A relay energy-saving control method and system

Abstract

The present invention relates to the field of relay control technology, and in particular to a relay energy-saving control method and system, the method comprising: S1, based on the high and low levels of the input, controlling the opening and closing of the relay, and after the relay is closed, introducing a preset pulse width modulation control signal to maintain the intermittent power supply of the relay; S2, real-time monitoring of the vibration data of the relay operation process, and based on the vibration prediction model, analyzing and predicting the impact of the vibration data on the relay, and generating a vibration feedback signal; S3, dynamically controlling the power supply state of the relay according to the vibration feedback signal, switching the relay back to the continuous power supply controlled by the high and low levels, and adjusting the input terminal voltage. The present invention realizes intermittent power supply to the relay by comprehensively utilizing the high and low levels and pulse width modulation control, effectively reducing energy consumption; at the same time, realizing intelligent control, it is more sensitive to environmental vibrations and can adapt to various working conditions more flexibly.

Description

Technical Field

[0001] The present invention relates to the technical field of relay control, and in particular to a relay energy-saving control method and system. Background Art

[0002] A relay is an electrical switching device whose main function is to open or close the current in a circuit. By controlling the state of the relay, remote or automatic control of the circuit in the power system can be achieved. Relays usually include a control end and an execution end. By applying an electrical signal to the control end, the state of the execution end can be controlled. Relays are widely used in power systems and automated control systems, and their reliability and flexibility make them key electrical components. There are two main types of relay control methods: traditional high and low level control and more advanced digital control. Traditional high and low level control controls the action of the relay by applying a high or low level to the control end. This method is simple and reliable, but may not be flexible enough in complex automated systems. Digital control is implemented through a microprocessor or controller, which can achieve more complex logic and precise control, and is suitable for scenarios that require intelligence and adaptability.

[0003] In practical applications, the control of relays is particularly critical for the company's intelligent power distribution unit. Although the control of relays in the company's intelligent power distribution unit has made significant progress in improving the efficiency and reliability of power systems, there are still some potential defects in energy conservation and environmental protection. First of all, for the traditional high and low level control method, it may appear too simple and crude in some scenarios and cannot fully adapt to complex changes in power demand. This control method may cause the system to maintain a high power supply even when it is not under high load, wasting precious energy resources. In addition, relays may generate additional energy consumption and electromagnetic interference when switching states, especially in the case of frequent switching, which may have a negative impact on the energy efficiency performance of the entire power system.

[0004] At present, there are some technologies that try to use vibration detection to control the power supply mode or opening and closing of relays. These technologies usually install vibration sensors near the relay to monitor the vibration state of the relay during operation, which is used as the basis for judging the working state of the relay. When abnormal vibration is detected, the system can take corresponding measures, such as adjusting the power supply mode or performing opening and closing operations. However, traditional technologies still have some shortcomings: First, some technologies are difficult to accurately predict the future vibration state because the vibration signal may be affected by a variety of complex factors, such as external environmental interference, changes in vibration sources, etc. Secondly, controlling the relay based on vibration detection leads to a certain time lag in the response speed, and it is impossible to achieve real-time monitoring and timely feedback of the vibration state. In some cases, it may not be possible to take quick measures before the relay is adversely affected, which may lead to system instability or failure.

[0005] Therefore, although vibration detection technology has potential applications in relay control, further improvements are still needed to improve the accurate prediction ability and real-time performance of future vibration states, so as to more effectively cope with the challenges in relay operation. Summary of the invention

[0006] Based on this, it is necessary to provide a relay energy-saving control method and system to address the above technical problems.

[0007] In a first aspect, the present invention provides a relay energy-saving control method, the method comprising:

[0008] S1, based on the input high and low levels, controls the opening and closing of the relay, and after the relay is closed, introduces a preset pulse width modulation control signal to maintain the intermittent power supply of the relay;

[0009] S2, real-time monitoring of the vibration data of the relay operation process, analysis and prediction of the impact of the vibration data on the relay, and generation of a vibration feedback signal;

[0010] S3. Dynamically control the power supply state of the relay according to the vibration feedback signal, switch the relay back to continuous power supply controlled by high and low levels, and adjust the input voltage.

[0011] Further, based on the input high and low levels, the opening and closing of the relay is controlled, and after the relay is closed, a preset pulse width modulation control signal is introduced to maintain the intermittent power supply of the relay, including:

[0012] S11, monitoring and receiving the control level signal sent from the relay control terminal, and analyzing the control level signal, if the control level signal is a low level, the relay is triggered to open, if the control level signal is a high level, the relay is triggered to close;

[0013] S12. Monitor the closing state of the relay in real time. When the relay is triggered to close under high-level control, introduce a pulse width modulation signal with a preset frequency and duty cycle. Use the pulse width modulation signal to dynamically change the voltage across the relay coil, and switch the power supply state of the relay from continuous power supply controlled by high level to intermittent power supply.

[0014] Furthermore, real-time monitoring of the vibration data of the relay operation process, analysis and prediction of the impact of the vibration data on the relay, and generation of vibration feedback signals include:

[0015] S21, using a vibration sensor of a vibration shaft type to monitor the vibration phenomenon in the operation process of the relay in real time. If there is no vibration phenomenon, the vibration sensor is used as a conductor, the controller does not detect the circuit breakage, and no timing signal is generated. If there is a vibration phenomenon, the vibration sensor is short-circuited. At this time, the controller detects the circuit breakage, generates a timing signal, and serves as the timing starting point of the piezoelectric sensor;

[0016] S22, using a piezoelectric sensor to detect a charge signal of the relay during vibration;

[0017] S23, analyzing the vibration data based on the charge signal, analyzing and predicting the predicted value of the vibration data at the next moment through a vibration prediction model, and generating a vibration feedback signal.

[0018] Further, analyzing the vibration data based on the charge signal, analyzing and predicting the predicted value of the vibration data at the next moment through a vibration prediction model, and generating a vibration feedback signal includes:

[0019] S231, converting the charge signal into vibration data, preferentially comparing the relationship between the vibration data and the safety threshold in the vibration protection standard, and determining whether to directly output a vibration feedback signal;

[0020] S232, based on the convolutional neural network, extracting effective features from the vibration data, and fusing the effective features in chronological order to form an effective feature sequence;

[0021] S233, processing the effective feature sequence based on the frequency domain decomposition method, inputting it into the vibration prediction model, predicting the vibration data at the next moment, and generating a vibration feedback signal based on the prediction result.

[0022] Further, converting the charge signal into vibration data, preferentially comparing the relationship between the vibration data and the safety threshold in the vibration protection standard, and determining whether to directly output the vibration feedback signal includes:

[0023] S2311, extracting the amplitude of the charge signal as the vibration intensity;

[0024] S2312, obtaining vibration frequency based on spectrum analysis of charge signal;

[0025] S2313, extracting the time-varying characteristics of the charge signal to obtain the time domain data of the vibration;

[0026] S2314, combining the vibration intensity, vibration frequency and time domain data as the vibration data of the relay, and establishing a vibration protection standard, respectively setting the safety thresholds of the vibration intensity, vibration frequency and time domain data;

[0027] S2315. Prioritize determining the size relationship between the three types of data in the current vibration data and their respective safety thresholds. If at least one type of data is greater than or equal to the safety threshold, directly generate a vibration feedback signal. If all data are less than the safety threshold, continue to perform analysis and prediction of the vibration data.

[0028] Furthermore, based on the convolutional neural network, effective features in the vibration data are extracted, and the effective features are fused in chronological order to form an effective feature sequence including:

[0029] S2321, preprocessing vibration data, wherein the preprocessing includes denoising, filtering and normalization;

[0030] S2322, construct a convolutional neural network and use the existing vibration data training set to learn feature extraction;

[0031] S2323, adding a feature extraction layer between the convolutional layer and the fully connected layer of the convolutional neural network, and using the output of the feature extraction layer as the input of the non-negative matrix factorization model;

[0032] S2324. Integrate the goals of convolutional neural network and non-negative matrix factorization model, set the overall objective function, and minimize the overall objective function through end-to-end training;

[0033] S2325. Execute the back propagation algorithm in the entire convolutional neural network, update the parameters of the two models, the convolutional neural network and the non-negative matrix factorization model, and fuse the features of the two to form a unified feature representation as a valid feature sequence.

[0034] Furthermore, the effective feature sequence is processed based on the frequency domain decomposition method, inputted into the vibration prediction model, the vibration data at the next moment is predicted, and the vibration feedback signal is generated according to the prediction result, including:

[0035] S2331. Based on the frequency domain decomposition method, decompose the effective feature sequence into three parts in the frequency domain, wherein the three parts include an intensity part, a frequency part and a time domain part;

[0036] S2332. Based on the back propagation neural network, a vibration prediction model is constructed, and the model is trained using the existing vibration data training set;

[0037] S2333, decomposing the intensity part by lifting wavelet to obtain a two-layer lifting wavelet, then inputting the low-frequency part of the two-layer lifting wavelet into the vibration prediction model, and outputting the intensity prediction value of the intensity part at the next moment;

[0038] S2334, analyzing and calculating the frequency prediction value of the frequency part at the next moment based on the method of predicting the next point value from two points of the univariate linear regression equation;

[0039] S2335, input the time domain part into the vibration prediction model, and output the period prediction value of the time domain part at the next moment;

[0040] S2336. Count the intensity prediction value, frequency prediction value and period prediction value at the next moment, and compare them with their respective safety thresholds. When one of the prediction values ​​is greater than the safety threshold, generate a vibration feedback signal.

[0041] Furthermore, based on the frequency domain decomposition method, the effective feature sequence frequency domain is decomposed into three parts including:

[0042] S23311. Perform Fourier decomposition on the effective feature sequence. The formula for Fourier decomposition is:

[0043]

[0044] Where, P(t) represents the effective feature sequence; N represents the length of the effective feature sequence; ωi represents the angular frequency of the i-th component; t represents the time domain data; a0 represents the initial constant term; ai and bi both represent the Fourier coefficients of the i-th component;

[0045] S23312. According to the periodicity of vibration changes, the effective feature sequence is decomposed into an intensity part, a frequency part and a time domain part, wherein the intensity part corresponds to the initial constant term a0, the frequency part corresponds to the Fourier coefficients ai and bi, and the time domain part corresponds to the time domain data t. Further, dynamically controlling the power supply state of the relay according to the vibration feedback signal, switching the relay back to continuous power supply controlled by high and low levels, and adjusting the input terminal voltage include:

[0046] S31, the control end receives the vibration feedback signal, and immediately switches the voltage across the relay coil to a power supply state of continuous power supply controlled by a high level;

[0047] S32, within the rated voltage range of the relay, linearly enhance the input voltage of the control terminal;

[0048] S33: After the safety setting time has passed, if the vibration feedback signal is no longer received, the power supply state of the relay is switched again to intermittent power supply controlled by the pulse width modulation signal.

[0049] In a second aspect, the present invention further provides a relay energy-saving control system, which includes: an input switching module, a vibration feedback module and an adjustment switching module, wherein:

[0050] An input switching module is used to control the opening and closing of the relay based on the high and low levels of the input, and after the relay is closed, introduces a preset pulse width modulation control signal to maintain the intermittent power supply of the relay;

[0051] The vibration feedback module is used to monitor the vibration data of the relay operation process in real time, and analyze and predict the impact of the vibration data on the relay based on the vibration prediction model to generate a vibration feedback signal;

[0052] The regulating switching module is used to dynamically control the power supply state of the relay according to the vibration feedback signal, switch the relay back to continuous power supply controlled by high and low levels, and adjust the input voltage.

[0053] The beneficial effects of the present invention are:

[0054] 1. Through the comprehensive use of high and low levels and pulse width modulation control, intermittent power supply to the relay is realized, effectively reducing energy consumption; at the same time, by real-time monitoring of the vibration data of the relay during operation, and analyzing its impact on the relay based on the prediction model, an accurate vibration feedback signal is generated. Based on this feedback signal, the system dynamically adjusts the power supply state of the relay, intelligently switches it back to continuous power supply controlled by high and low levels, and adjusts the input voltage when necessary, thereby maximizing the working efficiency and stability of the relay; at the same time, by predicting the vibration state at future times, the relay control can achieve a balance between real-time and accuracy, and generate corresponding vibration feedback signals in advance, thereby achieving precise control and protection of relays and other equipment, minimizing potential failures and improving system stability.

[0055] 2. By introducing a pulse width modulation signal after the relay is closed, the relay power supply state is switched from continuous power supply controlled by a high level to intermittent power supply, which effectively improves energy utilization efficiency, reduces unnecessary energy consumption, and achieves energy conservation and environmental protection; through dynamic power supply state adjustment, the energy consumption is effectively reduced, providing a more efficient and environmentally friendly solution for the relay application field.

[0056] 3. Through real-time monitoring and analysis of vibration phenomena during the operation of the relay, the vibration conditions of the relay during operation, including information such as intensity and frequency, can be sensitively captured. After the vibration occurs, the vibration state at the next moment can be quickly analyzed and predicted, and an accurate feedback signal can be generated, which helps the system to adjust the power supply state of the relay in real time and switch it to continuous power supply controlled by high and low levels to cope with possible abnormal vibration conditions. Therefore, the present invention not only improves the intelligence and stability of the system, but also helps to prevent potential problems and further improves the reliability and safety of the relay. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0058] Figure 1 is a flow chart of a relay energy-saving control method according to an embodiment of the present invention;

[0059] Figure 2 is a schematic diagram of continuous power supply and intermittent power supply according to an embodiment of the present invention;

[0060] Figure 3 is a schematic structural diagram of a vibration sensor of a vibration shaft type according to an embodiment of the present invention;

[0061] Figure 4 The present invention is a system principle block diagram of a relay energy-saving control system according to an embodiment of the present invention.

[0062] Figure numbers: 1. Input switching module; 2. Vibration feedback module; 3. Adjustment switching module. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0064] See also Figure 1 , provides a relay energy-saving control method, characterized in that the method comprises:

[0065] S1, based on the input high and low levels, controls the opening and closing of the relay, and after the relay is closed, introduces a preset pulse width modulation control signal to maintain the intermittent power supply of the relay.

[0066] In the description of the present invention, Figure 2 As shown, based on the high and low levels of the input, the opening and closing of the relay is controlled, and after the relay is closed, a preset pulse width modulation control signal is introduced to maintain the intermittent power supply of the relay, including:

[0067] S11, monitor and receive the control level signal sent from the relay control terminal, and analyze the control level signal. If the control level signal is at a low level, the relay is triggered to open; if the control level signal is at a high level, the relay is triggered to close.

[0068] S12. Monitor the closing state of the relay in real time. When the relay is triggered to close under high-level control, introduce a pulse width modulation signal with a preset frequency and duty cycle. Use the pulse width modulation signal to dynamically change the voltage across the relay coil, and switch the power supply state of the relay from continuous power supply controlled by high level to intermittent power supply.

[0069] Among them, after the relay is closed, a pulse width modulation signal with a preset frequency and duty cycle is introduced. Pulse width modulation is a modulation technology that transmits information by changing the pulse width of the signal. The preset frequency indicates the frequency of occurrence of the pulse, and the duty cycle indicates the proportion of the high level time of the pulse to the entire cycle. The pulse width modulation signal dynamically adjusts the voltage across the relay coil through the corresponding circuit or control system. During the high level of the pulse, a relatively high voltage is provided to keep the relay closed. During the low level of the pulse, the voltage is reduced, which can cause the relay to be disconnected during this period. In this way, the power supply state of the relay is dynamically switched between continuous power supply controlled by a high level and intermittent power supply controlled by a low level.

[0070] Through the regulation of the pulse width modulation signal, the power supply state of the relay is switched from continuous power supply under high-level control to intermittent power supply. This switching process makes the current in the relay coil intermittent in time, thus achieving the effect of intermittent power supply.

[0071] S2. Real-time monitoring of the vibration data of the relay operation process, analysis and prediction of the impact of the vibration data on the relay, and generation of a vibration feedback signal.

[0072] In the description of the present invention, real-time monitoring of vibration data of the relay operation process, analyzing and predicting the impact of the vibration data on the relay, and generating a vibration feedback signal include:

[0073] S21. Use a vibration shaft-type vibration sensor to monitor the vibration phenomenon during the operation of the relay in real time. If there is no vibration phenomenon, the vibration sensor serves as a conductor, the controller does not detect the circuit break, and does not generate a timing signal. If there is a vibration phenomenon, the vibration sensor is short-circuited. At this time, the controller detects the circuit break, generates a timing signal, and serves as the timing starting point of the piezoelectric sensor.

[0074] Among them, Figure 3 As shown, the vibration shaft type vibration sensor includes a plastic shell (A), a vibration shaft (B), a first guide needle (C) and a second guide needle (D). During use, when there is no vibration, the two ends of the vibration sensor are in a conductive state, and the conductive resistance is less than 10Ω. When there is vibration at any angle, the two ends of the vibration sensor are disconnected and a trigger signal is generated.

[0075] S22. Use a piezoelectric sensor to detect the charge signal of the relay during vibration.

[0076] Specifically, a piezoelectric sensor is a sensor that can convert mechanical vibration into a charge signal. During the vibration of the relay, the piezoelectric sensor can be used to detect the charge signal generated by the relay during the vibration, thereby realizing the monitoring of the working state of the relay.

[0077] Place the piezoelectric sensor at an appropriate location on the relay, usually on the relay housing or base. When the relay vibrates, the piezoelectric sensor can sense these vibrations. When the relay is subjected to external vibration or shock during operation, the piezoelectric sensor will sense these mechanical vibrations and convert them into corresponding charge signals. The piezoelectric effect causes the crystal structure of the sensor to deform slightly, generating charges, which can then be measured. Using a charge amplifier or related circuit to detect and amplify the charge signal output by the piezoelectric sensor, the weak signal can be converted into a voltage signal that is easier to process.

[0078] In addition, the vibration sensor and the piezoelectric sensor cooperate with each other, and the vibration sensor serves as a timing switch for the piezoelectric sensor. When the vibration sensor of the vibration shaft detects a vibration change, it will be disconnected. At this time, the controller will detect the disconnection change and generate a timing signal. The timing signal serves as the timing starting point of the piezoelectric sensor in this vibration detection process.

[0079] The piezoelectric sensor will always keep running, but when it does not receive the timing signal sent by the controller, the charge signal it detects does not need to be processed, stored, or analyzed. Only when it receives the timing signal, it indicates that there is obvious vibration, and the timing starts at this time, which serves as the starting point of the time domain data. After this timing signal starts, the charge signal detected by the piezoelectric sensor begins to be processed and analyzed, and finally forms vibration data.

[0080] Because the vibration sensor adopts a guide needle design, when there is no vibration, it is equivalent to a wire with negligible resistance, and no additional loss will be generated in the circuit. The piezoelectric sensor will only process the charge signal after detecting the vibration and starting the timing, so it can effectively avoid additional power loss and system occupation. When the vibration sensor is closed and no longer disconnected, and maintains the set safety time, it can be regarded as a safe state. At this time, the controller generates a stop timing signal as the end point of the time domain data of the piezoelectric sensor, realizing the complete detection and prediction of this vibration.

[0081] S23, analyzing the vibration data based on the charge signal, analyzing and predicting the predicted value of the vibration data at the next moment through a vibration prediction model, and generating a vibration feedback signal.

[0082] In the description of the present invention, analyzing vibration data based on the charge signal, analyzing and predicting the predicted value of the vibration data at the next moment through a vibration prediction model, and generating a vibration feedback signal includes:

[0083] S231, converting the charge signal into vibration data, preferentially comparing the relationship between the vibration data and the safety threshold in the vibration protection standard, and determining whether to directly output a vibration feedback signal.

[0084] In the description of the present invention, converting the charge signal into vibration data, preferentially comparing the relationship between the vibration data and the safety threshold in the vibration protection standard, and determining whether to directly output the vibration feedback signal includes:

[0085] S2311. Extract the amplitude of the charge signal as the vibration intensity.

[0086] Among them, the vibration intensity information is obtained by measuring the amplitude of the charge signal.

[0087] S2312. Obtain vibration frequency based on spectrum analysis of the charge signal.

[0088] Among them, the frequency information of vibration is obtained by performing spectrum analysis on the charge signal. The time domain signal is converted into a spectrum in the frequency domain by performing Fourier transform on the charge signal. In the spectrum diagram obtained, the horizontal axis represents the frequency and the vertical axis represents the amplitude of the signal at the corresponding frequency. In the spectrum diagram, the various frequency components in the signal can be clearly seen.

[0089] Frequency component identification: By analyzing the spectrum, the frequency components corresponding to the relay vibration are identified. The peaks on the spectrum correspond to the main frequencies in the signal.

[0090] S2313. Extract the time-varying characteristics of the charge signal and obtain the time domain data of the vibration.

[0091] Among them, the charge signal is converted from the frequency domain to the time domain, and the time domain analysis can provide a waveform diagram of the signal on the time axis. A time domain waveform diagram is drawn, in which the horizontal axis represents time and the vertical axis represents the amplitude of the signal. The time domain waveform diagram shows the changes of the signal at different time points. The time domain data related to the vibration is extracted from the waveform diagram, including parameters such as amplitude, period, pulse width, etc., depending on the required analysis information. Finally, the time domain data is analyzed to extract the characteristics of the relay vibration, including the fluctuation, periodicity, pulse width and other information of the vibration.

[0092] S2314. Combine the vibration intensity, vibration frequency and time domain data as the vibration data of the relay, establish a vibration protection standard, and set the safety thresholds of the vibration intensity, vibration frequency and time domain data respectively.

[0093] S2315. Prioritize determining the size relationship between the three types of data in the current vibration data and their respective safety thresholds. If at least one type of data is greater than or equal to the safety threshold, directly generate a vibration feedback signal. If all data are less than the safety threshold, continue to perform analysis and prediction of the vibration data.

[0094] S232. Based on a convolutional neural network, effective features are extracted from the vibration data, and the effective features are fused in chronological order to form an effective feature sequence.

[0095] In the description of the present invention, based on the convolutional neural network, effective features in the vibration data are extracted, and the effective features are fused in time sequence to form an effective feature sequence, including:

[0096] S2321. Preprocess the vibration data, wherein the preprocessing includes denoising, filtering and normalization.

[0097] S2322. Construct a convolutional neural network and use the existing vibration data training set to learn feature extraction.

[0098] Specifically, the existing vibration data training set is divided into a training set and a validation set. Make sure that the data set contains enough normal and abnormal vibration samples so that the model can learn various vibration patterns. Convert the vibration data into a format suitable for CNN processing. Usually, vibration data can be represented as a time series, with the vibration intensity at each time point as a feature. Then design the network structure of CNN, including convolutional layers, pooling layers, fully connected layers, etc. The convolutional layer is used to extract local features, the pooling layer is used for downsampling, and the fully connected layer is used to map the extracted features to the output category.

[0099] Define the loss function, optimizer, and evaluation metric. The loss function measures the difference between the model output and the true label. The optimizer is used to minimize the loss function. The evaluation metric is used to measure the model performance. Use the training set data to train the CNN model. During the training process, the model continuously adjusts the weights and biases through the back-propagation algorithm to minimize the loss function.

[0100] S2323. Add a feature extraction layer between the convolutional layer and the fully connected layer of the convolutional neural network, and use the output of the feature extraction layer as the input of the non-negative matrix decomposition model.

[0101] Specifically, adding a feature extraction layer between the convolutional layer and the fully connected layer of the convolutional neural network (CNN) and using the output of the feature extraction layer as the input of the non-negative matrix factorization model is a strategy that combines deep learning and traditional machine learning methods. It aims to comprehensively utilize the deep learning network's ability to learn high-level features and the advantages of non-negative matrix factorization in decomposing data.

[0102] In the convolutional neural network, an additional feature extraction layer is added, usually the Global Average Pooling Layer. The purpose of this layer is to reduce the dimension of the output of the convolutional layer and obtain the average value on each feature channel.

[0103] Non-negative matrix factorization is a traditional machine learning method for decomposing a matrix into the product of two non-negative matrices. In the present invention, the output of the feature extraction layer in the convolutional neural network is used as the input of the non-negative matrix factorization model.

[0104] S2324. Integrate the goals of the convolutional neural network and the non-negative matrix factorization model, set a comprehensive overall objective function, and minimize the overall objective function through end-to-end training.

[0105] Among them, the goals of the integrated convolutional neural network (CNN) and non-negative matrix factorization (NMF) model usually require the definition of an overall objective function so that it can simultaneously consider the training goals of the two models. Set the overall objective function (JointObjective Function), the expression is as follows:

[0106] TotalLoss=λ1CNNLoss+λ2NMFLoss

[0107] Where λ1 and λ2 are the weights of the two loss terms, which are used to balance the training contributions of the two models.

[0108] CNNLoss is the loss function of a convolutional neural network, usually the cross entropy loss or other appropriate loss functions in classification tasks, used to measure the performance of a CNN model.

[0109] NMFLoss is the loss function of the non-negative matrix factorization model, which can be a reconstruction error or other indicators of model performance. It is used to measure the modeling ability of the NMF model for the output of the convolutional neural network.

[0110] The purpose of the overall objective function is to minimize the combined loss of the CNN model and the NMF model. The entire joint model is trained by balancing the contributions of the two. Minimizing this overall objective function is achieved by using gradient descent or other optimization algorithms.

[0111] In practical applications, the specific objective function form and weights need to be adjusted according to the characteristics of the specific problem and data. The design of this comprehensive objective function needs to take into account the overall performance of the model and the requirements of the task in order to achieve the goal of effective feature extraction and modeling of vibration data.

[0112] S2325. Execute the back propagation algorithm in the entire convolutional neural network, update the parameters of the two models, the convolutional neural network and the non-negative matrix factorization model, and fuse the features of the two to form a unified feature representation as a valid feature sequence.

[0113] Specifically, the back propagation algorithm is executed in the entire convolutional neural network (CNN), the parameters of the two models, the convolutional neural network and the non-negative matrix factorization (NMF) model, are updated, and the features of the two are fused to form a unified feature representation, which is achieved through the following steps:

[0114] 1. Calculate the overall loss: Use the overall objective function to calculate the overall loss of the CNN model and the NMF model. The overall loss is the weighted sum of the CNN Loss and the NMF Loss.

[0115] 2. Back propagation: Back propagate the overall loss and calculate the gradients of the CNN model and the NMF model.

[0116] 3. Update parameters: Use gradient descent or other optimization algorithms to update the parameters of the CNN model and NMF model respectively.

[0117] 4. Feature fusion: During the training process, we choose to add an additional fully connected layer to the CNN model to fuse the output of the last convolutional layer of the CNN model with the output of the NMF model. The output of this fully connected layer is a unified feature representation.

[0118] S233, processing the effective feature sequence based on the frequency domain decomposition method, inputting it into the vibration prediction model, predicting the vibration data at the next moment, and generating a vibration feedback signal based on the prediction result.

[0119] In the description of the present invention, the effective feature sequence is processed based on the frequency domain decomposition method, inputted into the vibration prediction model, the vibration data of the next moment is predicted, and the vibration feedback signal is generated according to the prediction result, including:

[0120] S2331. Based on the frequency domain decomposition method, the effective feature sequence is decomposed into three parts in the frequency domain, wherein the three parts include an intensity part, a frequency part and a time domain part.

[0121] The intensity part represents the amplitude or energy of each frequency component in the frequency domain.

[0122] The frequency part represents the different frequency components existing in the frequency domain.

[0123] The time domain portion represents the periodicity of the vibration.

[0124] In the description of the present invention, based on the frequency domain decomposition method, the effective feature sequence frequency domain is decomposed into three parts including:

[0125] S23311. Perform Fourier decomposition on the effective feature sequence. The formula for Fourier decomposition is:

[0126]

[0127] Where P(t) represents the effective feature sequence, N represents the length of the effective feature sequence, ωi represents the angular frequency of the i-th component, t represents the time domain data, a0 represents the initial constant term, and ai and bi both represent the Fourier coefficients of the i-th component.

[0128] S23312. According to the periodicity of vibration changes, the effective feature sequence is decomposed into an intensity part, a frequency part and a time domain part, wherein the intensity part corresponds to the initial constant term a0, the frequency part corresponds to the Fourier coefficients ai and bi, and the time domain part corresponds to the time domain data t. The intensity part is represented by the initial constant term a0, which represents the DC component of the signal, i.e., the part without periodic changes; the frequency part is represented by ai and bi, which represents the sine and cosine components of the signal at different frequencies, i.e., the spectrum information of periodic vibration, ai represents the amplitude of the sine wave component at a frequency of n·fi, and bi represents the amplitude of the cosine wave component at a frequency of n·fi.

[0129] S2332. Based on the back propagation neural network, a vibration prediction model is constructed, and the model is trained using the existing vibration data training set.

[0130] Specifically, priority is given to collecting vibration data, including vibration intensity, vibration frequency, and time domain data, and preprocessing the vibration data, including denoising, normalization, feature extraction, etc., to finally construct a training set for the vibration prediction model.

[0131] Secondly, a vibration prediction model is constructed based on a back propagation neural network (BPNN). The BP neural network can simulate an artificial neural network and repeatedly train and simulate the data samples in the training set, thereby establishing a prediction model based on known information, that is, the vibration prediction model used in the present invention.

[0132] According to the setting principle of the number of nodes in the hidden layer of the BP neural network, the number of nodes in the hidden layer is R=2n+1, where n is the number of endpoints in the input layer. Therefore, the number of nodes in the hidden layer is set to R=5. According to this design idea, the vibration prediction model adopts a three-layer structure, namely the input layer, the hidden layer and the output layer, and sets the vibration data at the current moment as the input layer and the vibration data at the next moment as the output layer.

[0133] After the construction is completed, the model is trained. For each sample in the training set, the following steps are performed: input the vibration data at the current moment into the input layer of the network; calculate the model output through forward propagation, and calculate the loss function between the output value and the actual output; use the back propagation algorithm to calculate the gradient and update the network parameters, including weights and biases. Repeat the above steps until convergence or the predetermined training rounds are reached.

[0134] S2333. Decompose the intensity part by lifting wavelet to obtain a two-layer lifting wavelet, then input the low-frequency part of the two-layer lifting wavelet into the vibration prediction model, and output the intensity prediction value of the intensity part at the next moment.

[0135] Among them, the intensity part is decomposed by wavelet to obtain the wavelet coefficient. Wavelet decomposition is a signal processing technology that decomposes the signal into components of different frequencies to obtain the frequency domain information of the signal. The wavelet coefficient is lifted by wavelet operation. The low-frequency part is extracted from the two-layer lifting wavelet. In wavelet decomposition, the low-frequency part usually contains the overall trend and slowly changing information of the signal. The extracted low-frequency part is used as input to the vibration prediction model. The trained model is used to predict the intensity part at the future moment. The output of the model is the intensity prediction value at the next moment.

[0136] S2334. Based on the method of predicting the next value from two points of the univariate linear regression equation, analyze and calculate the frequency prediction value of the frequency part at the next moment.

[0137] S2335. Input the time domain part into the vibration prediction model, and output the period prediction value of the time domain part at the next moment.

[0138] S2336. Count the intensity prediction value, frequency prediction value and period prediction value at the next moment, and compare them with their respective safety thresholds. When one of the prediction values ​​is greater than the safety threshold, generate a vibration feedback signal.

[0139] S3. Dynamically control the power supply state of the relay according to the vibration feedback signal, switch the relay back to continuous power supply controlled by high and low levels, and adjust the input voltage.

[0140] In the description of the present invention, dynamically controlling the power supply state of the relay according to the vibration feedback signal, switching the relay back to continuous power supply controlled by high and low levels, and adjusting the input terminal voltage include:

[0141] S31, the control end receives the vibration feedback signal, and immediately switches the voltage across the relay coil to a power supply state of continuous power supply controlled by a high level.

[0142] S32: Within the rated voltage range of the relay, linearly increase the input voltage of the control terminal.

[0143] S33: After the safety setting time has passed, if the vibration feedback signal is no longer received, the power supply state of the relay is switched again to intermittent power supply controlled by the pulse width modulation signal.

[0144] See also Figure 4, a relay energy-saving control system is also provided, the system comprises: an input switching module 1, a vibration feedback module 2 and an adjustment switching module 3, wherein:

[0145] The input switching module 1 is used to control the opening and closing of the relay based on the high and low levels of the input, and after the relay is closed, introduce a preset pulse width modulation control signal to maintain the intermittent power supply of the relay.

[0146] The vibration feedback module 2 is used to monitor the vibration data of the relay operation process in real time, and based on the vibration prediction model, analyze and predict the impact of the vibration data on the relay to generate a vibration feedback signal.

[0147] The regulating switching module 3 is used to dynamically control the power supply state of the relay according to the vibration feedback signal, switch the relay back to the continuous power supply controlled by high and low levels, and adjust the input voltage.

[0148] To sum up, with the help of the above-mentioned technical scheme of the present invention, by comprehensively utilizing high and low levels and pulse width modulation control, intermittent power supply to the relay is realized, thereby effectively reducing energy consumption; at the same time, by real-time monitoring of the vibration data during the operation of the relay, and analyzing its impact on the relay based on the prediction model, an accurate vibration feedback signal is generated, and based on this feedback signal, the system dynamically adjusts the power supply state of the relay, intelligently switches it back to continuous power supply controlled by high and low levels, and adjusts the input voltage when necessary, thereby maximizing the working efficiency and stability of the relay; at the same time, by predicting the vibration state at future times, the relay control can achieve a balance between real-time and accuracy, and generate corresponding vibration feedback signals in advance, thereby realizing precise control and protection of equipment such as relays, minimizing potential failures and improving system stability.

[0149] By introducing a pulse width modulation signal after the relay is closed, the relay power supply state is switched from continuous power supply controlled by a high level to intermittent power supply, which effectively improves energy utilization efficiency, reduces unnecessary energy consumption, and achieves energy conservation and environmental protection; through dynamic power supply state adjustment, the energy consumption is effectively reduced, providing a more efficient and environmentally friendly solution for the relay application field.

[0150] By real-time monitoring and analysis of vibration phenomena during the operation of the relay, the vibration conditions of the relay during operation, including information such as intensity and frequency, can be sensitively captured. After the vibration occurs, the vibration state at the next moment can be quickly analyzed and predicted, and an accurate feedback signal can be generated, which helps the system to adjust the power supply state of the relay in real time and switch it to continuous power supply controlled by high and low levels to cope with possible abnormal vibration conditions. Therefore, the present invention not only improves the intelligence and stability of the system, but also helps to prevent potential problems, further improving the reliability and safety of the relay.

[0151] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

Claims

1. A relay energy-saving control method, characterized in that: The method includes: S1, based on the input high and low levels, control the opening and closing of the relay, and after the relay is closed, introduce a preset pulse width modulation control signal to maintain the intermittent power supply of the relay; S2, real-time monitoring of the vibration data of the relay operation process, analyzing and predicting the impact of the vibration data on the relay, and generating a vibration feedback signal; S3, dynamically controlling the power supply state of the relay according to the vibration feedback signal, switching the relay back to continuous power supply controlled by high and low levels, and adjusting the input terminal voltage; The real-time monitoring of the vibration data of the relay operation process, analyzing and predicting the influence of the vibration data on the relay, and generating a vibration feedback signal comprises: S21, using a vibration sensor of a vibration shaft type to monitor the vibration phenomenon in the operation process of the relay in real time. If there is no vibration phenomenon, the vibration sensor is used as a wire, the controller does not detect the circuit breakage, and no timing signal is generated. If there is a vibration phenomenon, the vibration sensor is short-circuited. At this time, the controller detects the circuit breakage, generates a timing signal, and uses it as the timing starting point of the piezoelectric sensor; S22, using the piezoelectric sensor to detect the charge signal of the relay during the vibration process; S23, analyzing the vibration data based on the charge signal, analyzing and predicting the predicted value of the vibration data at the next moment through a vibration prediction model, and generating a vibration feedback signal; Analyzing the vibration data based on the charge signal, analyzing and predicting the predicted value of the vibration data at the next moment through a vibration prediction model, and generating a vibration feedback signal includes: S231, converting the charge signal into vibration data, preferentially comparing the relationship between the vibration data and the safety threshold in the vibration protection standard, and determining whether to directly output a vibration feedback signal; S232, extracting effective features from the vibration data based on a convolutional neural network, and fusing the effective features in chronological order to form an effective feature sequence; S233, processing the effective feature sequence based on the frequency domain decomposition method, inputting it into the vibration prediction model, predicting the vibration data at the next moment, and generating a vibration feedback signal according to the prediction result; The converting the charge signal into vibration data, preferentially comparing the relationship between the vibration data and the safety threshold in the vibration protection standard, and determining whether to directly output the vibration feedback signal comprises: S2311, extracting the amplitude of the charge signal as the vibration intensity; S2312, acquiring a vibration frequency based on a spectrum analysis of the charge signal; S2313, extracting the time-varying characteristics of the charge signal to obtain time domain data of vibration; S2314, combining the vibration intensity, the vibration frequency and the time domain data as vibration data of the relay, and establishing a vibration protection standard, and setting safety thresholds for the vibration intensity, the vibration frequency and the time domain data respectively; S2315, preferentially determining the magnitude relationship between the three types of data in the current vibration data and their respective safety thresholds, and directly generating a vibration feedback signal if at least one type of data is greater than or equal to the safety threshold; and continuing to perform the analysis and prediction of the vibration data if all the data are less than the safety threshold; The extracting effective features from the vibration data based on a convolutional neural network and fusing the effective features in chronological order to form an effective feature sequence comprises: S2321, preprocessing the vibration data, wherein the preprocessing includes denoising, filtering and normalization; S2322, construct a convolutional neural network and use the existing vibration data training set to learn feature extraction; S2323, adding a feature extraction layer between the convolutional layer and the fully connected layer of the convolutional neural network, and using the output of the feature extraction layer as the input of the non-negative matrix decomposition model; S2324, integrating the objectives of the convolutional neural network and the non-negative matrix decomposition model, setting a comprehensive overall objective function, and minimizing the overall objective function through end-to-end training; S2325. Execute a back propagation algorithm in the entire convolutional neural network, update the parameters of the two models, the convolutional neural network and the non-negative matrix decomposition model, and fuse the features of the two to form a unified feature representation as a valid feature sequence.

2. A relay energy-saving control method according to claim 1, characterized in that: The step of controlling the opening and closing of the relay based on the input high and low levels, and introducing a preset pulse width modulation control signal after the relay is closed to maintain the intermittent power supply of the relay includes: S11, monitoring and receiving a control level signal sent from a relay control terminal, and analyzing the control level signal, if the control level signal is at a low level, the relay is triggered to open, and if the control level signal is at a high level, the relay is triggered to close; S12. Monitor the closing state of the relay in real time. When the relay is triggered to close under high-level control, introduce a pulse width modulation signal with a preset frequency and duty cycle. Use the pulse width modulation signal to dynamically change the voltage across the relay coil, and switch the power supply state of the relay from continuous power supply controlled by high level to intermittent power supply.

3. A relay energy-saving control method according to claim 1, characterized in that: The processing of the effective feature sequence based on the frequency domain decomposition method, inputting it into the vibration prediction model, predicting the vibration data at the next moment, and generating a vibration feedback signal according to the prediction result includes: S2331. Based on a frequency domain decomposition method, decompose the effective feature sequence into three parts in the frequency domain, wherein the three parts include an intensity part, a frequency part, and a time domain part; S2332. Based on the back propagation neural network, a vibration prediction model is constructed, and the model is trained using the existing vibration data training set; S2333, decomposing the intensity part by lifting wavelet to obtain a two-layer lifting wavelet, then inputting the low-frequency part of the two-layer lifting wavelet into the vibration prediction model, and outputting the intensity prediction value of the intensity part at the next moment; S2334, analyzing and calculating the frequency prediction value of the frequency part at the next moment in a method of predicting the next point value from two points of a univariate linear regression equation; S2335, inputting the time domain part into the vibration prediction model, and outputting a period prediction value of the time domain part at the next moment; S2336. Count the intensity prediction value, the frequency prediction value and the period prediction value at the next moment, and compare them with their respective safety thresholds. When one of the prediction values ​​is greater than the safety threshold, generate a vibration feedback signal.

4. A relay energy-saving control method according to claim 3, characterized in that: The frequency domain decomposition method is based on the frequency domain decomposition of the effective feature sequence into three parts, including: S23311. Perform Fourier decomposition on the effective feature sequence. The Fourier decomposition formula is: ； In the formula, P ( t ) represents a valid feature sequence; N Indicates the length of the valid feature sequence; ω i Indicates i The angular frequency of the components; t Represents time domain data; a 0 represents the initial constant term; a i and b i Both indicate i The Fourier coefficients of the components; S23312. According to the periodicity of vibration changes, the effective feature sequence is decomposed into an intensity part, a frequency part and a time domain part, wherein the intensity part corresponds to the initial constant term. a 0 , the frequency part corresponds to the Fourier coefficient a i and b i , the time domain part corresponds to the time domain data t .

5. A relay energy-saving control method according to claim 2, characterized in that: The dynamically controlling the power supply state of the relay according to the vibration feedback signal, switching the relay back to continuous power supply controlled by high and low levels, and adjusting the input terminal voltage includes: S31, the control end receives the vibration feedback signal, and immediately switches the voltage across the relay coil to a power supply state of continuous power supply controlled by a high level; S32, linearly enhancing the input voltage of the control terminal within the rated voltage range of the relay; S33: After the safety setting time has passed, if the vibration feedback signal is no longer received, the power supply state of the relay is switched again to intermittent power supply controlled by the pulse width modulation signal.

6. A relay energy-saving control system, used to implement the relay energy-saving control method according to any one of claims 1 to 5, characterized in that: The system includes: an input switching module, a vibration feedback module and an adjustment switching module, wherein: The input switching module is used to control the opening and closing of the relay based on the high and low levels of the input, and after the relay is closed, introduce a preset pulse width modulation control signal to maintain the intermittent power supply of the relay; The vibration feedback module is used to monitor the vibration data of the relay operation process in real time, and analyze and predict the impact of the vibration data on the relay based on the vibration prediction model to generate a vibration feedback signal; The regulating and switching module is used to dynamically control the power supply state of the relay according to the vibration feedback signal, switch the relay back to continuous power supply controlled by high and low levels, and adjust the input terminal voltage.

Images