Method and apparatus for compensation of intrinsic drift of electrochemical sensors

Abstract

The application provides an intrinsic drift compensation method and device of an electrochemical sensor, and relates to the technical field of electrochemical sensors. The intrinsic drift compensation method of the electrochemical sensor comprises the following steps: obtaining electrochemical response signal data generated by the electrochemical sensor in a measurement process of a target analyte; and extracting a static physical parameter from the electrochemical response signal data; wherein the static physical parameter is used to quantify an intrinsic characteristic parameter of an electrochemical system of the electrochemical sensor, which quantifies the mutual conversion of internal electrical energy and chemical energy when the electrochemical sensor measures the target analyte; and according to the electrochemical response signal data and the static physical parameter, an intrinsic drift compensation model is used to obtain a concentration estimation value of the target analyte; wherein the intrinsic drift compensation model is a machine learning model trained by using a training data set. The application can efficiently and reliably compensate the intrinsic drift of the electrochemical sensor.

Description

Technical Field

[0001] This invention relates to the field of electrochemical sensor technology, and in particular to an intrinsic drift compensation method and apparatus for electrochemical sensors. Background Technology

[0002] In the application of electrochemical sensors, accurate and reliable detection results are crucial. However, current electrochemical sensor technology faces a major challenge – intrinsic drift of the sensor signal. This phenomenon is caused by the coupling of various uncontrollable factors, such as differences in sensor manufacturing, inconsistencies in detection adapters, and fluctuations in measurement cycles. The existence of intrinsic drift seriously affects the reliability of detection results, causing measurement data to deviate and failing to accurately reflect the actual situation. Simultaneously, it shortens the maintenance-free cycle of the equipment and increases operating costs and maintenance workload.

[0003] To address the intrinsic drift problem, existing technologies have explored solutions from both hardware and software perspectives. In terms of hardware optimization, differential electrode structures and chemically modified electrodes are employed. While these methods can suppress some noise and slow down the drift process to a certain extent, they have significant limitations. For example, they increase sensor size and power consumption, complicate the fabrication process, and make it difficult to ensure product consistency during mass production, thus failing to fundamentally eliminate intrinsic drift. In the field of software signal processing, traditional methods such as Gaussian filtering, wavelet transform, and adaptive filtering generally suffer from low compensation accuracy, weak generalization ability, high computational complexity, strong parameter dependence, and the need for frequent manual parameter tuning. In real-world, complex, and variable monitoring scenarios, they struggle to achieve robust and efficient signal compensation, failing to meet the demands of the widespread application of electrochemical sensors.

[0004] Therefore, how to efficiently and reliably compensate for the intrinsic drift of electrochemical sensors is a technical problem that urgently needs to be solved. Summary of the Invention

[0005] This invention provides an intrinsic drift compensation method and apparatus for electrochemical sensors to overcome the aforementioned deficiencies in the prior art, thereby efficiently and reliably compensating for the intrinsic drift of electrochemical sensors.

[0006] This invention provides an intrinsic drift compensation method for an electrochemical sensor, comprising the following steps.

[0007] The process involves acquiring electrochemical response signal data generated by an electrochemical sensor during the measurement of a target analyte; extracting static physical parameters from the electrochemical response signal data; wherein the static physical parameters are used to quantify the intrinsic characteristics of the electrochemical system in which electrical and chemical energy are interconverted when the electrochemical sensor measures the target analyte; obtaining an estimated concentration of the target analyte using an intrinsic drift compensation model based on the electrochemical response signal data and the static physical parameters; wherein the intrinsic drift compensation model is a machine learning model trained using a training dataset; the training dataset is obtained by extracting a predetermined number of combined data from a full combination of data of various electrochemical sensor types, various electrode adapter types, and various analyte sample concentrations to obtain multiple sets of test conditions; and constructing multiple test loops based on the multiple sets of test conditions. The test involves inducing different drift effects of an electrochemical sensor under various testing environments. Electrochemical response signal data are collected under these multiple testing environments to obtain multiple sets of first sample data. Static physical parameters are extracted from each set of first sample data. Based on an optimization objective function, a target optimization algorithm is used to select target static physical parameters with drift intensity higher than a preset threshold from all the static physical parameters corresponding to the first sample data. The objective function includes a physical drift intensity reward term, which is used to quantify and score the drift intensity of at least one static physical parameter. Second sample data is obtained based on the first sample data corresponding to the target static physical parameters. The known concentration value of the analyte sample corresponding to the second sample data is used as the label of the second sample data. The training dataset is obtained based on the second sample data and its corresponding label.

[0008] According to the intrinsic drift compensation method for an electrochemical sensor provided by the present invention, the intrinsic drift compensation model includes a feature extraction layer and an output layer; the step of obtaining the concentration estimate of the target analyte based on the electrochemical response signal data and the static physical parameters using the intrinsic drift compensation model includes: obtaining a fused feature using the feature extraction layer based on the electrochemical response signal data and the static physical parameters; and obtaining the concentration estimate of the target analyte using the output layer based on the fused feature.

[0009] According to the intrinsic drift compensation method for an electrochemical sensor provided by the present invention, the step of obtaining fused features based on the electrochemical response signal data and the static physical parameters using a feature extraction layer includes: extracting intrinsic response features from the electrochemical response signal data using a convolutional neural network; and fusing the intrinsic response features and the static physical parameters to obtain the fused features.

[0010] According to the intrinsic drift compensation method for an electrochemical sensor provided by the present invention, the electrochemical response signal data is a bidirectional pulse scan time-series electrical signal sequence; the static physical parameters include at least one of differential peak current, peak potential corresponding to the differential peak current, forward peak current, forward peak potential, reverse peak current, reverse peak potential, and electrode interface charge-discharge hysteresis; the extraction of static physical parameters from the electrochemical response signal data includes performing at least one of the following operations: extracting the differential peak current and the peak potential corresponding to the differential peak current from the bidirectional pulse scan time-series electrical signal sequence using a global optimization operator; extracting the forward peak current and the forward peak potential corresponding to the forward scan from the bidirectional pulse scan time-series electrical signal sequence; extracting the reverse peak current and the reverse peak potential corresponding to the reverse scan from the bidirectional pulse scan time-series electrical signal sequence; and determining the electrode interface charge-discharge hysteresis based on the difference between the forward peak potential and the reverse peak potential.

[0011] According to the intrinsic drift compensation method for an electrochemical sensor provided by the present invention, the static physical parameters further include at least one of the following: left half-peak potential, right half-peak potential, half-peak width, peak symmetry factor, and half-peak potential mean; the extraction of static physical parameters from the electrochemical response signal data includes performing at least one of the following operations: extracting the left half-peak potential and right half-peak potential from the bidirectional pulse scan time-series electrical signal sequence; wherein, the left half-peak potential is the peak potential at the rising edge of the differential current when the current reaches half of its peak value; the right half-peak potential is the peak potential at the falling edge of the differential current when the current reaches half of its peak value; and the extraction of static physical parameters from the bidirectional pulse scan time-series electrical signal sequence... In the differential current-potential curve, for the target differential peak, the potential of the left half-peak and the potential of the right half-peak are extracted; wherein, the potential of the left half-peak is the potential corresponding to the current reaching half of the peak value of the target differential peak at the rising edge of the target differential peak; the potential of the right half-peak is the potential corresponding to the current reaching half of the peak value of the target differential peak at the falling edge of the target differential peak; based on the difference between the potential of the left half-peak and the potential of the right half-peak, the half-peak width is obtained; using the half-peak width, the difference between the differential peak potential and the potential of the left half-peak is normalized to obtain the peak symmetry factor; the mean of the potential of the left half-peak and the potential of the right half-peak is calculated to obtain the mean half-peak potential.

[0012] According to the intrinsic drift compensation method for an electrochemical sensor provided by the present invention, the static physical parameters further include at least one of the first derivative maxima and the first derivative minima of the signal function at the rising edge and the falling edge of the target difference peak, and the second derivative minima of the signal function; the extraction of static physical parameters from the electrochemical response signal data includes: establishing the signal function based on the bidirectional pulse scan time-series electrical signal sequence; the signal function has voltage as the independent variable and current as the dependent variable; performing a first-order derivative operation on the signal function to obtain the first derivative maxima and the first derivative minima of the target difference peak at the rising edge and the falling edge; and performing a second-order derivative operation on the signal function to obtain the second derivative minima.

[0013] According to the intrinsic drift compensation method for an electrochemical sensor provided by the present invention, the static physical parameters further include at least one of the mean current and standard deviation of current within the baseline interval of the bidirectional pulse scan time-series electrical signal sequence, net peak current, and peak area within the potential window during the measurement process for the target analyte; the extraction of static physical parameters from the electrochemical response signal data includes: calculating the mean current and standard deviation of current within the baseline interval; obtaining the net peak current based on the difference between the peak current and the mean current; and obtaining the peak area within the potential window during the measurement process for the target analyte based on the definite integral of the difference between the signal function and the baseline current function.

[0014] According to the intrinsic drift compensation method for an electrochemical sensor provided by the present invention, the step of acquiring electrochemical response signal data under multiple test environments to obtain multiple sets of first sample data includes: performing the following operations for a predetermined number of cycles under each test environment to obtain multiple sets of first sample data under each test environment: placing the electrochemical sensor into an analyte sample, wherein the electrochemical sensor is connected to an electrochemical workstation via an electrode adapter; acquiring the electrochemical response signal data of the electrochemical sensor through the electrochemical workstation; and obtaining a set of first sample data based on the acquired electrochemical response signal data.

[0015] The intrinsic drift compensation method for electrochemical sensors provided by the present invention has the following optimization objective function: ; in, To optimize the objective function, This represents the set of the first sample data. Indicates the number of data in the first sample. Indicates the first i Evaluation of static physical parameters of the first sample data Indicates the first jEvaluation of static physical parameters based on the first sample data; For the first i The evaluation of the first sample data's static physical parameters and the first j The weighted Mahalanobis distance between the evaluation static physical parameters of the first sample data; Indicates the first i The probability of occurrence of static physical parameters is evaluated for each sample data; and These are the weighting coefficients; in, The physical drift intensity bonus is calculated using the following formula: ; in, The standard deviation of the current within the baseline interval. The degree of charge-discharge hysteresis at the electrode interface. It is the peak symmetry factor; , , These are the weighting coefficients.

[0016] According to the intrinsic drift compensation method for electrochemical sensors provided by the present invention, the construction of multiple test environments includes: determining multiple sets of test conditions based on multiple types of electrochemical sensors, multiple types of electrode adapters, and multiple concentrations of analyte samples; and constructing the multiple test environments based on the multiple sets of test conditions.

[0017] The present invention also provides an intrinsic drift compensation device for an electrochemical sensor, comprising: The first acquisition module is used to acquire electrochemical response signal data generated by the electrochemical sensor during the measurement of the target analyte; the extraction module is used to extract static physical parameters from the electrochemical response signal data; wherein, the static physical parameters are used to quantify the intrinsic characteristic parameters of the electrochemical system in which electrical energy and chemical energy are interconverted when the electrochemical sensor measures the target analyte; the second acquisition module is used to obtain the concentration estimate of the target analyte based on the electrochemical response signal data and the static physical parameters using an intrinsic drift compensation model; wherein, the intrinsic drift compensation model is a machine learning model trained using a training dataset.

[0018] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intrinsic drift compensation method of any of the electrochemical sensors described above.

[0019] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intrinsic drift compensation method for the electrochemical sensor as described above.

[0020] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the intrinsic drift compensation method for any of the electrochemical sensors described above.

[0021] The intrinsic drift compensation method and apparatus for electrochemical sensors provided by this invention acquire electrochemical response signal data when the electrochemical sensor measures a target analyte and extract static physical parameters. This allows for the quantification of the intrinsic characteristic parameters of the electrochemical system, which involves the interconversion of electrical and chemical energy, during the measurement of the target analyte by the electrochemical sensor, providing a crucial basis for accurate sensor performance analysis. Furthermore, by combining the electrochemical response signal data and static physical parameters, an intrinsic drift compensation model trained on a training dataset is used to obtain an estimated concentration of the target analyte. This effectively compensates for the intrinsic drift that may occur during the electrochemical sensor measurement process, improving the accuracy and reliability of the target analyte concentration estimation. Consequently, this provides strong support for the precise measurement and practical application of electrochemical sensors. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a schematic flowchart of the intrinsic drift compensation method for the electrochemical sensor provided by the present invention.

[0024] Figure 2 This is a flowchart illustrating the training method of the intrinsic drift compensation model provided by the present invention.

[0025] Figure 3 This is a schematic diagram illustrating the training and inference processes of the intrinsic drift compensation model provided by this invention.

[0026] Figure 4 This is a schematic diagram of the fusion feature extraction process provided by the present invention.

[0027] Figure 5 This is a schematic diagram of the output layer provided by the present invention.

[0028] Figure 6 This is a schematic diagram of the bidirectional pulse scanning timing electrical signal sequence provided by the present invention.

[0029] Figure 7 A schematic diagram illustrating the process of extracting static physical parameters provided by this invention.

[0030] Figure 8 This is a schematic diagram of the intrinsic drift compensation device for the electrochemical sensor provided by the present invention.

[0031] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0033] The following is combined Figures 1-6 The intrinsic drift compensation method of the electrochemical sensor of the present invention is described.

[0034] Figure 1 This is a schematic flowchart of the intrinsic drift compensation method for the electrochemical sensor provided by the present invention. Figure 1 As shown, the method includes the following: Step 101: Acquire the electrochemical response signal data generated by the electrochemical sensor during the measurement of the target analyte.

[0035] The target analyte is a specific chemical substance or mixture whose concentration needs to be measured, such as a mixture of potassium ferricyanide and potassium ferrocyanide.

[0036] Electrochemical response signal data are signals of current or voltage changes over time generated by the sensor during the measurement process. These signals reflect the electrochemical reactions that occur between the analyte and the sensor surface.

[0037] In practice, an electrochemical workstation can be used to connect to the electrochemical sensor, and appropriate measurement parameters (including scan rate, potential range, etc.) can be set. The measurement program can then be started, and the electrochemical response signal data generated by the sensor during the measurement process can be recorded.

[0038] Step 102: Extract static physical parameters from electrochemical response signal data.

[0039] Static physical parameters are used to quantify the intrinsic characteristics of an electrochemical system in which electrical and chemical energy are interconverted when an electrochemical sensor measures a target analyte.

[0040] In some embodiments, the electrochemical response signal data is as follows: Figure 6 The bidirectional pulse scanning timing electrical signal sequence is shown.

[0041] In the specific implementation process, static physical parameters are obtained through methods such as... Figure 7 The electrochemical mechanism shown was obtained through calculation.

[0042] In some embodiments, the static physical parameters include at least one of the following: differential peak current, peak potential corresponding to the differential peak current, forward peak current, forward peak potential, reverse peak current, reverse peak potential, and electrode interface charge-discharge hysteresis. These parameters are used to characterize the apparent redox and interfacial hysteresis of the electrochemical system. Perform at least one of the following operations: Using a global optimization operator, the bidirectional pulse scan time-series electrical signal sequence is obtained. Extracting differential peak current and the peak potential corresponding to the differential peak current And the peak potential corresponding to the differential peak current. This can be expressed by the following formula: ; ; in,[ [ ] represents the effective potential window range for bidirectional pulse scanning.

[0043] Extract the forward peak current and forward peak potential corresponding to the forward scan from the bidirectional pulse scan timing electrical signal sequence: ; Extract the reverse peak current and reverse peak potential corresponding to the reverse scan from the bidirectional pulse scan timing electrical signal sequence: ; The degree of charge-discharge hysteresis at the electrode interface is determined based on the difference between the forward and reverse peak potentials. The formula is as follows: .

[0044] In some embodiments, the static physical parameters further include at least one of the following: left half-peak potential, right half-peak potential, half-peak width, peak symmetry factor, and mean half-peak potential. These parameters are used to characterize the electron transfer barrier and reversibility of the electrochemical system.

[0045] Perform at least one of the following operations: The left and right half-peak potentials are extracted from the bidirectional pulse scanning timing electrical signal sequence. The left half-peak potential is the peak potential when the differential current reaches half of its peak value at the rising edge of the differential current; the right half-peak potential is the peak potential when the differential current reaches half of its peak value at the falling edge of the differential current.

[0046] From the differential current-potential curve constructed based on the bidirectional pulse scanning time-series electrical signal sequence, the potential of the left half peak and the potential of the right half peak are extracted for the target differential peak. The potential of the left half peak is the potential corresponding to the current reaching half of the peak value of the target differential peak at the rising edge of the target differential peak. The potential of the right half peak is the potential corresponding to the current reaching half of the peak value of the target differential peak at the falling edge of the target differential peak.

[0047] Based on the left half-peak potential and right half peak potential The difference yields the full width at half maximum (FWHM). The formula is as follows: ; in, It depicts the irreversibility of the absolute amplitude drift process caused by electrode surface aging.

[0048] Using the full width at half maximum (FWHM) to determine the differential peak potential and left half peak potential Normalize the difference to obtain the peak symmetry factor. The formula is as follows: ; The closer the value is to 0.5, the stronger the electrochemical reversibility of the system.

[0049] Calculate the left half-peak potential and right half peak potential The mean of the half-peak potential is obtained by taking the average value. The formula is expressed as follows: ; The mean half-peak potential can be used as a thermodynamic intrinsic characterization.

[0050] In some embodiments, the static physical parameters further include at least one of the first derivative maxima and the first derivative minima of the signal function at the rising edge and falling edge of the target difference peak, and the second derivative minima of the signal function. These parameters can be used to characterize the charge transfer rate.

[0051] In the specific implementation process, a signal function can be established based on the bidirectional pulse scanning timing electrical signal sequence. The signal function has voltage as the independent variable and current as the dependent variable; by performing a first-order derivative operation on the signal function, the maximum value of the first-order derivative at the rising edge of the target difference peak is obtained. Minimum of the first derivative at the falling edge , representing the highest current abrupt change rates when the system enters and leaves the electron transfer region, respectively. This is expressed by the following formula: ; ; By performing a second derivative operation on the signal function, we obtain the minimum value of the second derivative. The sharpness of the voltammetric peak and the kinetic focusing effect of the electrochemical reaction are measured based on the minimum value of the second derivative, and the formula is expressed as follows: ; In some embodiments, the static physical parameters also include at least one of the mean current and standard deviation of current within the baseline interval of the bidirectional pulse scan time-series electrical signal sequence, the net peak current, and the peak area within the potential window during the measurement process for the target analyte. These parameters are used to characterize the global charge quantity and noise level.

[0052] In practice, the mean current and standard deviation of current within the baseline interval can be calculated to assess background drift caused by side reactions and conduction resistance.

[0053] Based on peak current and average current The difference yields the net peak current. To purely characterize the reaction of the analyte. The formula is expressed as follows: ; Based on signal function and baseline current function The definite integral of the difference yields the total Faraday charge within the potential window during the measurement process of the target analyte. This refers to the peak area. It can be expressed by the following formula: ; Through the above methods, a total of 19 static physical parameters in four categories can be obtained. These static physical parameters are then cascaded in a structured manner, ultimately constructing an explicit physical feature vector F that possesses prior knowledge of electrochemical drift resistance. ; In the specific implementation process, the bidirectional pulse scanning timing electrical signal sequence and static physical parameters need to be normalized and Z-score standardized before the following operations are performed.

[0054] For a detailed description of the normalization and Z-score standardization of the bidirectional pulse scan timing electrical signal sequence and static physical parameters, please refer to the relevant content in step 201, which will not be repeated here.

[0055] Step 103: Based on the electrochemical response signal data and static physical parameters, the concentration estimate of the target analyte is obtained using the intrinsic drift compensation model.

[0056] The intrinsic drift compensation model is a machine learning model trained using a training dataset. It is used to compensate for the drift component carried by the electrochemical response signal data, thereby obtaining an accurate concentration estimate of the target analyte.

[0057] Drift components are signal baseline shifts or amplitude changes caused by factors such as aging of electrochemical sensors and environmental changes. These changes can interfere with the accurate measurement of analyte concentrations.

[0058] In practice, the intrinsic drift compensation model can be constructed in various ways, without being limited by the description in this manual.

[0059] In some embodiments, such as Figure 3 As shown, the intrinsic drift compensation model includes a feature extraction layer and an output layer. In this embodiment, based on electrochemical response signal data and static physical parameters, the feature extraction layer can be used to obtain fused features; based on the fused features, the output layer can be used to obtain the concentration estimate of the target analyte.

[0060] In some embodiments, the feature extraction layer includes a convolutional neural network, which can be used to extract intrinsic response features from electrochemical response signal data; the intrinsic response features and static physical parameters are then fused to obtain fused features. This is merely an example. Figure 4 As shown, the convolutional neural network consists of three one-dimensional convolutional layers and an average pooling layer.

[0061] In practical implementation, intrinsic response features and static physical parameters can be fused in various ways to obtain fused features. For example, intrinsic response features and static physical parameters can be orthogonally concatenated to obtain fused features in tensor representation. Alternatively, convolutional neural networks can be used to fuse intrinsic response features and static physical parameters to obtain fused features in tensor representation.

[0062] By integrating the fusion features into the output layer, the concentration estimate of the target analyte is obtained.

[0063] In the specific implementation process, such as Figure 5 As shown, when downstream applications require accurate and continuous values ​​of the target analyte, the output layer uses a linear activation function, has 1 output node, and the concentration estimate is a continuous value.

[0064] like Figure 5 As shown, when downstream applications focus on status warnings (such as the "low / medium / high" concentration range), the number of output layer nodes is set to the preset total number of range categories, and the output layer combines the Softmax activation function to output the probability distribution of the concentration of the target analyte belonging to each concentration range.

[0065] For the structure and training method of the output layer of the intrinsic drift compensation model, see [link to relevant documentation]. Figure 2 The relevant content will not be repeated here.

[0066] Figure 2 This is a flowchart illustrating the training method of the intrinsic drift compensation model provided by this invention. Figure 2 As shown, the method includes the following: Step 201: Obtain the training dataset.

[0067] The training dataset is a collection of data used to train the intrinsic drift compensation model, and it includes multiple sets of sample data.

[0068] In the specific implementation process, the training dataset can be constructed as follows: A predetermined number of combined data sets are extracted from a full set of data including various electrochemical sensor types, electrode adapter types, and analyte sample concentrations to obtain multiple sets of test conditions; based on these multiple test conditions, multiple test environments are built to induce different drift effects of the electrochemical sensors under various test conditions; electrochemical response signal data are collected under these multiple test environments to obtain multiple sets of first sample data; static physical parameters are extracted from each set of first sample data; based on an optimization objective function, a target optimization algorithm is used to select target static physical parameters with drift intensity higher than a predetermined threshold from the static physical parameters corresponding to all first sample data; the objective function includes a physical drift intensity reward term, which is used to quantify and score the drift intensity of at least one static physical parameter; second sample data is obtained based on the first sample data corresponding to the target static physical parameters; the known concentration values ​​of the analyte samples corresponding to the second sample data are used as the labels of the second sample data; and the training dataset is obtained based on the second sample data and its corresponding labels.

[0069] Drift intensity is a comprehensive index used to quantify the degree of non-ideal physical disturbance experienced by the electrochemical system, which involves the interconversion of electrical and chemical energy within an electrochemical sensor during measurement.

[0070] The objective function to be optimized is shown below: ; in, To optimize the objective function, Represents the set of the first sample data. Indicates the number of data in the first sample. Indicates the first i Evaluation of static physical parameters of the first sample data Indicates the first j Evaluation of static physical parameters based on the first sample data; For the first iThe evaluation of the first sample data's static physical parameters and the first j The weighted Mahalanobis distance between the evaluation static physical parameters of the first sample data; Indicates the first i The probability of occurrence of static physical parameters is evaluated for each sample data; and These are the weighting coefficients; in, The physical drift intensity bonus is calculated using the following formula: ; in, The standard deviation of the current within the baseline interval. The degree of charge-discharge hysteresis at the electrode interface. It is the peak symmetry factor; , , The weighting coefficients are as follows: The larger the current standard deviation within the baseline interval, the more severe the charge / discharge hysteresis at the electrode interface, and the further the peak symmetry factor deviates from the ideal symmetry value of 0.5, the higher the physical drift intensity score. This is achieved by introducing hyperparameters. A reward will be given.

[0071] pass This can drive algorithms to find rare samples with extremely low probability of occurrence; through This allows the selected test combination to maintain maximum repulsion in terms of electrochemical topological characteristics; through It can drive algorithms to find extreme electrochemical physical conditions with severe degradation.

[0072] In the specific implementation process, under each test environment, the following operations are performed for a pre-set number of cycles to obtain multiple sets of first sample data under each test environment: the electrochemical sensor is placed in the analyte sample, and the electrochemical sensor is connected to the electrochemical workstation through the electrode adapter; the electrochemical response signal data of the electrochemical sensor is collected through the electrochemical workstation; and a set of first sample data is obtained based on the collected electrochemical response signal data.

[0073] The analyte sample can be a specific chemical substance or a mixture. In some embodiments, the analyte sample includes a mixed solution prepared by uniformly mixing potassium ferrocyanide, potassium ferrocyanide, and a buffer solution.

[0074] In practice, the number of cycles can be determined through experience or experimentation, for example, 20 times.

[0075] In the specific implementation process, information on electrochemical sensor types can be collected: identify all available electrochemical sensor types, such as current-type sensors and potential-type sensors, and assign a unique identification number to each sensor type.

[0076] Collect electrode adapter type information: Organize different types of electrode adapters, such as adapters with different interface specifications and different signal amplification factors, and assign a number to each adapter type.

[0077] Collect analyte sample concentration information: Determine multiple concentration values ​​for analyte samples. For example, for a potassium ferricyanide-potassium ferrocyanide mixed solution, set a series of different concentration gradients, such as 0.1mM, 0.5mM, 1mM, etc., and number each concentration value.

[0078] Using combinatorial mathematics, the types of electrochemical sensors, electrode adapters, and analyte sample concentrations are fully combined to obtain the complete combination data. Assuming there are m types of electrochemical sensors, n types of electrode adapters, and k types of analyte sample concentrations, the total number of test conditions for the complete combination is m × n × k.

[0079] Each set of combined data is treated as a test condition, and the corresponding sensor type number, adapter type number, and concentration number are recorded.

[0080] In the specific implementation process, a preset number can be set according to experience or algorithm requirements. For example, if the full set of data includes 1,000 sets of combined data, 50 sets can be randomly selected from them to obtain multiple sets of test conditions.

[0081] In some embodiments, the static physical parameters extracted from each set of first sample data include at least one of the following: current standard deviation within the baseline interval, electrode interface charge / discharge hysteresis, and peak symmetry factor. In specific implementations, various objective optimization algorithms can be employed, without being limited by the description herein.

[0082] In some embodiments, simulated annealing can be used as the target optimization algorithm: To optimize the target, a simulated annealing algorithm is used for optimization. Using the degradation of static physical parameters such as baseline distortion and half-peak broadening caused by aging of electrochemical sensors and hardware interference as a compass, the most valuable orthogonal physical boundary combinations for "anti-drift" training are precisely extracted.

[0083] In the embodiments provided by this invention, a physical constraint boundary sampling algorithm based on maximizing information entropy is used to selectively extract specific combinations of data from a full range of data including various electrochemical sensor types, electrode adapter types, and analyte sample concentrations. This avoids the problem of blindly and redundantly exhaustively enumerating physical experiments in traditional random sampling, effectively reducing the cost of real data acquisition. The algorithm deeply integrates the extracted static physical parameters into the optimization mechanism, accurately selecting target static physical parameters whose drift intensity exceeds a preset threshold. This results in a second sample dataset containing combinations of severe test conditions and the corresponding training dataset. With a minimal number of carefully selected experimental groups, the model built based on this training dataset ensures strong robustness and accurate compensation capabilities.

[0084] Prepare the corresponding experimental environment according to the requirements of each set of test conditions, and build multiple sets of test environments to conduct orthogonal experiments with multi-source perturbations: Equipment Connection and Configuration: Connect the electrochemical sensor to the electrochemical workstation using the appropriate electrode adapter specified in the test conditions. Set the parameters relevant to the test conditions on the electrochemical workstation, such as scan rate and pulse amplitude.

[0085] Analyte sample preparation: According to the analyte sample concentration specified in the test conditions, accurately prepare the corresponding mixed solution (such as a mixed solution prepared by uniformly mixing potassium ferricyanide, potassium ferrocyanide and buffer solution), and ensure the homogeneity and stability of the sample.

[0086] For any given measurement, the acquired bidirectional pulse scan timing electrical signal sequence From a physical mechanism perspective, it can be decoupled into the following nonlinear superposition model: ; in, For concentration Ideal intrinsic response signal under driving; For the sensor Device-level spatial perturbation drift introduced by manufacturing deviations; For adapter Switching introduces hardware interface inconsistency drift; To follow the test loop Accumulated non-stationary time-varying drift error (such as sensor aging); This is due to environmental noise drift.

[0087] In the embodiments provided by this invention, by constructing the aforementioned various test environments, multi-source systematic drift in real test environments can be comprehensively captured and decoupled, resulting in a high-dimensional orthogonal time-series dataset D containing multi-dimensional real perturbations. This dataset covers multi-source time-varying features such as sensor manufacturing deviations, adapter hardware inconsistencies, and long-term cyclic aging, and its mathematical definition is as follows: ; in, This represents a bidirectional pulse scan timing electrical signal sequence acquired from a single measurement. Number the sensor. The total number of independent sensors (in this embodiment) ), used to characterize cross-device differences; To detect the adapter number, Total number of adapters (in this embodiment) ), used to characterize the discreteness of hardware contact and transmission; For the concentration gradient of the target analyte, Number of concentration nodes (in this embodiment) This provides a baseline true value for nonlinear fitting; To continuously test the number of loops, The total number of cycles (in this embodiment) ), used to characterize time-varying disturbances such as electrode aging.

[0088] The high-dimensional orthogonal time-series dataset D constructed in this invention has a size of (A total of 1000 real multi-source disturbance samples) cover the systematic drift of sensor cross-device differences and time-varying disturbances, laying the physical data foundation for subsequent feature extraction models.

[0089] In the specific implementation process, for each set of bidirectional pulse scanning time-series electrical signal sequences in the high-dimensional orthogonal time-series dataset D, static physical parameters are extracted from the bidirectional pulse scanning time-series electrical signal sequences using the method in step 102. The bidirectional pulse scanning time-series electrical signal sequences and their corresponding static physical parameters are used as a set of sample data, and the known concentration values ​​of the analyte samples are used as the labels of the sample data, thereby constructing a training dataset containing multiple sets of sample data.

[0090] To meet the requirements of end-to-end batch training of deep learning networks, this embodiment of the invention performs global structured integration of sample data in the training dataset. The global fusion feature tensor and the corresponding global orthogonal truth label vector are defined as follows: ; in: : A global fusion feature tensor for input to a deep learning model; This represents the total number of samples in the training dataset. The total dimension of features for a single sample data (where The length of the data points in the bidirectional pulse scan timing electrical signal sequence. (for the static physical parameter dimension).

[0091] : for the first i (Values ​​range from 1 to...) N ( ) sample data. This vector is essentially a single measurement representation of a specific multi-source perturbation state p coupled with the true value c of the target analyte.

[0092] This is a global label vector, the data type of which depends on the downstream task configuration. When configured for quantitative evaluation mode, the vector elements belong to the set of real numbers. When configured in qualitative state mode, the vector elements belong to the set of natural numbers. (Representation category index).

[0093] : for the first i (Values ​​range from 1 to...) N The true value of the orthogonal experiment corresponding to each sample, i.e., the true concentration value of the analyte sample, is compared with... This constitutes a strict one-to-one mapping and supervision relationship.

[0094] In the specific implementation process, the original acquisition files generated by the multi-source perturbation orthogonal experiment are automatically parsed, and redundant index columns such as equipment serial numbers and timestamps are accurately removed. Secondly, the bidirectional pulse scan time-series electrical signal sequence and static physical parameters obtained by electrochemical mechanism calculation are extracted simultaneously, and features are concatenated at the data channel level to form a single-sample fusion vector. Simultaneously, based on the experimental orthogonal design matrix, pre-calibrated orthogonal experimental truth values ​​are uniformly assigned to all test cycle samples within the current file. Finally, tensor vertical stacking operations are performed along the batch dimension to merge all discrete sample vectors with labels, ultimately generating a global feature matrix suitable for GPU parallel computing. With the corresponding label vector This completes the structured closed-loop construction of the model composite training dataset.

[0095] In the specific implementation process, it is necessary to perform feature standardization processing on the sample data in the training dataset established above, as follows.

[0096] Since the dimensions and orders of magnitude of current, potential, and various static physical parameters in electrochemical response signal data vary greatly, it is necessary to normalize and Z-score standardize the above data to ensure that the deep network can converge smoothly.

[0097] In the specific implementation process, the global feature tensor can be... Perform a double normalization mapping along the feature dimension (column direction). Let... For tensor The Middle i The first sample j The original values ​​of each feature dimension, where , .

[0098] Determine the physical extreme value benchmark for a single feature channel in the tensor across all sample distributions, and calculate the mapping ratio between any data point in that feature column and the benchmark value. Normalized transition eigenvalues. The calculation is as follows: ; in, This represents the maximum response value of the j-th feature dimension in the N samples of the global training set. This step effectively compresses the absolute values ​​with different physical meanings to a reference space of approximately [0,1].

[0099] Building upon extremum normalization, the Z-Score algorithm is further used to reconstruct the feature tensor using an independent channel-level normal distribution. This ensures that the features across all dimensions not only have uniform dimensions but also variances of the same scale, and their standardized eigenvalues... The calculation is as follows: ; in, and Let be the mean and standard deviation of the j-th feature dimension after the first-order normalization, respectively. After the above two-order mapping, the final output is the standardized feature value. Strictly following a standard normal distribution with a mean of 0 and a standard deviation of 1, the standard input tensor is reconstructed for direct ingestion by a one-dimensional convolutional neural network (1D-CNN).

[0100] Step 202: Using the initial intrinsic drift compensation model, compensate for the drift component carried by the sample data to obtain the predicted concentration value of the analyte sample.

[0101] In practice, sample data from the training dataset can be input into the initial intrinsic drift compensation model, which then outputs the predicted concentration of the analyte sample.

[0102] Step 203: Calculate the difference between the concentration prediction value and the label of the sample data using the loss function.

[0103] Step 204: Adjust the parameters of the initial intrinsic drift compensation model using a preset optimization algorithm to reduce the difference value until the training termination condition is met, and obtain the trained intrinsic drift compensation model.

[0104] The preset optimization algorithms include, but are not limited to, gradient descent and Adam optimizer.

[0105] By adjusting the parameters of the initial intrinsic drift compensation model using a preset optimization algorithm, the difference between the concentration prediction value and the label of the sample data can be gradually reduced until the training termination condition is met. The model obtained at this time is the trained intrinsic drift compensation model, which has good prediction performance and stability.

[0106] Training termination conditions may include the convergence of the intrinsic drift compensation model, reaching a preset number of iterations, etc.

[0107] In some embodiments, when downstream applications require precise continuous values ​​for the target analyte, the output layer of the intrinsic drift compensation model employs a linear activation function (Linear) with only one output node, directly outputting continuous estimates of the target feature parameters (such as concentration). Correspondingly, the model training uses the mean squared error loss function (MSE) to calculate the Euclidean distance between the predicted value and the standard experimental true value. Through backpropagation and iteration with an optimizer (such as Adam), the absolute prediction error is gradually reduced, achieving precise quantitative compensation across the entire measurement range.

[0108] In this mode, the output layer uses a linear activation function to directly map the intrinsic response feature tensor extracted from the front end to the target feature parameters. For the i-th sample in batch training (Batch Size = N), its predicted value is... The calculation formula is: ; in, The fused features output by the feature extraction layer are input to the i-th sample data; : The continuous estimate of the i-th sample data output by the model; It is a linear activation function; Weight matrix W Transpose of; b This is a bias term.

[0109] In this mode, the loss function model training uses mean squared error (MSE) as the optimization objective, and drives the iteration of network parameters by calculating the Euclidean distance between the predicted values ​​and the true values ​​of orthogonal experiments. ; This represents the number of samples in each training iteration. The formula represents the pre-calibrated true value (e.g., standard concentration value) of the orthogonal experiment for the i-th sample; this formula forces the network to eliminate the influence of the perturbation state p in the high-dimensional space, making... infinitely close ; Let be the mean squared error loss function.

[0110] In some embodiments, when the downstream application focuses on state warning (e.g., concentration "low / medium / high" intervals), the number of output nodes of the intrinsic drift compensation model is set to the preset total number of interval categories, and the output layer combines the Softmax activation function to output the probability distribution of the current sample belonging to each concentration interval.

[0111] Correspondingly, the model training employs a multi-class cross-entropy loss function, which measures the difference between the probability distribution of the model output and the true values ​​of the orthogonal experiments (One-Hot encoded labels). This cross-entropy loss is used as the core optimization objective, and gradient descent guides parameter updates to achieve implicit filtering of perturbation drift and accurate classification of the target state.

[0112] In this mode, the output layer contains K nodes, and the Softmax activation function is used to convert the original feature map scores generated by the network into a normalized probability distribution. For the i-th sample, it belongs to the k-th state interval ( The predicted probability is: ; in, The score of the unnormalized linear mapping at the k-th output node of the output layer of the intrinsic drift compensation model (which is determined by...) (Calculated) : To predict the probability value of the i-th sample belonging to the k-th discrete state in the network ( =1).

[0113] In this mode, model training uses a multi-class cross-entropy loss function to measure the information divergence between the probability distribution of the model output and the true value distribution of orthogonal experiments. ; in, This represents the global cross-entropy loss of a deep learning model; Batch Size is the number of samples in each training iteration. The total number of categories of characteristic parameters for the target analyte (e.g., the number of preset concentration gradients). Let be the true label indicator variable. When the true value of the orthogonal experiment of the i-th sample belongs to the k-th class, ,otherwise ; During the model parameter optimization phase, the Adam (Adaptive Moment Estimation) optimizer can be used to optimize the aforementioned loss function. Perform gradient descent and weight update.

[0114] The Adam optimizer uses an adaptive learning rate that dynamically calculates gradients, combining first-order momentum (accumulating historical gradients to accelerate convergence) with second-order momentum (adjusting parameter update step size to smooth out fluctuations). During iterative training, the Adam optimizer aims to minimize the loss. To achieve this goal, the backpropagation algorithm is used to simultaneously update the convolutional kernel weights in the feature extraction module and the fully connected layer parameters in the intrinsic drift compensation model. This joint optimization process enables the model to adaptively filter out multi-source uncontrollable perturbations in the high-dimensional manifold space, ultimately establishing a highly robust nonlinear mapping relationship between intrinsic response features and orthogonal experimental ground truth.

[0115] Figure 8 This is a schematic diagram of the intrinsic drift compensation device for the electrochemical sensor provided by the present invention. Figure 8 As shown, the intrinsic drift compensation device 800 of the electrochemical sensor includes the following modules: The first acquisition module 810 is used to acquire electrochemical response signal data generated by the electrochemical sensor during the measurement of the target analyte.

[0116] Extraction module 820 is used to extract static physical parameters from the electrochemical response signal data; wherein, the static physical parameters are used to quantify the intrinsic characteristic parameters of the electrochemical system in which electrical energy and chemical energy are interconverted when the electrochemical sensor measures the target analyte.

[0117] The second acquisition module 830 is used to obtain the concentration estimate of the target analyte based on the electrochemical response signal data and the static physical parameters using an intrinsic drift compensation model.

[0118] The intrinsic drift compensation model is a machine learning model trained using a training dataset.

[0119] Figure 9 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 9As shown, the electronic device may include a processor 910, a communication interface 920, a memory 930, and a communication bus 940, wherein the processor 910, the communication interface 920, and the memory 930 communicate with each other via the communication bus 940. The processor 910 can call logic instructions in the memory 930 to execute an intrinsic drift compensation method for the electrochemical sensor. This method includes: acquiring electrochemical response signal data generated by the electrochemical sensor during the measurement of a target analyte; extracting static physical parameters from the electrochemical response signal data; wherein the static physical parameters are used to quantify the intrinsic characteristic parameters of the electrochemical system in which electrical energy and chemical energy are interconverted when the electrochemical sensor measures the target analyte; and obtaining an estimated concentration of the target analyte using an intrinsic drift compensation model based on the electrochemical response signal data and the static physical parameters; wherein the intrinsic drift compensation model is a machine learning model trained using a training dataset.

[0120] Furthermore, the logical instructions in the aforementioned memory 930 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0121] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the intrinsic drift compensation method for the electrochemical sensor provided by the above methods. The method includes: acquiring electrochemical response signal data generated by the electrochemical sensor during the measurement of a target analyte; extracting static physical parameters from the electrochemical response signal data; wherein the static physical parameters are used to quantify the intrinsic characteristic parameters of the electrochemical system in which electrical energy and chemical energy are interconverted when the electrochemical sensor measures the target analyte; and obtaining an estimated concentration of the target analyte using an intrinsic drift compensation model based on the electrochemical response signal data and the static physical parameters; wherein the intrinsic drift compensation model is a machine learning model trained using a training dataset.

[0122] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements an intrinsic drift compensation method for an electrochemical sensor provided by the methods described above. This method includes: acquiring electrochemical response signal data generated by the electrochemical sensor during measurement of a target analyte; extracting static physical parameters from the electrochemical response signal data; wherein the static physical parameters are used to quantify the intrinsic characteristic parameters of the electrochemical system in which electrical energy and chemical energy are interconverted when the electrochemical sensor measures the target analyte; and obtaining an estimated concentration of the target analyte using an intrinsic drift compensation model based on the electrochemical response signal data and the static physical parameters; wherein the intrinsic drift compensation model is a machine learning model trained using a training dataset.

[0123] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0124] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for compensating the intrinsic drift of an electrochemical sensor, characterized in that, include: Acquire electrochemical response signal data generated by the electrochemical sensor during the measurement of the target analyte; Static physical parameters are extracted from the electrochemical response signal data; wherein, the static physical parameters are used to quantify the intrinsic characteristic parameters of the electrochemical system in which electrical energy and chemical energy are interconverted when the electrochemical sensor measures the target analyte; Based on the electrochemical response signal data and the static physical parameters, the concentration estimate of the target analyte is obtained using the intrinsic drift compensation model; The intrinsic drift compensation model is a machine learning model trained using a training dataset; the training dataset is obtained in the following way: Multiple sets of test conditions are obtained by extracting a predetermined number of combined data from a full combination of data of various electrochemical sensor types, various electrode adapter types and various analyte sample concentrations; Based on the aforementioned multiple sets of test conditions, multiple sets of test environments were constructed to induce different drift effects of the electrochemical sensor under various test environments. Under the aforementioned multiple test environments, electrochemical response signal data were collected to obtain multiple sets of first sample data. Extract static physical parameters from the first sample data of each group; Based on the objective function, a target optimization algorithm is used to select target static physical parameters with drift intensity higher than a preset threshold from all static physical parameters corresponding to the first sample data; wherein, the objective function includes a physical drift intensity reward term, which is used to quantify and score the drift intensity of at least one static physical parameter; Based on the first sample data corresponding to the target static physical parameters, the second sample data is obtained; Use the known concentration value of the analyte sample corresponding to the second sample data as the label of the second sample data; The training dataset is obtained based on the second sample data and its corresponding label.

2. The intrinsic drift compensation method for electrochemical sensors according to claim 1, characterized in that, The intrinsic drift compensation model includes a feature extraction layer and an output layer; The step of obtaining the concentration estimate of the target analyte based on the electrochemical response signal data and the static physical parameters using an intrinsic drift compensation model includes: Based on the electrochemical response signal data and the static physical parameters, a fused feature is obtained using a feature extraction layer; Based on the fusion characteristics, the concentration estimate of the target analyte is obtained using the output layer.

3. The intrinsic drift compensation method for electrochemical sensors according to claim 2, characterized in that, The step of obtaining fused features based on the electrochemical response signal data and the static physical parameters using a feature extraction layer includes: Intrinsic response features are extracted from the electrochemical response signal data using a convolutional neural network. The fused feature is obtained by fusing the intrinsic response characteristics and the static physical parameters.

4. The intrinsic drift compensation method for an electrochemical sensor according to any one of claims 1 to 3, characterized in that, The electrochemical response signal data is a bidirectional pulse scan time-series electrical signal sequence; the static physical parameters include at least one of the following: differential peak current, peak potential corresponding to the differential peak current, forward peak current, forward peak potential, reverse peak current, reverse peak potential, and electrode interface charge-discharge hysteresis. The extraction of static physical parameters from the electrochemical response signal data includes: Perform at least one of the following operations: The differential peak current and the peak potential corresponding to the differential peak current are extracted from the bidirectional pulse scan timing electrical signal sequence using a global optimization operator. Extract the forward peak current and the forward peak potential corresponding to the forward scan from the bidirectional pulse scan timing electrical signal sequence; Extract the reverse peak current and the reverse peak potential corresponding to the reverse scan from the bidirectional pulse scan timing electrical signal sequence; The degree of charge / discharge hysteresis at the electrode interface is determined based on the difference between the forward peak potential and the reverse peak potential.

5. The intrinsic drift compensation method for an electrochemical sensor according to claim 4, characterized in that, The static physical parameters also include at least one of the following: left half-peak potential, right half-peak potential, half-peak width, peak symmetry factor, and half-peak potential mean. The extraction of static physical parameters from the electrochemical response signal data includes: Perform at least one of the following operations: The left half-peak potential and the right half-peak potential are extracted from the bidirectional pulse scanning timing electrical signal sequence; wherein, the left half-peak potential is the peak potential when the current reaches half of the peak value at the rising edge of the differential current; and the right half-peak potential is the peak potential when the current reaches half of the peak value at the falling edge of the differential current. From the differential current-potential curve constructed based on the bidirectional pulse scanning timing electrical signal sequence, the left half-peak potential and the right half-peak potential are extracted for the target differential peak; wherein, the left half-peak potential is the potential corresponding to the current reaching half of the peak value of the target differential peak at the rising edge of the target differential peak; and the right half-peak potential is the potential corresponding to the current reaching half of the peak value of the target differential peak at the falling edge of the target differential peak. The half-peak width is obtained based on the difference between the potential of the left half-peak and the potential of the right half-peak. Using the half-peak width, the difference between the differential peak potential and the left half-peak potential is normalized to obtain the peak symmetry factor; The average value of the left half-peak potential and the right half-peak potential is calculated to obtain the average half-peak potential.

6. The intrinsic drift compensation method for an electrochemical sensor according to claim 5, characterized in that, The static physical parameters also include at least one of the following: the maximum value of the first derivative of the signal function at the rising edge and the minimum value of the first derivative at the falling edge of the target difference peak; and the minimum value of the second derivative of the signal function. The extraction of static physical parameters from the electrochemical response signal data includes: Based on the bidirectional pulse scanning timing electrical signal sequence, the signal function is established; the signal function uses voltage as the independent variable and current as the dependent variable. Perform a first-order derivative operation on the signal function to obtain the first-order derivative maxima at the rising edge and the first-order derivative minima at the falling edge of the target difference peak; Perform a second-order derivative operation on the signal function to obtain the minimum value of the second derivative.

7. The intrinsic drift compensation method for an electrochemical sensor according to claim 6, characterized in that, The static physical parameters also include at least one of the mean current and standard deviation of current in the baseline interval of the bidirectional pulse scan time-series electrical signal sequence, net peak current, and peak area in the potential window during the measurement process of the target analyte. The extraction of static physical parameters from the electrochemical response signal data includes: Calculate the mean current and the standard deviation of the current within the baseline interval; The net peak current is obtained based on the difference between the peak current and the average current. Based on the definite integral of the difference between the signal function and the baseline current function, the peak area within the potential window during the measurement process of the target analyte is obtained.

8. The intrinsic drift compensation method for an electrochemical sensor according to claim 1, characterized in that, The process involves collecting electrochemical response signal data under multiple test environments to obtain multiple sets of first sample data, including: In each test environment, the following operation is performed for a pre-set number of loops to obtain multiple sets of first sample data for each test environment: An electrochemical sensor is placed in the analyte sample, and the electrochemical sensor is connected to an electrochemical workstation via an electrode adapter. The electrochemical response signal data of the electrochemical sensor are acquired through the electrochemical workstation; Based on the collected electrochemical response signal data, a set of first sample data was obtained.

9. The intrinsic drift compensation method for an electrochemical sensor according to claim 1, characterized in that, The optimization objective function is as follows: ； in, To optimize the objective function, This represents the set of the first sample data. Indicates the number of data in the first sample. Indicates the first i Evaluation of static physical parameters of the first sample data Indicates the first j Evaluation of static physical parameters based on the first sample data; For the first i The evaluation of the first sample data's static physical parameters and the first j The weighted Mahalanobis distance between the evaluation static physical parameters of the first sample data; Indicates the first i The probability of occurrence of static physical parameters is evaluated for each sample data; and These are the weighting coefficients; in, The physical drift intensity bonus is calculated using the following formula: ； in, The standard deviation of the current within the baseline interval. The degree of charge-discharge hysteresis at the electrode interface. It is the peak symmetry factor; , , These are the weighting coefficients.

10. An intrinsic drift compensation device for an electrochemical sensor, characterized in that, include: The first acquisition module is used to acquire electrochemical response signal data generated by the electrochemical sensor during the measurement of the target analyte; An extraction module is used to extract static physical parameters from the electrochemical response signal data; wherein, the static physical parameters are used to quantify the intrinsic characteristic parameters of the electrochemical system in which electrical energy and chemical energy are interconverted when the electrochemical sensor measures the target analyte; The second acquisition module is used to obtain the concentration estimate of the target analyte based on the electrochemical response signal data and the static physical parameters using an intrinsic drift compensation model. The intrinsic drift compensation model is a machine learning model trained using a training dataset.

Images