activation patterns are detected by exceeding a threshold value, for example, at least 1 ampere for at least 100 milliseconds. Such an activation pattern is then identified as an event.

[0051] To minimize false entries, debouncing mechanisms and noise filters can be used. The waveform of the electronic fuse signal can also be analyzed. In this case, the waveform can be segmented into defined events with metadata such as start and end times, duration, and average and peak current values.

[0052] Following this, the candidate load is identified. This involves retrieving all electrical loads assigned to a specific electronic fuse, such as "Fuse01" for "Seat Heating" and "Rear Window Defrosting". The assignment of loads to electronic fuses is performed either statically based on the vehicle configuration or dynamically through learning processes.

[0053] A further process step is the assignment of the BUS signal context. This involves retrieving assigned BUS signals and examining state changes and transitions within the context of functions or internal software component protocols within a time window of ± 2 seconds around the event. Signals that correlate with the current increase are marked as context-related to identify why certain loads rise or fall.

[0054] In a subsequent sub-step, electrical signatures are analyzed. Waveform functions are calculated, and parameters such as inrush current peak, rise time, RMS value, and shape classification are analyzed. The waveform is then compared to known templates from a signature library for each charging process. Signature expectations can be adjusted based on contextual factors such as reduced consumption during ADAS activation and environmental modifiers like increased inrush current during cold starts.

[0055] As a final step, a confidence score and load allocation are performed. A weighted score is calculated, potentially incorporating a CAN match score, a signature match score, and ecological and contextual multipliers. The overall confidence score, for example, is calculated as 0.5 * CAN confidence + 0.3 * signature confidence + 0.2 * context confidence. The best matches are then saved based on their scores.

[0056] Each event is ultimately tagged with metadata such as "efuse_id", associated "load_types", contextual information, environmental data, average current values, confidence level, and timestamp. Changes in vehicle functions are interpreted as the result of passenger interaction or internal protocols.

[0057] Depending on the vehicle configuration, there is a one-to-many relationship between functions and software components.

[0058] The storage format is designed to support analyses and simulations, making the data efficiently available for later investigations. Optional: VS4: Machine learning outlier detection and load behavior modeling

[0059] This optional step uses machine learning (ML) to predict the behavior of current loads at electronic fuses. This helps identify outliers that may originate both inside and outside the vehicle.

[0060] The approach is based on using vehicle operating states to predict electronic fuse behavior. These predictions utilize CAN signals and environmental conditions and may also take specific software components into account.

[0061] This process step has several sub-steps, which can proceed as follows: Entrances

[0062] The inputs include time-series data from electronic fuses, such as current waveforms, timestamps, and fuse IDs. Vehicle metadata, such as model series and hardware and software versions, is also included. Furthermore, active functions and software components, for example, switching the air conditioning on and off, are taken into account. Additional input data includes environmental data such as interior and exterior temperatures. Model training

[0063] Model training is performed through supervised learning using techniques such as neural networks or XGBoost to predict the normal current range. Various algorithms can be employed, including machine learning-based tagging, rule-based tagging, probabilistic classifiers, and signature decomposition. Forecast & Evaluation

[0064] The forecasting and evaluation process involves comparing the actual current pattern with the predicted pattern to identify anomalies. Marked cases are used to improve forecast accuracy and diagnose potential problems.

[0065] The use of machine learning algorithms enables the detection of anomalies in both testing and real-world operation. It improves the accuracy of simulations, supports the identification of specific scenarios, and enables onboard inference as well as post-hoc analysis. VS5: Exporting and saving the load profile

[0066] In this step, the tagged events are exported into various formats to make them usable for different applications. Export formats can include JSON for web dashboards and visualizations, CSV for data processing in batch pipelines, and MATLAB (.mat) for MATLAB simulations.

[0067] The exported fields can include, for example: electronic fuse ID (efuse_id), load assignment (load_tag), average current (avg_current), peak load (peak_current), confidence level, context (e.g., software activation or button press), and environmental context such as weather, state of charge, cold start, etc. (environment_context). The export process can be modular and automated, and can include data validation.

[0068] Although the invention has been further illustrated and explained in detail by means of preferred embodiments, the invention is not limited by the disclosed examples, and other variations can be derived from them by a person skilled in the art without departing from the scope of protection of the invention. It is therefore clear that a multitude of possible variations exist. It is also clear that the embodiments mentioned as examples are truly only examples and are not to be understood in any way as limiting, for example, the scope of protection, the possible applications, or the configuration of the invention.Rather, the preceding description and the description of the figures enable the person skilled in the art to implement the exemplary embodiments in concrete terms, whereby the person skilled in the art, with knowledge of the disclosed inventive concept, can make various changes, for example with regard to the function or the arrangement of individual elements mentioned in an exemplary embodiment, without leaving the scope of protection defined by the claims and their legal equivalents, such as a further explanation in the description. Reference symbol list 2 Device for load detection and assignment 4 Motor vehicle 5 data loggers 6 Data bus signal loggers 8 telemetry loggers 10 environmental data loggers 12 Time synchronization module 14 processor 16 raw data storage 18 central computers 20 Export module 22 Storage medium with allocation database

Claims

Method for load detection and allocation in an electrical vehicle infrastructure, comprising a plurality of consumers, wherein more than one of these consumers is protected by an electronic fuse, wherein the electronic fuse is configured to detect a current load state of the electrical fuse, wherein a current consumption state of at least one consumer protected by the electronic fuse is changed via the electronic fuse, wherein a change in the load state is recorded, characterized in that signals relating to at least one change in a vehicle function are detected via a data bus of a vehicle (4), wherein at least one environmental metadata is also detected, wherein the change in the load state of the electronic fuse, the change in the vehicle function and the at least one detected environmental metadata are time-synchronized.where the change in the load state is correlated with the change in the vehicle function and stored in a data set. Method according to claim 1, characterized in that the correlation of the change in the load state with the change in the motor vehicle function is mapped using a confidence algorithm. Method according to claim 1 or 2, characterized in that a time course of the load state is recorded and evaluated. Method according to one of the preceding claims, characterized in that the load condition comprises electric current, voltage and / or operating temperature. Method according to one of the preceding claims, characterized in that a machine learning algorithm is used to detect outliers and / or ambiguous scenarios. Method according to one of the preceding claims, characterized in that the machine learning algorithm is configured to reject detected outliers. Method according to one of the preceding claims, characterized in that the machine learning algorithm is trained by means of supervised learning. Computer program product comprising a computer-readable storage medium on which instructions are embedded which, when executed by at least one computing unit, cause the at least one computing unit to be configured to execute the method according to one of the preceding claims. Device for load detection and allocation in an electrical motor vehicle infrastructure (2), comprising a telemetry logger (8) connectable to at least one electronic fuse, at least one data bus signal logger (6), at least one time synchronization module (12) and at least one storage medium comprising an allocation database (22). Device according to claim 9, characterized in that a control unit is provided which has at least one of the following modules: - waveform function analysis module, - confidence calculation module, - reference signature library, - table with environment and context modifiers, - outlier detection and data validation, - load profile generator module and / or - export module.

Images