A time-delay fuse based on microprocessor control and control method thereof
By introducing microprocessor control and long-term memory network models into the delay fuse, real-time monitoring and analysis of current signals and dynamically adjusting protection strategies, the problem of traditional delay fuses cannot be preventively protected and the response speed of circuit protection is improved and accuracy enhances the adaptability and safety of the system.
Patent Information
- Application Number
- CN202410578777.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-11
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-05-11
AI Technical Summary
Traditional delay fuses lack the ability to predict the change trend of the circuit state, cannot achieve preventive protection, and can only passively respond to circuit abnormalities, and the protection parameters are fixed and difficult to adjust, and their applicability is limited.
The delay fuse system based on microprocessor control is adopted to monitor the current value in real time through the current detection unit, and intelligently analyze the current signal change trend using the long and short-term memory network model, generate fuse instructions, dynamically adjust protection strategies, and improve response speed and accuracy.
It realizes early prediction and accurate response to circuit abnormalities, enhances the adaptability and preventiveness of protection, avoids malfunctions caused by instantaneous current impact, and improves the accuracy of circuit protection and system safety.
Smart Images

Figure CN118431999B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of circuit protection technology, and more particularly to a time-delay fuse based on microprocessor control and a control method thereof. Background Art
[0002] As electrical systems become increasingly complex, higher requirements are placed on circuit protection technology. Although traditional time-delay fuses can provide overload protection to a certain extent, this mainly relies on the thermal sensitivity of physical materials to achieve protection functions. However, the thermal response of physical materials has a certain hysteresis, which may cause the fuse to fail to respond in time when the circuit undergoes drastic changes, such as short circuit events, thereby causing damage to the circuit system. The protection parameters of traditional time-delay fuses (such as melting current value and delay time) are fixed and difficult to adjust, which limits their applicability under different circuit conditions. Traditional time-delay fuses lack the ability to predict the trend of circuit state changes, cannot achieve preventive protection, and can only passively respond to circuit anomalies that have already occurred.
[0003] For example, a Chinese patent with the authorization announcement number CN109327010B discloses an electronic fuse circuit, which belongs to the field of protection circuit technology. It includes a main circuit and a control circuit, and is connected to a DC power supply. The main circuit includes a current detection resistor RSENSE and a switch circuit. The DC power supply, the current detection resistor RSENSE, the switch circuit and the load are connected in sequence; the control circuit includes a current detection circuit and a power supply circuit. The current detection circuit is connected to the power supply circuit, the current detection resistor RSENSE and the switch circuit respectively. The current detection circuit and the voltage detection circuit are combined to achieve high-precision control of the circuit current and voltage. When the circuit current is too large or the power supply voltage is too small, the power supply can be cut off normally.
[0004] The above methods have the problem raised by this background technology, that is, they lack the ability to predict the trend of circuit state changes, cannot achieve preventive protection, and can only passively respond to circuit anomalies that have already occurred. In this context, a time-delay fuse system based on microprocessor control and integrated machine learning model is developed, which can monitor current changes in real time, intelligently analyze the trend of current signal changes, predict potential circuit anomalies, and dynamically adjust the protection strategy according to the prediction results, improve the response speed and accuracy of circuit protection, and enhance the adaptability and preventive protection capabilities of the system. Summary of the invention
[0005] In order to solve the above problems, the present application provides a time-delay fuse based on microprocessor control and a control method thereof, which is used to monitor current changes in real time, intelligently analyze the changing trend of current signals, predict potential circuit abnormalities, and dynamically adjust the protection strategy according to the prediction results.
[0006] In a first aspect, the present application provides a time-delay fuse based on microprocessor control, comprising:
[0007] A current detection unit, used for monitoring the current value flowing through the fuse in real time, and converting the monitored current value into a first current sequence;
[0008] A microprocessor control unit, configured to receive a first current sequence from a current detection unit, determine a current current state based on a long short-term memory network model, and generate a first fuse instruction;
[0009] The time delay trigger unit is used to control the heating element to perform a first processing operation on the wire inside the fuse based on the first fusing instruction generated by the microprocessor control unit.
[0010] As a possible implementation, the current detection unit includes:
[0011] The sensor is arranged near the fuse and is used to monitor the current value flowing through the fuse in real time and generate time information matching the current value;
[0012] A signal conversion device, connected to the sensor, for receiving the current value flowing through the fuse and time information matching the current value, and adding a timestamp to the current value to convert its sequence into a first current sequence;
[0013] The data interface is used to provide a data communication function with the microprocessor control unit to transmit the converted first current sequence.
[0014] As a possible implementation, the signal conversion device, when adding a timestamp to the current value and converting its sequence into a first current sequence, includes:
[0015] A signal enhancement module is used to amplify the current value flowing through the fuse collected by the sensor and filter the amplified current value;
[0016] The analog-to-digital conversion module is used to convert the current value represented by the analog signal into a digital signal.
[0017] As a possible implementation, the microprocessor control unit includes:
[0018] receiving the first current sequence from a current detection unit, and performing a preprocessing operation on the received first current sequence;
[0019] Inputting the preprocessed first current sequence into a pretrained long short-term memory network model, determining a change trend and a potential abnormal mode of the current value flowing through the fuse, and generating first identification data; wherein the change trend and the potential abnormal mode of the current value flowing through the fuse include: overload, short circuit, current fluctuation and starting current;
[0020] Based on the first identification data, generating a first fuse instruction;
[0021] The real-time monitored current value and fuse operation results are used to update the long short-term memory network model to optimize the prediction accuracy.
[0022] As a possible implementation, the preprocessed first current sequence is input into a pretrained long short-term memory network model to determine the change trend and potential abnormal mode of the current value flowing through the fuse, and generate first identification data, including:
[0023] Get historical current data;
[0024] Based on the historical current data, determining a first target feature;
[0025] Based on the first target features and the event labels corresponding to the first target features, construct the pre-trained long short-term memory network model;
[0026] Deploying the pre-trained long-short term network model into a microprocessor control unit;
[0027] The pre-trained long short-term memory network model receives the pre-processed current data in real time, determines the change trend and potential abnormal mode of the current signal, and generates first recognition data.
[0028] As a possible implementation, the historical current data includes: time series current data, current events under different load conditions, abnormal event annotation data, transient current data, long-term operation data, and environmental factor impact data;
[0029] The first target characteristics include: current magnitude, current change speed, current fluctuation frequency, current waveform characteristics, short-term and long-term current changes, and historical current abnormal events.
[0030] As a possible implementation, the pre-trained long short-term memory network model receives the pre-processed current data in real time, determines the change trend and potential abnormal mode of the current signal, and generates the first recognition data, including:
[0031] The function expression for determining whether information is retained or discarded from the memory cell is:
[0032] f t =σ(W f *[h t-1 ,x t ]+b f )
[0033] Among them, f trepresents the output of the forget gate, which is used to decide which information is retained or discarded; σ is the activation function, which ensures that the output value is between 0 and 1, where 0 means that no information is allowed to pass, and 1 means that all information is allowed to pass; W f is the weight matrix of the forget gate; h t-1 Represents the previous state of the current data; x t is the input at the current time point, that is, the current data monitored in real time; b f is the bias term of the forget gate; the input gate is used to update the memory cell information and determine the current data of the new input at the current time point. The function expression is:
[0034]
[0035] Among them, t and t-1 represent the current time point and the previous time point respectively, i t Represents new information added to memory cells; Indicates the candidate value for updating the memory cell, reflecting the new anomaly or trend that the current current data may indicate; W i and W C are the weight matrices of the input gate and candidate memory cells respectively; b i and b C is the corresponding bias term;
[0036] Based on the information of the forget gate and the input gate, the cell state is updated, and the function expression is:
[0037]
[0038] Among them, C t is the memory cell state at the current time step, combined with past information C t-1 and current candidate information
[0039] The output gate is used to determine the output based on the current memory cell, predict the upcoming state of the current and identify potential abnormal patterns. The function expression is:
[0040]
[0041] Among them, t It is the output gate, which also controls the state of the memory cell C t To output state h t The conversion degree function, h t Represents the output at the current time point, that is, the model's final recognition and judgment of the current state; W o is the weight matrix of the output gate; b o is the bias term of the output gate;
[0042] The first identification data is generated based on the upcoming state of the electric current and the potential abnormal pattern.
[0043] As a possible implementation, the first processing operation includes: heating to the target temperature immediately, heating to the target temperature slowly, heating to the target temperature immediately after a preset time interval, and heating to the target temperature slowly after a preset time interval; the first fusing instruction includes: an immediate fusing instruction, an immediate slow fusing instruction, a pre-fusing instruction, a delayed immediate fusing instruction, and a delayed slow fusing instruction; the generating of the first fusing instruction based on the first identification data includes:
[0044] In response to the presence of preset abnormal data in the first identification data, the first fuse instruction is generated based on the preset abnormal data type.
[0045] As a possible implementation, the delay trigger unit, when controlling the heating element to perform a first processing operation on the wire inside the fuse based on the first fusing instruction generated by the microprocessor control unit, includes:
[0046] In response to the first fusing instruction generated by the microprocessor control unit being an immediate fusing instruction, controlling the heating element to immediately raise the temperature to a first target temperature;
[0047] In response to the first fusing instruction generated by the microprocessor control unit being a slow fusing instruction, controlling the heating element to gradually increase the temperature to a first target temperature based on a preset heating mode;
[0048] In response to the first fusing instruction generated by the microprocessor control unit being a pre-fusing instruction, controlling the heating element to immediately raise the temperature to a second target temperature;
[0049] In response to the first fusing instruction generated by the microprocessor control unit being a delayed immediate fusing instruction, controlling the heating element to immediately raise the temperature to a first target temperature after a preset time interval;
[0050] In response to the first fusing instruction generated by the microprocessor control unit being a delayed slow fusing instruction, the heating element is controlled to gradually increase the temperature to a first target temperature based on a preset heating mode after a preset time interval.
[0051] In a second aspect, the embodiment of the present disclosure further provides a time-delay fuse control method based on microprocessor control, comprising:
[0052] Real-time monitoring of the current value flowing through the fuse, and conversion of the monitored current value into a first current sequence;
[0053] For the first current sequence, determining a current current state based on a long short-term memory network model, and generating a first fuse instruction;
[0054] Based on the first fusing instruction, the heating element is controlled to perform a first processing operation on the wire inside the fuse.
[0055] Compared with the prior art, the beneficial effects of the present application are as follows: by real-time monitoring and intelligent analysis of current changes, the present application can predict abnormal conditions in the circuit faster and more accurately than traditional fuses, and dynamically adjust the protection strategy according to the prediction results, thereby avoiding malfunctions caused by instantaneous current shocks and improving the accuracy and response speed of circuit protection. Users can flexibly set the delay time, current protection method and threshold according to actual needs, which increases the flexibility and adaptability of the fuse. The addition of the delay function enables the circuit to obtain more sophisticated protection, reduces the risk of equipment damage due to current anomalies, and improves the safety and reliability of the entire system. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present application and are used together with the specification to illustrate the technical solutions of the present application. It should be understood that the following drawings only illustrate certain embodiments of the present application and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can also be obtained based on these drawings without creative work.
[0057] Figure 1 A schematic diagram of the structure of a time-delay fuse based on microprocessor control provided in Example 1 of the present application;
[0058] Figure 2 A schematic diagram of the structure of a current detection unit provided in Example 1 of the present application;
[0059] Figure 3 A flow chart of a control method for a time-delay fuse based on microprocessor control provided in Example 2 of the present application.
[0060] The corresponding relationship between the illustrations and component names in the figure is as follows:
[0061] 10. Current detection unit; 11. Sensor; 12. Signal conversion device; 13. Data interface; 20. Microprocessor control unit; 30. Delay trigger unit. DETAILED DESCRIPTION
[0062] Some example embodiments will be described below with reference to the accompanying drawings. Although example embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.
[0063] As used herein, the term "including" and its variations mean open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "based at least in part on". The terms "example embodiment" and "certain embodiments" mean "at least one example embodiment". Other explicit and implicit definitions may also be included below.
[0064] The present application provides a time-delay fuse based on microprocessor control and a control method thereof. The embodiment of the present application is composed of three cores: a current detection unit 10, a microprocessor control unit 20, and a time-delay trigger unit 30. The current detection unit 10 is responsible for capturing the current data passing through the fuse and converting it into serialized information; the microprocessor control unit 20 receives these current data sequences and analyzes the data through a pre-trained long short-term memory network to identify the current change trend and possible abnormal mode, and generates a variety of fuse instructions including immediate fuse, pre-fuse, delayed fuse, etc. These instructions will directly guide the time-delay trigger unit 30 to perform corresponding temperature control actions to ensure that the circuit is effectively protected.
[0065] See also Figure 1-Figure 2 As shown, Figure 1 A schematic diagram of a time-delay fuse based on microprocessor control provided in an embodiment of the present application, including:
[0066] The current detection unit 10 is used to monitor the current value flowing through the fuse in real time and convert the monitored current value into a first current sequence;
[0067] The current value refers to the measurement of the flow of charge through a conductor, and its unit is ampere (Ampere, A). The current value reflects the amount of charge passing through the cross section of the conductor per unit time, and is one of the basic parameters for describing the state of a circuit. The information that the current value can contain includes: the intensity of the current, which is expressed as the amount of charge passing through a certain point in the circuit per unit time, which can be used to determine whether the circuit is within the normal working range; the direction of the current, which indicates the direction of the flow of the charge, and is used to determine the power flow and signal transmission of the circuit; the stability or fluctuation of the current, which indicates whether there are interference or unstable factors in the circuit; the current waveform, the current value of alternating current (AC) is not constant, but changes with time, and its waveform can be in the form of sine wave, square wave, sawtooth wave, etc.; the frequency, which describes the rate of change of the current waveform; the current timing information, in complex circuits or pulse circuits, the current timing information indicates the change of the current size over time.
[0068] The device used to monitor the current value is called a current sensor or current detector. Common current detection devices may include: current transformer, Hall effect sensor, shunt resistor and current clamp, etc.
[0069] It is understandable that the above devices and sensors can be selected in different types and specifications in different application scenarios to meet the accuracy and range of actual monitoring requirements.
[0070] As a possible implementation, the current detection unit 10 includes: a sensor 11, which is arranged near the fuse and is used to monitor the current value flowing through the fuse in real time and generate time information matching the current value; depending on the usage scenario, the type and specification of the selected sensor 11 and fuse will also be different. In a specific implementation, the selection of the sensor 11 can be based on the following parameters: range, sensitivity, speed, stability, and anti-interference ability.
[0071] For example, in a scenario where the current value varies over a large range, the parameter of range can be given priority when selecting sensor 11; in a scenario where the electromagnetic environment is complex, the parameter of anti-interference ability can be given priority when selecting sensor 11; in a high-precision scenario, parameters such as sensitivity and speed can be given priority when selecting sensor 11.
[0072] The signal conversion device 12 is electrically connected to the sensor 11 and is used to receive the current value flowing through the fuse and the time information matching the current value, and convert it into a first current sequence;
[0073] In the embodiment of the present application, the first current sequence is a time-current sequence. In a specific implementation, the change of the current over time can be recorded, the time can be divided into equally spaced moments, and the current value can be measured at each moment.
[0074] For example, Table 1 is a simplified current sequence example, where the current value is in ampere (A):
[0075] Table 1 A simplified example of current sequence
[0076] Time (seconds) Current (Ampere) 0 0.5 1 0.6 2 0.4 3 0.7 4 0.8 5 0.6 6 0.7
[0077] As a possible implementation, the signal conversion device 12, when adding a timestamp to the current value and converting its sequence into a first current sequence, includes: a signal enhancement module for amplifying the current value flowing through the fuse collected by the sensor 11 and filtering the amplified current value; an analog-to-digital conversion module for converting the current value represented in the form of an analog signal into a digital signal form. In a specific implementation, a non-inverting amplifier can be used to amplify the tiny current signal flowing through the fuse. The operational amplifier has a high input impedance and a low output impedance, ensuring that subtle current changes can be captured without introducing additional loads or distortion. Among them, in the non-inverting amplifier configuration, the input signal is fed into the non-inverting input terminal (+ terminal) of the operational amplifier, and the inverting input terminal (- terminal) is connected to the output through a feedback network. The characteristics and advantages of this configuration include: high input impedance, controllable gain, stability, and easy implementation and integration. During the amplification process, the gain of the operational amplifier can be set to a specific value, the size of which depends on the expected current range and the requirements of subsequent digital processing. For example, if the maximum current signal output by the sensor 11 is 1 mA, and the minimum effective signal processed by the microprocessor is 1 μA, the gain needs to be set to at least 1000 times to ensure that the signal can be effectively detected and processed after amplification.
[0078] After the current value flowing through the fuse is amplified, the signal enters a low-pass filter, which is used to remove possible high-frequency noise to ensure the purity of the signal. The low-pass filter may include, but is not limited to: a simple RC circuit or a more complex filter design, such as a Butterworth filter, a Chebyshev filter, and an elliptic filter. In a specific implementation, the filter can be selected according to the characteristics of the noise and the frequency range that needs to be filtered.
[0079] The amplified and filtered current value flowing through the fuse can be sent to an analog-to-digital converter (ADC) to convert the analog current signal into a digital signal that can be processed by the microprocessor.
[0080] Among them, an analog-to-digital converter (ADC) is an electronic device or integrated circuit used to convert an analog signal (usually a continuous, variable voltage signal) into a digital signal (usually a discrete, binary value). This conversion is important for computers and digital circuits to process signals in the analog world, because computers and other digital devices cannot directly understand and process analog signals. In a specific implementation, the selected ADC should have sufficient resolution and sampling rate to ensure that the converted digital signal can accurately represent the original analog signal. In the fuse, a slight increase in current is detected before an overload or short circuit, so high resolution is the first parameter to be considered when selecting an ADC. Exemplarily, if the resolution of the ADC is not enough to distinguish the slight increase in current before the fuse is about to blow, the system may not be able to respond in time, resulting in the failure of the protection measure. For the sampling rate of the ADC, the embodiment of the present application monitors the current change in real time or quasi-real time so as to respond quickly when an abnormality is detected. In this case, the sampling rate of the ADC needs to be high enough to capture the rapid changes in current. If the sampling rate is too low, critical changes may be missed, such as a sudden increase in current, causing the system to delay blowing the fuse. According to the Nyquist theorem, the ADC sampling rate should be at least twice the highest frequency of the monitored current changes to avoid aliasing.
[0081] For example, if the maximum frequency of the current signal of the current value flowing through the fuse to be detected is 1kHz, according to the Nyquist theorem, in order to avoid aliasing and accurately reconstruct the signal, the ADC sampling rate needs to be at least twice the maximum frequency of the signal, that is, 2kHz. In order to achieve effective monitoring, the ADC in the system should be set to a sampling rate higher than 2kHz, such as 2.5kHz or higher, to provide a certain margin. In this way, even in the most unfavorable conditions, it can be ensured that the signal will not be distorted and all important characteristics of the signal can be accurately captured.
[0082] Finally, the signal conversion device 12 adds a timestamp to the digital current value of each sampling point. This timestamp can be generated by a high-precision clock to ensure that each current value can be accurately associated with a specific time point; the data interface 13 is used to provide a data communication function with the microprocessor control unit 20 to transmit the converted first current sequence.
[0083] Through the above method, the signal enhancement module ensures accurate amplification and clear filtering of the current signal, and converts the analog signal into a digital signal, providing the microprocessor with high-quality, time-serialized current data, so that the microprocessor-based time-delay fuse system can monitor and analyze current changes and improve the efficiency and reliability of circuit protection.
[0084] The microprocessor control unit 20 is used to receive the first current sequence from the current detection unit 10, determine the current current state based on the long short-term memory network model, and generate a first fuse instruction;
[0085] As a possible implementation, the microprocessor control unit 20 includes: receiving a first current sequence from the current detection unit 10, and after the current sequence is processed by the signal enhancement module, it is also necessary to perform data normalization, signal segmentation, detrending, outlier detection and processing operations. Since the above are common means in the field of data preprocessing, this application will not be expanded here. By performing these preprocessing operations, the microprocessor control unit 20 can optimize the current sequence data to prepare for subsequent analysis, pattern recognition or decision making. These operations help to improve the system's response speed and accuracy to current changes and ensure timely and effective circuit protection.
[0086] The preprocessed first current sequence will be processed by the pre-trained long short-term memory network model to identify the changing trend and potential abnormal mode of the current value flowing through the fuse, generate first identification data, and generate a first fuse instruction based on the first identification data; for the update of the network model, the long short-term memory network model can be updated using the real-time monitored current value and the fuse operation result to optimize the prediction accuracy.
[0087] Among them, the change trend and potential abnormal modes of the current value flowing through the fuse may include: overload, short circuit, current fluctuation and starting current.
[0088] As a possible implementation, when the preprocessed first current sequence is input into the pretrained long short-term memory network model, the change trend and potential abnormal mode of the current value flowing through the fuse are determined, and the first identification data is generated, it includes: determining the first target feature based on the historical current data; as a possible implementation, the historical current data may include: time series current data, current events under different load conditions, abnormal event annotation data, transient current data, long-term operation data, and environmental factor impact data. The first target feature includes: current size, current change speed, current fluctuation frequency, current waveform characteristics, short-term and long-term current changes, and historical current abnormal events.
[0089] In a specific implementation, the data used to train the long short-term memory network model may not be collected from scratch in the usage scenario, because this will face insufficient data volume and poor prediction of the change trend and potential abnormal mode of the current value flowing through the fuse in the early stage of system operation. To address this problem, the embodiment of the present application adopts a pre-trained model in the early stage of the operation of the entire example system to ensure that there is a good change trend and potential abnormal mode prediction effect in the early stage of system operation; wherein, the pre-trained long short-term memory network model is trained with historical data.
[0090] The historical current data includes common change trends and abnormal patterns, which can be the operating data of other similar power systems or the operating data of a power system and its fuse before the embodiment of the present application is adopted. By collecting this data, a long short-term memory network model for predicting the change trend and potential abnormal patterns of the current value flowing through the fuse can be pre-trained.
[0091] When training a long short-term memory network (LSTM) model, feature selection is a key step because it directly affects the performance and accuracy of the model. Feature selection refers to the process of selecting relevant data as input variables from raw data. In the embodiment of the present application, features that are helpful for predicting future current states are extracted from historical current data. For time series data of current values, features can be created by defining a time window.
[0092] In a specific implementation, the maximum value, minimum value, average value, standard deviation, etc. of the current within a certain period of time in the past can be calculated and input into the LSTM model as features.
[0093] Exemplarily, the size of the time window needs to be defined first. For example, if one minute is selected as the time window and the sampling frequency of the current value data is once per second, the window size is 60 data points. Statistical features are calculated in each time window, which may include but are not limited to: average current value, maximum current value, minimum current value, standard deviation, and range of variation. The above statistics are used to form a feature vector. For example, for a specific time window, if the calculated average current value is 0.5A, the maximum value is 0.8A, the minimum value is 0.3A, the standard deviation is 0.1A, and the range of variation is 0.5A, then the feature vector of this time window can be expressed as [0.5, 0.8, 0.3, 0.1, 0.5]. Finally, in order to predict the current value of each second, the time window needs to slide according to a certain step size. For example, the window moves forward 1 second each time, and the statistical features of the window are recalculated and used to predict the current value of the next second.
[0094] Through the above steps, feature vectors reflecting the current changes in the past minute can be extracted from the time series current data, and then these feature vectors can be used as inputs to the LSTM model to predict future current values. This feature extraction method effectively captures the short-term trends and fluctuations of current data, which helps improve the performance of the LSTM model in time series prediction tasks.
[0095] It is understandable that the above exemplary description is intended to provide the basic operation of the feature vector formation. For the current magnitude, current change speed, current fluctuation frequency, current waveform characteristics, short-term and long-term current changes, and historical current abnormal events, the feature vector formation of these characteristics can refer to the above examples. The selection of the feature vector can be determined according to the actual application scenario of the system, and the embodiments of the present disclosure are not limited here.
[0096] It can be understood that for the selection of feature vectors, the rate of change of current or the first-order difference of current value (the difference between current values at two consecutive moments) can also be used as features to capture the trend of current change; if the current signal contains periodic changes, frequency domain features can be extracted by methods such as Fourier transform; various feature selection algorithms can be used to automatically select important features, such as model-based feature selection (for example, using feature importance of random forests), recursive feature elimination (RFE) or statistical-based methods.
[0097] Through the above method, the most useful feature set for the prediction model can be extracted from the historical current data. These features can not only help the LSTM model learn and predict the future state of the current more accurately, but also improve the interpretability of the model.
[0098] After determining the feature vector required for model training, it is necessary to match the event label corresponding to the feature vector. For example, if the current surges at a certain moment, it can be marked as a short circuit; if the current is zero at a certain moment, it can be marked as an open circuit; if the current fluctuates too much during a certain period of time, it can be marked as a current fluctuation. In this way, a data set for long-term and short-term memory network model training is made and trained. The pre-trained long-term and short-term memory network model will be deployed to the microprocessor control unit 20. After the model is deployed, the control unit of the operating microprocessor can receive the first current sequence sent by the current detection unit 10. The pre-trained long-term and short-term memory network model deployed in the control unit of the microprocessor receives the pre-processed current data, i.e., the first current sequence, in real time, analyzes the changing trend and potential abnormal pattern of the current signal, and generates the first identification data.
[0099] As a possible implementation method, the pre-trained long short-term memory network model receives the preprocessed current data in real time, determines the changing trend and potential abnormal pattern of the current signal, and generates the first identification data by utilizing the network model characteristics of LSTM, combined with the usage scenario of this application, to adapt the network model to the usage scenario, and modify some parameter characteristics.
[0100] In the Long Short-Term Memory Network (LSTM), memory cells are the core component that enables LSTM to capture long-term dependencies on time series data. Memory cells are designed to allow the network to store, update, and access long-term state information during training. Memory cells contain a long-term state line called the cell state, which runs through the entire chain of LSTM units with only minor linear interactions, allowing information to flow unimpeded along the chain. The cell state is similar to a conveyor belt where information can flow and is only affected by a few operations, which allows information to be retained for a long time.
[0101] LSTM controls the information flow of the cell state through three types of gates: forget gate, input gate, and output gate. These gates are a special structure that outputs a value between 0 and 1 through the σ activation function to determine how much information each gate allows to pass through. 0 means "no information is allowed to pass", while 1 means "all information is allowed to pass". Among them, the forget gate determines which information is discarded from the cell state. It looks at the state at the previous time point and the current input, and outputs a value between 0 and 1 for each number in the cell state. The input gate determines what value is updated to the cell state. It consists of two parts: the σ layer determines which values will be updated, and the tanh layer is used to create a new candidate value vector, which will be added to the cell state. The output gate determines the output at the next time point based on the cell state. Contains information about previous inputs for prediction or decision-making.
[0102] For the update of memory cells, during the forgetting process, the forget gate determines how much of the previous cell state to retain. This is achieved by multiplying the cell state with the output of the forget gate. If the forget gate output is close to 0, the information in the state will be "forgotten"; Information update: The input gate determines which parts of the new information will be stored in the cell state. The tanh layer creates a new candidate vector, which is multiplied by the output of the input gate and then added to the cell state; During the output generation process: The output gate controls which information will be used as the output of this LSTM unit. This is achieved by passing the cell state through the tanh function (ensuring that the output value is between -1 and 1) and multiplying it with the output of the output gate.
[0103] Through the above structure and process, the memory cells of LSTM can effectively maintain and transmit long-term information, while also having the flexibility to remove or add information, which makes LSTM very effective in processing long sequence data.
[0104] Based on the above introduction to LSTM, in the specific implementation, since the fuse device or the fuse module of the power system usually needs to monitor the current flowing through the system with high frequency and high precision, this means that a large amount of data will be generated, of which the data generated by the state of stable operation of the system will account for the majority. The amount of this information is huge, and if the network model used for prediction makes use of a large amount of data without learning value, it will also affect the learning and prediction effect of the network model. Based on this, information can be retained or discarded from the memory cell to ensure the high efficiency of the data information for the prediction performance of the network model. The function expression is:
[0105] f t =σ(W f *[h t-1 ,x t ]+b f )
[0106] Among them, f t represents the output of the forget gate, which is used to decide which information is retained or discarded; σ is the activation function, which ensures that the output value is between 0 and 1; W f is the weight matrix of the forget gate; h t-1 Represents the previous state of the current data; x t is the input at the current time point, that is, the current data monitored in real time; b f is the bias term of the forget gate;
[0107] The input gate is used to update the memory cell information and determine the current data of the new input at the current time point. The function expression is:
[0108]
[0109] Among them, t and t-1 represent the current time point and the previous time point respectively, i t Represents new information added to memory cells; Indicates the candidate value for updating the memory cell, reflecting the new anomaly or trend that the current current data may indicate; W i and W C are the weight matrices of the input gate and candidate memory cells respectively; b i and b C is the corresponding bias term;
[0110] Based on the information of the forget gate and the input gate, the cell state is updated, and the function expression is:
[0111]
[0112] Among them, C t is the memory cell state at the current time step, combined with past information C t-1 and current candidate information This ensures that the LSTM network model can maintain a comprehensive understanding of current changes at consecutive time points;
[0113] The output gate is used to determine the output based on the current memory cell, predict the upcoming state of the current and identify potential abnormal patterns. The function expression is:
[0114]
[0115] Among them, t To control the state of the memory cell C t To output state h t The conversion degree function, h t Represents the output at the current time point, that is, the model's final recognition and judgment of the current state; W o is the weight matrix of the output gate; b o is the bias term of the output gate;
[0116] The first identification data is generated based on the upcoming state of the electric current and the potential abnormal pattern.
[0117] For example, assuming the current values (in amperes) shown in Table 2, these values are collected and recorded by the sensor 11 every second.
[0118] Table 2 An example of a current sequence including an overload signal
[0119] Time (seconds) Current value (ampere) 0 0.5 1 0.6 2 0.7 3 0.8 4 1.5 (overload signal) 5 0.6 6 0.7
[0120] The following features can be extracted from the above data: Average current value: the average of the current values within a certain time window (for example, the past 5 seconds); Maximum current value: the maximum current value within the same time window. Current increment: the difference between the current value and the current value in the previous second. Current fluctuation frequency: the number of times the current value fluctuates (rises and falls) within a time window.
[0121] The features extracted above are used as the input of the LSTM network. For example, at the 5th second, the feature vector can be shown in Table 3:
[0122] Table 3. Example table of a feature vector and its corresponding value
[0123] feature value Average current value 0.72 Maximum current value 1.5 Current increment 0.7 Current fluctuation frequency 1
[0124] The above features are input into the LSTM network for processing. The LSTM model processes the feature vector and generates output through the following steps:
[0125] 1. Forget gate (f t ): Determines what information will be forgotten from the cell state. For example, if the current value did not fluctuate much over the past few seconds, the frequency of the previous current fluctuations may be forgotten.
[0126] 2. Input gate (i t ) and candidate cell states The cell state is updated based on the current value and the current increment. The input gate can choose to make the current increment have a greater impact on the cell state because this may indicate that there is an abnormality in the circuit.
[0127] 3. Update cell status (C t ): Updated by combining the decision of the forget gate and the new candidate cell state.
[0128] 4. Output gate (o t ) and output (h t ): The output gate determines what to output based on the current cell state. If the cell state shows abnormal current flow, such as an average current value higher than the normal range, the output may trigger a warning signal.
[0129] If the LSTM model predicts that the current state may cause circuit abnormality (such as predicting overload), the microprocessor control unit 20 will generate a corresponding preset fuse instruction. In response to this instruction, the delay trigger unit 30 controls the heating element to heat the fuse, thereby disconnecting the circuit when necessary.
[0130] Understandably, the above example ignores the complexity of model training and the need for a large amount of historical data. In the specific implementation, the LSTM network is trained with a large amount of historical current data to identify various patterns of current values, including normal operation and various fault conditions. The accuracy of the prediction will increase with the increase of data volume and the deepening of model training.
[0131] For the delay trigger unit 30, based on the first fusing instruction generated by the microprocessor control unit 20, the heating element is controlled to perform a first processing operation on the wire inside the fuse.
[0132] In the embodiment of the present application, after the microprocessor control unit 20 makes a state judgment and a future state prediction for the current value of the current passing through the fuse, a fuse instruction for controlling the delay trigger unit 30 can be generated based on the judgment and prediction results. After receiving the fuse instruction, the delay trigger unit 30 can perform preset related protection circuits or fuse-related operations on the wires inside the fuse according to the content of the instruction.
[0133] Among them, the embodiment of the present application presets fuse instructions with different contents in combination with the actual operation conditions of the power system. These fuse instructions with different contents will control the delay trigger unit 30 to perform corresponding preset operations in the target scenario.
[0134] As a possible implementation, the first processing operation includes: heating to the target temperature immediately, heating to the target temperature slowly, heating to the target temperature immediately after a preset time interval, and heating to the target temperature slowly after a preset time interval; the first fusing instruction includes: an immediate fusing instruction, an immediate slow fusing instruction, a pre-fusing instruction, a delayed immediate fusing instruction, and a delayed slow fusing instruction; the generating of the first fusing instruction based on the first identification data includes:
[0135] In response to the presence of preset abnormal data in the first identification data, the first fuse instruction is generated based on the preset abnormal data type.
[0136] In this application, the first fuse instruction includes a variety of different response types, which allows the system to respond adaptively to different abnormal conditions or predicted results. Each fuse instruction corresponds to a specific processing method, the purpose of which is to provide a flexible and detailed protection mechanism so that the fuse can provide appropriate responses in different situations. The following are the situations in which these fuse instructions may be applicable:
[0137] 1. Immediate fuse instruction: This is applicable to emergency situations, such as when a serious circuit fault is detected (such as a large short circuit event), the system must immediately cut off the current to protect the circuit.
[0138] 2. Immediate slow fuse instruction: used in less urgent situations, allowing the system to transition to the fuse state more smoothly. It may be suitable for overload situations, but immediate fuse is not required.
[0139] 3. Pre-fuse instruction: used when the system predicts problems that may occur in the future, allowing the system to respond in advance to reduce the potential impact of sudden fuses.
[0140] 4. Delayed immediate fuse instruction: This instruction is suitable for scenarios where the current needs to be cut off immediately after a short delay. It may be used under certain specific conditions, such as waiting for the system to run to a safer or more suitable time point for fuse.
[0141] 5. Delayed slow fuse instruction: Similar to delayed immediate fuse, this is to provide an additional time buffer after identifying the potential problem, and then gradually introduce the fuse state. It is used in non-urgent scenarios where gradual intervention is predicted.
[0142] The above different instructions make the system more intelligent and flexible, and can provide customized responses according to different operating conditions and predicted results. It can not only protect the circuit from immediate damage, but also reduce equipment downtime and repair costs caused by fuse errors or overly aggressive reactions. In addition, it also provides a preventive strategy that can take measures before circuit failure occurs, thereby optimizing the performance and reliability of the entire system.
[0143] As a possible real-time method, the delay trigger unit 30, when controlling the heating element to perform the first processing operation on the wire inside the fuse based on the first fusing instruction generated by the microprocessor control unit 20, includes:
[0144] In response to the first fusing instruction generated by the microprocessor control unit 20 being an immediate fusing instruction, controlling the heating element to immediately raise the temperature to a first target temperature;
[0145] In response to the first fusing instruction generated by the microprocessor control unit 20 being a slow fusing instruction, the heating element is controlled to gradually increase the temperature to a first target temperature based on a preset heating mode;
[0146] In response to the first fusing instruction generated by the microprocessor control unit 20 being a pre-fusing instruction, controlling the heating element to immediately raise the temperature to a second target temperature;
[0147] In response to the first fusing instruction generated by the microprocessor control unit 20 being a delayed immediate fusing instruction, the heating element is controlled to immediately raise the temperature to a first target temperature after a preset time interval;
[0148] In response to the first fusing instruction generated by the microprocessor control unit 20 being a delayed slow fusing instruction, the heating element is controlled to gradually increase the temperature to the first target temperature based on a preset heating mode after a preset time interval.
[0149] In a specific implementation, when the delay trigger unit 30 receives an "immediate fuse instruction", the delay trigger unit 30 will drive the heating element to output maximum power instantly, and quickly heat the wire inside the fuse to the fuse temperature; when the delay trigger unit 30 receives an "immediate slow fuse instruction", the delay trigger unit 30 controls the heating element to gradually heat up according to a predetermined temperature rise curve, and may slowly reach the target temperature within a few minutes or seconds, and fuse the wire in a progressive manner; when the delay trigger unit 30 receives a "delayed immediate fuse instruction", the heating element is delayed for a certain period of time, such as a preset few minutes or hours, and then performs the operation of immediately heating to the target temperature. This method can be used for scheduled maintenance or non-emergency fuse; similarly, when the delay trigger unit 30 receives a "delayed slow fuse instruction", the heating element will begin to slowly heat up to the fuse point after a predetermined time delay. This operation can be used for preventive maintenance or when the system predicts that the circuit will be overloaded at a certain point in the future.
[0150] These different heating strategies and operations give the system greater flexibility in handling current anomalies, allowing the fuse to respond to a wide range of scenarios from immediate emergency response to predetermined delayed response, thereby ensuring the safety of the circuit and minimizing interference with overall system operation.
[0151] Example 2
[0152] Based on the same inventive concept, the embodiment of the present application also provides a control method corresponding to a time-delay fuse based on microprocessor control. Since the principle of solving the problem by the control method in the embodiment of the present application is similar to that of the above-mentioned time-delay fuse based on microprocessor control in the embodiment of the application, the implementation of the control method can refer to the implementation of Example 1, and the repeated parts will not be repeated.
[0153] See also Figure 3 , is a flow chart of a control method for a time-delay fuse based on microprocessor control provided in Example 2 of the present application.
[0154] A time-delay fuse control method based on microprocessor control, comprising:
[0155] S101, monitoring the current value flowing through the fuse in real time, and converting the monitored current value into a first current sequence;
[0156] S102, for the first current sequence, determining a current current state based on a long short-term memory network model, and generating a first fuse instruction;
[0157] S103: Based on the first fusing instruction, control the heating element to perform a first processing operation on the wire inside the fuse.
[0158] Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present application.
[0159] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0160] In the several embodiments provided in the present application, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the coupling or communication connection between devices or units can be electrical or other forms.
[0161] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A time-delay fuse based on microprocessor control, characterized in that: include: A current detection unit, used for monitoring the current value flowing through the fuse in real time, and converting the monitored current value into a first current sequence; A microprocessor control unit, configured to receive a first current sequence from a current detection unit, determine a current current state based on a long short-term memory network model, and generate a first fuse instruction; A time delay trigger unit, for controlling the heating element to perform a first processing operation on the wire inside the fuse based on a first fusing instruction generated by the microprocessor control unit; The current detection unit comprises: The sensor is arranged near the fuse and is used to monitor the current value flowing through the fuse in real time and generate time information matching the current value; A signal conversion device, connected to the sensor, for receiving the current value flowing through the fuse and time information matching the current value, and adding a timestamp to the current value to convert its sequence into a first current sequence; A data interface, used to provide a data communication function with a microprocessor control unit to transmit the converted first current sequence; The signal conversion device, when adding a timestamp to the current value and converting its sequence into a first current sequence, includes: A signal enhancement module is used to amplify the current value flowing through the fuse collected by the sensor and filter the amplified current value; An analog-to-digital conversion module, used for converting the current value represented by the analog signal into a digital signal form; The microprocessor control unit comprises: receiving the first current sequence from a current detection unit, and performing a preprocessing operation on the received first current sequence; Inputting the preprocessed first current sequence into a pretrained long short-term memory network model, determining a change trend and a potential abnormal mode of the current value flowing through the fuse, and generating first identification data; wherein the change trend and the potential abnormal mode of the current value flowing through the fuse include: overload, short circuit, current fluctuation and starting current; Based on the first identification data, generating a first fuse instruction; The real-time monitored current value and fuse operation results are used to update the long short-term memory network model to optimize the prediction accuracy.
2. The time-delay fuse based on microprocessor control according to claim 1, characterized in that: Inputting the preprocessed first current sequence into the pretrained long short-term memory network model, determining the change trend and potential abnormal mode of the current value flowing through the fuse, and generating first identification data, including: Get historical current data; Based on the historical current data, determining a first target feature; Based on the first target features and the event labels corresponding to the first target features, construct the pre-trained long short-term memory network model; Deploying the pre-trained long-short term network model into a microprocessor control unit; The pre-trained long short-term memory network model receives the pre-processed current data in real time, determines the change trend and potential abnormal mode of the current signal, and generates first recognition data.
3. The time-delay fuse based on microprocessor control according to claim 2, characterized in that: The historical current data includes: time series current data, current events under different load conditions, abnormal event annotation data, transient current data, long-term operation data and environmental factor impact data; The first target characteristics include: current magnitude, current change speed, current fluctuation frequency, current waveform characteristics, short-term and long-term current changes, and historical current abnormal events.
4. The time-delay fuse based on microprocessor control according to claim 2, characterized in that: The pre-trained long short-term memory network model includes: a forget gate, an input gate, an output gate and a memory cell; the pre-trained long short-term memory network model receives the pre-processed current data in real time, determines the change trend and potential abnormal mode of the current signal, and generates first recognition data, including: The function expression for determining whether information is retained or discarded from the memory cell is: ; in, Represents the output of the forget gate, which is used to decide which information to keep or discard; is an activation function that ensures that the output value is between 0 and 1, where 0 means no information is allowed to pass through and 1 means all information is allowed to pass through; is the weight matrix of the forget gate; Represents the previous state of the current data; It is the input at the current time point, that is, the current data monitored in real time; is the bias term of the forget gate; The input gate is used to update the memory cell information and determine the current data of the new input at the current time point. The function expression is: ; Among them, t and t-1 represent the current time point and the previous time point respectively. Represents new information added to memory cells; represents the candidate value for updating the memory cell, reflecting the new anomaly or trend that the current current data may indicate; and are the weight matrices of the input gate and candidate memory cells respectively; and is the corresponding bias term; Based on the information of the forget gate and the input gate, the cell state is updated, and the function expression is: ; in, is the state of the memory cell at the current time point, combined with past information and current candidate information ; The output gate is used to determine the output based on the current memory cell, predict the upcoming state of the current and identify potential abnormal patterns. The function expression is: ; in, It is the output gate and also controls the state of the memory cell. To output state The conversion degree function, Represents the output at the current time point, that is, the model's final recognition and judgment of the current state; is the weight matrix of the output gate; is the bias term of the output gate; The first identification data is generated based on the upcoming state of the electric current and the potential abnormal pattern.
5. The time-delay fuse based on microprocessor control according to claim 1, characterized in that: The first processing operation includes: heating to the target temperature immediately, heating to the target temperature slowly, heating to the target temperature immediately after a preset time interval, and heating to the target temperature slowly after a preset time interval; the first fusing instruction includes: an immediate fusing instruction, an immediate slow fusing instruction, a pre-fusing instruction, a delayed immediate fusing instruction, and a delayed slow fusing instruction; the generating of the first fusing instruction based on the first identification data includes: In response to the presence of preset abnormal data in the first identification data, the first fuse instruction is generated based on the preset abnormal data type.
6. The time-delay fuse based on microprocessor control according to any one of claims 1 to 5, characterized in that: The delay trigger unit, when controlling the heating element to perform a first processing operation on the wire inside the fuse based on the first fusing instruction generated by the microprocessor control unit, includes: In response to the first fusing instruction generated by the microprocessor control unit being an immediate fusing instruction, controlling the heating element to immediately raise the temperature to a first target temperature; In response to the first fusing instruction generated by the microprocessor control unit being a slow fusing instruction, controlling the heating element to gradually increase the temperature to a first target temperature based on a preset heating mode; In response to the first fusing instruction generated by the microprocessor control unit being a pre-fusing instruction, controlling the heating element to immediately raise the temperature to a second target temperature; In response to the first fusing instruction generated by the microprocessor control unit being a delayed immediate fusing instruction, controlling the heating element to immediately raise the temperature to a first target temperature after a preset time interval; In response to the first fusing instruction generated by the microprocessor control unit being a delayed slow fusing instruction, the heating element is controlled to gradually increase the temperature to a first target temperature based on a preset heating mode after a preset time interval.
7. A time-delay fuse control method based on microprocessor control, which is implemented based on the time-delay fuse based on microprocessor control according to any one of claims 1 to 6, characterized in that: include: Real-time monitoring of the current value flowing through the fuse, and conversion of the monitored current value into a first current sequence; For the first current sequence, determining a current current state based on a long short-term memory network model, and generating a first fuse instruction; Based on the first fusing instruction, the heating element is controlled to perform a first processing operation on the wire inside the fuse.
