Methanol fuel filling detection and alarm system for vehicle

Abstract

The invention discloses a methanol fuel filling detection and alarm system for a vehicle, and particularly relates to the technical field of data acquisition and correction, comprising a data acquisition module for acquiring environmental data and sensor data in real time; the fluctuation analysis module analyzes data fluctuation and determines whether to activate an evaluation mechanism; the feature evaluation module performs environmental adaptability and compensation correction evaluation when the evaluation mechanism is activated; the decision classification module inputs an evaluation result into a machine learning model and divides job types; the alarm module gives an alarm when the height interference operation occurs; the switching optimization module adjusts a compensation model according to the interference level to ensure the filling precision and stability; according to the method, environment adaptability evaluation and compensation correction evaluation are combined, so that the system can evaluate and reflect the condition in the filling process in time when the environment condition changes, the response capability to the environment change is enhanced, the compensation coefficient can be automatically adjusted under the condition that the adaptability of the fixed compensation model is insufficient, and the compensation efficiency is improved. And the accuracy of sampling data in the filling process is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of data acquisition and correction, and more specifically, to a methanol fuel filling detection and alarm system for vehicles. Background Art

[0002] With the gradual widespread application of methanol as a clean energy source, the filling technology of methanol fuel for vehicles has become increasingly important. Especially during the filling process, how to monitor the quality of filling data in real time, avoid data distortion, and ensure filling accuracy has become an urgent problem to be solved.

[0003] Traditional filling detection systems often rely on fixed compensation models. However, in complex environmental conditions (such as when temperature, humidity, air pressure, etc. fluctuate greatly), the fixed compensation models in the prior art often face problems of insufficient adaptability and low accuracy.

[0004] In addition, traditional fluctuation analysis and environmental adaptability evaluation methods usually lack dynamic response and real-time adjustment capabilities, and cannot fully meet the requirements for accuracy and stability during the filling process. Therefore, the present invention proposes a methanol fuel filling detection and alarm system for vehicles in order to improve the accuracy and reliability of filling detection. Summary of the Invention

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A methanol fuel filling detection and alarm system for vehicles, including a data collection module, a fluctuation analysis module, a feature evaluation module, a decision classification module, an alarm module, and a switching optimization module;

[0007] The data collection module is used to collect various preset environmental data and sensor data during the filling process in real time, and summarize the data to obtain a collection of acquired data;

[0008] The fluctuation analysis module is used to perform volatility analysis on various preset environmental data based on the collection of acquired data to determine whether to activate the evaluation mechanism;

[0009] The feature evaluation module is used to perform feature extraction and conduct environmental adaptability evaluation and compensation correction evaluation respectively when the evaluation mechanism is in an activated state;

[0010] The decision classification module is used to input the results of environmental adaptability evaluation and compensation correction evaluation into a pre-trained machine learning model together, and classify the data sampling operation in the current time window as a highly interfering operation or a lowly interfering operation;

[0011] The alarm module is used to give an alarm during a highly interfering operation;

[0012] The switching optimization module is used to use the pre-trained dynamic correction model to correct the compensation coefficients in the preset fixed compensation model in real time during highly interfering operations. During low-interference operations or when the evaluation mechanism is not activated, the preset fixed compensation model is used for sensor data compensation.

[0013] In a preferred embodiment, the volatility analysis refers to:

[0014] Under the currently corresponding preset time window, the maximum value, minimum value, and standard deviation of various types of environmental data are respectively obtained. The maximum and minimum values corresponding to the environmental data type are compared with the preset normal fluctuation range of this type, and an indicator function f(Vi) is set; i represents the type of environmental data. When both the maximum and minimum values corresponding to the environmental data type i are within the preset normal fluctuation range of this type, the value of the indicator function f(Vi) is 0. When the maximum and minimum values corresponding to the environmental data type i do not satisfy being within the preset normal fluctuation range of this type, the value of the indicator function f(Vi) is 1. At the same time, the standard deviations of all types of environmental data are averaged to obtain the average fluctuation value BD.

[0015] In a preferred embodiment, determining whether to activate the evaluation mechanism refers to:

[0016] The value result of the indicator function f(Vi) and the average fluctuation value BD are substituted into the following formula together:

[0017] Both w1 and w2 are preset non-zero proportional coefficients, and the sum of w1 and w2 is one. P represents the total number of types of environmental data, and JH represents the risk value. When the risk value JH is greater than the preset activation threshold, the evaluation mechanism is activated. When the risk value JH is less than or equal to the preset activation threshold, the evaluation mechanism is not activated.

[0018] In a preferred embodiment, the feature evaluation module is used to perform feature extraction and respectively perform environmental adaptability evaluation and compensation correction evaluation when the evaluation mechanism is in an activated state, which refers to:

[0019] When the evaluation mechanism is in an activated state, the preset environmental condition feature group and the compensation feature group of the fixed compensation model are respectively extracted. Based on the environmental conditions described by the environmental condition feature group, an environmental adaptability evaluation is performed to generate an environmental adaptability index. Based on the compensation conditions described by the compensation feature group of the fixed compensation model, a compensation correction evaluation is performed to generate a compensation correction index.

[0020] In a preferred embodiment, the acquisition logic of the environmental adaptability index is:

[0021] Under the current corresponding preset time window, obtain various environmental condition characteristic data at time t from the preset environmental condition characteristic group and calculate the environmental impact factor:

[0022] Hi(t) represents the environmental condition characteristic data of environmental data type i at time t, Hi(ref) represents the environmental condition characteristic data of environmental data type i under standard conditions, and Fenv(t) represents the environmental impact factor of environmental data at time t;

[0023] Introduce the dynamic decay coefficient I(t), and substitute it into the environmental adaptation index calculation formula together with the environmental impact factor Fenv(t):

[0024] t0 represents the start time of the current corresponding preset time window, t1 represents the end time of the current corresponding preset time window, and IFI represents the environmental adaptation index.

[0025] In a preferred embodiment, the calculation formula of the dynamic decay coefficient I(t) is:

[0026] I(t) = e -λ·(t-t0)·(1+α·Fenv(t)) ; λ is a preset time decay coefficient, and α is a preset non-zero control coefficient.

[0027] In a preferred embodiment, the calculation formula of the compensation correction index is:

[0028] Under the current corresponding preset time window, obtain the compensation coefficient characteristic data and error characteristic data at time j from the compensation characteristic group of the fixed compensation model, and then calculate the compensation correction factor:

[0029] CAF(j) represents the compensation correction factor at time j, C(j) represents the compensation coefficient characteristic data at time j, E(j) represents the error characteristic data at time j, and both θ1 and θ2 are preset non-zero adjustment coefficients;

[0030] The compensation correction index calculation formula is:

[0031] γ represents a preset non-zero adjustment factor, which controls the response degree of the compensation correction index to the compensation correction factor, and CMI represents the compensation correction index.

[0032] In a preferred embodiment, both the machine learning model and the dynamic correction model are convolutional neural network models.

[0033] In a preferred embodiment, the fixed compensation model is a linear compensation model.

[0034] The technical effects and advantages of the present invention:

[0035] Through the dynamic compensation correction model, the present invention can adjust the compensation coefficient in real time according to environmental changes (such as temperature, humidity, air pressure, etc.), ensuring the accuracy of filling data. Compared with the traditional fixed compensation model, the dynamic correction method of the present invention can adapt to the dynamics of environmental changes, making the compensation process more flexible and accurate. By adopting a convolutional neural network (CNN) deep learning model, the relationship between environmental data and compensation requirements is automatically extracted to achieve automatic adjustment of the compensation coefficient, reducing the need for manual intervention. Through machine learning, the system can continuously optimize the model, improve its adaptability under different environmental conditions, and enhance the intelligent level of the system.

[0036] The present invention combines environmental adaptability assessment and compensation correction assessment, enabling the system to timely evaluate and reflect the situation during the filling process when environmental conditions change, thereby enhancing the system's response ability to environmental changes. The system can promptly activate the assessment mechanism to evaluate the correction situation of filling data when environmental fluctuations are large, so as to reflect the correction accuracy. The fixed compensation model often fails to provide sufficient correction in the face of drastic environmental changes, while the dynamic correction model of the present invention can adapt to these changes in real time, avoiding the limitations of traditional methods and improving the overall filling accuracy. Through real-time monitoring and compensation, the system can trigger the warning mechanism when environmental conditions fluctuate greatly, timely adjust the compensation parameters, and prevent measurement errors caused by environmental factors and potential safety problems during the filling process. In high-risk environments, the dynamic correction model can monitor the operation quality in real time, reduce errors caused by external factors, and ensure the safety of fuel filling.

[0037] The compensation correction mechanism of the present invention can cope with various environmental changes, improving the reliability of the system. In addition, due to the dynamics of the compensation mechanism, the system has stronger flexibility and adaptability, and can be widely applied to different operating environments. Through the automated compensation correction mechanism, the system can automatically adjust the compensation parameters without frequent manual intervention and maintenance. The system has self-adaptability and can adjust the working state according to actual needs, thereby reducing the operation and maintenance costs. Using a pre-trained machine learning model further simplifies the system deployment and maintenance process, reduces the complexity of system debugging, and can be called at any time to ensure real-time use. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] For the convenience of those skilled in the art to understand, the present invention will be further described below in conjunction with the accompanying drawings;

[0039] Figure 1 It is a schematic diagram of a methanol fuel filling detection and alarm system for vehicles in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0040] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0041] Referring to Figure 1 the following embodiments are obtained:

[0042] Embodiment 1:

[0043] As methanol, a clean energy source, is gradually widely used, the refueling technology for vehicle methanol fuel has become increasingly important. Especially during the refueling process, how to monitor the quality of refueling data in real time, avoid data distortion, and ensure refueling accuracy has become an urgent problem to be solved. Traditional refueling detection systems often rely on fixed compensation models. However, in complex environmental conditions (such as when temperature, humidity, air pressure, etc. fluctuate greatly), the fixed compensation models in the prior art often face problems of insufficient adaptability and low accuracy. In addition, traditional fluctuation analysis and environmental adaptability evaluation methods usually lack dynamic response and real-time adjustment capabilities and cannot fully meet the requirements for accuracy and stability during the refueling process. Therefore, an innovative technology is needed to overcome the limitations of the prior art and improve the accuracy and reliability of refueling detection.

[0044] The object of the present invention is to provide an improved vehicle methanol fuel refueling detection and alarm system, which improves the data accuracy and safety during the refueling process through the following aspects:

[0045] Dynamic activation of the volatility analysis and evaluation mechanism: By collecting environmental data in real time, analyzing volatility, and judging whether to activate the evaluation mechanism according to a preset threshold, corresponding evaluation and compensation correction are triggered when the environment changes greatly.

[0046] Environmental adaptability evaluation and compensation correction evaluation: When the evaluation mechanism is activated, environmental adaptability evaluation is carried out to generate an environmental adaptation index, which reflects the impact of environmental changes on sensor data, and compensation correction evaluation is used to reflect the errors caused by environmental changes.

[0047] Introduction of a machine learning model: Using a pre-trained convolutional neural network (CNN) model, by classifying and judging the results of environmental adaptability evaluation and compensation correction evaluation, the refueling operation is divided into low-interference or high-interference operations, and the compensation model is adjusted according to the classification results.

[0048] Switching mechanism of the dynamic correction model: During highly interfering operations, switch to the dynamic correction model for real-time compensation to ensure the accuracy of filling data; during low-interference operations, use the fixed compensation model for data compensation.

[0049] The invention aims to achieve high-precision and high-stability methanol fuel filling detection, reduce errors caused by environmental changes, improve safety during the filling process, and ultimately meet the growing demand of modern intelligent transportation systems for filling detection technology. The invention provides a methanol fuel filling detection and alarm system for vehicles, including a data collection module, a fluctuation analysis module, a feature evaluation module, a decision classification module, an alarm module, and a switching optimization module; the data collection module, the fluctuation analysis module, the feature evaluation module, the decision classification module, the alarm module, and the switching optimization module are communicatively connected.

[0050] The data collection module is used to collect various preset environmental data and sensor data in real time during the filling process, and summarize the data to obtain a collection of collected data; this module is used to collect various environmental data and sensor data in real time during the methanol fuel filling process. By summarizing different types of data, a comprehensive data set is generated, providing a basis for subsequent analysis and evaluation. Real-time monitor relevant data such as temperature, humidity, air pressure, fuel flow, and pressure. Integrate the original data collected by different sensors to form a data set. Provide raw data support for subsequent volatility analysis, feature evaluation, and compensation correction.

[0051] The fluctuation analysis module is used to perform volatility analysis on various preset environmental data based on the collection of data sets to determine whether to activate the evaluation mechanism; the core function of this module is to analyze the collected data set, especially the volatility of environmental data. By identifying the fluctuation of the data, it is decided whether to activate the evaluation mechanism to ensure that the system can respond in a timely manner to unstable or violently fluctuating environmental changes. Perform volatility analysis on environmental data, calculate the maximum value, minimum value, and standard deviation of each environmental data. Compare with the preset normal fluctuation range to determine whether the current environmental data has abnormal fluctuations. If the fluctuation exceeds the preset activation threshold, activate the evaluation mechanism; if the fluctuation is within the normal range, maintain the existing operating state.

[0052] The feature evaluation module is used to extract features and perform environmental adaptability evaluation and compensation correction evaluation respectively when the evaluation mechanism is in the active state. When the evaluation mechanism is activated, this module is responsible for extracting features from environmental data and sensor data, and performing environmental adaptability evaluation and compensation correction evaluation. Its purpose is to calculate the corresponding correction value by evaluating the impact of environmental changes on sensor data for compensation. Environmental adaptability evaluation: Analyze the impact of changes in environmental conditions on the sensor, generate an environmental adaptation index, and measure the adaptability and impact of the current environment on the measurement data. Compensation correction evaluation: On the basis of the environmental adaptability evaluation, evaluate whether the fixed compensation model is effective, calculate the compensation correction index, and evaluate the error caused by environmental changes.

[0053] The decision classification module is used to input the results of environmental adaptability evaluation and compensation correction evaluation into a pre-trained machine learning model together, and classify the data sampling operation of the current time window as a highly interfering operation or a lowly interfering operation. This module inputs the results of environmental adaptability evaluation and compensation correction evaluation into a pre-trained machine learning model, automatically determines the interference degree of the filling operation using machine learning algorithms, and decides on subsequent compensation measures. Input the results of environmental adaptability evaluation and compensation correction evaluation (such as environmental adaptation index and compensation correction index) into the machine learning model. Use the trained model to automatically classify the current filling operation as a "highly interfering operation" or a "lowly interfering operation". Determine whether the operation needs to switch the compensation mode or adjust the compensation strategy.

[0054] The alarm module is used to give an alarm during highly interfering operations. When the interference degree of the filling operation is classified as a "highly interfering operation", the alarm module will promptly trigger an alarm to remind the operator of potential problems during the filling process, ensuring that the system can take appropriate countermeasures. When a highly interfering operation is detected, an alarm signal is issued to indicate possible errors or instability factors during the filling operation. Ensure that the operator can promptly discover and handle possible failures or problems during the filling process.

[0055] The switching optimization module is used to use the pre-trained dynamic correction model to correct the compensation coefficients in the preset fixed compensation model in real time during highly interfering operations. During low-interference operations or when the evaluation mechanism is not activated, the preset fixed compensation model is used for sensor data compensation. The function of this module is to dynamically adjust the compensation model according to the classification results of the operation interference level. During highly interfering operations, it switches to the pre-trained dynamic correction model for real-time compensation; while during low-interference operations or when the evaluation mechanism is not activated, it continues to use the fixed compensation model for data compensation. According to the output of the decision classification module, it is judged whether the current operation is a "highly interfering operation". If it is a highly interfering operation, the system will automatically switch to the dynamic correction model to correct the compensation coefficients in the compensation model in real time to ensure the accuracy of the filling data. If it is a low-interference operation or the evaluation mechanism is not activated, the existing fixed compensation model is continued to be used for data compensation to maintain the stability of the system.

[0056] Volatility analysis means that under the current preset time window, the maximum value, minimum value, and standard deviation of various types of environmental data are obtained respectively. The maximum and minimum values corresponding to the environmental data type are compared with the preset normal fluctuation range of this type, and the indicator function f(Vi) is set; i represents the type of environmental data. When both the maximum and minimum values corresponding to the environmental data type i are within the preset normal fluctuation range of this type, the value of the indicator function f(Vi) is 0. When the maximum and minimum values corresponding to the environmental data type i do not satisfy being within the preset normal fluctuation range of this type, the value of the indicator function f(Vi) is 1. At the same time, the average value of the standard deviations of all types of environmental data is calculated to obtain the average fluctuation value BD. Determining whether to activate the evaluation mechanism means substituting the value result of the indicator function f(Vi) and the average fluctuation value BD into the following formula: Both w1 and w2 are preset non-zero proportionality coefficients, and the sum of w1 and w2 is one. P represents the total number of types of environmental data, JH represents the risk value. When the risk value JH is greater than the preset activation threshold, the evaluation mechanism is activated. When the risk value JH is less than or equal to the preset activation threshold, the evaluation mechanism is not activated.

[0057] By calculating the maximum value, minimum value, and standard deviation, the range of variation and the fluctuation of environmental data within a preset time window can be comprehensively understood. This provides the basic data for subsequent volatility analysis. Comparing the obtained maximum and minimum values with the preset normal fluctuation range to determine whether the environmental data is within the normal range. If the data exceeds the normal fluctuation range, it indicates that there are abnormal changes, which may affect the accuracy of the filling data. By defining an indicator function, when both the maximum and minimum values of the environmental data are within the preset normal fluctuation range, the value of the indicator function is 0, indicating no abnormal fluctuations; if the maximum and minimum values do not meet the normal fluctuation range, the value of the indicator function is 1, indicating abnormal fluctuations. This operation can provide a clear identifier for subsequent decision-making. By calculating the average value of the standard deviations of various environmental data, the average fluctuation value is obtained. This can help the system evaluate the degree of overall environmental fluctuation and further determine whether it is necessary to activate the evaluation mechanism.

[0058] Based on the results of the volatility analysis, the system can generate a risk value, which reflects the degree of influence of the current environmental conditions on the filling process. If the risk value exceeds the preset threshold, it indicates that the environmental fluctuation is large, which may affect the accuracy and stability of the filling, and thus it is necessary to activate the evaluation mechanism for further compensation and correction. The risk value is calculated by combining the value of the indicator function with the average fluctuation value, taking into account the contribution of each environmental data type. The value (0 or 1) of the indicator function determines whether the environmental data is within the normal fluctuation range, while the average fluctuation value reflects the overall degree of environmental fluctuation. The proportionality coefficient is used to adjust the contribution of each environmental data type to the risk value. Their sum is 1, ensuring the balance of the comprehensive role of all environmental data types in the risk assessment. When the calculated risk value exceeds the preset activation threshold, the evaluation mechanism is activated, indicating that the environmental fluctuation has exceeded the normal range and it may be necessary to adjust the measurement compensation strategy of the system; if the risk value is less than or equal to the threshold, the evaluation mechanism does not need to be activated, indicating that the environmental fluctuation is within the acceptable range.

[0059] The feature evaluation module is used to extract features and conduct environmental adaptability evaluation and compensation correction evaluation respectively when the evaluation mechanism is in the activated state, which refers to:

[0060] When the evaluation mechanism is in the activated state, the preset environmental condition feature group and the compensation feature group of the fixed compensation model are extracted respectively. Based on the environmental situation described by the environmental condition feature group, the environmental adaptability evaluation is carried out to generate an environmental adaptation index. Based on the compensation situation described by the compensation feature group of the fixed compensation model, the compensation correction evaluation is carried out to generate a compensation correction index.

[0061] The core purpose of environmental adaptability assessment is to analyze the impact of environmental changes on sensor data during the filling process based on the current environmental conditions, ensure that the filling data can adapt to different environmental changes, and then adjust the compensation strategy. By extracting the environmental condition feature group, the impact of the current environment on the sensor performance is evaluated. The environmental condition feature group may include data such as temperature, humidity, air pressure, wind speed, etc., which describe the changing trend of the environment. By comparing with the preset normal range, it is judged whether the current environment has an adverse impact on the sensor, and an environmental adaptation index is generated, which reflects the degree of adaptability to environmental changes. After generating the environmental adaptation index, it can quantitatively reflect whether the current environmental conditions are within the working range of the sensor. If the environmental adaptation index deviates from the normal range, it indicates that the accuracy of the preset fixed compensation model may decline. This index can provide a basis for subsequent compensation correction to ensure that the system can automatically adjust the compensation strategy according to environmental changes.

[0062] Compensation correction assessment is based on the compensation feature group (such as compensation coefficient, error feature, etc.), and analyzes and evaluates whether the existing fixed compensation model can effectively eliminate the errors brought by environmental changes, ensuring that the measured value of the sensor is as close as possible to the true value. The specific significance includes: extracting the compensation feature group of the fixed compensation model (for example: compensation coefficient, sensor error), and based on these features, evaluating the correction ability of the compensation model. The compensation feature group is used to describe how the fixed compensation model responds to environmental changes, especially how to adjust the compensation parameters to eliminate the errors caused by environmental factors when environmental changes such as temperature and humidity are drastic. Based on the compensation correction assessment, a compensation correction index is generated, which represents the correction effect of the compensation model and the effectiveness of the current compensation correction. The generation of the compensation correction index can help judge whether the current compensation mechanism is sufficient and whether it is necessary to switch to a dynamic correction model for more refined adjustment.

[0063] The acquisition logic of the environmental adaptation index is as follows:

[0064] Under the current corresponding preset time window, from the preset environmental condition feature group, obtain various environmental condition feature data at time t and calculate the environmental impact factor:

[0065] Hi(t) represents the environmental condition feature data of environmental data type i at time t, Hi(ref) represents the reference data of the environmental condition feature data of environmental data type i under standard conditions (usually the preset standard value or the value under ideal conditions). Fenv(t) represents the environmental impact factor of the environmental data at time t, P represents the type number of environmental data features, that is, the type number of environmental condition data, such as temperature, humidity, etc.

[0066] The dynamic decay coefficient I(t) is introduced and substituted into the environmental adaptation index calculation formula together with the environmental impact factor Fenv(t):

[0067] t0 represents the starting time of the currently corresponding preset time window, t1 represents the ending time of the currently corresponding preset time window, and IFI represents the environmental adaptation index. It is obtained by integrating the product of the environmental impact factor and the corresponding refueling operation impact function. This index is used to reflect the impact of the environment on the refueling operation. The environmental adaptation index is used to reflect the overall impact of environmental changes on the refueling operation within a specific time window.

[0068] The calculation formula for the dynamic decay coefficient I(t) is as follows:

[0069] I(t) = e -λ·(t-t0)·(1+α·Fenv(t)) ; λ is the preset time decay coefficient, which represents the decay rate of the environmental impact. As time increases, the environmental impact will decay exponentially, and the time decay coefficient controls the speed of this decay. α is the preset non-zero control coefficient, which represents the adjustment intensity of the environmental impact factor on the dynamic characterization coefficient. This coefficient determines the degree of impact of environmental changes on the operation. The dynamic decay coefficient I(t) reflects the degree of impact of the environmental conditions on the refueling operation at that moment. It changes with time and depends on the environmental impact factor and the time decay factor.

[0070] The decay term e -λ·(t-t0) in the formula indicates that as time goes by, the environmental impact will gradually decrease. The longer the time, the more significant the decay, indicating that the impact of the environment gradually decreases. The product of the environmental impact factor and the control coefficient represents the adjustment effect of the environmental conditions on the dynamic coefficient. The greater the environmental fluctuation, the stronger the impact, and correspondingly, the greater the compensation correction. The dynamic decay coefficient I(t) represents the impact of environmental changes on the current refueling operation. A higher value indicates a greater environmental impact, and higher compensation correction may be required, while a lower value indicates a smaller environmental impact, and less significant correction may be needed.

[0071] The calculation formula for the compensation correction index is as follows:

[0072] Under the currently corresponding preset time window, from the compensation characteristic group of the fixed compensation model, obtain the compensation coefficient characteristic data and error characteristic data at time j, and then calculate the compensation correction factor:

[0073] CAF(j) represents the compensation correction factor at time j, which reflects the influence of the current environmental conditions on the compensation system and determines the intensity of the compensation correction. C(j) represents the compensation coefficient characteristic data at time j, which is the correction parameter adjusted by the compensation system according to the real-time environmental data. E(j) represents the error characteristic data at time j, which measures the error or deviation of the system at this time. Both θ1 and θ2 are preset non-zero adjustment coefficients, which respectively control the influence degree of the compensation system characteristic data and the error characteristic data on the compensation correction factor. The hyperbolic tangent function represents the non-linear calculation method of the compensation correction factor. This function is used to adjust the amplitude of the compensation correction to avoid overcorrection.

[0074] The calculation formula for the compensation correction index is:

[0075] γ represents a preset non-zero adjustment factor, which is used to control the response degree of the compensation correction index to the compensation correction factor. This coefficient determines the sensitivity of the correction. To control the response degree of the compensation correction index to the compensation correction factor, CMI represents the compensation correction index, which is used to quantify the total effect of the compensation correction factor. This index reflects the intensity of the compensation system correction within a given time window. The compensation correction factor is used to dynamically adjust the compensation intensity of the compensation system, taking into account the environmental data and error characteristics, and is non-linearly adjusted through the hyperbolic tangent function. By performing a time integral on the compensation correction factor, the compensation correction index is generated to quantify the overall effect of the compensation correction.

[0076] The fixed compensation model is a linear compensation model. The linear compensation model means that the relationship between the correction factor and environmental variables (such as temperature, humidity, etc.) during the compensation process is linear, that is, the change of the correction factor is a linear function of the change of environmental factors. In other words, the linear compensation model assumes that there is a simple proportional relationship between the change of environmental conditions and the adjustment amount of the compensation. Specifically, the linear compensation model usually has the following form: y(t) = a * x(t) + b. y(t) is the data after compensation correction; x(t) is the original data of the sensor; a is the compensation coefficient, which represents the influence degree of environmental conditions on the compensation correction; b is the bias term, which usually represents a fixed correction amount. In this model, the compensation value y(t)) is linearly calculated from the sensor data x(t)) and the preset compensation coefficient (a).

[0077] By setting a fixed compensation model, it is possible to ensure that the system has a stable compensation mechanism in most cases, reducing measurement errors caused by environmental changes. This is particularly effective for routine operations and applications with little environmental change. The fixed compensation model does not require complex adaptive mechanisms or real-time adjustments, making it easy to maintain and manage. Especially when the environmental changes are relatively stable or the system operating conditions are relatively constant, the fixed compensation model is sufficient to meet the requirements. For systems with relatively small changes in environmental conditions, the fixed compensation model can provide sufficient accuracy to ensure data accuracy during the filling process. It helps to eliminate or reduce measurement errors caused by environmental factors. The linear compensation model is applicable to most ordinary environmental condition changes (such as regular temperature fluctuations, humidity changes, etc.). It can effectively compensate for the impacts caused by environmental factors, especially when the environmental changes are relatively slow. The fixed compensation model is usually easier to implement than the dynamic compensation model, especially when the understanding of environmental impact factors is relatively simple. The linear model helps the system to quickly adjust data through preset compensation coefficients. In some cases, the fixed compensation model can serve as a basic framework, and if the environmental conditions change significantly, the system can gradually introduce a more complex dynamic compensation model.

[0078] Both the machine learning model and the dynamic correction model are convolutional neural network models.

[0079] In this system, the results of environmental adaptability assessment and compensation correction assessment (such as the environmental adaptation index and the compensation correction index) will be used as inputs and passed to a pre-trained convolutional neural network (CNN) model. The task of this model is to determine whether the data sampling operation in the current time window belongs to a highly interfering operation or a lowly interfering operation by learning and classifying these input features.

[0080] Environmental adaptation index: It represents the degree of influence of the current environmental conditions on sensor data. When the environmental fluctuations are large, this index will increase, indicating that the system needs stronger compensation and correction.

[0081] Compensation correction index: It represents the effectiveness of the current compensation model correction, reflecting whether the compensation model is accurate enough and whether more refined dynamic adjustments are needed.

[0082] These two indexes are obtained through the feature evaluation module and used as inputs to the machine learning model. The CNN model will use these input data to learn the relationship between environmental changes and compensation requirements, so as to predict the interference degree of the filling operation.

[0083] The process of training a Convolutional Neural Network (CNN) includes the following steps: Data Preparation: During training, a large amount of labeled sample data needs to be prepared. Each sample data should include an environmental adaptation index and a compensation correction index, as well as the corresponding job type label (such as "high-interference job" or "low-interference job"). This data may come from historical records or be obtained through real-time acquisition.

[0084] Forward Propagation: Each sample data is input into the CNN and passes through the various layers of the network (such as convolutional layers, pooling layers, fully connected layers, etc.) to generate an output result (i.e., the prediction result of the model). The convolutional layer extracts local features of the input data, and the pooling layer downsamples the features to reduce the computational amount and the number of parameters. Finally, the classification result of the job type is output through the fully connected layer.

[0085] Loss Calculation: A loss function (such as the cross-entropy loss function) is used to calculate the gap between the predicted value and the true label. The purpose of the loss function is to measure the accuracy of the model prediction. For classification problems, the cross-entropy loss function can calculate the difference between the predicted probability of the model and the actual class label.

[0086] Backward Propagation: Through the backpropagation algorithm, the gradients of the loss function with respect to each weight and bias in the network are calculated. The weights of the network are updated through the gradient descent algorithm to minimize the loss function.

[0087] Optimization: During training, optimization algorithms (such as gradient descent, Adam optimizer, etc.) are used to adjust the parameters of the model to minimize the loss function. After each iteration, the model gradually optimizes the weights, making the model's prediction closer to the true label.

[0088] Iterative Training: The training process is usually completed through multiple iterations. Each iteration uses a batch of data to perform forward propagation, loss calculation, backward propagation, and optimization steps. When the number of training times is sufficient, the parameters of the model will converge, and finally, it can accurately predict new data.

[0089] Model Evaluation: During training, a validation set is usually used to evaluate the generalization ability of the model to prevent overfitting. Once the model training is completed, a test set is used for the final evaluation to obtain the performance of the model on unknown data.

[0090] The core task of the dynamic correction model is to adjust the compensation coefficient a according to real-time environmental change data, such as environmental conditions like temperature and humidity. These input data can reflect changes in environmental conditions and affect the accuracy and stability of sensor data.

[0091] Environmental change data group: Changes in environmental data (such as temperature, humidity, air pressure, wind speed, etc.) will directly affect the measurement accuracy of sensors. Input data includes but is not limited to the following aspects: Temperature: Affects the physical properties of sensors and fuel filling systems, which may lead to measurement errors. Humidity: Humidity changes may affect the electronic components or reaction rates of sensors. Air pressure: Air pressure changes may affect the density or flow rate of fluids. Other environmental factors: Such as wind speed, light, pollutant concentration, etc. These factors may also affect the filling system, especially in special environments. These environmental change data are collected in real time through sensors and used as inputs to the dynamic correction model.

[0092] Output result: The goal of the dynamic correction model is to output a compensation coefficient a, which will be used to correct the calculation process in the fixed compensation model. The adjustment of the compensation coefficient is based on the current environmental condition changes, so it can dynamically respond to the impact of different environmental factors on filling accuracy. The compensation coefficient a is a calibrated correction coefficient that controls the intensity of compensation. Based on the current environmental change data, the dynamic correction model will predict a new a, which reflects the compensation requirements under the current environmental conditions. This coefficient will directly affect the data calibration during the filling process and ensure the adaptability of the system to environmental changes.

[0093] Working process of the dynamic correction model:

[0094] Data collection: Environmental data (such as temperature, humidity, air pressure, etc.) are collected in real time through sensors and form an input data set. Data input: These real-time collected environmental data will be passed as inputs to the pre-trained convolutional neural network (CNN) model. Convolutional neural network processing: The convolutional layer in the CNN model will automatically extract local features in the environmental data (such as rapid changes in temperature and humidity) and learn the relationship between these features and the compensation coefficient. The activation function (such as ReLU) performs a non-linear transformation on the extracted features to increase the expression ability of the model. Compensation coefficient prediction: The fully connected layer of the CNN model will combine the features extracted by the convolutional layer and output a correction coefficient a according to the learned pattern, that is, the compensation coefficient adapted at the current moment. This coefficient a represents the intensity of compensation or calibration requirements under the current environmental conditions. Correct the compensation coefficient: Apply the output compensation coefficient a to the fixed compensation model. Adjust the compensation coefficient in the fixed compensation model according to the current environmental data to ensure the adaptability of the system to different environmental changes. Feedback mechanism: The prediction results of the dynamic correction model can be continuously optimized through real-time feedback. Over time, the system will continuously adjust the compensation coefficient to ensure the accuracy and stability of the filling data.

[0095] Application of the training process: During the training process, the dynamic correction model optimizes its prediction ability through a large amount of historical environmental data and known compensation coefficients (as target values). The training steps generally include the following parts:

[0096] Data preparation: Collect historical data, including changes in temperature, humidity, air pressure, etc. under different environmental conditions, as well as the compensation coefficient a under these environmental conditions. Model training: Use this data to train the CNN model. Through multiple iterations, the model learns the relationship between different environmental factors and the compensation coefficient. During the training process, a loss function (such as mean square error) is used to evaluate the gap between the model output and the true compensation coefficient, and the model weights are optimized through backpropagation.

[0097] Forward propagation and feedback: After training, new environmental data is input into the model to predict the corresponding compensation coefficient a, and it is applied to the compensation adjustment in the system.

[0098] This dynamic correction model can effectively improve the data accuracy and stability during the filling process, especially when the environmental conditions change drastically, ensuring that the system can adapt in real time and make compensation.

[0099] The above formulas are all dimensionless and take their numerical calculations. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0100] It should be understood that in various embodiments of the present application, the magnitude of the sequence numbers of the above processes does not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0101] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0102] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0103] As described above, it is only the specific implementation manner of the present application. However, the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claimed rights.

Claims

1. A methanol fuel filling detection and alarm system for vehicles, characterized in that, It includes a data collection module, a fluctuation analysis module, a feature evaluation module, a decision classification module, an alarm module, and a switching optimization module; The data collection module is used to collect various types of environmental data and sensor data preset during the filling process in real time, and summarize the data to obtain a collection of collected data; The fluctuation analysis module is used to perform volatility analysis on various types of preset environmental data based on the collection of collected data to determine whether to activate the evaluation mechanism; The feature evaluation module is used to extract features and perform environmental adaptability evaluation and compensation correction evaluation respectively when the evaluation mechanism is in an activated state; The decision classification module is used to input the results of environmental adaptability evaluation and compensation correction evaluation into a pre-trained machine learning model together, and classify the data sampling operation in the current time window as a highly interfering operation or a lowly interfering operation; The alarm module is used to give an alarm during a highly interfering operation; The switching optimization module is used to use a pre-trained dynamic correction model to correct the compensation coefficient in the preset fixed compensation model in real time during a highly interfering operation, and use the preset fixed compensation model for sensor data compensation during a lowly interfering operation or when the evaluation mechanism is not activated.

2. The methanol fuel filling detection and alarm system for a vehicle according to claim 1, characterized in that, The volatility analysis refers to: Under the current corresponding preset time window, obtain the maximum value, minimum value, and standard deviation of various types of environmental data respectively, compare the maximum and minimum values corresponding to the environmental data type with the preset normal fluctuation range of this type, and set up an indicator function f(Vi); i represents the type of environmental data. When both the maximum and minimum values corresponding to the environmental data type i are within the preset normal fluctuation range of this type, the value of the indicator function f(Vi) is 0. When the maximum and minimum values corresponding to the environmental data type i do not satisfy being within the preset normal fluctuation range of this type, the value of the indicator function f(Vi) is 1. At the same time, calculate the average value of the standard deviations of all types of environmental data to obtain the average fluctuation value BD.

3. The methanol fuel filling detection and alarm system for vehicles according to claim 2, characterized in that, Determining whether to activate the evaluation mechanism refers to: Substitute the value result of the indicator function f(Vi) and the average fluctuation value BD into the following formula: Both w1 and w2 are preset non-zero proportionality coefficients, and the sum of w1 and w2 is one. P represents the total number of types of environmental data, and JH represents the risk value. When the risk value JH is greater than the preset activation threshold, the evaluation mechanism is activated; when the risk value JH is less than or equal to the preset activation threshold, the evaluation mechanism is not activated.

4. A methanol fuel filling detection and alarm system for a vehicle according to claim 3, characterized in that, The feature evaluation module is used to extract features and perform environmental adaptability evaluation and compensation correction evaluation respectively when the evaluation mechanism is in an activated state refers to: When the evaluation mechanism is in an activated state, extract the preset environmental condition feature group and the compensation feature group of the fixed compensation model respectively, perform environmental adaptability evaluation based on the environmental situation described by the environmental condition feature group to generate an environmental adaptation index, and perform compensation correction evaluation based on the compensation situation described by the compensation feature group of the fixed compensation model to generate a compensation correction index.

5. The methanol fuel filling detection and alarm system for a vehicle according to claim 4, characterized in that, The acquisition logic of the environmental adaptation index is: Under the current corresponding preset time window, obtain various types of environmental condition feature data at time t from the preset environmental condition feature group and calculate the environmental impact factor: Hi(t) represents the environmental condition characteristic data of environmental data type i at time t, Hi(ref) represents the environmental condition characteristic data of environmental data type i under standard conditions, and Fenv(t) represents the environmental impact factor of environmental data at time t; Introduce a dynamic decay coefficient I(t), and substitute it into the environmental adaptation index calculation formula together with the environmental impact factor Fenv(t): t0 represents the starting time of the currently corresponding preset time window, t1 represents the ending time of the currently corresponding preset time window, and IFI represents the environmental adaptation index.

6. The methanol fuel filling detection and alarm system for a vehicle according to claim 5, characterized in that, The calculation formula of the dynamic decay coefficient I(t) is: I(t) = e -λ·(t-t0)·(1+α·Fenv(t)) ; λ is a preset time decay coefficient, and α is a preset non-zero control coefficient.

7. A methanol fuel filling detection and alarm system for a vehicle according to claim 6, characterized in that, The calculation formula of the compensation correction index is: Under the currently corresponding preset time window, obtain the compensation coefficient feature data and error feature data at time j from the compensation feature group of the fixed compensation model, and then calculate the compensation correction factor: CAF(j) represents the compensation correction factor at time j, C(j) represents the compensation coefficient characteristic data at time j, E(j) represents the error characteristic data at time j, and both θ1 and θ2 are preset non-zero adjustment coefficients; The calculation formula for the compensation correction exponent is: γ represents a preset non-zero adjustment factor that controls the response degree of the compensation correction index to the compensation correction factor, and CMI represents the compensation correction index.

8. The methanol fuel filling detection and alarm system for a vehicle according to claim 7, wherein Both the machine learning model and the dynamic correction model are convolutional neural network models.

9. The methanol fuel filling detection and alarm system for a vehicle according to claim 8, wherein, The fixed compensation model is a linear compensation model.

Images