Ring main unit internal arc fault rapid detection pressure relief system

Abstract

The invention discloses a rapid detection and pressure relief system for an arc fault in a ring main unit. The rapid detection and pressure relief system comprises a multi-channel sensing signal acquisition unit, a characteristic function calculation unit and a parallel short-time average and long-time average recursive calculation unit, the multi-channel fusion triggering judgment unit is used for calculating the ratio of the short-time average value to the long-time average value of each channel, dividing the ratio of each channel by a respective preset triggering threshold value to obtain the normalized triggering intensity of each channel, and carrying out comprehensive judgment according to preset multi-channel fusion logic; outputting an arc fault trigger signal when the fusion logic judgment is established; and a graded pressure relief control unit. According to the system, through a mode of combining multi-channel sensing fusion and graded pressure relief control, the problem that a traditional protection device is insufficient in response speed is solved, and the pressure relief response intensity can be adjusted in a self-adaptive mode according to the severity degree of an arc fault.

Description

Technical Field

[0001] This invention relates to the field of safety protection technology for medium-voltage switchgear, and in particular to a rapid detection and pressure relief system for internal arc faults in ring main units. Background Technology

[0002] Ring main units (RNBs) are widely used switching equipment in medium-voltage power distribution networks, responsible for functions such as segmentation, interconnection, and load distribution of power lines. RNBs are typically installed at various nodes in urban underground cable ring networks. Their operating environment is enclosed and compact, integrating high-voltage current-carrying components such as busbars, load switches, circuit breakers, and cable terminals. Due to prolonged operation under energized conditions and exposure to various factors such as temperature and humidity, condensation, insulation aging, and foreign object intrusion, there is a risk of arcing faults within the RNB.

[0003] Internal arcing is one of the most serious safety threats faced by ring main units. When insulation breakdown or phase-to-phase short circuits trigger an arc, the core temperature of the arc column can reach thousands or even tens of thousands of degrees Celsius. The arc energy causes the gas inside the chamber to expand rapidly, resulting in a pressure surge to several times the atmospheric pressure within tens of milliseconds. If the pressure exceeds the structural limits of the cabinet, it will cause the cabinet to burst, high-temperature gas to be ejected, and molten metal to splash, directly threatening the lives of nearby maintenance personnel and potentially causing fires and large-scale power outages.

[0004] Existing arc protection technologies for ring main units mainly fall into the following categories. First, there are traditional overcurrent protection schemes, which utilize relay protection devices to detect short-circuit current and trip the circuit breaker. However, their response time is typically 30-60 milliseconds or even longer, during which time the arc has already released a large amount of energy, and the cabinet structure may have been severely damaged. Second, there are arc detection schemes, which use photoelectric sensors to detect the strong light generated by the arc to quickly identify faults. However, single-light detection channels are easily affected by external strong light sources, leading to false trips. Furthermore, most existing arc protection devices only provide simple threshold comparison judgments and lack adaptive capabilities to background light drift. Third, there are passive protection schemes that involve setting up pressure relief channels on the cabinet. However, most operating equipment does not have dedicated pressure relief channels or the pressure relief channels are poorly designed. The internal arc release capacity has not been rigorously verified, and the pressure relief response lacks graded control capabilities that match the severity of the fault.

[0005] The aforementioned existing technologies share the following common shortcomings. Regarding detection speed, the response time of traditional overcurrent protection is on the same order of magnitude as the time window from the occurrence of an arc to structural damage, resulting in extremely limited safety margin. Regarding detection reliability, single-sensor channel detection schemes cannot simultaneously meet the contradictory requirements of speed and resistance to false tripping, and fixed threshold criteria cannot adapt to changes in operating conditions. Regarding pressure relief control, there is a lack of ability to provide graded responses based on the severity of the arc; either insufficient response leads to structural damage, or excessive pressure relief causes the escape of insulating gas.

[0006] Therefore, developing a ring main unit arc fault protection system that integrates multi-channel arc detection and graded pressure relief control is of great significance for improving the safe operation level of the power distribution network and ensuring the personal safety of operation and maintenance personnel. Summary of the Invention

[0007] The purpose of this invention is to provide a rapid detection and pressure relief system for arc faults inside a ring main unit. This system, through a combination of multi-channel sensor fusion and graded pressure relief control, not only solves the problem of insufficient response speed of traditional protection devices, but also adaptively adjusts the pressure relief response intensity according to the severity of the arc fault.

[0008] The above-mentioned technical objective of this invention is achieved through the following technical solution: The system includes a multi-channel sensor signal acquisition unit, a characteristic function calculation unit, a parallel short-time averaging and long-time averaging recursive calculation unit, a multi-channel fusion trigger determination unit, and a graded pressure relief control unit. The multi-channel sensor signal acquisition unit includes an arc light sensor installed inside each functional chamber of the ring main unit, a pressure sensor installed on the chamber wall, and a current sensor installed in each circuit. These three sensors respectively constitute an arc light sensing channel, a pressure sensing channel, and a current sensing channel. The three sensing channels are synchronously sampled by an analog-to-digital converter equipped with a synchronous sampling clock to obtain the discrete time sequence signal of each channel. The characteristic function calculation unit receives discrete time series signals from three channels and performs energy characterization transformation operations on the physical characteristics of each channel to generate characteristic function value sequences for each channel. The energy characterization transformation of the arc light sensing channel uses the square value of the signal to amplify the energy contrast of sudden changes in arc light intensity. The energy characterization transformation of the pressure sensing channel superimposes the weighted differential square term of adjacent sampling points on the square value of the signal to enhance the sensitivity to rapid pressure changes. The energy characterization transformation of the current sensing channel uses the square value of the difference between the current value and the current value in phase of the previous power frequency cycle to eliminate the normal power frequency fundamental component and extract the fault transient component. Parallel short-time and long-time average recursive calculation units perform moving average recursive calculations on the characteristic function value sequences of each channel using independently configured short-time and long-time windows. The number of sampling points in the short-time window is much smaller than that in the long-time window to achieve scale separation. The time scale of the short-time window matches the time scale of the signal rise edge after arc ignition to capture transient energy changes. The time scale of the long-time window is much larger than the expected rise time of the arc signal to continuously track the background signal baseline level of each channel. Furthermore, the long-time average value is frozen and not updated in the triggered state to prevent the background baseline from being pulled up by the event signal during the arc's duration. The multi-channel fusion trigger determination unit calculates the ratio of the short-time average value to the long-time average value of each channel. The ratio of each channel is divided by its respective preset trigger threshold to obtain the normalized trigger intensity of each channel. A comprehensive determination is made according to the fusion logic, which requires that the normalized trigger intensity of the arc channel reaches or exceeds the unit value as a necessary condition and that the normalized trigger intensity of at least one of the pressure channel and current channel also reaches or exceeds the unit value as an auxiliary confirmation condition. When the fusion logic determination is valid, an arc fault trigger signal is output. The graded pressure relief control unit receives the normalized trigger intensity of each channel and calculates the arc severity index by weighted combination. The arc severity index is compared with the preset multi-level response threshold arranged from low to high. Based on the highest level reached by the arc severity index, the circuit breaker tripping mechanism and pressure relief actuator corresponding to that level are driven to match the pressure relief response intensity with the severity of the arc fault.

[0009] The present invention is further configured such that: the energy characterization transformation of the arc light sensing channel is achieved by the following formula: CFopt (i)=V opt (i)^2; where CF opt (i) represents the characteristic function value of the arc channel at the i-th sampling time, in V. 2 V opt (i) is the output voltage value of the arc sensor at the i-th sampling time, in V, and its value is proportional to the illuminance incident on the photosensitive surface of the sensor.

[0010] The present invention is further configured such that the energy representation transformation of the pressure sensing channel is achieved by the following formula: CF prs (i)=V prs (i)^2+K dp *(V prs (i)-V prs (i-1))^2; where CF prs (i) represents the characteristic function value of the pressure channel at the i-th sampling time, in V. 2 V prs (i) represents the output voltage value of the pressure sensor at the i-th sampling time, in V, and its value is linearly related to the absolute pressure in the air chamber; V prs (i-1) represents the output voltage value at the (i-1)th sampling time, in V; K dp This is a dimensionless differential weighting coefficient used to adjust the algorithm's sensitivity to the rate of pressure change, with a value ranging from 5 to 20.

[0011] The present invention is further configured such that the energy characterization transformation of the current sensing channel is achieved by the following formula: CF cur (i)=(I(i)-I(iN T ))^2; where CF cur (i) represents the characteristic function value of the current channel at the i-th sampling time, in A. 2 I(i) is the instantaneous value of the loop current at the i-th sampling time, in A; I(iN) T ) is the iNth T The instantaneous value of the loop current at each sampling moment is the in-phase current value exactly one power frequency cycle prior to the current sampling moment, in A; N T The number of sampling points corresponding to one power frequency cycle is calculated as N. T =f s / f0,f s f0 is the sampling frequency in Hz, and f0 is the power supply frequency in Hz.

[0012] The present invention is further configured such that the short-time average value of each channel is calculated using the following recursive formula: STA k (i)=STA k(i-1)+(1 / N sk )*(CF k (i)-CF k (iN sk )); Among them, STA k (i) represents the short-time average value of the k-th channel at the i-th sampling time, with the subscript k indicating the channel type and taking values ​​of k respectively. opt , prs ,cur corresponds to the arc light channel, pressure channel, and current channel; STA k (i-1) is the short-time average value of the k-th channel at the (i-1)-th sampling time; N sk CF is the number of sampling points in the short-time window of the k-th channel and is a dimensionless positive integer; k (i) represents the characteristic function value of the k-th channel at the i-th sampling time; CF k (iN sk ) represents the characteristic function value corresponding to the oldest data point that is about to slide out of the short-time window.

[0013] The present invention is further configured such that the long-term average value of each channel is calculated using a recursive formula with a freezing strategy. When R k (i-1)<θ onk At that time, LTA k (i)=LTA k (i-1)+(1 / N lk )*(CF k (i)-CF k (iN lk When R k (i-1)>=θ onk At that time, LTA k (i)=LTA k (i-1). Wherein, LTA k (i) represents the long-term average value of the k-th channel at the i-th sampling time; N lk R is the number of sampling points in the long-term window of the k-th channel and is a dimensionless positive integer; k (i-1) is the ratio of the short-time average to the long-time average of the k-th channel at the (i-1)-th sampling time, and is a dimensionless quantity; θ onk The preset trigger threshold for the k-th channel is a dimensionless positive real number. The freeze strategy uses the ratio of the previous sampling step as a criterion to avoid circular dependencies and ensure that the protection action is not erroneously interrupted during the arc fault due to the long-term average value being gradually increased by the event signal.

[0014] The present invention is further configured such that: the multi-channel fusion trigger determination is implemented through the following formula and logical expression. The calculation formula for the normalized trigger intensity of each channel is S. k (i)=R k (i) / θonk The fusion trigger condition is Trigger(i) = [S opt (i)>=1]AND[S prs (i)>=1 OR S cur (i)>=1]. Where S k (i) is the normalized trigger strength of the k-th channel and is dimensionless; R k (i) represents the ratio of the short-term average value to the long-term average value of the k-th channel; θ onk is the trigger threshold for the k-th channel; Trigger(i) is the comprehensive trigger determination result. When this value is logically true, the system outputs an arc fault trigger signal.

[0015] The present invention further specifies that the arc severity index is calculated using the following formula: ASI(i) = w opt *S opt (i)+w prs *S prs (i)+w cur *S cur (i); where ASI(i) is the arc severity index and is a dimensionless positive real number; w opt w prs w cur The weights of the three channels are respectively, and w satisfies opt +w prs +w cur =1. The multi-level response thresholds include the first-level threshold θ1, the second-level threshold θ2, and the third-level threshold θ3 arranged from low to high. When ASI(i) reaches or exceeds θ1, the fast circuit breaker is tripped and the first-level pressure relief valve is opened. When it reaches or exceeds θ2, the large-area pressure relief plate is released and the cabinet door is interlocked. When it reaches or exceeds θ3, all pressure relief channels are opened and the isolation protection action of adjacent cabinets is activated.

[0016] The present invention is further configured such that: the sampling rate of the analog-to-digital converter is not less than 100 kHz and the resolution is not less than 12 bits; the short-time window length corresponding to the arc light sensing channel is 0.1-0.5 ms and the long-time window length is 50-200 ms; the short-time window length corresponding to the pressure sensing channel is 0.5-2 ms and the long-time window length is 100-500 ms; the short-time window length corresponding to the current sensing channel is 0.2-1 ms and the long-time window length is 100-500 ms; and the ratio of the number of sampling points in the short-time window to the number of sampling points in the long-time window of each channel is not greater than 1:100.

[0017] The present invention is further configured such that: the characteristic function calculation unit, the parallel short-time average and long-time average recursive calculation unit, the multi-channel fusion trigger determination unit, and the graded pressure relief control unit are all implemented in the same field programmable gate array chip with hardware logic circuits. The recursive calculation of the three sensing channels corresponds to three independent parallel computing pipelines that are executed simultaneously. The multi-channel fusion trigger determination unit completes the fusion determination operation in a single clock cycle with combinational logic circuits. The arc fault trigger signal is output through an optocoupler isolation circuit to drive the circuit breaker tripping mechanism and the pressure relief execution mechanism. The calculation and processing delay from the occurrence of the arc fault to the output of the arc fault trigger signal does not exceed 10 microseconds.

[0018] In summary, the present invention has the following beneficial effects: The present invention creatively transfers the STA / LTA (Short-Time Average and Long-Time Average) earthquake event first arrival detection algorithm, which originates from the fields of geophysics and seismology, to the field of rapid detection of arc faults inside ring main units. By combining multi-channel sensor fusion and graded pressure relief control, a complete protection system from sub-millisecond arc detection to graded pressure relief response is constructed, compressing the arc fault detection response time from tens of milliseconds in traditional overcurrent protection to about 1 millisecond.

[0019] Regarding detection speed, the recursive calculation form of the STA / LTA ratio algorithm requires only constant-level operations for each sampling step. The computational latency implemented on FPGA hardware is no more than 10 microseconds. Combined with the sub-millisecond short-time window setting of the arc channel, the system can complete the detection and judgment within approximately 1 millisecond after the arc ignition, securing a critical time window for the effective operation of the pressure relief system. In terms of detection reliability, the multi-channel fusion criterion uses the arc channel as a necessary condition, supplemented by confirmation from the pressure or current channel, balancing response speed and anti-maloperation performance. The long-term average freezing strategy ensures that the protection action is not erroneously interrupted during the arc's duration. Regarding pressure relief control, the graded response strategy based on the arc severity index ensures that the pressure relief action is precisely matched to the fault severity, avoiding the problem of excessive pressure relief leading to the escape of insulating gas during minor faults.

[0020] The core inventiveness of this invention lies in cross-domain algorithm transfer. The STA / LTA algorithm is used in seismology to detect the first arrival of P-waves from continuous seismic waveforms. However, engineers in the ring main unit (RNB) field typically seek solutions based on the inherent knowledge framework of power system protection, such as improving the sensitivity of overcurrent protection or optimizing the threshold setting for arc detection. They rarely consider applying the principles of seismic wave detection to arc signal detection. This invention identifies the core signal characteristic of transient high-energy mutations sharing a steady-state background between seismic events and arc faults. Overcoming the cognitive barriers between the two fields, by redefining the characteristic function, remapping the timescale window, and constructing a multi-channel fusion criterion, the STA / LTA algorithm is successfully adapted to the sub-millisecond timescale and multi-physical quantity fusion detection of arcs in RNBs.

[0021] In summary, this invention achieves algorithm migration and provides a high-speed arc protection system for ring main units that can be retrofitted onto existing equipment. This system, through a combination of multi-channel sensor fusion and graded pressure relief control, not only solves the problem of insufficient response speed of traditional protection devices, but also adaptively adjusts the pressure relief response intensity according to the severity of the arc fault. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the rapid detection and pressure relief system for internal arc faults in a ring main unit according to an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the internal functional structural units of the ring main unit in this invention. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to the accompanying drawings.

[0025] 1. System Overall Structure Description like Figure 1 As shown, the system in this embodiment consists of five functional units, which are arranged in order of signal flow: a multi-channel sensor signal acquisition unit, a feature function calculation unit, a parallel short-time average and long-time average recursive calculation unit, a multi-channel fusion trigger determination unit, and a graded pressure relief control unit.

[0026] A multi-channel sensor signal acquisition unit constitutes the system's sensing layer. The arc light sensor employs a wide-spectrum response photoelectric sensor, covering the 300 nm to 1100 nm wavelength range, with a rise time of less than 50 microseconds, and is installed inside each functional air chamber of the ring main unit. The pressure sensor uses a piezoresistive or piezoelectric high-frequency response sensor with a natural frequency of at least 10 kHz and a measurement range covering standard atmospheric pressure to 5 times atmospheric pressure, and is installed on the air chamber wall. The current sensor uses a Rogowski coil with a bandwidth of at least 100 kHz, installed in each circuit. The arc light sensor, pressure sensor, and current sensor constitute the arc light sensing channel, pressure sensing channel, and current sensing channel, respectively. The analog signals from the three channels are converted into digital discrete-time series signals by an analog-to-digital converter equipped with a synchronous sampling clock, with a sampling rate of at least 100 kHz and a resolution of at least 12 bits.

[0027] like Figure 2As shown, in one possible implementation, the ring main unit is typically divided into three main functional compartments: a busbar compartment, a circuit breaker compartment, and a cable compartment. These compartments are separated by metal or insulating partitions. At least one arc flash sensor is installed in each functional compartment, positioned on the top wall of the compartment away from live components, with the photosensitive surface facing the center of the compartment to ensure sufficient optical field coverage for arcs occurring at any location within the compartment. For the busbar compartment, the arc flash sensor is installed on the inner side of the top wall above the busbar; for the circuit breaker compartment, it is installed on the inner wall above or to the side of the circuit breaker; and for the cable compartment, it is installed on the inner wall above the cable termination. At least one pressure sensor is installed on the wall of each functional gas chamber. The installation location is chosen on the side or top wall of the gas chamber, away from the preset pressure relief port. The pressure-sensing diaphragm of the sensor is directly connected to the internal space of the gas chamber. During installation, a sealed threaded interface or flange sealing method is used to fix it to the cabinet wall panel to ensure that the pressure sensed by the sensor is the true pressure inside the gas chamber. The current sensor uses a Rogowski coil, and one is installed at the cable terminal of each circuit, sleeved on the outside of the cable insulation layer.

[0028] The characteristic function calculation unit receives discrete-time series signals from three sensing channels and performs energy characterization transformation operations on the physical characteristics of each channel to generate characteristic function value sequences for each channel. Parallel short-time and long-time average recursive calculation units perform moving average recursive calculations on the characteristic function value sequences of each channel using independently configured short-time and long-time windows. The number of sampling points in the short-time window is much smaller than that in the long-time window to achieve scale separation, and a long-time average freezing strategy is introduced. The multi-channel fusion trigger determination unit calculates the ratio of the short-time average to the long-time average for each channel, divides the ratio by its respective preset trigger threshold to obtain the normalized trigger intensity for each channel, and performs a comprehensive determination based on preset multi-channel fusion logic. When the fusion logic determination is valid, an arc fault trigger signal is output. The graded pressure relief control unit receives the normalized trigger intensity of each channel and calculates the arc severity index. It compares the arc severity index with the preset multi-level response thresholds arranged from low to high, and drives the circuit breaker tripping mechanism and pressure relief actuator corresponding to the highest level reached by the arc severity index, so that the pressure relief response intensity matches the severity of the arc fault.

[0029] 2. Calculation of characteristic function The energy characterization transformation operations performed by the characteristic function calculation unit for the three sensing channels follow the following design principles: For the arc light sensing channel, the energy characterization transformation uses the square of the signal to amplify the energy contrast of sudden changes in arc light intensity; for the pressure sensing channel, the energy characterization transformation superimposes a weighted squared difference term between adjacent sampling points on the square of the signal to enhance sensitivity to rapid pressure changes; for the current sensing channel, the energy characterization transformation uses the square of the difference between the current value and the in-phase current value of the previous power frequency cycle to eliminate the normal power frequency fundamental component and extract the fault transient component. The specific calculation formulas for the characteristic functions of each channel are given below.

[0030] The characteristic function of the arc light sensing channel uses the squared value of the signal: CF opt (i)=V opt (i)^2. Where V opt (i) represents the output voltage value of the arc sensor at the i-th sampling moment, in V, and its value is proportional to the illuminance incident on the sensor's photosensitive surface. CF opt (i) is in units of V 2 The squaring operation maps changes in the signal amplitude domain to changes in the energy domain, amplifying the signal contrast of the arc event from approximately 500 times in the amplitude domain to approximately 250,000 times in the energy domain, significantly improving detection sensitivity.

[0031] The characteristic function of the pressure sensing channel introduces a difference term based on the square of the signal: CF prs (i)=V prs (i)^2+K dp *(V prs (i)-V prs (i-1))^2. Where V prs (i) represents the output voltage value of the pressure sensor at the i-th sampling time, in V, and its value is linearly related to the absolute pressure in the air chamber. V prs (i-1) is the output voltage value of the pressure sensor at the (i-1)th sampling time, in V. K dp This is a dimensionless differential weighting coefficient used to adjust the algorithm's sensitivity to the rate of pressure change; its value ranges from 5 to 20. CF prs (i) is in units of V 2 Difference term V prs (i)-V prs (i-1) reflects the rate of pressure change between adjacent sampling points, making the algorithm more sensitive to rapid pressure changes, while remaining insensitive to slow pressure drift caused by temperature.

[0032] The characteristic function of the current sensing channel adopts the power frequency period difference method: CF cur (i)=(I(i)-I(iN T))^2. Where I(i) is the instantaneous value of the loop current at the i-th sampling time, in A. I(iN T ) is the iNth T The instantaneous value of the loop current at each sampling moment, that is, the in-phase current value exactly one power frequency cycle before the current sampling moment, in amperes (A). T The number of sampling points corresponding to one power frequency cycle is calculated as N. T =f s / f0,f s Where N is the sampling frequency in Hz, and f0 is the power supply frequency in Hz. In a 50Hz power supply system using a 100kHz sampling rate, N... T =100000 / 50=2000. CF cur (i) is in units of A 2 During normal steady-state operation, the phase difference is close to zero. When a fault occurs, this difference is equal to the non-periodic transient short-circuit component, thus effectively separating the normal load current from the fault current.

[0033] 3. Recursive calculation of short-term and long-term averages The short-time average value of each channel is calculated using a sliding window recursive formula: STA k (i)=STA k (i-1)+(1 / N sk )*(CF k (i)-CF k (iN sk The subscript k indicates the channel type and is respectively taken as... opt , prs ,cur corresponds to the arc light channel, pressure channel, and current channel, STA k (i) represents the short-time average value of the k-th channel at the i-th sampling time, STA k (i-1) is the short-time average value of the k-th channel at the (i-1)-th sampling time, N sk CF is the number of sampling points in the short-time window of the k-th channel, and is a dimensionless positive integer. k (i) represents the characteristic function value of the k-th channel at the i-th sampling time, CF k (iN sk ) represents the characteristic function value corresponding to the oldest data point that is about to slide out of the short-time window. Each recursive step requires only one addition, one subtraction, and one multiplication, with a computational complexity of O(1). The short-time window length for the arc channel is set to 0.1-0.5 milliseconds, for the pressure channel to 0.5-2 milliseconds, and for the current channel to 0.2-1 milliseconds.

[0034] The long-term average value of each channel is calculated using a recursive formula with a freeze strategy. The specific rule is: when R... k (i-1)<θonk At that time, LTA k (i)=LTA k (i-1)+(1 / N lk )*(CF k (i)-CF k (iN lk When R k (i-1)>=θ onk At that time, LTA k (i)=LTA k (i-1). Where LTA k (i) represents the long-term average value of the k-th channel at the i-th sampling time, LTA k (i-1) is the long-term average value of the k-th channel at the (i-1)-th sampling time, N lk CF is the number of sampling points in the long-term window of the k-th channel, and is a dimensionless positive integer. k (i) represents the characteristic function value of the k-th channel at the i-th sampling time, CF k (iN lk ) represents the k-th channel at the iNth position. lk The characteristic function values ​​at each sampling time, R k (i-1) is the ratio of the short-time average to the long-time average of the k-th channel at the (i-1)-th sampling time, and is a dimensionless quantity, θ onk The preset trigger threshold for the k-th channel is a dimensionless positive real number. The freeze strategy uses the ratio R from the previous sampling step. k (i-1) is used instead of the current step ratio as the freeze condition criterion to avoid circular dependencies in the computation.

[0035] The freezing strategy allows arc faults to last from tens to hundreds of milliseconds. If the long-time average is continuously updated, high-energy arc signals will gradually seep into the long-time window, raising the baseline and causing the ratio to drop, prematurely releasing the protection. When the ratio reaches or exceeds the trigger threshold, the long-time average remains unchanged, ensuring that the protection action is not erroneously interrupted during the arc fault due to the long-time average being gradually raised by event signals. The long-time window length for the arc channel is set to 50-200 milliseconds, and for the pressure and current channels, it is set to 100-500 milliseconds. The ratio of the number of sampling points in the short-time window to the number of sampling points in the long-time window for each channel is no greater than 1:100 to ensure sufficient scale separation between the short-time and long-time windows.

[0036] 4. Ratio calculation and multi-channel fusion trigger judgment The ratio of short-time average to long-time average for each channel is: R k (i)=STA k (i) / LTA k(i). This ratio is dimensionless. Under normal conditions, the ratio fluctuates around 1, but it jumps sharply after an electric arc occurs.

[0037] The normalized trigger strength for each channel is: S k (i)=R k (i) / θ onk Among them, S k (i) is the normalized trigger intensity of the k-th channel at the i-th sampling time, and it is a dimensionless quantity, R k (i) represents the ratio of the short-time average to the long-time average of the k-th channel at the i-th sampling time, θ onk The trigger threshold is a dimensionless positive real number preset for the k-th channel. When the normalized trigger strength reaches or exceeds 1, it indicates that the channel has reached the trigger threshold.

[0038] The fusion trigger condition is: Trigger(i) = [S opt (i)>=1]AND[S prs (i)>=1 OR S cur (i)>=1]. Where S opt (i) represents the normalized trigger intensity of the arc channel, S prs (i) represents the normalized trigger strength of the pressure channel, S cur (i) represents the normalized trigger strength of the current channel, and Trigger(i) represents the comprehensive trigger determination result at the i-th sampling time. When this value is logically true, the system outputs an arc fault trigger signal.

[0039] This logic requires that the normalized trigger intensity of the arc light channel reach or exceed a unit value as a necessary condition, while the normalized trigger intensity of at least one of the pressure and current channels must also reach or exceed a unit value as an auxiliary confirmation condition. The arc light channel is set as a necessary condition because the optical signal is the most specific physical representation of an arc fault, and confirmation of the auxiliary channels effectively eliminates malfunctions caused by interference from non-arc light sources. The trigger threshold values ​​for each channel are: Arc light channel θ onopt =20, pressure channel θ onprs =10, current path θ oncur =15.

[0040] The trigger thresholds mentioned above are determined based on the following: Arc channel trigger threshold θ onoptThe threshold value of 20 is determined based on the energy contrast analysis of the arc sensor output under normal operating conditions and arc fault conditions. Under normal operating conditions, the background illuminance inside the ring mains cabinet is extremely low, and the root mean square value of the arc sensor output voltage typically ranges from 0.005V to 0.02V. At this time, the STA / LTA ratio fluctuates around 1, with a fluctuation range usually not exceeding ±3. Under arc fault conditions, the arc sensor output voltage can reach several volts to tens of volts, and the STA / LTA ratio will rapidly jump to tens of thousands or even higher. Therefore, the arc channel trigger threshold is set to 20. This value is much higher than the upper limit of normal fluctuation to avoid false triggering, while being much lower than the ratio jump caused by actual arc events to ensure detection sensitivity, maintaining a sufficient safety margin between anti-false triggering and sensitivity. Pressure channel trigger threshold θ onprs The threshold value of θ is determined based on a comparison of the pressure variations within the gas chamber under normal temperature fluctuations and under arc fault conditions. During normal operation, the gas chamber pressure changes slowly with ambient temperature, resulting in a typical STA / LTA ratio fluctuation of no more than ±2. An arc fault causes a sudden pressure surge within the gas chamber, causing the ratio to reach tens to hundreds. Setting the threshold to 10 ensures that interference from slow changes such as temperature drift can be excluded. Current channel trigger threshold θ oncur The threshold of 15 is determined based on the amplitude ratio of normal load fluctuations to short-circuit fault currents. The characteristic function after power frequency periodic differentiation outputs close to zero during normal steady-state operation. The STA / LTA ratio caused by normal load fluctuations is usually no more than 5, while the ratio caused by short-circuit fault currents will jump significantly. Therefore, the threshold is set to 15 to ensure reliable differentiation between load fluctuations and fault currents.

[0041] The above threshold values ​​apply to typical ring main units with rated voltages of 10kV to 12kV, rated currents below 630A, and SF6 gas or dry air as the insulating medium. For ring main units of different models and operating conditions, the threshold values ​​can be adjusted according to the following principles: First, under normal operating conditions of the target ring main unit, background data is collected for no less than 72 hours. The mean value μ and standard deviation σ of the STA / LTA ratio of each channel are statistically analyzed. The trigger threshold of each channel is set to a value not less than μ + 6σ to ensure that the false trip rate meets the requirements. When conditions permit, a controlled arc event is generated by a test arc generator to verify the detection sensitivity under the selected threshold.

[0042] 5. Staged pressure relief control The arc severity index is calculated by weighted combination of the normalized trigger intensity of each channel: ASI(i) = w opt *S opt (i)+w prs *S prs (i)+w cur *S cur(i). Where ASI(i) is the arc severity index at the i-th sampling time and is a dimensionless positive real number, S opt (i), S prs (i), S cur (i) represents the normalized trigger intensity of the arc channel, pressure channel, and current channel at the i-th sampling time, respectively, w opt w prs w cur The weighting coefficients for the arc light channel, pressure channel, and current channel are respectively, and the three satisfy w opt +w prs +w cur =1. A value of w is recommended. opt =0.4, w prs =0.35, w cur =0.25. The larger the value of ASI(i), the more severe the arc fault.

[0043] The hierarchical response rule is that the multi-level response thresholds include the first-level threshold θ1, the second-level threshold θ2, and the third-level threshold θ3, arranged from low to high.

[0044] When ASI(i) reaches or exceeds θ1, the fast circuit breaker is tripped and the first-stage pressure relief valve is opened. When ASI(i) reaches or exceeds θ2, a large-area pressure relief plate release and cabinet door interlocking are added to the first-level response; When ASI(i) reaches or exceeds θ3, all pressure relief channels are opened and adjacent cabinets are isolated for protection.

[0045] The preferred values ​​are θ1=1.0, θ2=5.0, and θ3=20.0. The tiered strategy avoids the problem of excessive dissipation of insulating gas caused by applying the highest level of response to minor triggering events, and matches the pressure relief response intensity with the severity of the arc fault.

[0046] The following is one preferred structure for the pressure relief valves at each of the above levels: like Figure 2As shown, the primary pressure relief valve is an electromagnetically driven quick-opening pressure relief valve, installed at a pre-designed circular pressure relief port on the top or back wall panel of each functional air chamber of the ring main unit. The diameter of the pressure relief port is 80 mm to 120 mm. The primary pressure relief valve consists of a valve seat, valve disc, return spring, and electromagnetic actuator. The valve seat is a stainless steel flange structure, bolted to the pressure relief port flange face on the air chamber wall panel for sealing. The sealing surface is equipped with a high-temperature resistant silicone rubber sealing ring. Under normal conditions, the valve disc is pressed against the valve seat sealing surface by the return spring to maintain the airtightness of the air chamber. The electromagnetic actuator is a push-pull electromagnet. Upon receiving an arc fault trigger signal, it is energized and actuates, overcoming the force of the return spring to push open the valve disc, opening the pressure relief port. The high-pressure gas inside the air chamber is discharged through the pressure relief port into the exhaust channel on the top or back of the cabinet. The response time of the electromagnetic actuator does not exceed 3 milliseconds. The effective pressure relief area of ​​the primary pressure relief valve is 50 square centimeters to 113 square centimeters.

[0047] The large-area pressure relief plate is a mechanically released, flip-type pressure relief plate, installed at the large pressure relief port on the top or back of the ring main unit. The pressure relief port is a rectangular opening with an area of ​​0.1 to 0.3 square meters. The pressure relief plate is a steel plate structure, connected to the cabinet frame on one side by a hinge, and locked to the cabinet frame on the other side by an electromagnetic locking mechanism. Under normal conditions, the electromagnetic locking mechanism is in the locked position to maintain the seal of the pressure relief plate. Upon receiving a secondary response trigger signal, the electromagnetic lock is energized and released, causing the pressure relief plate to flip open around the hinge under the pressure inside the gas chamber, forming a large-area pressure relief channel. A high-temperature resistant sealing strip is provided on the outer edge of the pressure relief plate to ensure airtightness during normal operation. After the pressure relief plate releases, the gas is guided to a safe direction for discharge through the exhaust pipe at the top of the cabinet.

[0048] "Opening all pressure relief channels" refers to, under Level 3 response conditions, in addition to the simultaneous opening of the primary pressure relief valve and the large-area pressure relief plate, the additional opening of auxiliary pressure relief ports on the side and back of the cabinet. The auxiliary pressure relief ports utilize the same electromagnetic latch release structure as the large-area pressure relief plate, and are installed on the side and back panels of the cabinet. Each auxiliary pressure relief port has an area of ​​not less than 0.05 square meters. The total pressure relief area when all pressure relief channels are open is not less than 0.5 square meters. Simultaneously, the Level 3 response will also send an isolation command signal to adjacent cabinets, triggering the circuit breakers of the adjacent cabinets to trip via fiber optic cable or dedicated hardwiring to achieve fault isolation.

[0049] The characteristic function calculation unit, the parallel short-time and long-time average recursive calculation unit, the multi-channel fusion trigger determination unit, and the graded pressure relief control unit are all implemented in hardware logic circuits within the same field-programmable gate array (FPGA) chip. The recursive calculations for the three sensing channels correspond to three independent parallel computing pipelines executed simultaneously. The multi-channel fusion trigger determination unit completes the fusion determination operation within a single clock cycle using combinational logic circuits. The arc fault trigger signal is output via an optocoupler isolation circuit to drive the circuit breaker tripping mechanism and the pressure relief actuator. On a 100 MHz FPGA, the computational processing delay from the occurrence of an arc fault to the output of the arc fault trigger signal does not exceed 10 microseconds.

[0050] 6. STA / LTA Algorithm Conversion Process and Analysis The STA / LTA algorithm was originally designed by seismologists to solve the specific problem of automatically extracting the first arrival time of seismic P-waves from continuous waveform data from seismic stations. Its core principle is to construct two sliding energy windows with different time scales on the continuous time series, and detect transient changes in signal energy by the ratio of the short-time average to the long-time average.

[0051] In the field of seismology, short-time windows are typically set to 0.5-2 seconds to match the periodic characteristics of seismic waves, while long-time windows are set to 10-60 seconds to establish an environmental noise baseline. Trigger thresholds are set to 3-10 to balance detection sensitivity and false trigger rate. This invention's migration of the algorithm to the field of ring main unit arc fault detection involves three levels: conceptual mapping, model modification, and parameter redefinition.

[0052] At the conceptual mapping level, the arrival of seismic waves at the station is mapped as an arc fault occurring in the gas chamber, background environmental noise is mapped as the baseline signal of the sensor under normal operating conditions, and the transient rise of seismic energy is mapped as transient changes in arc light, pressure, and current signals.

[0053] At the model modification level, the original single-channel detection model was extended into a three-channel parallel detection plus fusion judgment architecture. The feature function was transformed from a single signal square to multiple forms customized for the physical characteristics of each channel. In particular, the current channel introduced power frequency period difference to remove the fundamental component, and the update rule of the long-term average value was updated with a freezing strategy to adapt to the characteristics of the arc duration.

[0054] At the parameter redefinition level, the window length is compressed from the second level to the millisecond and sub-millisecond level to match the time scale of arc faults, and the trigger threshold is increased from 3-10 to 10-50 to take advantage of the extremely high energy contrast of the arc signal to reduce the false alarm rate.

[0055] This shift is not obvious because there are deep disciplinary barriers between the field of ring main units and the field of seismology. Technical personnel in the field of ring main unit protection are constrained by the professional knowledge framework of power systems, and their usual paths to seeking solutions include improving the sensitivity of relay protection algorithms, optimizing the photoelectric threshold criterion for arc detection, and increasing the redundancy channels of protection devices, all of which fall within the scope of power system protection.

[0056] The STA / LTA algorithm in seismology is documented in professional literature in geophysics and seismology, which is typically not consulted or familiar to technicians in the power protection field. Furthermore, the signal timescales of the two fields differ by approximately three orders of magnitude; seismology deals with waveforms on the order of seconds, while ring main units (RNBs) need to handle signals on the order of microseconds to milliseconds. This scale difference further increases the difficulty of cross-disciplinary association. Technicians also need to overcome the obstacle of characteristic function adaptation, recognizing that current signals must undergo power frequency periodic differential preprocessing to eliminate the fundamental component for the algorithm to function correctly, and that pressure signals require superposition of differential terms to achieve sufficient sensitivity to rapid changes. These adaptation tasks require both a deep understanding of the mathematical mechanisms of the STA / LTA algorithm and a thorough understanding of the physical characteristics of the electric arc in RNBs, two disciplines distinct from each other.

[0057] The feasibility verification method after conversion includes: generating a controlled arc event in a laboratory environment using a small current test arc simulator, synchronously recording the original waveforms and algorithm processing results of each channel, and verifying the detection delay, trigger reliability and false trigger rate; Historical arc fault waveform data was replayed to the detection system for offline verification; and long-term stability tests were conducted under different ambient temperatures, humidity and background noise conditions to evaluate the adaptive baseline tracking capability.

[0058] The performance evaluation indicators include: the detection delay from arc ignition to trigger signal output should not exceed 2 milliseconds; the arc fault detection rate under standard test conditions should not be less than 99.9%; and the number of malfunctions during one year of continuous operation should not exceed once.

[0059] 7. Calculation Examples To fully demonstrate the entire operation process of the algorithm of this invention, this application provides the following detailed calculation example using an indoor phase-to-phase short-circuit arc fault at a cable terminal as a scenario. The calculation is explained in six stages: establishing a baseline for normal operation, recursive calculation and fusion judgment of each channel at the first sampling moment of arc ignition, gradual filling of short-time windows for each channel, summarizing the state after all three channels are triggered, calculating the arc severity index and outputting the graded pressure relief response, and estimating the detection delay.

[0060] The initial parameters are set as follows: Sampling rate f s =100000Hz, sampling interval Ts=1 / f s=10 microseconds. Power frequency f0 = 50Hz, number of sampling points N corresponding to a single power frequency cycle. T =f s / f0=100000 / 50=2000. Arc channel parameters: Number of sampling points N in the short-time window. s,opt =20 (corresponding to a window duration of 0.2ms), N is the number of sampling points in the long-term window. l,opt =10000 (corresponding to a window duration of 100ms), trigger threshold θ onopt =20. Pressure channel parameters: Number of short-time window sampling points N s,prs =100 (corresponding to a window duration of 1.0ms), N is the number of sampling points in the long-term window. l,prs =50000 (corresponding to a window duration of 500ms), differential weighting coefficient K dp =10, trigger threshold θ onprs =10. Current path parameters: Number of short-time window sampling points N s,cur =50 (corresponding to a window duration of 0.5ms), number of sampling points N for the long window l,cur =50000 (corresponding to a window duration of 500ms), trigger threshold θ oncur =15. Channel weight coefficient w opt =0.4、w prs =0.35、w cur =0.25. The graded response thresholds are θ1=1.0, θ2=5.0, and θ3=20.0.

[0061] The fault scenario is set as follows: A phase-to-phase short-circuit arc fault occurs in the indoor cable terminal of the ring main unit. The moment of arc ignition is recorded as t=0, and the corresponding sampling point number is recorded as i. arc After the arc ignition, the output voltage of the arc sensor instantly jumped to 5.0V and remained stable; the instantaneous short-circuit current of the fault circuit was 2500A, while the normal instantaneous current of this phase before the fault was 100A; the pressure in the gas chamber had not yet changed at the sensor installation location due to the delay in sound wave propagation at the moment of arc ignition, from i arc Then, starting from the first sampling point, a linear model V is used. prs (k) = 1.00 + 0.05 * k rises, where k is the sampling point number after arc ignition. This linear rise model is a simplified calculation model. In reality, the sensor output will gradually saturate when it approaches the upper limit of its range. However, the pressure channel trigger determination is established before the pressure reaches the upper limit of the sensor's range (e.g., at t = 0.5 ms, V). prs =3.50V is still within the linear region), and sensor saturation does not affect the system's trigger timeliness.

[0062] Phase 1: Normal operation steady-state baseline (t<0, background steady state after each channel's STA and LTA have been fully initialized) Arc light channel background: During normal operation of the ring main unit, there is no strong light source in the air chamber, and the root mean square value of the arc light sensor output voltage is approximately 0.01V. The characteristic function values ​​of the arc light channel at each sampling time are: CF optbg = V opt ^2 = (0.01)^2 = 0.0001 V 2 After sufficient initialization, both the short-term and long-term averages stabilized at the background level. STA optbg = LTA optbg = 0.0001 V 2 Ratio and Normalized Trigger Strength: R opt =STA optbg / LTA optbg = 0.0001 / 0.0001 = 1.0 S opt = R opt / θ onopt = 1.0 / 20 = 0.05 S opt =0.05 is much smaller than the unit value, indicating that the arc channel is under normal monitoring.

[0063] Pressure channel background: During normal operation, the sensor output voltage corresponding to the insulating gas pressure inside the gas chamber remains stable at 1.00V, and the voltage difference between adjacent sampling points due to slow changes in ambient temperature is approximately 0.001V. The characteristic function value of the pressure channel is: CF prsbg = V prs ^2 + K dp * (V prs (i) - V prs (i-1))^2 = (1.00)^2 + 10 * (0.001)^2 = 1.0000 + 0.00001 = 1.00001 (approximately 1.0000 V) 2 ) STA prsbg = LTA prsbg = 1.0000 V 2 R prs = 1.0000 / 1.0000 = 1.0 S prs = R prs / θ onprs= 1.0 / 10 = 0.10 S prs =0.10 is much smaller than the unit value, and the pressure channel is under normal monitoring.

[0064] Current path background: During normal steady-state operation, the loop current is a power frequency sine wave. The difference between the current sampling point and the sampling point in the same phase of the previous power frequency cycle mainly comes from slight load fluctuations, and the root mean square difference is about 2A.

[0065] The characteristic function value of the current channel is: CF curbg = (I(i) - I(iN T ))^2 = (2)^2 = 4.0 A 2 STA curbg = LTA curbg = 4.0 A 2 R cur = 4.0 / 4.0 = 1.0 S cur = R cur / θ oncur = 1.0 / 15 = 0.067 S cur =0.067 is much smaller than the unit value, and the current channel is under normal monitoring.

[0066] Comprehensive trigger determination: Three-channel normalized trigger strength S opt =0.05, S prs =0.10、S cur =0.067 is much smaller than the unit value, the fusion trigger condition is not met, Trigger=FALSE. The system is in normal monitoring state, and the background baseline of each channel has been established.

[0067] Second stage: Arc ignition moment (t=0, the first sampling moment containing the arc signal i) arc ) (1) Arc light channel in i arc Recursive calculation of time The arc sensor output voltage instantly jumps to 5.0V. Calculate the characteristic function value: CF opt (i arc ) = (5.0)^2 = 25.0 V 2 Perform a long-term average freeze condition check, using the ratio from the previous sampling step: R opt (i arc -1)=1.0, satisfying Ropt (i arc -1)<θ onopt =20, long-term average updated normally: LTA opt (i arc ) = LTA opt (i arc -1) + [CF opt (i arc ) - CF opt (i arc - N l,opt )] / N l,opt = 0.0001 + (25.0 - 0.0001) / 10000 = 0.0001 + 2.500 * 10^(-3) = 0.0026 V 2 Among them CF opt (i arc - N l,opt The value is 0.0001 V, representing the oldest data point that has been slid out of the long-term window. It belongs to the background signal. 2 .

[0068] Perform short-term average recursive update: STA opt (i arc ) = STA opt (i arc -1) + [CF opt (i arc ) - CF opt (i arc - N s,opt )] / N s,opt = 0.0001 + (25.0 - 0.0001) / 20 = 0.0001 + 1.2500 = 1.2501 V 2 Among them CF opt (i arc - N s,opt The oldest data point that has slipped out of the short-time window is also part of the background signal, with a value of 0.0001 V. 2 .

[0069] Calculate the ratio and normalized trigger strength: R opt (i arc) = STA opt (i arc ) / LTA opt (i arc = 1.2501 / 0.0026 = 480.8 S opt (i arc ) = R opt (i arc ) / θ onopt = 480.8 / 20 = 24.0 S opt =24.0>= 1, the arc channel trigger determination is valid.

[0070] Update the frozen state flag: R opt (i arc )=480.8>= θ onopt =20, therefore from i arc LTA from +1 moment opt Maintain 0.0026 V 2 The update is now frozen and will no longer be updated.

[0071] (2) Current path in i arc Recursive calculation of time Fault current I(i) arc =2500A, the normal current I(i) in phase during the previous power frequency cycle arc -N T =100A, the power frequency period difference is 2500-100=2400A. Calculate the characteristic function value: CF cur (i arc ) = (I(i arc ) - I(i arc - N T ))^2 = (2400)^2 = 5760000 A 2 Perform long-term average freeze condition check: R cur (i arc -1)=1.0<θ oncur =15, long-term average updated normally: LTA cur (i arc ) = LTA cur (i arc -1) + [CF cur (i arc ) - CF cur (i arc - N l,cur )] / Nl,cur = 4.0 + (5760000 - 4.0) / 50000 = 4.0 + 115.2 = 119.2 A 2 Perform short-term average recursive update: STA cur (i arc ) = STA cur (i arc -1) + [CF cur (i arc ) - CF cur (i arc - N s,cur )] / N s,cur = 4.0 + (5760000 - 4.0) / 50 = 4.0 + 115200 = 115204 A 2 Calculate the ratio and normalized trigger strength: R cur (i arc = 115204 / 119.2 = 966.5 S cur (i arc = 966.5 / 15 = 64.4 S cur =64.4>= 1, the current channel trigger determination is valid.

[0072] Update the frozen state flag: R cur (i arc )=966.5>= θ oncur =15, therefore from i arc LTA from +1 moment cur Maintain 119.2 A 2 The update is now frozen and will no longer be updated.

[0073] (3) Pressure channel in i arc Recursive calculation of time The pressure wave inside the chamber has not yet reached the sensor installation location, and the pressure sensor output voltage remains at V. prs (i arc The value is 1.000V, the same as the previous sampling time, with a difference of zero. Calculate the characteristic function value: CF prs (i arc) = (1.000)^2 + 10 * (1.000 - 1.000)^2 = 1.0000 + 0 = 1.0000V 2 The feature function value is the same as the background level and remains unchanged.

[0074] Freeze condition check: R prs (i arc -1)=1.0<θ onprs =10, long-term average updated normally: LTA prs (i arc ) = 1.0000 + (1.0000 - 1.0000) / 50000 = 1.0000 V 2 STA prs (i arc ) = 1.0000 + (1.0000 - 1.0000) / 100 = 1.0000 V 2 R prs (i arc = 1.0000 / 1.0000 = 1.0 S prs (i arc = 1.0 / 10 = 0.10 S prs =0.10<1, pressure channel not triggered.

[0075] (4) i arc Moment fusion trigger judgment Substitute the normalized trigger strength of the three channels into the fusion trigger condition: Trigger(i arc ) = [S opt (i arc )>= 1] AND [S prs (i arc )>= 1 OR S cur (i arc )>= 1] = [24.0>= 1] AND [0.10>= 1 OR 64.4>= 1] = TRUE AND [FALSE OR TRUE] = TRUE AND TRUE = TRUE Fusion triggering conditions based on arc light channel (S) opt=24.0) is a necessary condition, current path (S) cur =64.4) is an auxiliary confirmation channel, which is established after only one sampling interval following arc ignition. The system outputs an arc fault trigger signal at t=Ts=10 microseconds.

[0076] Phase 3: Gradual filling process of short-time windows for each channel (cross-section calculation at times t=0.2ms, t=0.5ms, and t=1.0ms) Note: The short-time window length for the arc channel is 0.2 ms (20 sampling points), the short-time window length for the current channel is 0.5 ms (50 sampling points), and the short-time window length for the pressure channel is 1.0 ms (100 sampling points). The short-time windows of the three channels will be completely filled with the arc event sampling values ​​at different times. The states of each channel at three typical moments are calculated below.

[0077] (1) At time t=0.2ms (i=i arc +20, arc channel short-time window fully filled) Arc light channel: After 20 recursive steps, all 20 sampling points in the short time window have been sampled by the arc light event (CF) value. opt =25.0 V 2 Fill it in. From i arc to i arc In each step of the +19 iteration, a background value (CF=0.0001) slides out of the short-time window, and an arc value (CF=25.0) slides in. The STA increment is (25.0-0.0001) / 20=1.250. After 20 steps of accumulation: STA opt (i arc +20) = 25.0 V 2 LTA opt Keep the freeze value at 0.0026 V. 2 constant.

[0078] R opt = 25.0 / 0.0026 = 9615 S opt = 9615 / 20 = 480.8 The arc channel STA has reached its steady-state peak value. Thereafter, as long as the arc continues, STA... opt It will remain at 25.0 V 2 .

[0079] Current channel: 50 sampling points in a short-time window, current time i arc +20 corresponds to a window range of i. arc -29 to i arc+20, which includes 29 background sampling points (CF). cur =4.0 A 2 ) and 21 fault current sampling points (CF cur =5760000A 2 ).

[0080] STA cur (i arc +20) = (29 * 4.0 + 21 * 5760000) / 50 = (116 + 120960000) / 50 = 2419202 A 2 LTA cur Keep the frozen value at 119.2 A. 2 constant.

[0081] R cur = 2419202 / 119.2 = 20295 S cur = 20295 / 15 = 1353 Pressure channel: 100 sampling points in a short-time window, current window range is i arc -79 to i arc +20, which includes 80 background horizontal sampling points (i arc -79 to i arc , including i arc Since the pressure level itself remained unchanged, it was also at the background level. CF prs =1.0000V 2 ) and 20 pressure rise sampling points (i arc +1 to i arc +20).

[0082] For 20 pressure rise sampling points, according to model V prs (k) = 1.00 + 0.05 * k (k = 1 to 20), and the adjacent difference between each sampling point is always ΔV. prs =0.05V, calculate the sum of the characteristic functions respectively: The sum of the squared voltage terms = ∑{k=1}^{20} (1+0.05k)^2 =∑{k=1}^{20} (1 + 0.1k + 0.0025k^2) = 20 + 0.1 * [2021 / 2] + 0.0025 * [202141 / 6] = 20 + 0.1 * 210 + 0.0025 * 2870 = 20 + 21.0 + 7.175 = 48.175 The sum of the squared differences = 20 * K dp * (0.05)^2 = 20 * 10 * 0.0025 = 0.5 The sum of the characteristic functions of the 20 rising sampling points = 48.175 + 0.5 = 48.675 STA prs (i arc +20) = (80 * 1.0000 + 48.675) / 100 = 128.675 / 100 = 1.287V 2 Long-term averages are continuously updated (due to R) prs (always less than 10), from i arc to i arc The cumulative change over 20 steps (21 steps in total) is: △LTA prs = [CF prs (i arc ) + ∑{k=1}^{20} CF prs (k) - 21 * CF prsbg ] / N l,prs = [1.0000 + 48.675 - 21 * 1.0000] / 50000 = 28.675 / 50000 = 0.000574 LTA prs (i arc +20) = 1.0000 + 0.000574 = 1.0006 V 2 R prs = 1.287 / 1.0006 = 1.286 S prs = 1.286 / 10 = 0.129 S prs =0.129<1, the pressure channel has not yet been triggered. At this time, the pressure sensor output voltage has risen to V. prs (20) = 1.00 + 0.05 * 20 = 2.00V, but since 80% of the data points in the short window are still at the background level, the STA / LTA ratio increases only slightly.

[0083] (2) At time t=0.5ms (i=iarc +50, the current channel window is fully filled in a short time. Current path: All 50 sampling points in the short-time window have been filled with fault current sampling values.

[0084] STA cur (i arc +50) = 5760000 A 2 LTA cur Keep the frozen value at 119.2 A. 2 constant.

[0085] R cur = 5760000 / 119.2 = 48322 S cur = 48322 / 15 = 3221 The current channel STA has reached its steady-state peak value.

[0086] Arc light channel maintains stable state: STA opt =25.0 V 2 LTA opt =0.0026 V 2 R opt =9615, S opt =480.8.

[0087] Pressure channel: Short-time window range is i arc -49 to i arc +50, which includes 50 background horizontal sampling points (i arc -49 to i arc ) and 50 pressure rise sampling points (i arc +1 to i arc +50).

[0088] For 50 pressure rise sampling points (k=1 to 50), V prs (k) = 1.00 + 0.05 * k, at this time V prs (50) = 3.50V: The sum of the squared voltage terms = ∑{k=1}^{50} (1+0.05k)^2 = ∑{k=1}^{50} (1 + 0.1k + 0.0025k^2) = 50 + 0.1 * [5051 / 2] + 0.0025 * [5051101 / 6] = 50 + 0.1 * 1275 + 0.0025 * 42925 = 50 + 127.5 + 107.3 = 284.8 The sum of the squared differences = 50 * 10 * (0.05)^2 = 50 * 0.025 = 1.25 The sum of the characteristic functions of 50 rising sampling points = 284.8 + 1.25 = 286.05 STA prs (i arc +50) = (50 * 1.0000 + 286.05) / 100 = 336.05 / 100 = 3.361V 2 Long-term average cumulative change (from i) arc to i arc (+50 steps total) △LTA prs = [1.0000 + 286.05 - 51 * 1.0000] / 50000 = 236.05 / 50000 = 0.00472 LTA prs (i arc +50) = 1.0000 + 0.00472 = 1.005 V 2 R prs = 3.361 / 1.005 = 3.344 S prs = 3.344 / 10 = 0.334 S prs =0.334<1, the pressure channel has not yet been triggered. Although the chamber pressure has risen to 3.5 times the atmospheric pressure, half of the data points in the short window are still at the background level, and the cumulative effect of the pressure rise is not enough to make the ratio reach the trigger threshold.

[0089] (3) At time t=1.0ms (i=i arc +100, pressure channel short-term window fully filled) Pressure channel: Short-time window range is i arc +1 to i arc +100, all 100 sampling points are pressure rise sampling points. At this time, V prs (100) = 1.00 + 0.05 * 100 = 6.00V, which corresponds to a chamber pressure of approximately 6 atmospheres.

[0090] For 100 rising sampling points (k=1 to 100): The sum of the squared voltage terms = ∑{k=1}^{100} (1+0.05k)^2 = ∑{k=1}^{100} (1 + 0.1k + 0.0025k^2) = 100 + 0.1 * [100101 / 2] + 0.0025 * [100101201 / 6] = 100 + 0.1 * 5050 + 0.0025 * 338350 = 100 + 505.0 + 845.9 = 1450.9 The sum of the squared differences = 100 * 10 * (0.05)^2 = 100 * 0.025 = 2.5 The sum of the characteristic functions of 100 rising sampling points = 1450.9 + 2.5 = 1453.4 STA prs (i arc +100) = 1453.4 / 100 = 14.534 V 2 Explanation of long-term average update: In i arc to i arc During the recursion of +100, the ratio R of the pressure channels prs As it gradually increases from 1.0, it is necessary to check R. prs Is it upon arrival at i? arc The freeze threshold θ has been reached before +100. onprs =10. According to step (2), at t=0.5ms, R prs The calculated result of 3.34, and the trend of STA accelerating in the latter half of the window filling, and the ratio R prs Around i arc It first exceeded 10 near +87, and then LTA prs Freeze. However, due to the extremely large long-term window (50,000 points), in i arc to i arc The cumulative change in LTA during the +87 period was only about 0.019 V. 2 The difference in LTA before and after freezing does not exceed 0.008 V. 2 The impact on the final ratio is less than 6%. To simplify the calculation and clearly illustrate the algorithm principle, the following calculation is based on an approximation of the cumulative change of LTA over the entire 100 steps: △LTA prs= [1453.4 - 100 * 1.0000] / 50000 = 1353.4 / 50000 = 0.0271 LTA prs (i arc +100) is approximately equal to 1.0000 + 0.0271 = 1.027 V 2 Calculate the ratio and normalized trigger strength: R prs (i arc +100) = 14.534 / 1.027 = 14.15 S prs (i arc +100) = 14.15 / 10 = 1.42 S prs =1.42>= 1, the pressure channel trigger determination is valid.

[0091] Update the frozen state flag: R prs =14.15>= θ onprs =10, LTA prs Frozen at approximately 1.027 V 2 No longer being updated.

[0092] Phase 4: Summary and fusion determination of the three-channel status at time t=1.0ms The normalized trigger strength of the three channels at t=1.0ms is summarized as follows: S opt (i arc +100) = 480.8 (Arc channel, continuously meeting trigger conditions from t=10 microseconds) S cur (i arc +100) = 3221 (Current channel, continuously meets triggering conditions from t=10 microseconds) S prs (i arc +100) = 1.42 (Pressure channel, trigger condition is met at t=1.0ms) Fusion trigger determination: Trigger = [S opt =480.8>= 1] AND [S prs =1.42>= 1 OR S cur =3221>= 1] = TRUE AND [TRUE OR TRUE] = TRUE All three channels met their respective trigger conditions at t=1.0ms, and the system was in the highest confidence operating state with all three channels confirmed, with the trigger signals continuously and effectively output. The long-term average freezing mechanism for each channel ensures that the LTA will not be pulled up due to the infiltration of high-energy event signals during the arc duration, and the protection action will not be erroneously interrupted.

[0093] Phase 5: Calculation of Arc Severity Index and Graded Pressure Relief Response (1) The earliest time when fusion triggers is established (t=10 microseconds, i arc The arc severity index is calculated using three-channel data. ASI(i arc ) = w opt * S opt (i arc ) + w prs * S prs (i arc ) + w cur * S cur (i arc ) = 0.4 * 24.0 + 0.35 * 0.10 + 0.25 * 64.4 = 9.60 + 0.035 + 16.10 = 25.74 Compare ASI=25.74 with each level of response threshold: Level 1: 25.74>= θ1=1.0, the condition is met, start the fast circuit breaker to trip and open the first-level pressure relief valve.

[0094] Level 2: 25.74>= θ2=5.0, the condition is met, additionally activate the large-area pressure relief plate release and cabinet door interlocking.

[0095] Level 3: 25.74>= θ3=20.0, the condition is met, all pressure relief channels are opened and isolation protection commands are sent to adjacent cabinets.

[0096] The system immediately identifies itself as a Level 3 response (highest level) at the earliest trigger moment, and promptly activates all pressure relief channels to open synchronously and sends isolation commands to adjacent cabinets.

[0097] (2) At time t=1.0ms (i arc The arc severity index is calculated using three-channel data (+100). ASI(i arc+100) = 0.4 * 480.8 + 0.35 * 1.42 + 0.25 * 3221 = 192.32 + 0.497 + 805.25 = 998.1 At this point, ASI = 998.1, far exceeding θ3 = 20.0, and the system maintains the highest level of response (Level 3). The continuous increase in the ASI value reflects the worsening severity of the arc fault over time.

[0098] (3) Applicability description of the graded response strategy In the high-energy phase-to-phase short-circuit arc fault scenario corresponding to this calculation example, due to the extremely high contrast between the arc and current signals, the ASI exceeds the third-level threshold at the first trigger moment. However, for minor arc events, such as weak flashovers caused by partial discharge, the normalized trigger intensity of each channel is low. Assuming that under a certain operating condition S... opt =1.5, S prs =0.8, S cur =1.2, then ASI = 0.41.5 + 0.35 * 0.8 + 0.25 * 1.2 = 0.60 + 0.28 + 0.30 = 1.18, which only exceeds θ1 = 1.0. The system will only execute a first-level response, namely the fast circuit breaker tripping and the first-level pressure relief valve opening, thus avoiding unnecessary large-area pressure relief that could lead to the escape of insulating gas.

[0099] Phase 6: Detection Delay Estimation The fusion triggering condition is met at the first sampling moment after arc ignition (t=Ts=10 microseconds). The delay analysis of each stage from arc ignition to the system output trigger signal is as follows: Arc sensor hardware response time (including broadband photoelectric conversion and analog conditioning circuit signal setup time): T sensor <50 microseconds Time required for an analog-to-digital converter to complete one synchronous conversion (one sampling period at a sampling rate of 100kHz): T adc =1 / f s = 10 microseconds Pipeline delay for all algorithm processing within the FPGA (including characteristic function calculation, parallel recursion of three-channel STA and LTA, ratio calculation, normalized trigger strength calculation, and fused trigger combinational logic determination, based on a 100MHz clock frequency): T fpga <10 microseconds The maximum total detection delay from arc ignition to trigger signal output is: T detect = T sensor + T adc + Tfpga <50 + 10 + 10 = 70 microseconds This value is much less than 0.1ms. Adding the electromagnet response time of the circuit breaker tripping mechanism and the mechanical action time of the pressure relief valve (approximately 3ms), the total protection action time is approximately: T action = T detect + T mechanism <0.07 + 3 = 3.07ms Compared to the 30 to 60 ms response time of traditional overcurrent protection, the total protection action time is shortened by about an order of magnitude, providing ample time window to prevent the continuous accumulation of arc energy in the limited gas chamber volume and to prevent structural damage to the cabinet.

[0100] Taking this calculation example as a reference, the three channels can be fully confirmed and a steady-state trigger signal can be continuously output within 1.0ms after the arc ignition. The cabinet structure can withstand the time limit of tens to hundreds of milliseconds under typical arc fault conditions. The system detection response speed has sufficient safety margin.

[0101] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A rapid detection and pressure relief system for internal arc faults in a ring main unit, characterized in that, include: A multi-channel sensor signal acquisition unit includes an arc light sensor installed inside each functional chamber of the ring main unit, a pressure sensor installed on the chamber wall, and a current sensor installed in each circuit. The arc light sensor, pressure sensor, and current sensor respectively constitute an arc light sensing channel, a pressure sensing channel, and a current sensing channel. The three sensing channels are synchronously sampled by an analog-to-digital converter configured with a synchronous sampling clock to obtain discrete time sequence signals of each channel. The feature function calculation unit receives the discrete time series signals output by the three sensing channels respectively, and performs energy characterization transformation operations on the physical characteristics of each channel to generate a sequence of feature function values ​​for each channel. Parallel short-time average and long-time average recursive calculation unit performs sliding average recursive calculation on the feature function value sequence of each channel using independently configured short-time windows and long-time windows respectively. The number of sampling points in the short-time window is much smaller than the number of sampling points in the long-time window to achieve scale separation. A multi-channel fusion trigger determination unit calculates the ratio of the short-time average value to the long-time average value of each channel, divides the ratio of each channel by its respective preset trigger threshold to obtain the normalized trigger intensity of each channel, and makes a comprehensive determination based on the preset multi-channel fusion logic. When the fusion logic determination is valid, an arc fault trigger signal is output. The graded pressure relief control unit receives the normalized trigger intensity of each channel and calculates the arc severity index. It compares the arc severity index with a preset multi-level response threshold arranged from low to high, and drives the circuit breaker tripping mechanism and pressure relief actuator corresponding to the highest level reached by the arc severity index, so that the pressure relief response intensity matches the severity of the arc fault.

2. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 1, characterized in that: Specifically, the energy characterization transformation operation performed by the characteristic function calculation unit for each channel is as follows: the energy characterization transformation of the arc light sensing channel uses the square value of the signal to amplify the energy contrast of sudden changes in arc light intensity; the energy characterization transformation of the pressure sensing channel superimposes a weighted differential square term of adjacent sampling points on the square value of the signal to enhance the sensitivity to rapid pressure changes; and the energy characterization transformation of the current sensing channel uses the square value of the difference between the current value and the current value in phase of the previous power frequency cycle to eliminate the normal power frequency fundamental component and extract the fault transient component.

3. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 2, characterized in that: The energy representation transformation of the arc light sensing channel is achieved through the following formula: CF opt (i)=V opt (i)^2; Among them, CF opt (i) represents the characteristic function value of the arc channel at the i-th sampling time, in V. 2 V opt (i) is the output voltage value of the arc sensor at the i-th sampling time, in V, and its value is proportional to the illuminance incident on the photosensitive surface of the sensor.

4. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 2, characterized in that: The energy representation transformation of the pressure sensing channel is achieved through the following formula: CF prs (i)=V prs (i)^2+K dp *(V prs (i)-V prs (i-1))^2; Among them, CF prs (i) represents the characteristic function value of the pressure channel at the i-th sampling time, in V. 2 V prs (i) represents the output voltage value of the pressure sensor at the i-th sampling time, in V, and its value is linearly related to the absolute pressure in the air chamber; V prs (i-1) is the output voltage value of the pressure sensor at the (i-1)th sampling time, in V; K dp This is a dimensionless differential weighting coefficient used to adjust the algorithm's sensitivity to the rate of pressure change, with a value ranging from 5 to 20.

5. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 2, characterized in that: The energy representation transformation of the current sensing channel is achieved through the following formula: CF cur (i)=(I(i)-I(i-N T ))^2; Among them, CF cur (i) represents the characteristic function value of the current channel at the i-th sampling time, in A. 2 I(i) is the instantaneous value of the loop current at the i-th sampling time, in A; I(iN) T ) is the iNth T The instantaneous value of the loop current at each sampling moment, in amperes (A), i.e., the in-phase current value exactly one power frequency cycle prior to the current sampling moment; N T The number of sampling points corresponding to one power frequency cycle is calculated as N. T =f s / f0,f s f0 is the sampling frequency in Hz, and f0 is the power supply frequency in Hz.

6. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 1, characterized in that: The short-time average value of each channel in the parallel short-time average and long-time average recursive calculation unit is calculated using the following recursive formula: STA k (i)=STA k (i-1)+(1 / N sk )*(CF k (i)-CF k (i-N sk )); Among them, STA k (i) represents the short-time average value of the k-th channel at the i-th sampling time, with the subscript k indicating the channel type and taking opt, prs, and cur respectively to correspond to the arc light channel, pressure channel, and current channel; STA k (i-1) is the short-time average value of the k-th channel at the (i-1)-th sampling time; N sk CF is the number of sampling points in the short-time window of the k-th channel and is a dimensionless positive integer; k (i) represents the characteristic function value of the k-th channel at the i-th sampling time; CF k (iN sk ) represents the k-th channel at the iNth position. sk The feature function value at each sampling time corresponds to the feature function value of the oldest data point that is about to slide out of the short-time window.

7. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 1, characterized in that: The long-term average value of each channel in the parallel short-term average and long-term average recursive calculation unit is calculated using a recursive formula with a freezing strategy. The specific rules are as follows: When R k (i - 1)<theta onk then, LTA k (i) = LTA k (i - 1)+(1 / N lk )*(CF k (i)-CF k (i - N lk )); When R k (i - 1) >= theta onk then, LTA k (i) = LTA k (i - 1); Among them, LTA k (i) represents the long-term average value of the k-th channel at the i-th sampling time; LTA k (i-1) is the long-term average value of the k-th channel at the (i-1)-th sampling time; N lk CF is the number of sampling points in the long-term window of the k-th channel and is a dimensionless positive integer; k (i) represents the characteristic function value of the k-th channel at the i-th sampling time; CF k (iN lk ) represents the k-th channel at the iNth position. lk The characteristic function values ​​at each sampling time; R k (i-1) is the ratio of the short-time average to the long-time average of the k-th channel at the (i-1)-th sampling time, and is a dimensionless quantity; theta onk The preset trigger threshold for the k-th channel is a dimensionless positive real number; the freezing strategy uses the ratio R from the previous sampling step. k (i-1) serves as a freeze condition criterion to avoid circular dependencies in the current step. When the ratio reaches or exceeds the trigger threshold, the long-term average remains unchanged, thereby ensuring that the protection action will not be erroneously interrupted during the duration of the arc fault due to the long-term average being gradually increased by the event signal.

8. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 1, characterized in that: The multi-channel fusion trigger determination unit implements fusion determination through the following formula and logical expression: The formula for calculating the normalized trigger strength of each channel is: S k (i)=R k (i) / theta onk ; The fusion trigger condition is: Trigger(i) = [S opt (i)>=1]AND[S prs (i)>=1ORS cur [(i)>=1]; Among them, S k (i) represents the normalized trigger intensity of the k-th channel at the i-th sampling time, and is a dimensionless quantity; R k (i) represents the ratio of the short-time average to the long-time average of the k-th channel at the i-th sampling time; theta onk The preset trigger threshold for the k-th channel is a dimensionless positive real number; S opt (i) represents the normalized trigger intensity of the arc channel; S prs (i) represents the normalized trigger strength of the pressure channel; S cur (i) represents the normalized trigger strength of the current channel; Trigger(i) represents the comprehensive trigger determination result at the i-th sampling time. When this value is logically true, the system outputs an arc fault trigger signal.

9. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 1, characterized in that: The graded pressure relief control unit calculates the arc severity index using the following formula: ASI(i)=w opt *S opt (i)+w prs *S prs (i)+w cur *S cur (i); Where ASI(i) is the arc severity index at the i-th sampling time and is a dimensionless positive real number; S opt (i), S prs (i), S cur (i) represents the normalized trigger intensity of the arc channel, pressure channel, and current channel at the i-th sampling time, respectively; w opt w prs w cur The weighting coefficients for the arc light channel, pressure channel, and current channel are respectively, and the three satisfy w opt +w prs +w cur =1; The multi-level response thresholds include the first-level threshold theta1, the second-level threshold theta2, and the third-level threshold theta3 arranged from low to high. When ASI(i) reaches or exceeds theta1, the fast circuit breaker is tripped and the first-level pressure relief valve is opened. When ASI(i) reaches or exceeds theta2, the large-area pressure relief plate is released and the cabinet door is interlocked on the basis of the first-level response. When ASI(i) reaches or exceeds theta3, all pressure relief channels are opened and the isolation protection action of adjacent cabinets is activated.

10. The rapid detection and pressure relief system for internal arc faults in a ring main unit according to claim 1, characterized in that: The analog-to-digital converter in the multi-channel sensor signal acquisition unit has a sampling rate of no less than 100 kHz and a resolution of no less than 12 bits; the short-time window length for the arc light sensing channel is 0.1-0.5 ms and the long-time window length is 50-200 ms; the short-time window length for the pressure sensing channel is 0.5-2 ms and the long-time window length is 100-500 ms; the short-time window length for the current sensing channel is 0.2-1 ms and the long-time window length is 100-500 ms; the ratio of the number of short-time window sampling points to the number of long-time window sampling points for each channel is no greater than 1:100; the characteristic function calculation... The unit, the parallel short-time average and long-time average recursive calculation unit, the multi-channel fusion trigger determination unit, and the graded pressure relief control unit are all implemented in hardware logic circuits within the same field-programmable gate array chip. The recursive calculations of the three sensing channels correspond to three independent parallel computing pipelines that are executed simultaneously. The multi-channel fusion trigger determination unit completes the fusion determination operation in a single clock cycle using combinational logic circuits. The arc fault trigger signal is output through an optocoupler isolation circuit to drive the circuit breaker tripping mechanism and the pressure relief execution mechanism. The calculation and processing delay from the occurrence of the arc fault to the output of the arc fault trigger signal does not exceed 10 microseconds.

Images