Intelligent fuse

Abstract

The invention provides an intelligent fuse, and belongs to the technical field of circuit protection, and the intelligent fuse is characterized in that an upper housing is provided with at least one excitation chamber, the excitation chamber is internally provided with at least one excitation part and at least one impact part, and the excitation part releases high-pressure gas when receiving any one of an external trigger signal and an internal trigger signal; the lower shell is provided with at least one cooling cavity; the current-conducting plate is arranged between the upper shell and the lower shell, and the impact part moves towards the direction where the cooling cavity is located under the thrust action of the high-pressure gas so as to cut off the current-conducting plate; the self-control module is arranged outside the shell and comprises a current detection unit and a processing unit, and the current detection unit is used for detecting a current signal on the conductive plate; the processing unit is used for generating an internal trigger signal according to the current signal. The beneficial effects of the invention are that a dual-trigger mechanism and redundancy design of an internal trigger signal and an external trigger signal are adopted, so that normal action can still be ensured when any signal fails, and the breaking reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of circuit protection technology, and in particular to an intelligent fuse. Background Technology

[0002] With the rapid development of high-voltage, high-capacity power electronics technologies such as electric vehicles and energy storage systems, circuit systems are placing higher demands on protection devices. These devices must provide reliable protection over a wide current range, capable of rapidly interrupting extremely large short-circuit currents while also effectively protecting against small overload current faults.

[0003] Currently, fuses are a widely used protective device. They mainly rely on the thermal accumulation effect to achieve their protective function. When the current in the circuit exceeds a certain value, the fuse element melts due to heat, thereby cutting off the circuit. However, fuses have limited protection range, low breaking reliability, poor system fault tolerance, and are difficult to adapt to complex and changing operating conditions. Summary of the Invention

[0004] To address the above technical problems, this invention provides an intelligent fuse to solve the problems of slow response speed, low breaking reliability, and lack of intelligent control capabilities in existing high-voltage circuit protection devices. By integrating intelligent detection, adaptive judgment, and rapid mechanical cutting technology, it achieves rapid, reliable, and intelligent cutting off of abnormal currents in high-voltage circuits, thereby improving the safety and stability of the power system.

[0005] The technical problem solved by this invention can be achieved by the following technical solutions: A smart fuse, comprising: The housing includes an upper housing and a lower housing. The upper housing is provided with at least one excitation chamber, and each excitation chamber is provided with at least one excitation component and at least one impact component. The excitation component is used to release high-pressure gas when it receives either an external trigger signal or an internal trigger signal. The lower housing is provided with at least one cooling chamber. A conductive plate is disposed between the upper housing and the lower housing, and the impact component is used to move toward the direction of the cooling chamber under the thrust of the high-pressure gas to cut off the conductive plate; The self-control module is located outside the housing. The self-control module includes a current detection unit and a processing unit. The current detection unit is used to detect the current signal on the conductive plate. The processing unit is connected to the current detection unit and is used to generate the internal trigger signal based on the current signal.

[0006] Preferably, the self-control module further includes: A signal preprocessing unit is connected between the current detection unit and the processing unit. It is used to filter and debouncing the current signal and transmit the filtered and debouncing current signal to the processing unit.

[0007] Preferably, the processing unit includes: The feature extraction unit is used to extract features from the current signal to obtain key feature parameters; A prediction unit is connected to the feature extraction unit. The prediction unit has at least one mathematical model pre-set to fit the key feature parameters and generate a predicted current. The decision unit, connected to the prediction unit, is used to generate the internal trigger signal when either the real-time current or the predicted current exceeds the corresponding preset current threshold.

[0008] Preferably, the prediction unit includes: The first prediction subunit is pre-set with a first mathematical model based on autoregressive integral moving average, which is used to perform differential processing on the key feature parameters and use the moving average term to suppress random noise and generate the first predicted current within a preset time period in the future. The decision unit is connected to the first prediction subunit and is used to generate the internal trigger signal when the first predicted current exceeds the corresponding first preset current threshold.

[0009] Preferably, the prediction unit further includes: The second prediction subunit is pre-configured with a second mathematical model based on machine learning, which is used to generate a second predicted current for the key feature parameters; the second mathematical model based on machine learning has pre-learned the correspondence between the key feature parameters of the current signal and the predicted current. The decision unit is connected to the second prediction subunit and is used to perform weighted processing on the first prediction current and the second prediction current to obtain a third prediction current. When the third prediction current exceeds the corresponding second preset current threshold, the internal trigger signal is generated.

[0010] Preferably, the second mathematical model is an attention-based machine learning model, wherein the attention layer of the machine learning model assigns a higher weight coefficient to the key feature parameters during periods of sudden current changes than to the key feature parameters during periods of stability.

[0011] Preferably, the conductive plate has at least two weak points, and the impact component is disposed above the weak points.

[0012] Preferably, the cooling chamber is provided with an arc-extinguishing melt, which is connected in parallel with the conductive plate, and the arc-extinguishing melt is surrounded by an arc-extinguishing medium, which is used to receive the current on the conductive plate after the conductive plate is cut off to generate heat and melt when the temperature is higher than a preset temperature. The arc-extinguishing medium is used to cool the electric arc.

[0013] Preferably, a metal wire mesh is provided in the cooling chamber to absorb the heat generated during the cutting of the conductive plate.

[0014] Preferably, it further includes: An electrical connection module is provided, wherein the input terminals of the electrical connection module are respectively connected to an external signal trigger terminal and the self-control module, and the output terminals of the electrical connection module are respectively connected to at least one excitation component, for receiving the external trigger signal or the internal trigger signal and transmitting it to all the excitation components.

[0015] The advantages or beneficial effects of the technical solution of this invention are as follows: This invention employs a dual triggering mechanism, constructing a redundant design for internal and external triggering signals to ensure normal operation even if either signal fails, thereby improving the reliability of the disconnection. At the same time, it adopts a modular design, which facilitates maintenance and expansion. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the structure of a smart fuse in a preferred embodiment of the present invention; Figure 2 The circuit logic block diagram of the intelligent fuse in the preferred embodiment 1 of the present invention; Figure 3 The circuit logic block diagram of the intelligent fuse in the preferred embodiment 2 of the present invention; Figure 4 The circuit logic block diagram of the intelligent fuse in the preferred embodiment 3 of the present invention is shown. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0020] In a preferred embodiment of the present invention, based on the above-mentioned problems existing in the prior art, an intelligent fuse for high-voltage circuit systems is provided, which is suitable for electric vehicles, wind, solar and energy storage new energy scenarios. By integrating intelligent monitoring and programmable control, it realizes adaptive and predictive circuit protection and can quickly and reliably disconnect and protect against faults such as overload and short circuit.

[0021] like Figure 1 As shown, the smart fuse includes: The housing includes an upper housing 3 and a lower housing 5; the upper housing 3 is provided with at least one excitation chamber 9; the lower housing 5 is provided with at least one cooling chamber 10 for heat dissipation and maintaining stable operating temperature of the equipment. A conductive plate 4 is disposed between the upper housing 3 and the lower housing 5. The conductive plate 4 is used as the main current channel to be electrically connected to the protected circuit system. Each excitation chamber 9 is provided with at least one excitation component 1 and at least one impact component 2. The excitation component 1 is used to release high-pressure gas when receiving either an external trigger signal or an internal trigger signal. The impact component 2 is used to move toward the direction of the cooling chamber 10 under the thrust of the high-pressure gas to cut off the conductive plate 4. The self-control module 100 is located outside the housing. The self-control module 100 includes a current detection unit 12 and a processing unit 13. The current detection unit 12 is used to detect the current signal on the conductive plate. The processing unit 13 is connected to the current detection unit 12 and is used to generate an internal trigger signal based on the current signal.

[0022] Specifically, in this embodiment, at least one excitation component 1, at least one impact component 2, a conductive plate 4, and at least one cooling chamber 10 are sequentially arranged from top to bottom between the upper housing 3 and the lower housing 5. The excitation component can be a miniature gas generator (ignition tube), which can release high-pressure gas within milliseconds after receiving an electrical signal. The impact component 2 can be made of high-strength plastic, such as PA66 or PA6T, and can move linearly in the vertical direction within the excitation chamber 9. The conductive plate 4 can be made of T2 copper, is disposed between the upper housing 3 and the lower housing 5, and is electrically connected to an external circuit system via bolts. An excitation component 1 is disposed above each impact component 2. When the excitation component 1 receives a trigger signal, it releases high-pressure gas, driving the impact component 2 to break the conductive plate 4, thereby cutting off the main current path and achieving disconnection.

[0023] The trigger signal can be an external trigger signal. The external trigger signal can be an emergency cut-off command controlled by the Battery Management System (BMS) 15. When the intelligent fuse of this invention is used in high-voltage, high-capacity power electronics technologies such as electric vehicles and energy storage systems, the BMS system built into the electric vehicle or energy storage system detects the current on the conductive plate 4. When the current exceeds a preset current threshold, an external trigger signal is output to excite the impact component 2 to break the conductive plate 4.

[0024] The trigger signal can also be an internal trigger signal output by the aforementioned self-control module 100. When the self-control module 100 determines that the current signal on the conductive plate 4 exceeds a preset current threshold, it outputs an internal trigger signal to the excitation component 1. Compared to the BMS system, the detection and signal output delays within the self-control module 100 are lower, thus accelerating the response speed of the smart fuse.

[0025] The embodiments of the present invention employ a dual-start method using both external and internal trigger signals, which improves the reliability and fault tolerance of the disconnection process.

[0026] In this embodiment, the self-control module 100 is integrated into the side of the upper and lower housings, and mainly includes a current detection unit 12 and a processing unit 13. The self-control module 100 can be packaged in a modular housing independent of the upper housing 3 and the lower housing 5. This modular housing can be located above the conductive plate 4; it can also be located below the conductive plate 4; or it can be divided into two housings, with the first modular housing located above the conductive plate 4 and the processing unit 13 located inside the first modular housing, and the second modular housing located below the conductive plate 4 and the current detection unit 12 located inside the second modular housing. The positions of the current detection unit 12 and the processing unit 13 can also be interchanged.

[0027] The current detection unit 12 is connected between the conductive plate 4 and at least one load (LOAD1, LOAD2). In this embodiment, the current detection unit 12 includes a current sensor, preferably a Hall current sensor 121. The power supply terminal of the Hall current sensor 121 is connected to the power supply interface 16 for real-time monitoring of the current state on the conductive plate and acquisition of current signals. The current signal includes parameters such as the amplitude and waveform of the current, and the detected current signal is transmitted to the processing unit 13. The sampling frequency of the current detection unit 12 can reach over 100kHz, which can accurately capture transient current changes. The processing unit 13 receives the current signal detected by the current sensor, analyzes and processes the current signal according to the preset protection logic or the user-defined mathematical model, and determines whether there are abnormal conditions such as overcurrent or short circuit. The output of the internal trigger signal is controlled by a threshold judgment method. When the real-time current exceeds the set current threshold, such as 1.35 times the rated current for 5ms, the processing unit 13 immediately generates an internal trigger signal to start the excitation component 1.

[0028] Furthermore, the processing unit 13 can also predict the current at the next moment based on the current waveform fitting through the algorithm model. When it is predicted that the current is about to exceed the limit, an internal trigger signal is immediately generated to realize the pre-start function and complete the main path cutoff before the fault is fully formed.

[0029] To improve detection reliability and suppress interference, the smart fuse in this embodiment has signal filtering and debouncing functions. Furthermore, the self-control module 100 also includes: The signal preprocessing unit is connected between the current detection unit 12 and the processing unit 13. It is used to filter and debouncing the current signal and transmit the filtered and debouncing current signal to the processing unit 13.

[0030] Specifically, since existing fuse signal processing often uses a single filtering method, which is susceptible to electromagnetic interference, in this embodiment, the self-control module 100 also integrates a signal preprocessing unit, which is configured to perform digital filtering and debouncing processing on the raw current signal output by the current detection unit 12.

[0031] To improve signal quality, the filtering process includes both hardware-level and digital-level filtering. At the hardware level, an RC low-pass filter circuit with an adjustable cutoff frequency is configured at the output of the current detection unit 12 to suppress high-frequency noise. At the digital level, two stages of digital filtering are implemented within the processing unit 13 to eliminate high-frequency noise and transient interference. The first stage is a moving average filter with a window length L = 5-10 sampling points, used to smooth random noise fluctuations. The second stage is a dynamic limiting filter, setting the maximum rate of change between adjacent sample values ​​to ±5%, with any deviations considered interference and corrected, for example, by replacing it with the historical average.

[0032] Furthermore, the current detection unit 12 may also include a shunt 122, replacing the aforementioned current sensor. Compared to the Hall current sensor 121, the shunt 122 does not require an external power supply, making it suitable for applications without external interfaces. In particular, the filtering process for the shunt detection signal further includes median filtering to resist pulse interference.

[0033] Debouncing is achieved through a software debouncing counter combined with a majority voting strategy within a window. For example, a fault is triggered only when 80% of the sampling points meet the fault condition within a 5ms window. Unlike traditional timer-based debouncing, this avoids response lag caused by fixed delays. Specifically, when the real-time or predicted current exceeds a preset current threshold, the debouncing counter is activated. The counter accumulates the sampling results that continuously meet the fault condition. Only when the accumulated count reaches a preset number or time window is a valid fault identified, and an internal trigger signal is output. This effectively avoids malfunctions caused by signal jitter or electromagnetic interference, significantly reducing the risk of false triggers due to load switching, electromagnetic pulses, or instantaneous sensor drift.

[0034] In this embodiment, the preset number of times N is preferably 5 to 10. Of course, it is not limited to this, and the specific value of the preset number of times N can be customized by programming.

[0035] Furthermore, the aforementioned filtering and debouncing parameters, such as the filter window size and amplitude limiting threshold, can be dynamically and adaptively adjusted according to the current environment. For example, in high-noise scenarios such as electric vehicle motor drives, the moving average window is automatically increased to improve stability.

[0036] The embodiments of the present invention employ multi-level collaborative filtering to improve the signal-to-noise ratio, and combined with a debouncing mechanism, greatly reduce the false alarm rate; at the same time, the response delay of filtering and debouncing is controlled at the sub-millisecond level, providing a foundation for supporting the implementation of the pre-start mechanism.

[0037] Furthermore, the processing unit 13 includes: The feature extraction unit 132 is used to extract features from the current signal to obtain key feature parameters; The prediction unit is connected to the feature extraction unit 132. The prediction unit has at least one mathematical model pre-set to fit the key feature parameters and generate the predicted current. The decision unit, connected to the prediction unit, is used to generate an internal trigger signal when either the real-time current or the predicted current exceeds the corresponding preset current threshold.

[0038] In this embodiment, the processing unit 13 includes a processor, which is a high-performance embedded microcontroller, such as an ARM Cortex-M7 or DSP chip, with built-in non-volatile memory, supporting users to upload custom protection programs via communication interfaces (such as CAN bus or UART). The processing unit 13 can run various mathematical models, such as: performing spectral analysis on periodic load currents using Fourier transform; predicting peak values ​​for sudden short-circuit currents using exponential fitting; and predicting future trends for gradual overloads using an Autoregressive Integrated Moving Average (ARIMA) time series model.

[0039] The self-control module 100 can be customized according to circuit protection requirements. The processing unit 13 can fit the current waveform according to the programmed mathematical model, predict the current change trend, determine the current state according to the corresponding predicted current value, and control the output of the internal trigger signal through threshold judgment. When the predicted current is about to exceed the set current threshold, the internal trigger signal is output to pre-start the excitation component 1. By pre-starting the interruption, intervention is carried out before the fault occurs, shortening the response time.

[0040] Furthermore, the prediction unit includes: The first prediction subunit is pre-set with a first mathematical model based on autoregressive integral moving average, which is used to perform differential processing on key feature parameters and use the moving average term to suppress random noise and generate the first predicted current within a preset time period in the future. The decision unit is connected to the first prediction subunit and is used to generate an internal trigger signal when the first predicted current exceeds the corresponding first preset current threshold.

[0041] Specifically, compared to existing smart fuses that typically rely on simple threshold comparisons or thermal accumulation effects, this embodiment achieves predictive protection based on a mathematical model. The first mathematical model is an autoregressive integral moving average (ARIMA) model. This invention is the first to integrate the ARIMA model into the fuse's self-control module 100 for real-time current trend prediction.

[0042] This ARIMA model is designed for high-voltage circuit characteristics, such as the pulse current of electric vehicles and the gradual overload of energy storage systems. It dynamically calibrates model parameters, such as the autoregressive order p, the difference order d, and the moving average order q, using historical fault data. For example, the model employs a short-term window, such as 5-10 sampling points, for fitting to balance response speed and accuracy.

[0043] The model performs differential processing on the current waveform to eliminate non-stationarity and uses a moving average term to suppress random noise, thereby predicting the current change trend within the next few milliseconds. When the ARIMA model predicts that the current is about to exceed the set current threshold, it predicts the peak short-circuit current through exponential fitting. The processing unit 13 generates an internal trigger signal in advance, triggering the pre-start of the excitation component 1. Compared with the passive response of traditional fuses, this achieves active intervention.

[0044] By using the predictive interruption of the ARIMA model, this invention reduces the fuse response time from the traditional millisecond level to the microsecond level, preventing the complete formation of fault current and significantly reducing arc energy and system damage. For example, under the acceleration conditions of electric vehicles, the model can distinguish between normal pulse current and fault current, preventing false tripping, and simultaneously interrupting potential short circuits in advance, thus improving system safety.

[0045] Furthermore, the prediction unit also includes: The second prediction subunit is pre-set with a second mathematical model based on machine learning, which is used to generate a second predicted current for key feature parameters; the second mathematical model based on machine learning has pre-learned the correspondence between the key feature parameters of the current signal and the predicted current; The decision unit is connected to the second prediction subunit, which is used to weight the first prediction current and the second prediction current to obtain the third prediction current, and generate an internal trigger signal when the third prediction current exceeds the corresponding second preset current threshold.

[0046] Specifically, in this embodiment, the machine learning model 134 is used for real-time current fault identification of high-voltage fuses, and is fused with other models, such as the ARIMA model, to achieve accurate identification of complex fault modes (such as intermittent arcing and gradual overload).

[0047] The machine learning model 134 employs a lightweight neural network, such as a one-dimensional convolutional neural network (CNN). The model input consists of multi-dimensional features of the current waveform, such as amplitude, slope, and harmonic components, which are extracted in real-time by the signal processing unit 131. Model training utilizes domain-adaptive techniques, leveraging historical fault data from high-voltage circuits, such as electric vehicle BMS logs, for transfer learning. Training data includes various scenarios such as normal current, overcurrent, and short circuits, and electromagnetic interference noise is added to enhance generalization ability. The loss function combines classification accuracy and response time penalties to ensure the model minimizes latency while identifying faults.

[0048] The outputs of machine learning model 134 and ARIMA time series model 133 are input in parallel into weighted decision unit 135 (using fuzzy logic or weighted average algorithm), and the weights are dynamically adjusted according to the confidence level. For example, for periodic loads, the ARIMA model has higher weights; for sudden failures, the machine learning model has higher weights.

[0049] Machine learning model 134 can identify complex patterns that traditional methods cannot detect, such as distinguishing between instantaneous overcurrent during motor startup and actual faults, thus reducing the false alarm rate. Furthermore, through model fusion, the system improves prediction accuracy and reliability in gradual overload scenarios at energy storage power stations, far exceeding that of single models or threshold comparison methods.

[0050] Furthermore, the second mathematical model is an attention-based machine learning model 134, in which the attention layer of the machine learning model 134 assigns higher weight coefficients to the key feature parameters during periods of sudden current changes than to the key feature parameters during periods of stability.

[0051] Specifically, compared with the traditional model, this embodiment introduces an attention mechanism to prioritize the processing of data during periods of sudden current changes, thereby improving the sensitivity to sudden faults.

[0052] In particular, the self-control module 100 supports programmable functionality, allowing users to set parameters such as the protection current threshold, response time, and current change trend prediction algorithm according to different application scenarios. This enables highly customized circuit protection strategies, improving the adaptability and intelligence level of the fuse. Programmability allows the smart fuse to adapt to complex operating conditions, such as pulse current or gradual overload, enhancing system adaptability.

[0053] Furthermore, the conductive plate 4 has at least two weak points 6, and the impact component 2 is disposed above the weak points.

[0054] Specifically, in this embodiment, the conductive plate 4 is provided with at least two pre-designed weak points 6 to ensure that it can break quickly when the excitation component 1 is activated, thereby achieving controllable breakage.

[0055] Each impact component 2 is positioned directly above the weak point 6 of the conductive plate 4 to ensure concentrated impact force and reliable operation. Furthermore, the bottom cutter of the impact component 2 faces the weak point 6 on the conductive plate 4. Upon receiving a trigger signal, the excitation component 1 rapidly releases high-pressure gas, driving the impact component 2 downwards in a piston-like motion to strike the weak point 6 on the conductive plate 4, causing it to break instantly and thus cutting off the main current path.

[0056] In this embodiment, the thickness of the weak point 6 is 10%-30% of the thickness of the normal area of ​​the conductive plate 4. For example, four V-shaped grooves are pre-machined on the conductive plate 4 as weak points 6, and their thickness is only 10%-30% of the normal area, ensuring that it can break quickly when subjected to impact. Of course, in practical applications, the number and shape of the weak points 6 can be set according to specific needs, such as square grooves, U-shaped grooves, etc.

[0057] Furthermore, an arc-extinguishing melt 8 is provided in the cooling chamber 10. The arc-extinguishing melt 8 is connected in parallel with the conductive plate 4, and the arc-extinguishing melt 8 is surrounded by an arc-extinguishing medium 7. This medium is used to receive the current on the conductive plate 4 after the conductive plate 4 is cut off, generate heat, and melt when the temperature is higher than the preset temperature. The arc-extinguishing medium 7 is used to cool the electric arc.

[0058] Specifically, in this embodiment, arc-extinguishing melts 8 are connected in parallel at both ends of the conductive plate 4, and surrounded by arc-extinguishing medium 7, for rapid cooling and extinguishing of the electric arc.

[0059] The excitation component 1, impact component 2, conductive plate 4, and arc-extinguishing melt 8 are arranged sequentially from top to bottom between the upper housing 3 and the lower housing 5. When the excitation component 1 is activated, the high-pressure gas pushes the impact component 2 to strike the weak point 6 of the conductive plate 4, causing mechanical breakage and forcibly disconnecting the main circuit. At this time, the current is transferred to the arc-extinguishing melt 8 connected in parallel with the conductive plate 4. Under the action of the transferred current, the arc-extinguishing melt 8 heats up and melts rapidly, generating an electric arc. The arc-extinguishing medium 7 surrounding it cools rapidly and extinguishes the arc. At the same time, the cooling components in the lower housing 5 work together to further accelerate heat dissipation, prevent secondary arcing, and ensure that the circuit is completely and safely disconnected.

[0060] In this embodiment, the arc-extinguishing melt 8 is preferably made of copper-silver composite material.

[0061] In this embodiment, the arc-extinguishing medium 7 is preferably made of high-purity quartz sand or ceramic particles, etc.

[0062] Furthermore, a metal mesh 11 is provided inside the cooling chamber 10 to absorb the heat generated during the cutting of the conductive plate.

[0063] Specifically, in this embodiment, a metal wire mesh 11 is used as the heat dissipation structure of the fuse and is filled in the cooling chamber 10 to absorb the heat generated during the breaking process.

[0064] Furthermore, it also includes: Electrical connection module 14, the input terminal of electrical connection module 14 is connected to external signal trigger terminal and self-control module 100 respectively, and the output terminal of electrical connection module 14 is connected to at least one excitation component respectively, for receiving external trigger signal or internal trigger signal and transmitting it to all excitation components.

[0065] Specifically, the electrical connection module 14 includes an electrical connector, which has at least two input terminals and at least one output terminal. The number of output terminals of the electrical connection module 14 corresponds to the number of excitation components 1, and the number of output terminals is at least one.

[0066] When two or more excitation components 1 are provided, the electrical connection module 14 can simultaneously transmit the trigger signal to each excitation component 1. When the excitation component 1 receives either an external trigger signal or an internal trigger signal, it can be activated, thereby realizing a dual-start mechanism and significantly improving the reliability and fault tolerance of the fuse action.

[0067] Figure 1 The diagram shows two input terminals (IN1, IN2) and two output terminals (OUT1, OUT2). The first input terminal, IN1, receives external trigger signals from the battery management system 15 (BMS) or other higher-level control systems, such as transistor-transistor logic (TTL) or 24V level signals. The second input terminal, IN2, is connected to the output of the processing unit 13 of the self-control module 100 and receives internal trigger signals. The first output terminal, OUT1, and the second output terminal, OUT2, are connected to the corresponding excitation components 1 via wires. When either input terminal receives a valid trigger signal, the signal is synchronously transmitted to one or more excitation components 1 in parallel, ensuring that multiple impact components 2 operate simultaneously, improving disconnection consistency and reliability.

[0068] The working principle of the intelligent fuse of the present invention is as follows: When the circuit is abnormal, the external trigger signal output by the battery management system 15 and the internal trigger signal output by the self-control module 100 are transmitted to the excitation component 1 through the electrical connection module 14. After receiving any trigger signal, the excitation component 1 releases high-pressure gas, which pushes the impact component 2 to move perpendicularly to the conductive plate 4. After the impact, the weak point 6 on the conductive plate 4 breaks, and the current surges into the arc-extinguishing fuse 8 connected in parallel on the conductive plate 4. The arc-extinguishing fuse 8 overheats and melts, and the arc-extinguishing medium 7 cools the arc, thus completing the safe disconnection of the circuit.

[0069] This invention employs a dual triggering mechanism (internal / external trigger signals) to ensure operation even when either signal fails, combined with redundant design to improve disconnection reliability; a predictive model enables pre-start disconnection, intervening before a fault occurs and shortening response time; programmability allows the fuse to adapt to complex operating conditions, improving system adaptability; through a multi-level collaborative mechanism of hardware RC filtering, digital moving average / limiting filtering, and software debouncing counter, electromagnetic interference, load transients, and sensor noise are significantly suppressed, significantly reducing the false trip rate, making it suitable for high-noise new energy application scenarios; the modular design facilitates maintenance and expansion, and the integrated design of the cooling chamber 10 and the arc-extinguishing fuse 8 reduces volume and improves space utilization.

[0070] The following is in conjunction with the appendix Figure 2-4 The technical solution of the present invention is further illustrated through Examples 1-3.

[0071] Example 1 like Figure 2 As shown, the complete intelligent decision-making and protection process of the high-voltage intelligent fuse under complex operating conditions is as follows: When the system is working, the high-voltage intelligent fuse is located between the positive terminal of the battery and the load. The current detection unit 12 uses a Hall current sensor 121 that does not require isolation operation and is powered by the low-voltage interface within the system. The processing unit 13 adopts a parallel multi-signal processing and decision-making mechanism: On the one hand, the current detection unit 12 monitors the current on the conductive plate 4 in real time. Its output signal passes through the signal processing unit 131 and the feature extraction unit 132 in sequence to obtain the key feature parameters in the current waveform. Before entering the feature extraction, the signal first passes through the preprocessing unit to perform moving average filtering (window length L=8) and ±5% amplitude limiting filtering, and the persistence of the fault state is verified by the debouncing counter. Only when all 8 consecutive samples are determined to be overcurrent or predicted overlimit, the weighted decision unit 135 outputs the disconnection command, which effectively isolates transient interference under common operating conditions such as electric vehicle motor start-stop and relay operation.

[0072] On the other hand, these characteristic parameters are simultaneously fed into the ARIMA time series model 133 and the machine learning model 134 for intelligent analysis. The ARIMA time series model 133 is used to predict short-term current trends, while the machine learning model 134 can identify complex fault modes. The prediction results of the two intelligent models are uniformly fed into the weighted decision unit 135 for fusion analysis. This unit performs a comprehensive evaluation based on a preset weighting algorithm to form the final disconnect command. Once the decision is made, an internal trigger signal is simultaneously sent to multiple excitation components, such as the first excitation component MGG1 and the second excitation component MGG2, causing them to act synchronously and impact component 2 to break the weak point 6 on the conductive plate 4. At the same time, external trigger signals from the battery management system 15 (BMS) are continuously received to achieve a dual protection mechanism of internal and external triggering. After the main path is forcibly cut off, the fault current is transferred to the arc-extinguishing fuse 8 for safe extinguishing, thereby achieving reliable protection for the downstream load.

[0073] This invention combines traditional overcurrent protection with cutting-edge predictive intelligent analysis. By integrating multiple information sources through weighted decision-making, it achieves a leap from passive response to proactive prediction, significantly improving the speed, accuracy, and reliability of high-voltage circuit protection.

[0074] Example 2 like Figure 3As shown, a working mode for intelligent protection is provided, which relies entirely on the self-control module 100. This system uses an internal trigger signal as the sole activation source, forming a complete closed-loop control system for self-monitoring, analysis, and execution. Throughout the operation, the current detection unit 12 continuously monitors the load current and sends the detection signal to the signal processing unit 131 for preprocessing. Preprocessing includes a dual mechanism of filtering and debouncing: filtering employs a 5-point moving average combined with a dynamic amplitude limiting algorithm; debouncing is achieved through a state-holding counter, requiring the fault state to persist for at least 5ms (corresponding to N=5@1kHz sampling rate) before protection is triggered, ensuring stable operation of the system in complex electromagnetic environments. Subsequently, feature extraction is performed to obtain key feature parameters from the current waveform. These feature parameters are fed in parallel into the machine learning model 134 and the ARIMA time series model 133 for intelligent analysis: the machine learning model 134 is responsible for identifying complex fault modes, while the ARIMA time series model 133 focuses on predicting short-term current trends. The outputs of the two models are sent to the weighted decision unit 135 for fusion analysis, and the resulting decision signal is fed back to the internal trigger signal source, thus forming a complete closed-loop self-control system.

[0075] The intelligent fuse of Embodiment 2 of this invention can perform real-time evaluation of the current state entirely without relying on external commands, solely through an internal intelligent algorithm. When an anomaly is detected or a potential fault is predicted, it autonomously generates a trigger signal and initiates protection actions, achieving intelligent protection for multiple loads. The entire system obtains operating power through the power supply interface 16, forming a complete internal loop from signal acquisition and intelligent analysis to protection execution, demonstrating a highly autonomous circuit protection capability.

[0076] Example 3 like Figure 4 As shown, a high-voltage intelligent fuse intelligent protection system is provided that uses shunt 122 detection technology and relies entirely on the self-control module 100. In this system, the current detection unit 12 adopts shunt 122, which is connected in series in the main circuit. Compared with Hall current sensor 121, shunt 122 does not require external power supply and is suitable for application scenarios without external interfaces.

[0077] The entire protection action is triggered by an internal trigger signal, which simultaneously controls multiple excitation components 1, including MGG1 and MGG2. Specifically, the shunt 122 generates a small voltage signal proportional to the measured current when the load current flows through it. This signal is directly sent to the subsequent signal processing unit 131 for amplification and filtering. The filtering here uses a second-order Butterworth low-pass filter (cutoff frequency f_c = 1kHz) to effectively suppress switching power supply noise and high-frequency harmonics. At the same time, a median value filter is introduced in the digital domain to assist the moving average, further improving the anti-pulse interference capability. The debouncing logic adopts a majority voting strategy within a window. If more than 80% of the sampling points meet the fault conditions within a 5ms window, it is determined to be a valid fault, balancing response speed and reliability. Subsequently, the key characteristic parameters of the current are obtained by the feature extraction unit 132.

[0078] These parameters are fed in parallel into the ARIMA time series model 133 and the machine learning model 134 for intelligent analysis. The former predicts short-term current trends, while the latter identifies complex fault modes. The outputs of the two models are fed into the weighted decision unit 135 for fusion analysis, and the final decision signal is fed back to the internal trigger signal source, forming a complete closed-loop automatic control system. This design eliminates the need for external trigger commands and the independent power supply interface 16 required for the Hall sensor. It can achieve autonomous monitoring, trend prediction, and fault judgment simply by sampling through the shunt 122 and combining with the internal intelligent algorithm. When the system detects overcurrent or predicts potential risks, it autonomously generates a trigger signal to drive the excitation component 1 to break the weak point 6 on the conductive plate 4 and safely cut off the circuit using the arc-extinguishing fuse 8, thereby achieving a highly integrated, self-sufficient, and reliable circuit protection function.

[0079] The present invention is not limited to the above embodiments and can be adjusted according to actual needs: for example, the excitation device can be equipped with more independent excitation chambers 9 and excitation components 1 for the protection of higher voltage level systems; the self-control module 100 can be integrated with a wireless communication module to realize remote status monitoring and parameter configuration, etc.

[0080] In summary, this invention provides a high-voltage intelligent fuse with a reasonable structure, rapid response, and high degree of intelligence, which is suitable for application scenarios with extremely high safety requirements, such as electric vehicles, energy storage power stations, rail transit, and data centers, and has significant technological progress and practical value.

[0081] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. An intelligent fuse, characterized in that, The application relates to a self-control module for a circuit breaker. The self-control module comprises a current detection unit and a processing unit. The current detection unit is configured to detect a current signal on a conductive plate. The processing unit is connected to the current detection unit and configured to generate an internal trigger signal according to the current signal.

2. The intelligent fuse of claim 1, wherein, The self-control module further comprises a signal preprocessing unit connected between the current detection unit and the processing unit. The signal preprocessing unit is configured to filter and de-bounce the current signal and transmit the filtered and de-bounced current signal to the processing unit.

3. The intelligent fuse of claim 1, wherein, The processing unit comprises a feature extraction unit configured to extract features from the current signal to obtain key feature parameters. The processing unit further comprises a prediction unit connected to the feature extraction unit. The prediction unit is pre-stored with at least one mathematical model and configured to perform fitting processing on the key feature parameters to generate a predicted current. The prediction unit further comprises a decision unit connected to the prediction unit.

4. The intelligent fuse of claim 3, wherein, The decision unit is configured to generate the internal trigger signal when either a real-time current or the predicted current exceeds a corresponding preset current threshold. The prediction unit further comprises a first prediction subunit pre-stored with a first mathematical model based on autoregressive integrated moving average. The first prediction subunit is configured to perform difference processing on the key feature parameters and suppress random noise by using a sliding average term to generate a first predicted current in a future preset time period.

5. The intelligent fuse of claim 4, wherein, The decision unit is connected to the first prediction subunit and configured to generate the internal trigger signal when the first predicted current exceeds a corresponding first preset current threshold. The prediction unit further comprises a second prediction subunit pre-stored with a second mathematical model based on machine learning. The second prediction subunit is configured to generate a second predicted current for the key feature parameters.

6. The intelligent fuse of claim 5, wherein, The second mathematical model based on machine learning has learned a corresponding relationship between the key feature parameters of the current signal and the predicted current in advance.

7. The intelligent fuse of claim 1, wherein, The decision unit is connected to the second prediction subunit and configured to perform weighted processing on the first predicted current and the second predicted current to obtain a third predicted current. The decision unit is configured to generate the internal trigger signal when the third predicted current exceeds a corresponding second preset current threshold. The second mathematical model is a machine learning model based on attention. An attention layer of the machine learning model gives a higher weight coefficient to the key feature parameters in a current mutation period than to the key feature parameters in a stable period. The conductive plate is provided with at least two weak positions. The impact components are arranged above the weak positions.

8. The intelligent fuse of claim 1, wherein, An arc-extinguishing melt is arranged in the cooling chamber, which is connected in parallel with the conductive plate, and the arc-extinguishing melt is surrounded by an arc-extinguishing medium, which is used to receive the current on the conductive plate to generate heat after the conductive plate is cut off and to melt when the temperature is higher than a preset temperature, and the arc-extinguishing medium is used to cool the electric arc.

9. The intelligent fuse of claim 1, wherein, A wire mesh is arranged in the cooling chamber to absorb the heat generated during the cutting process of the conductive plate.

10. The intelligent fuse of claim 1, wherein, Further comprising: An electrical connection module, input ends of the electrical connection module are connected with an external signal trigger end and the self-control module respectively, and output ends of the electrical connection module are connected with the at least one excitation component respectively, which are used to receive the external trigger signal or the internal trigger signal and transmit to all the excitation components.

Images