Method and system for improving precision of voltage and current acquisition module
Through the deep learning-based neural network model, the nonlinear relationship between voltage, current and temperature is modeled, and the problem of insufficient accuracy of the voltage and current acquisition module is solved, higher accuracy and reliability are achieved, and equipment fault detection capabilities are enhanced.
Patent Information
- Application Number
- CN202411507558.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-28
- Publication Date
- 2025-03-11
AI Technical Summary
The insufficient accuracy of the voltage and current acquisition module has affected the accuracy of ship energy management, power system control and equipment monitoring, and it is impossible to accurately detect equipment failures and abnormal conditions.
A neural network model based on deep learning is adopted to model the nonlinear relationship between voltage, current and temperature through the architecture of the input layer, hidden layer and output layer, combined with temperature data, and perform accuracy correction.
It improves the accuracy and reliability of the voltage and current acquisition module, can more accurately detect equipment failures and abnormal conditions, and enhances the stability and performance of the system.
Smart Images
Figure CN119595986A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of electronic technology, and in particular to a method and system for improving voltage and current acquisition accuracy. Background Art
[0002] Voltage and current acquisition modules are often used in the field of marine power system monitoring and control, energy management and optimization, etc. Due to the error coupling of components such as resistors, capacitors, amplifiers, and ADC analog-to-digital converters in the voltage and current acquisition modules, there may be a large deviation between the collected data and the true value.
[0003] The insufficient accuracy of the voltage and current acquisition module may affect the accuracy of ship energy management, power system control and equipment monitoring, thereby affecting the performance and efficiency of the ship. At the same time, the acquisition module with insufficient accuracy may not accurately detect equipment failures and abnormal conditions, causing further deterioration of the failure or damage to the equipment. Therefore, it is necessary to improve the accuracy of the voltage and current acquisition module.
[0004] In the prior art, the voltage and current acquisition module generally uses a piecewise linear correction method to improve the acquisition accuracy. The piecewise linear correction method usually assumes that the nonlinear error of the acquisition system can be approximated by multiple linear segments. However, the actual nonlinear error may be more complex and cannot be completely described by a linear model. Therefore, in some cases, the piecewise linear correction method may not provide sufficiently accurate correction. In addition, the piecewise linear correction method requires the selection of appropriate linear segments to establish a correction model. This involves determining the number of segments, the location of the segment points, and the slope and intercept of each segment. Improperly selected segments may result in poor correction effects and fail to effectively reduce system errors.
[0005] Neural networks have significant application potential and advantages in power systems, especially in voltage and current acquisition. Neural networks can approximate complex nonlinear relationships and have strong noise suppression capabilities to improve acquisition accuracy. Neural networks can process a large amount of voltage and current data in real time, and can automatically adjust calibration parameters by learning the characteristics of input data to achieve automatic calibration. They can also predict future voltage and current changes based on historical data to improve acquisition efficiency. Therefore, using neural networks to improve the acquisition accuracy of voltage and current acquisition modules is creative. Summary of the invention
[0006] The present invention aims to overcome the deficiencies of the prior art and provide a method and system for improving the accuracy of a voltage and current acquisition module.
[0007] In order to achieve the above object, the present invention provides the following technical solutions:
[0008] Data preparation:
[0009] The present invention uses a temperature sensor to obtain temperature data of a voltage and current signal acquisition module. At the same time, a high-precision signal source is used to generate a series of voltage and current signals, and the signals are input into the voltage and current signal acquisition module. The voltage and current signal source setting value here is the theoretical value of the voltage and current signal, and the output value of the voltage and current signal acquisition module is the sampling value of the voltage and current signal. The above sampling values are integrated into a data set, and the integrated original data set is preprocessed, including removing outliers, filtering and data normalization, etc., to improve the quality and reliability of the data.
[0010] Neural network model architecture design:
[0011] Design a neural network model architecture based on deep learning, including input layer, hidden layer and output layer. The input layer includes voltage, current and temperature feature sampling values, and the output layer is the theoretical value of voltage and current. Determine the appropriate number of neurons and connection mode, which can be adjusted according to the size of the data set. Select an appropriate activation function (such as ReLU, sigmoid or tanh) to introduce nonlinear characteristics to increase the expressive power of the neural network.
[0012] Data partitioning:
[0013] Furthermore, the data set is divided into training set, validation set and test set by cross-validation. The training set is used for the parameter update and training process of the neural network; the validation set is used to evaluate the performance of the model and make adjustments to prevent overfitting or underfitting; the test set is used to evaluate the performance of the trained, validated and tuned neural network model.
[0014] Network training:
[0015] The network is trained using the training set data, and the network weights and biases are updated through the back propagation algorithm, so that the neural network can learn the complex mapping relationship between the input signal and the output signal. The appropriate loss function and optimization algorithm are selected to minimize the difference between the predicted output and the theoretical value to improve the accuracy and stability of the model.
[0016] Model validation and tuning:
[0017] Use the validation set data to evaluate the performance of the model, monitor the model training process and make adjustments to prevent overfitting or underfitting. Tune the model parameters based on the results of the validation set, including adjusting the learning rate, adding regularization terms, or increasing the number of training rounds, to further improve the performance of the model.
[0018] Model evaluation and testing:
[0019] Use the test set data to evaluate the trained and tuned neural network model and calculate the error indicators between the predicted value and the theoretical value, such as the root mean square error (RMSE) or the mean absolute error (MAE). Based on the evaluation results, the structure and parameters of the model can be further adjusted to improve the performance and adaptability of the model.
[0020] Application and deployment:
[0021] The tested neural network model is applied to the actual voltage and current acquisition module to correct the actual value and provide more accurate estimation results. The new voltage, current and temperature signal sampling values are input into the tested neural network model to obtain the corrected voltage and current signal prediction values. Through the calculation and correction of the neural network model, a more accurate output signal can be obtained, which improves the accuracy and reliability of the voltage and current signal acquisition module.
[0022] Compared with the prior art, the present invention has the following beneficial effects:
[0023] 1. Nonlinear relationship modeling: Neural networks can model complex nonlinear relationships, including error coupling relationships between components such as resistors, capacitors, amplifiers, ADC analog-to-digital converters, and nonlinear relationships between voltage, current, and temperature.
[0024] 2. Adaptability: The correction method based on neural network is adaptive and can automatically adjust the correction model according to data at different temperatures.
[0025] 3. Generalization ability: The neural network model has good generalization ability, can learn general rules from limited training samples, and make accurate predictions on unseen data.
[0026] 4. Processing of system complexity: The neural network model can handle multiple factors and parameters in the system and automatically optimize weights and biases during model training. For example, temperature is one of the important factors affecting sensor accuracy. The neural network can perform temperature compensation on the output of the current sensor by fusing the data of the temperature sensor and the current sensor. This enables it to better adapt to complex circuit characteristics and temperature effects, providing more accurate correction results.
[0027] 5. In summary, the present invention uses a neural network to establish an accuracy correction model that takes temperature influence into account, realizes correction and compensation of the collected signal, and improves the accuracy and stability of the voltage and current acquisition module. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 Schematic diagram of the precision calibration system
[0029] Figure 2 Data preparation unit diagram
[0030] Figure 3 Schematic diagram of neural network building blocks
[0031] Figure 4 Schematic diagram of precision calibration unit
[0032] Figure 5 Schematic diagram of the process flow of the precision correction method DETAILED DESCRIPTION
[0033] The present invention is described in detail below with reference to the accompanying drawings and embodiments.
[0034] The present invention provides a voltage and current acquisition module accuracy correction system, which specifically comprises the following components: Figure 1 As shown. For ease of explanation, only the parts related to the present invention are shown. The system can be a software unit, a hardware unit or a combination of software and hardware units built into an electronic device. The system includes: a data preparation unit, a neural network construction unit and an accuracy correction unit. It is worth noting that in the above system embodiment, the various units included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.
[0035] The precision correction method implementation process is as follows Figure 2 As shown, connect the standard signal source and send the standard signal.
[0036] Data preparation unit: e.g. Figure 2 As shown, the present invention uses a temperature sensor to obtain temperature data of a voltage and current signal acquisition module. At the same time, a high-precision signal source is used to generate a series of voltage and current signals, and input into the voltage and current signal acquisition module. The voltage and current signal source setting value here is the theoretical value of the voltage and current signal, and the output value of the voltage and current signal acquisition module is the sampling value of the voltage and current signal. The above sampling values are integrated into a data set, and the integrated original data set is preprocessed, including removing outliers, filtering and data normalization, etc., to improve the quality and reliability of the data.
[0037] Among them, outliers can be removed by identifying them through statistical analysis and other methods, and then removed; filtering can adopt nonlinear filtering methods, such as median filtering, to smooth the data by calculating the median of the data; data normalization is shown in formula (1):
[0038]
[0039] x' represents the eigenvalue after data normalization preprocessing; x represents the original eigenvalue; min(x) represents the minimum value of the original eigenvalue, and max(x) represents the maximum value of the eigenvalue.
[0040] Neural network building blocks: e.g. Figure 3 As shown in the figure, the neural network model architecture based on deep learning is designed, including input layer, hidden layer and output layer. The input layer includes the sampled values of voltage, current and temperature. i (i=1, 2, ...n) represents input features;
[0041] The hidden layer needs to determine the appropriate number of neurons and the way they are connected, which can be adjusted according to the size of the data set. Select an appropriate activation function, such as ReLU, sigmoid, or tanh, to introduce nonlinear characteristics. As shown in formula (2):
[0042]
[0043] The output layer is the theoretical value of the voltage and current signal. The collected data samples are used as the training set and trained through the neural network. The difference between the predicted output and the set value is used as the target output, and the network parameters are adjusted to reduce the error between the predicted output and the set value. During the training process, indicators such as the root mean square error (RMSE) or the mean absolute error (MAE) can be used to evaluate the performance of the model and perform model tuning.
[0044] A feedforward neural network (FFNN) can process complex data relationships and has a high noise tolerance. Therefore, a feedforward neural network is selected to construct the voltage and current prediction model in this embodiment.
[0045] The specific neural network model architecture design is as follows:
[0046] 1. Input layer design
[0047] Input characteristics: voltage, current and temperature characteristic sampling values.
[0048] Number of neurons: 3, corresponding to the three input features mentioned above.
[0049] 2. Hidden Layer Design
[0050] Number of hidden layers:
[0051] According to the complexity of the problem and the scale of the data, the appropriate number of hidden layers is selected. In this embodiment, two hidden layers are used.
[0052] Number of neurons:
[0053] The number of neurons in each hidden layer is determined based on experience or experimentation. The first hidden layer can select more neurons to capture the complex features of the input data. The second hidden layer can select fewer neurons to further refine and compress the features.
[0054] In this embodiment, the first hidden layer has 100 neurons and the second hidden layer has 50 neurons.
[0055] Activation function:
[0056] A nonlinear activation function is selected to increase the nonlinear capability of the model. In this embodiment, ReLU is selected as the activation function because it has the advantages of fast calculation speed and is not prone to gradient disappearance.
[0057] 3. Output layer design
[0058] Output features: predicted voltage value and predicted current value.
[0059] Number of neurons: 2, corresponding to the predicted voltage and current values respectively.
[0060] Activation function: In this embodiment, the activation function of the output layer is usually a linear activation function (ie, no activation function is applied), because the linear activation function can retain the output value of the neuron, so that the output of the model can take any real number.
[0061] 4. Loss Function and Optimizer
[0062] Loss function: In this embodiment, mean square error (MSE) is selected as the loss function. MSE can measure the difference between the model prediction value and the true value, and is easy to calculate and optimize.
[0063] Optimizer: Select the Adam optimizer to update the model weights. The Adam optimizer has the advantages of adaptive learning rate adjustment, efficient calculation, and fast convergence.
[0064] Precision correction unit: Figure 4 As shown, the new voltage, current and temperature signal sampling values are input into the trained neural network model to obtain the corrected voltage and current signal prediction values. After calculation and correction by the neural network model, a more accurate output signal is obtained.
[0065] Data partitioning: The data set is divided into training set, validation set and test set, and the partitioning is performed by cross-validation. The training set is used for parameter update and training process of the neural network, and the validation set is used to evaluate the performance of the model and make adjustments to prevent overfitting or underfitting. Finally, the test set is used to evaluate the performance of the trained and tuned neural network model.
[0066] Network training: Use the training set data to train the network, and update the network weights and biases through the back propagation algorithm, so that the neural network can learn the complex mapping relationship between input signals and output signals. Select appropriate loss functions and optimization algorithms to minimize the difference between the predicted output and the theoretical value to improve the accuracy and stability of the model.
[0067] Model validation and tuning: Use validation set data to evaluate model performance, monitor the model training process and make adjustments to prevent overfitting or underfitting. Tune model parameters based on validation set results, including adjusting learning rate, adding regularization terms, or increasing training rounds, to further improve model performance.
[0068] Model evaluation and testing: Use the test set data to evaluate the trained and tuned neural network model and calculate the error indicators between the predicted value and the theoretical value, such as the root mean square error (RMSE) or mean absolute error (MAE). Based on the evaluation results, the structure and parameters of the model can be further adjusted to improve the performance and adaptability of the model.
[0069] Application and deployment: The trained model is applied to the actual voltage and current signal acquisition module to correct the actual value and provide more accurate estimation results. The new voltage, current and temperature signal sampling values are input into the trained neural network model to obtain the corrected voltage and current signal prediction values. Through the calculation and correction of the neural network model, a more accurate output signal can be obtained, which improves the accuracy and reliability of the voltage and current signal acquisition module.
[0070] The precision correction method implementation process is as follows Figure 5 shown.
[0071] The present invention is not limited to the embodiments described above. The above description of the specific embodiments is intended to describe and illustrate the technical solution of the present invention. The above specific embodiments are merely illustrative and not restrictive. Without departing from the scope of the present invention and the scope of protection of the claims, a person of ordinary skill in the art can also make many forms of specific changes under the guidance of the present invention, which all fall within the scope of protection of the present invention.
Claims
1. A method for improving the accuracy of a voltage and current acquisition module, characterized in that: A neural network is used to calibrate the voltage and current signals output by the voltage and current signal acquisition module under different temperature conditions. The training process of the neural network includes: inputting the voltage and current signals input and output by the voltage and current signal acquisition module and temperature data into the neural network so that the error between the voltage and current signals predicted by the neural network and the theoretical value is minimized.
2. A method for improving the accuracy of a voltage and current acquisition module according to claim 1, characterized in that: The theoretical value is the voltage and current signal of the input voltage and current signal acquisition module.
3. A method for improving the accuracy of a voltage and current acquisition module according to any one of claims 1 to 2, characterized in that: The preparation process of the training data set includes using a temperature sensor to obtain temperature data from the voltage and current signal acquisition module; at the same time, using a high-precision signal source to generate a series of voltage and current signals and inputting them into the voltage and current signal acquisition module; the voltage and current signals generated by the signal source are the theoretical values of the voltage and current signals, and the output values of the voltage and current signal acquisition module are the sampling values of the voltage and current signals; the above temperature data, theoretical values, and sampling values are integrated into a data set and preprocessed.
4. A method for improving the accuracy of a voltage and current acquisition module according to any one of claim 3, characterized in that: Preprocessing includes removing outliers, filtering and data normalization.
5. A voltage and current acquisition module accuracy correction system, characterized in that: It includes a data preparation unit, a neural network construction unit and an accuracy correction unit; the data preparation unit is used to obtain temperature data of the voltage and current signal acquisition module, theoretical values of the voltage and current signals, and sampling values of the voltage and current signals; the neural network construction unit uses the data obtained by the data preparation unit to train the neural network; The precision correction unit is used to input the real-time voltage, current and temperature signal sampling values into the trained neural network model to obtain the corrected voltage and current signal prediction values.