Environment energy self-adaptive collection method based on SF6 online monitoring

Abstract

This invention relates to the field of SF6 online monitoring technology and discloses an adaptive environmental energy harvesting method based on SF6 online monitoring. The method includes: dividing energy periods according to the natural valley value of energy storage and simultaneously collecting energy from two sources; extracting the magnetocaloric timing deviation of the two-source energy; using the deviation to correct the reference arrival time to obtain the effective arrival time, thereby constructing effective gas observability; generating dual-source harvesting weights based on the deviation to dynamically adjust the equivalent input load; combining instantaneous stored energy with effective gas observability to find the earliest moment that meets the total energy requirement as the sole sampling moment for monitoring; and finally, inverting the multi-component gas concentration values ​​and closing the loop to correct the observability of the next period. This invention solves the problem of mismatch between environmental energy enrichment and the arrival time of the target gas, achieving a high degree of synergy between energy scheduling and fault diagnosis windows.

Description

Technical Field

[0001] This invention relates to the field of SF6 online monitoring technology, and more specifically, to an adaptive environmental energy harvesting method based on SF6 online monitoring. Background Technology

[0002] With the development of the power Internet of Things, online monitoring of sulfur hexafluoride (SF6) decomposition products inside gas-insulated equipment has become an important means to ensure the safe operation of the power grid. In the existing deployment of monitoring nodes, a dual-source system with the same circuit but different locations is often used to power the sensor nodes. That is, a magnetic field energy harvesting unit is deployed near the current-carrying conductor, and a thermoelectric energy harvesting unit is deployed on the outer shell of the gas chamber. However, when faults such as local overheating or poor contact occur inside the equipment, the response of the magnetic field energy harvesting module to current fluctuations is almost instantaneous; while the heat conduction to the outer shell requires a long thermal inertia link, resulting in a significant lag in the thermoelectric response; in particular, the time required for the decomposition gas caused by the fault to reach the gas chamber sampling branch through natural convection and diffusion is even longer. Existing technologies, when performing self-powered energy management, usually follow the control logic of immediate wake-up when energy is abundant. This traditional strategy completely ignores the natural timing misalignment between electromagnetic transfer, heat conduction, and gas diffusion under co-source excitation. When a surge in transient current brings abundant magnetic field energy, the controller is often prematurely awakened and blindly performs energy-intensive actions such as preheating, sampling, and transmission. At this time, the truly representative fault gas has not yet reached the monitoring blind zone. By the time the mixed gas that best reflects the nature of the fault finally spreads to the sensor inlet, the system's energy storage has often already dropped to its lowest value and become inoperable. This severe timing mismatch between energy availability and gas observability leads to a complete disconnect between online monitoring actions and the actual fault evolution window, greatly weakening the actual diagnostic effectiveness of self-powered monitoring equipment and the validity of online monitoring data. Summary of the Invention

[0003] This invention provides an adaptive environmental energy harvesting method based on online SF6 monitoring, which solves the technical problems mentioned in the background art.

[0004] This invention provides an adaptive environmental energy harvesting method based on online SF6 monitoring, comprising:

[0005] The beneficial effects of this invention are as follows: This invention breaks through the limitations of traditional self-powered systems that blindly trigger monitoring based solely on instantaneous power surplus. By extracting the time lag deviation between the magnetic and thermal sources under the same fault excitation, and deeply integrating this deviation into the dual-source collection weight allocation and gas arrival time assessment, this invention, through the construction of a joint determination mechanism of energy storage status and gas effective observability, can accurately predict the gas sampling window with the highest diagnostic value. This effectively avoids premature energy consumption when magnetic field energy is temporarily enriched but the faulty gas has not yet arrived, and also prevents energy depletion during the optimal observation period of gas concentration. Ultimately, it achieves deep synchronization between energy metabolism rhythm and the actual fault evolution process, significantly improving the accuracy of online inversion of multi-component gases and the long-term operational reliability of the system. Attached Figure Description

[0006] Figure 1 This is a flowchart of an adaptive environmental energy harvesting method based on SF6 online monitoring according to the present invention. Detailed Implementation

[0007] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0008] like Figure 1 As shown, an adaptive environmental energy harvesting method based on SF6 online monitoring is applied to a dual-source system with co-loop and disparate locations, comprising a magnetic field energy harvesting unit, a thermoelectric energy harvesting unit, a gas sensing unit, and an energy storage supercapacitor. The method includes: The energy cycle is divided according to the energy storage natural valley value of the supercapacitor and dual-source energy is collected simultaneously. Extract the magnetocaloric timing deviation of the dual-source energy within the energy cycle; The reference arrival time is obtained, and the effective arrival time is obtained by correcting the reference arrival time using the magnetocaloric timing deviation, thereby constructing the effective observability of the gas; The dual-source collection weight is generated based on the magnetocaloric timing deviation, and the dual-source collection weight is converted into an equivalent input load. The instantaneous stored energy is obtained, and the instantaneous stored energy is multiplied by the effective observability of the gas to obtain a comprehensive evaluation index. Based on the comprehensive evaluation index, the earliest moment that meets the total energy requirement is found as the unique sampling moment. At the unique sampling time, the gas sensing unit is controlled to perform monitoring, and the concentration values ​​of multi-component gases are inverted using a comprehensive sensitivity model that incorporates the magnetocaloric timing deviation. The cumulative value of the gas effective observability is calculated to update the correction coefficient, and the gas effective observability for the next period is corrected using the updated correction coefficient.

[0009] Preferably, the energy cycle is divided according to the natural valley value of the energy storage supercapacitor, and dual-source energy is collected simultaneously, including: The instantaneous power of the magnetic field is obtained by simultaneously acquiring the voltage and current at the magnetic field energy harvesting terminal and multiplying them. Similarly, the instantaneous power of the thermoelectric energy harvesting terminal is obtained by simultaneously acquiring the voltage and current at the thermoelectric energy harvesting terminal and multiplying them. The instantaneous stored energy is calculated by collecting the terminal voltage of the supercapacitor and combining it with the capacitance value of the supercapacitor. The moment when the first derivative of the instantaneous stored energy with respect to time is equal to zero and the second derivative with respect to time is greater than zero is taken as the valley moment. : The time interval between two adjacent valley moments is divided into the energy cycle. : Within the energy cycle, the instantaneous power of the magnetic field and the instantaneous power of the thermoelectric field are integrated over time to obtain the dual-source energy, which includes magnetic field harvesting energy. and thermoelectric energy harvesting : in, The instantaneous power of the magnetic field. This is the voltage at the magnetic field energy harvesting terminal. This refers to the current at the magnetic field energy harvesting end. For thermoelectric instantaneous power, This is the voltage at the thermoelectric energy harvesting terminal. This refers to the current at the thermoelectric energy harvesting terminal. For the instantaneous energy storage, Let be the capacitance value of the energy storage supercapacitor. The terminal voltage of the energy storage supercapacitor; For the first At the lowest point, For the first A valley moment; The energy cycle; To collect energy for the magnetic field, To collect energy for thermoelectricity; It is a time variable.

[0010] The magnetic field energy acquisition terminal voltage is the actual potential difference between the two ends of the output terminal of the magnetic field energy acquisition unit at a certain sampling moment. It can be obtained directly through an isolated sampling circuit, a voltage divider sampling circuit, or a low-power voltage acquisition chip with analog-to-digital conversion function.

[0011] The magnetic field energy harvesting terminal current is the actual current value flowing through the output circuit of the magnetic field energy harvesting unit at a certain sampling moment. It can be obtained directly through a sampling resistor, a current detection amplifier, a Hall current sensor, or a magnetic modulation current acquisition circuit.

[0012] Instantaneous magnetic field power is the instantaneous power value obtained by multiplying the voltage and current at the magnetic field energy harvesting end at the same moment. It is used to characterize the ability of the magnetic field energy harvesting unit to provide energy to the system at that moment, and is also the basic input for subsequent calculations of magnetic field energy collection, average magnetic field power, and magnetocaloric timing deviation.

[0013] The thermoelectric energy harvesting terminal voltage is the actual potential difference between the two ends of the output terminal of the thermoelectric energy harvesting unit at a certain sampling moment. It can be obtained directly through a high input impedance voltage sampling circuit, a voltage divider network, or an analog-to-digital converter chip.

[0014] The thermoelectric energy harvesting terminal current is the actual current value flowing through the output circuit of the thermoelectric energy harvesting unit at a certain sampling moment. It can be obtained directly through a sampling resistor, a current sensing amplifier, or the current sensing terminal of a dedicated energy harvesting management chip.

[0015] Thermoelectric instantaneous power is the instantaneous power value obtained by multiplying the thermoelectric energy harvesting terminal voltage and the thermoelectric energy harvesting terminal current at the same moment. It is used to characterize the ability of the thermoelectric energy harvesting unit to provide energy to the system at that moment, and is also used for subsequent calculations of thermoelectric energy harvesting, average thermoelectric power, and effective observability of the gas.

[0016] The terminal voltage of a supercapacitor is the actual voltage value across its terminals at a given moment. It can be obtained directly using a high-impedance voltage follower sampling circuit, a voltage divider network with analog-to-digital conversion, or an ultra-low static power consumption voltage monitoring chip.

[0017] The capacitance value of an energy storage supercapacitor is its rated capacity for storing electrical charge. It is preferably between 10 and 200 Faradays. The preferred value is based on the power consumption level of the online monitoring node, the sampling and transmission interval, the allowable size, and the target lifetime. When a node requires high energy for a single complete monitoring operation, 50 to 200 Faradays is preferred. When the node size is limited and the operating frequency is low, 10 to 50 Faradays is preferred.

[0018] Instantaneous energy storage is the current available energy storage value calculated based on the capacitance value and terminal voltage of the supercapacitor.

[0019] The trough moment is the moment when the instantaneous stored energy reaches a local minimum on the time axis.

[0020] The energy cycle is a continuous time interval between two adjacent trough moments, which adapts to changes in environmental stimuli and node energy consumption behavior.

[0021] Dual-source energy is the collective term for the energy output from the magnetic field energy harvesting unit and the thermoelectric energy harvesting unit within the same energy cycle. It is used to characterize the total energy input obtained by the system from the two types of environmental energy sources within the current cycle.

[0022] Magnetic field energy harvesting is the energy value obtained by integrating the instantaneous power of the magnetic field over a period of time within an energy cycle. It is used to characterize the actual contribution of the magnetic field source to the system's energy storage during that period.

[0023] Thermoelectric energy harvesting is the energy value obtained by integrating the instantaneous thermoelectric power over time within an energy cycle. It is used to characterize the actual contribution of the thermoelectric power source to the system's energy storage during that cycle.

[0024] In detail, the hardware synchronization method for dual-source synchronous acquisition is as follows: a unified sampling clock is set inside the controller, and the sampling of magnetic field energy acquisition terminal voltage, magnetic field energy acquisition terminal current, thermoelectric energy acquisition terminal voltage, and thermoelectric energy acquisition terminal current are started simultaneously with the same trigger interrupt; if the number of analog-to-digital conversion channels is insufficient, the same start pulse drives multiple analog switches to complete polling sampling in a fixed order within a time window of no more than 1 millisecond, and the samples in this group are uniformly marked as the same sampling time; for magnetic field channels with rapid power changes, magnetic field voltage and magnetic field current are preferentially arranged in the first two sampling sub-slots; for thermoelectric channels with slow power changes, they can be arranged in subsequent sub-slots.

[0025] In detail, the noise reduction method for sampling the terminal voltage of the supercapacitor and converting the instantaneous stored energy is as follows: First, the terminal voltage of the supercapacitor is sampled at a rate not lower than the dual-source power sampling frequency. Then, the moving average or median filtering is performed on the terminal voltage samples from the most recent 3 to 7 points to obtain the noise-reduced terminal voltage sequence. Subsequently, the stored energy is converted using the nominal capacitance value or the effective capacitance value after temperature compensation of the supercapacitor. If there is a large current pulse load at the node, a recovery delay of 1 to 3 sampling cycles is added after the transmission and preheating actions to avoid voltage rebound causing false valley values. For example, when the transmission phase causes a momentary drop in terminal voltage, 200 milliseconds can be waited before using the average voltage of the most recent 5 points to calculate the instantaneous stored energy.

[0026] In detail, the discrete implementation rules for determining the valley moment are as follows: In the discrete sampling sequence, the instantaneous stored energy sequence is first smoothed, and then the difference between adjacent sampling points is used to approximate the first-order rate of change, and the difference between two consecutive differences is used to approximate the second-order rate of change; when a sampling point satisfies that the previous difference is less than or equal to 0, the subsequent difference is greater than or equal to 0, and the corresponding second-order difference is greater than a preset curvature threshold, the point is determined to be a candidate valley moment; the preset curvature threshold must be greater than the maximum background noise change rate caused by random fluctuations in the natural environment to filter high-frequency interference and ensure that the identified valley value represents the true energy cycle alternation; if the minimum value appears between two sampling points, the smaller value corresponding to the sampling point is taken or a more refined valley moment is obtained by quadratic interpolation; for example, when the sampling period is 100 milliseconds, if the value is still decreasing from point 25 to point 26, and starting to rise from point 26 to point 27, then point 26 can be used as a candidate valley moment.

[0027] In detail, the method for eliminating false valley values ​​and the minimum interval rule for adjacent valley values ​​is as follows: First, set a minimum valley value interval, which is preferably 1 to 3 times the duration of a complete monitoring action; if multiple candidate valley values ​​appear within the same minimum interval window, only the point with the smallest instantaneous stored energy is retained; at the same time, set a minimum valley depth threshold, and only when the decrease of the candidate valley value relative to the local average on both sides exceeds the threshold is it retained, otherwise it is regarded as a false valley value caused by noise disturbance; for example, when node standby jitter will cause several very shallow valleys, the minimum valley depth can be set to 1% to 3% of the average stored energy, and all valleys below this depth are eliminated.

[0028] In detail, the discrete algorithm for time integration and the sampling period are as follows: Within each energy cycle, the instantaneous power of the magnetic field and the instantaneous power of thermoelectricity are discretely integrated using the trapezoidal integral method; the sampling period is preferably between 10 milliseconds and 500 milliseconds, specifically determined by the rate of change of the dual-source power; when the magnetic field power changes rapidly, it is preferably between 10 milliseconds and 50 milliseconds; when the thermoelectric power changes gradually and the node processing capacity is limited, it is preferably between 50 milliseconds and 500 milliseconds; for example, between the nth and n+1th sampling points, the energy collected by the magnetic field can be accumulated by multiplying the average instantaneous power of the magnetic field at those two points by the sampling period, and the energy collected by the thermoelectricity is calculated in the same way.

[0029] Preferably, extracting the magnetocaloric timing deviation of the dual-source energy within the energy cycle includes: Calculate the mean and standard deviation of the instantaneous magnetic field power within the energy cycle, and the mean and standard deviation of the instantaneous thermoelectric power: The instantaneous magnetic field power is standardized using the mean magnetic field power, the standard deviation of the magnetic field power, and the numerical stability term to obtain the normalized magnetic field power; the instantaneous thermoelectric power is standardized using the mean thermoelectric power, the standard deviation of the thermoelectric power, and the numerical stability term to obtain the normalized thermoelectric power. The product of the normalized magnetic field power and the time derivative of the normalized thermoelectric power, minus the difference between the product of the normalized thermoelectric power and the time derivative of the normalized magnetic field power, is integrated over time within the energy period and halved to obtain the directed area of ​​the trajectory: The arc length of the trajectory is obtained by taking the square root of the sum of the squares of the time derivatives of the normalized magnetic field power and the normalized thermoelectric power, and then integrating over time within the energy period. The magnetocaloric timing deviation is obtained by dividing twice the directed area of ​​the trajectory by the square of the arc length of the trajectory and the sum of the numerical stability term. in, The duration of the energy cycle; This is the average magnetic field power. This represents the standard deviation of the magnetic field power. The average thermoelectric power. The standard deviation of thermoelectric power; It is a numerically stable term; To normalize the magnetic field power, Normalized thermoelectric power; Let the area of ​​the trajectory be the directed area. Let the arc length of the trajectory be denoted as 'arc'. This is the magnetocaloric timing deviation.

[0030] The duration of an energy cycle is the time from the previous trough to the current trough. It is used to normalize the average power, fluctuation amplitude, and integral results within the cycle, so that energy cycles of different lengths can be compared on the same scale.

[0031] The average magnetic field power is a representative value obtained by averaging the instantaneous magnetic field power over time within the current energy cycle. It is used to characterize the average level of magnetic field energy input in this cycle.

[0032] The standard deviation of magnetic field power is a measure of the degree to which the instantaneous power of the magnetic field fluctuates around the mean of the magnetic field power within the current energy cycle, and is used to characterize the strength of the fluctuations in the output of the magnetic field source.

[0033] The average thermoelectric power is a representative value obtained by averaging the instantaneous thermoelectric power over time within the current energy cycle, and is used to characterize the average level of thermoelectric energy input in this cycle.

[0034] Thermoelectric power standard deviation is a measure of the degree to which instantaneous thermoelectric power fluctuates around the thermoelectric power mean within the current energy cycle, and is used to characterize the strength of the fluctuations in thermoelectric power output.

[0035] The numerical stability term is a small positive number added during normalization, ratio calculation, arc length normalization, and subsequent load conversion. When this positive number is actually substituted into the calculation, it is by default assigned the same physical unit as the main variable in the denominator of the formula, to satisfy the homogeneity of units when adding physical formulas and to prevent numerical divergence when the denominator approaches zero. The preferred value for its pure numerical part is between one part per million and one part per thousand.

[0036] Normalized magnetic field power is a dimensionless sequence obtained by subtracting the mean magnetic field power from the instantaneous magnetic field power and then scaling it using the standard deviation of magnetic field power and the numerical stability term.

[0037] Normalized thermoelectric power is a dimensionless sequence obtained by subtracting the mean thermoelectric power from the instantaneous thermoelectric power and then scaling it using the standard deviation of thermoelectric power and the numerical stability term.

[0038] The normalized magnetic field power time derivative is the rate of change of normalized magnetic field power over time, used to reflect the rise and fall rates of the output rhythm of the magnetic field source.

[0039] The normalized thermoelectric power time derivative is the rate of change of normalized thermoelectric power over time, used to reflect the rise and fall rates of the thermoelectric power output rhythm.

[0040] The directional area of ​​a trajectory is a directional area enclosed by the normalized magnetic field power and the normalized thermoelectric power as two-dimensional plane trajectories. It is used to describe which of the two types of energy sources changes first and which changes later in the same period, as well as the direction of the deviation.

[0041] The trajectory arc length is the total path length of a two-dimensional trajectory composed of normalized magnetic field power and normalized thermoelectric power within one energy cycle, and is used to characterize the overall activity of the dual-source temporal variation.

[0042] The magnetocaloric timing deviation is a characterization quantity obtained by normalizing the directed area and arc length of the trajectory. It is used to measure the sequence, phase deviation, and degree of asymmetry between the magnetic field source and the thermal source under the same fault excitation. The closer the magnetocaloric timing deviation is to zero, the closer the rhythms of the two types of energy sources are to synchronization. The larger the absolute value of this parameter, the more significant the magnetocaloric timing deviation.

[0043] In detail, the discrete calculation method for the mean and standard deviation of magnetic field power, as well as the mean and standard deviation of thermoelectric power, is as follows: Collect all sampling points within the current energy cycle, and calculate the sample mean for the instantaneous magnetic field power and the instantaneous thermoelectric power respectively; then divide the sum of squares of the deviations between the sample and the mean by the number of sample points or the number of sample points minus 1 to obtain the variance, and then take the square root of the variance to obtain the standard deviation; when the number of sample points in the cycle is less than 3, pause the update of the time series deviation for the current cycle, and use the results of the previous valid cycle.

[0044] In detail, the normalized power time derivative is obtained and smoothed as follows: First, the normalized magnetic field power and normalized thermoelectric power are smoothed by a 3-point to 7-point moving average or local polynomial, and then the time derivative is obtained by central difference; the sampling points at the beginning and end of the sequence are obtained by forward difference and backward difference; if the absolute value of the derivative exceeds 5 times the median of the derivative within the period for two consecutive sampling points, it is regarded as spike noise and replaced by the neighborhood mean; for example, when the sampling period is 50 milliseconds, the normalized magnetic field power time derivative at a certain point can be obtained by dividing the difference between the next sampling point and the previous sampling point by 100 milliseconds.

[0045] In detail, the discrete integration method for the directed area and arc length of the trajectory is as follows: the normalized magnetic field power and normalized thermoelectric power of each sampling point are regarded as two-dimensional trajectory nodes; the directed area of ​​the trajectory is discretely accumulated by the trapezoidal integral of adjacent sampling segments; the arc length of the trajectory is discretely accumulated by taking the square root of the sum of the squares of the two derivatives of adjacent sampling segments and then multiplying it by the sampling period; to avoid high-frequency noise amplification, derivative smoothing can be performed before accumulation; for example, in the nth time segment, the integrand of the first and last points of the segment can be averaged and multiplied by the segment length, and then summed over all time segments to obtain the directed area and arc length of the trajectory in the current period.

[0046] In detail, the meaning of the sign and the effective range of the magnetocaloric timing deviation are as follows: a positive deviation is defined as the magnetic field leading the thermoelectric response, and a negative deviation is defined as the thermoelectric response leading the magnetic field. To ensure control stability, the calculated magnetocaloric timing deviation is limited, preferably to between -1 and 1. When the absolute value is less than 0.02, it can be judged as approximately synchronous; when the absolute value is greater than 0.3, it can be judged as significantly asynchronous. For example, if the calculated value for a certain period is 0.45, it indicates that the magnetic field response has a significant lead over the thermoelectric response, and a stronger correction effect should be reflected in the subsequent effective arrival time and dual-source weighting.

[0047] Preferably, the reference arrival time is obtained, and the effective arrival time is obtained by correcting the reference arrival time using the magnetocaloric timing deviation, thereby constructing the effective observability of the gas, including: The average shell temperature difference is obtained by averaging the difference between the shell temperature of the target chamber and the ambient temperature. The equivalent velocity of natural convection is calculated by combining gravitational acceleration, gas thermal expansion coefficient, gas kinematic viscosity, gas thermal diffusivity, the average shell temperature difference, and the equivalent convection characteristic length: Dividing the equivalent diffusion coefficient of this period by the equivalent propagation path length yields the equivalent diffusion velocity: Dividing the equivalent transmission path length by the sum of the natural convection equivalent velocity, the diffusion equivalent velocity, and the numerical stability term yields the reference arrival time: The absolute value integral of the difference between the normalized magnetic field power time derivative and the normalized thermoelectric power time derivative is calculated and divided by the sum of the trajectory arc length and the numerical stability term to obtain the asynchronous intensity correction term: The effective arrival time is obtained by multiplying the reference arrival time by the sum of the products of the asynchronous intensity correction term and the absolute value of the magnetocaloric timing deviation: An exponential decay function is constructed using the effective arrival time, and the exponential decay function is integrated with the instantaneous thermoelectric power to obtain the effective observability of the gas. in, For the shell temperature, For ambient temperature, This represents the average shell temperature difference. It is the acceleration due to gravity. The coefficient of thermal expansion of the gas is _____. For the kinetic viscosity of the gas, For gas thermal diffusivity, It is the equivalent convection characteristic length; The equivalent velocity for natural convection; This is the equivalent diffusion coefficient for that period. This is the equivalent transmission path length; For diffusion equivalent velocity; The reference arrival time; This is an asynchronous intensity correction term; The effective arrival time; For integration time; The effective observability of the gas.

[0048] The shell temperature is the actual temperature value of the target gas chamber shell at the corresponding measurement point during the sampling period, which is used to reflect the thermal state after the internal fault heat is conducted to the shell.

[0049] Ambient temperature is the actual temperature of the air surrounding the target chamber, used as a reference baseline for the heated state of the outer shell.

[0050] The average shell temperature difference is a representative value obtained by subtracting the shell temperature from the ambient temperature over one energy cycle and averaging the results. It is used to characterize the average degree of heating of the target chamber shell relative to the environment.

[0051] Gravitational acceleration is a physical constant characterizing the strength of the gravitational field, and its value is taken as 9.8 meters per second squared.

[0052] The coefficient of thermal expansion of a gas is a physical property parameter characterizing the sensitivity of a target gas to volume changes when heated. It can be obtained through volume change experiments under isothermal and isobaric conditions.

[0053] Gas kinematic viscosity is a physical property parameter that characterizes the ability of a gas to diffuse momentum within it, and it can be obtained through flow calibration tests.

[0054] Gas thermal diffusivity is a physical property parameter characterizing the ability of a gas to diffuse heat within its interior. It can be obtained through thermal response calibration tests.

[0055] The equivalent convection characteristic length is a parameter that converts the geometric dimensions of the heated area of ​​the target air chamber into a representative length for natural convection calculation. It is used to uniformly characterize the size of the buoyancy channel and is preferably 0.02 meters to 0.5 meters.

[0056] The natural convection equivalent velocity is a representative convective transport velocity calculated based on the average shell temperature difference and gas physical properties. It is used to characterize the speed at which fault gas migrates along the gas chamber and sampling branch due to the thermal buoyancy effect.

[0057] The equivalent diffusion coefficient is a comprehensive parameter that converts the diffusion behavior of multi-component fault gases in the current cycle into a single representative of diffusion capacity. It is used to characterize the ability of fault gases to migrate to the sampling location via molecular diffusion in the absence of significant directional flow. In the first cycle, it can be given by the typical diffusion coefficient of the target gas with empirical weights, and in subsequent cycles, it is updated by the concentration weights of the multi-component gases.

[0058] The equivalent transport path length is the representative migration distance between the main generation area of ​​the faulty gas and the effective sampling position of the gas sensing unit, used to uniformly characterize the average path of the gas from the source region to the sensing region. For the internal geometry of the gas chamber, the length of the sampling branch, the number of bends, and the location of the effective mixing region, a length of 0.05 meters to 2 meters is preferred. For structures where the sampling port is close to the main cavity of the gas chamber, a length of 0.05 meters to 0.5 meters is preferred. For structures with long drainage branches, a length of 0.5 meters to 2 meters is preferred.

[0059] The diffusion equivalent velocity is a measure obtained by dividing the equivalent diffusion coefficient by the equivalent propagation path length.

[0060] The baseline arrival time is the representative arrival time of the fault gas calculated based on the equivalent transport path length, the equivalent velocity of natural convection, and the equivalent velocity of diffusion, without considering the asynchronous effect of the two sources. Here, the equivalent velocity of convection and the equivalent velocity of diffusion are added as pure numerical scalars. This is an engineering equivalent conservative treatment based on the fact that the two mass transfer mechanisms are completely collinear and superimposed in the same direction on the main spatial path under the fluid a priori condition. The aim is to quickly obtain the shortest theoretical minimum lower limit of the arrival time of the real gas.

[0061] The asynchronous intensity correction term is obtained by integrating the absolute value of the difference between the normalized magnetic field power time derivative and the normalized thermoelectric power time derivative, and then normalizing it using the trajectory arc length and numerical stability term. It characterizes the degree of temporal asynchrony between the two sources. A larger asynchronous intensity correction term indicates a greater temporal inconsistency in the magnetothermal response.

[0062] The effective arrival time is the representative arrival time obtained after the reference arrival time has been corrected by the asynchronous intensity correction term and the magnetocaloric timing deviation. It is used to approximately reflect the arrival time of the real fault gas that is truly diagnostically significant for the current sensing node.

[0063] The exponential decay function is a weighted function constructed based on the effective arrival time, which gradually decreases with the lag time. It is used to give greater weight to the thermoelectric response that is closer to the current time, while gradually decaying the earlier thermal response, thereby highlighting the thermal information that is more relevant to the current arrival state of the fault gas.

[0064] Gas effective observability is a characterization quantity obtained by integrating the exponential decay function with the instantaneous thermoelectric power. It is used to reflect the degree to which the fault gas is effectively detectable by the gas sensing unit at the current moment.

[0065] In detail, the placement of the casing temperature and ambient temperature measurement points is as follows: at least 1 to 3 measurement points should be placed for the casing temperature, preferably near the thermoelectric energy collection unit, on the outer surface of the main path of the predicted fault heat conduction, and at a representative position in the middle of the air chamber; the ambient temperature measurement points should be placed within 50 mm to 300 mm of the casing, avoiding direct sunlight and forced air vents; if multiple casing temperature measurement points are placed, the temperature of each measurement point should be averaged first and then the difference should be calculated with the ambient temperature; for example, for a long air chamber, one casing temperature sensor can be placed at the left end, the middle, and the right end, while the ambient temperature sensor is placed in the unobstructed area on the outer side of the middle.

[0066] In detail, the method for determining the equivalent convection characteristic length is as follows: First, identify the main heated area that forms natural convection after the fault heat propagates to the shell, and then select the geometric scale of this area along the main buoyancy direction as the equivalent convection characteristic length; when the main buoyancy direction is not obvious, take the characteristic side length of the main heated surface as the equivalent convection characteristic length; if the structure is complex, the main convection channel can be determined by finite element thermal simulation or flue gas visualization test, and then a representative length can be selected; for example, the vertical height of the heated air chamber can be taken, and the longer side length of the heated surface can be taken for the flat box.

[0067] In detail, the equivalent transmission path length is determined as follows: from the main area where the fault gas is generated to the effective sampling position of the gas sensing unit, the gas is measured in segments along the main migration path that it is most likely to pass through, and then the results are summed; if there are multiple parallel branches in the path, the main path with the least flow resistance and the easiest way to reach the sensor is selected as the representative path; if there is a backflow or mixing cavity inside the equipment, the equivalent path length can be calculated based on the actual smoke tracing or flow field simulation; for example, if the fault point is located in the middle of the main cavity and the sensor is located at the end of the side branch, the path length should include the distance from the main cavity to the sampling port and the length of the side branch.

[0068] In detail, the initial value of the equivalent diffusion coefficient for the first period is obtained as follows: Before the concentration weights of multiple components are obtained in the first monitoring period, the proportion of empirical components is set according to the gas production spectrum of the most common fault of the target equipment, and the equivalent diffusion coefficient for the first period is obtained by weighted summation of the diffusion coefficients of each target gas; if the empirical spectrum is lacking, the arithmetic mean of the diffusion coefficients of each target gas can be taken as a conservative initial value; for example, for monitoring the decomposition of sulfur hexafluoride, the initial weighted diffusion coefficient can be given according to the empirical proportions of sulfur dioxide, hydrogen sulfide, carbon dioxide and hydrogen, and then updated by the inversion results in subsequent periods.

[0069] In detail, the natural convection equivalent velocity is applicable to operating conditions where natural convection is dominant and there is no external forced wind disturbance or the forced disturbance is negligible. If there is continuous fan purging or significant forced ventilation on site, this operating condition should be marked separately as an abnormal forced convection operating condition and the processing should be suspended. In terms of direction processing, the natural convection equivalent velocity only takes the effective component along the main migration direction of the fault gas to the sensor. If the calculated direction is opposite to the main migration direction, the component is treated as 0. For example, if the casing is heated and mainly drives the gas to rise, and the sensor is located in the upper branch, the upward component is taken. If the sensor is located in the lower branch and there is no downward mainstream, the upward floating velocity is not directly used to shorten the arrival time.

[0070] In detail, the effective arrival time is limited and anomaly handling methods are as follows: First, the original effective arrival time is obtained by multiplying the baseline arrival time by a correction factor, and then it is limited to a minimum of 0.1 seconds and a maximum of 3600 seconds; if the sum of the natural convection equivalent velocity, diffusion equivalent velocity and numerical stability term is too small, causing the baseline arrival time to be abnormally amplified, the upper limit value is directly taken and the anomaly flag is recorded; if the shell temperature difference is close to zero and the thermoelectric instantaneous power is extremely low for a long period of time in the current cycle, it can be considered that the effective arrival time cannot be reliably estimated for the time being. At this time, the result of the previous effective cycle is used or the conservative default value is switched; for example, if the temperature difference is abnormally large and negative due to the failure of the temperature sensor in a certain cycle, the effective arrival time can be frozen and not updated.

[0071] In detail, the discrete implementation of the exponential decay function is as follows: On the discrete time axis, with the current observation time as the endpoint, trace back the instantaneous thermoelectric power samples since the current valley time; for each historical sample, calculate the exponential decay weight according to its lag time relative to the current time, and then multiply the weight by the corresponding instantaneous thermoelectric power and accumulate it; finally, divide by the effective arrival time or perform uniform scaling according to the normalization coefficient in the formula; for example, when the effective arrival time is 20 seconds and the sampling period is 1 second, the most recent 20 to 40 samples can be summed according to the principle that the longer the lag, the smaller the weight, to obtain the effective observability of the gas at the current time.

[0072] Preferably, generating dual-source collection weights based on the magnetocaloric timing deviation and converting the dual-source collection weights into an equivalent input load includes: Divide the magnetic field harvesting energy by its sum with the thermoelectric harvesting energy and the numerical stability term, and then divide by the sum of the products of the asynchronous intensity correction term and the absolute value of the magnetocaloric timing deviation to obtain the initial magnetic field weight: Divide the thermoelectric harvested energy by its sum with the magnetic field harvested energy and the numerical stability term, and then multiply by the sum of the products of the asynchronous intensity correction term and the absolute value of the magnetocaloric timing deviation to obtain the initial thermoelectric weight: Dividing the initial magnetic field weight and the initial thermoelectric weight by their sum respectively yields the magnetic field collection weight and the thermoelectric collection weight that constitute the dual-source collection weight: The square of the open-circuit voltage of the corresponding energy source in the previous cycle is divided by four times the product of the maximum available power of the corresponding energy source in the previous cycle and the corresponding dual-source collection weight, plus the sum of the numerical stability terms, to convert it into the equivalent input load of the energy source: in, For the initial magnetic field weights, Initial thermoelectric weights; As an energy source identifier variable, Corresponding magnetic field source, Corresponding thermal power source; For source The corresponding magnetic field collection weights and thermoelectric collection weights; For source The corresponding open-circuit voltage of the previous cycle; For source The corresponding maximum available power in the previous cycle; This is the equivalent input load of the energy source.

[0073] The initial magnetic field weight is the initial allocation ratio of the magnetic field source obtained by combining the proportion of energy collected by the magnetic field with the asynchronous intensity correction term and the magnetocaloric timing deviation. It is used to reflect the initial priority of the magnetic field source for subsequent energy collection decisions under the current dual-source timing relationship.

[0074] The initial thermoelectric weight is the initial allocation ratio of the thermal power source obtained by combining the proportion of thermoelectric harvested energy with the asynchronous intensity correction term and the magnetocaloric timing deviation. It is used to reflect the initial priority of the thermal power source for subsequent energy harvesting decisions under the current dual-source timing relationship.

[0075] The magnetic field harvesting weight is the final magnetic field source allocation ratio obtained by normalizing the initial magnetic field weight and the initial thermoelectric weight. It is used to represent the actual preference of the energy management strategy for the magnetic field source in the next cycle. The larger the magnetic field harvesting weight, the more the control strategy tends to obtain energy from the magnetic field source.

[0076] The thermoelectric harvesting weight is the final thermoelectric power allocation ratio obtained by normalizing the initial magnetic field weight and the initial thermoelectric weight. It is used to represent the actual preference of the energy management strategy for the thermoelectric power source in the next cycle. The larger the thermoelectric harvesting weight, the more the control strategy tends to obtain energy from the thermoelectric power source.

[0077] Energy source identification variables are discrete identification parameters used to distinguish between magnetic field sources and thermal sources.

[0078] The open-circuit voltage of the corresponding energy source in the previous cycle is the terminal voltage measured at the end of the previous cycle under the condition that the corresponding energy source is approximately disconnected from the subsequent load. It is used to characterize the intrinsic voltage level of the energy source. It can be obtained by sampling through the energy acquisition management circuit under the condition of briefly disconnecting the load or switching to a high-impedance measurement channel.

[0079] The maximum available power of the corresponding energy source in the previous cycle is the maximum power that the corresponding energy source can theoretically or practically output under optimal load matching conditions in the previous cycle. It can be measured by multi-level load scanning or estimated from open-circuit voltage and short-circuit current.

[0080] The equivalent input load is the target load value obtained through conversion. The purpose of introducing collection weights is not to pursue the optimal power output of a single source, but to increase its input impedance by actively deviating from the optimal impedance matching point of low-priority energy sources to limit power extraction, thereby achieving the scheduling purpose of acquiring environmental energy according to weight quotas at the global system level.

[0081] In detail, the measurement method of the open-circuit voltage of the previous cycle is as follows: at the end of the previous cycle or within a short window before the weight update, the corresponding energy source is disconnected from the downstream load or switched to a high-impedance detection state by analog switch, and the voltage value is collected after waiting for 1 to 50 milliseconds for the terminal voltage to stabilize. In order to avoid affecting the continuous power supply, the disconnection measurement window can be limited to when the system has sufficient energy storage and the load fluctuation is small. For example, for thermoelectric energy harvesting unit, it can enter the high-impedance detection mode during the gap when the boost chip stops switching action and collect its approximate open-circuit voltage.

[0082] In detail, the method for determining the maximum available power in the previous cycle is as follows: Select several representative load points in the previous cycle, perform multi-level load scanning on the corresponding energy source, record the output voltage and output current of each load point, calculate the output power, and take the maximum value as the maximum available power in this cycle; if the system does not allow frequent scanning, the maximum available power can be approximated by the open-circuit voltage and short-circuit current estimation method; for example, for the magnetic field energy harvesting unit, 4 to 8 equivalent loads can be tried in one cycle, and the maximum value is taken after calculating the output power of each load.

[0083] In detail, the weight normalization method for single-source failure or dual-source extremely low energy is as follows: if the open-circuit voltage of a certain energy source is lower than the oscillation threshold or the maximum available power is lower than the minimum available power threshold in the current cycle, then the source is considered to be temporarily failed, and its final weight is directly set to 0, while the weight of the other source is reset to 1; if both sources are in an extremely low energy state at the same time, the weight of the previous effective cycle remains unchanged, or it reverts to the preset safety weight; for example, when the thermoelectric side does not oscillate for a long time under low temperature difference conditions, the thermoelectric collection weight can be temporarily set to 0, and only the magnetic field side is powered.

[0084] In detail, the implementation of mapping the equivalent input load to hardware control is as follows: if the system uses a programmable resistor network, the equivalent input load is quantized to the nearest resistance level; if the system uses a switching converter to achieve impedance transformation, the duty cycle, conduction time, or reference voltage threshold is calculated based on the equivalent input load; if the system uses an energy harvesting and management chip, the equivalent input load is mapped to the chip's maximum power point tracking target value or input adjustment threshold; for example, when the calculated equivalent input load on the thermoelectric side is 80 ohms, 82 ohms can be selected from discrete levels such as 64 ohms, 82 ohms, and 100 ohms for execution at 82 ohms.

[0085] In detail, the weight update cycle and load switching rhythm are as follows: the dual-source collection weight is updated once at the end of each energy cycle, and the equivalent input load is switched to the new target value at the beginning of the next cycle; if the difference between the old and new loads exceeds 30%, a two- or three-stage gradual switching is adopted to avoid the energy sampling voltage collapse caused by impedance abrupt change; for example, if it is calculated at the end of the previous cycle that the magnetic field side needs to be adjusted from 40 ohms to 100 ohms, it can be switched to 68 ohms first, stabilized for one sampling cycle, and then switched to 100 ohms.

[0086] Preferably, the instantaneous stored energy is obtained, and the instantaneous stored energy is multiplied by the effective observability of the gas to obtain a comprehensive evaluation index. Based on the comprehensive evaluation index, the earliest moment that satisfies the total energy requirement is found as the unique sampling moment, including: The total energy required is obtained by multiplying the preheating power consumption, sampling power consumption, processing power consumption, and transmission power consumption by their respective durations, summing the results, and then adding the product of standby power consumption and the effective arrival time. The comprehensive evaluation index is obtained by multiplying the instantaneous stored energy by the effective observability of the gas: The sets of local maxima are defined as the moments in the next period when the first derivative of the comprehensive evaluation index with respect to time is zero and the second derivative is less than zero. The net collected power is obtained by adding the product of the magnetic field power conversion efficiency, the magnetic field collection weight, and the instantaneous magnetic field power to the product of the thermoelectric power conversion efficiency, the thermoelectric collection weight, and the instantaneous thermoelectric power, and then subtracting the standby power consumption. From the set of local maxima, the earliest time in which the integral value of the net collected power from the valley point at the start of the next cycle to that point is greater than or equal to the total energy required is selected as the unique sampling time. in, The total energy required; These are preheating power consumption, sampling power consumption, processing power consumption, transmitting power consumption, and standby power consumption, respectively. These represent the durations corresponding to preheating power consumption, sampling power consumption, processing power consumption, and transmission power consumption, respectively. The comprehensive evaluation indicators are as described above; For the set of local maxima, For the next cycle; For the unique sampling time, These are the magnetic field power conversion efficiency and the thermoelectric power conversion efficiency, respectively. It is the integral variable.

[0087] Preheating power consumption is the power consumed by the gas sensing unit during heating or activation before entering a stable detection state. It is used to calculate the energy that must be pre-invested before completing a single detection. For micro-hotplate sensors, 50 milliwatts to 200 milliwatts are preferred. For larger sensors or sensors requiring higher operating temperatures, 200 milliwatts to 500 milliwatts are preferred.

[0088] The preheating duration is the time required for the gas sensing unit to transition from a cold or standby state to a stable sampling state. It is used in conjunction with preheating power consumption to calculate the energy consumption during the preheating phase. For low-heat-capacity micro-hotplate sensors, it is preferably 1 to 10 seconds. For thick-film heating or cavity preheating schemes, it is preferably 10 to 60 seconds.

[0089] Sampling power consumption is the power consumed by the gas sensing unit during the actual sampling and sensing response acquisition phases.

[0090] The sampling duration is the duration required for the gas sensing unit to maintain effective sampling and complete a stable response reading, preferably 0.5 seconds to 10 seconds.

[0091] Processing power consumption is the power consumed by the controller during the signal filtering, feature extraction, concentration inversion and decision calculation stages, preferably 5 milliwatts to 100 milliwatts.

[0092] The processing duration is the time required for the controller to complete one sample data processing and result generation, preferably 0.01 seconds to 2 seconds.

[0093] Transmit power consumption refers to the power consumed by the communication module when reporting monitoring results. For short-range, low-speed communication, it is preferably 30 milliwatts to 150 milliwatts. For long-range wireless transmission, it is preferably 150 milliwatts to 800 milliwatts.

[0094] The transmission duration is the time required for the communication module to complete the transmission of one monitoring result.

[0095] Standby power consumption is the power consumed by the system to maintain a monitoring standby state when it is not performing preheating, sampling, processing, and transmission tasks.

[0096] Total energy required is the sum of energy consumption in each working phase, plus the standby energy consumption within the effective arrival time. Here, the physical fluid delay time is used as the multiplier of the electrical standby time to reserve a conservative energy redundancy baseline covering the entire fluid transmission window during the evaluation phase. This baseline is used to determine whether the system still has the energy reserve to complete a monitoring under the most stringent conditions. In the specific execution judgment logic, to ensure that the system never loses power in extremely harsh power supply environments, the instantaneous stored energy remaining at the valley time is treated as a hard safety margin and frozen at the system level. It does not participate in the allocation and calculation of any available sampling energy. Therefore, it is required that the net collected energy integral newly accumulated from the valley time must independently and completely cover the total energy required.

[0097] The comprehensive evaluation index is a joint evaluation quantity obtained by multiplying the instantaneous stored energy by the effective observability of the gas. It is used to simultaneously consider two conditions: whether the current energy is sufficient and whether the current gas is worth sampling. The larger the index, the more likely that there is both sufficient energy and high diagnostic value at that moment.

[0098] The next cycle is the prediction time window from the current trough moment to the end of the next trough moment.

[0099] The set of local maxima is the set of all candidate moments in the next cycle when the comprehensive evaluation index reaches a local peak.

[0100] Magnetic field power conversion efficiency is the proportion of power that can be effectively utilized by the system after rectification, conversion, and management by the magnetic field energy harvesting unit, relative to the input magnetic field power. It is used to convert the instantaneous magnetic field power into usable net harvested power. For synchronous rectification and high-efficiency conversion schemes, it is preferably 0.75 to 0.95. For low-input, weak magnetic field schemes, it is preferably 0.5 to 0.75.

[0101] Thermoelectric power conversion efficiency is the proportion of power that can be effectively utilized by the system after the thermoelectric energy harvesting unit has undergone voltage boosting, rectification, and management, relative to the input thermoelectric power. For scenarios where oscillation is difficult due to small temperature differences, a value of 0.03 to 0.2 is preferred. For scenarios with relatively stable temperature differences and mature interface designs, a value of 0.2 to 0.6 is preferred.

[0102] Net harvested power is the real-time net power value obtained by adding the available power on the magnetic field side and the available power on the thermoelectric side, and then subtracting the standby power consumption.

[0103] The unique sampling time is the earliest time in the set of local maxima that satisfies both the comprehensive evaluation index being a local peak and the requirement that the net collected power integral from the trough of the current period is not less than the total energy required.

[0104] In detail, the prediction method for the comprehensive evaluation index in the next cycle is as follows: First, using the energy storage state at the end of the current cycle, the dual-source collection weight, the equivalent input load, and the power conversion efficiency, the instantaneous energy storage trajectory at each moment in the next cycle is predicted; then, using the currently updated effective arrival time and the basic observability model, the effective observability trajectory of the gas at each moment in the next cycle is predicted; finally, the two predicted trajectories are multiplied point by point to obtain the prediction sequence of the comprehensive evaluation index for the next cycle; for example, the predicted value can be calculated every second in the future to form a discrete comprehensive evaluation curve for searching the sampling time.

[0105] In detail, the discrete search method for the set of local maxima is as follows: First, perform a light smoothing on the prediction sequence of the comprehensive evaluation index for the next period, and then compare the size of a certain point with its adjacent points before and after it point by point; if the point is greater than the previous point and not less than the next point, or greater than the average value of the neighboring areas on both sides and exceeds the minimum peak threshold, it is recorded as a local maxima; when the interval between two adjacent local maxima is less than the minimum peak interval, only the one with the larger value is retained; for example, in a prediction sequence of 1 point per second, if the value corresponding to the 12th second is higher than that of the 11th second and not lower than that of the 13th second, then the 12th second can be included in the set of local maxima.

[0106] In detail, the predicted input sources for net harvested power are as follows: the predicted instantaneous magnetic field power is derived from the extrapolation of the final trend of the current cycle's instantaneous magnetic field power sequence or from matching historical templates with the same operating conditions; the predicted instantaneous thermoelectric power is derived from the joint extrapolation of the current cycle's instantaneous thermoelectric power sequence and the shell temperature difference change trend; the dual-source harvesting weight and power conversion efficiency use the latest updated values; the standby power consumption uses the equipment calibration value; for example, when the current cycle's magnetic field power is basically stable in the past 10 seconds, its average value in the last 10 seconds can be used as the short-term prediction input for the next cycle.

[0107] In detail, the rollback strategy when there is no time that meets the conditions in the next cycle is as follows: If the net collected power integral is always less than the total energy required throughout the next cycle, then no monitoring is performed in this cycle, and only standby is maintained and energy continues to accumulate; if there is a local maximum but the energy is insufficient, the search is postponed to the next cycle; if the conditions are not met for several consecutive cycles, the system can switch to a low-power emergency monitoring mode, which only performs simplified sampling and local caching, and does not perform high-power transmission; for example, in a continuous low load and low temperature difference environment, the node may not perform formal sampling for 2 to 5 consecutive cycles, and only retain standby and trend records.

[0108] Preferably, at the unique sampling time, the gas sensing unit is controlled to perform monitoring, and the concentration values ​​of the multi-component gas are inverted using a comprehensive sensitivity model that incorporates the magnetocaloric timing deviation, including: The sensor array response vector is obtained at the unique sampling time. ; The comprehensive sensitivity model is obtained by adding the product of the reference sensitivity matrix, the magnetocaloric timing deviation and the timing transmission correction matrix, the average shell temperature difference and the temperature correction matrix, and the estimated hydrogen concentration of the previous cycle and the hydrogen interference correction matrix. The sum of the observation variances of each channel is divided by the sum of the squared magnitude of the sensor array response vector and the numerical stability term to obtain the regularization parameter; a diagonal observation weight matrix is ​​constructed from the reciprocals of the observation variances of each channel. The transpose of the integrated sensitivity model, the observation weight matrix, and the integrated sensitivity model are multiplied consecutively, and the product of these multiplications with the regularization parameter and the identity matrix is ​​added to obtain the inverse matrix. This inverse matrix is ​​then multiplied sequentially by the transpose of the integrated sensitivity model, the observation weight matrix, and the sensor array response vector to invert and obtain the multi-component gas concentration values. The concentration values ​​of the multi-component gases for each type of gas are divided by the sum of the total concentration and the sum of the numerical stability terms to obtain the corresponding concentration weights. The diffusion coefficients of each target gas are multiplied by their corresponding concentration weights and then summed to update the equivalent diffusion coefficients for the next period. in, This is the response vector of the sensor array; This is the comprehensive sensitivity model. The reference sensitivity matrix, For timing transmission correction matrix, The average shell temperature difference, This is the temperature correction matrix. Estimate the hydrogen concentration for the previous cycle. This is the hydrogen interference correction matrix; For regularization parameters, The variance of observations for each channel, The square of the magnitude of the sensor array response vector; For the observation weight matrix; The concentration values ​​of the multi-component gas are... It is the identity matrix; Concentration weighting, and These are the multi-component gas concentration values ​​for various gases. The diffusion coefficients of each target gas are given. This is the equivalent diffusion coefficient for the next cycle.

[0109] The sensor array response vector is a vector composed of the response values ​​output simultaneously by multiple gas sensing channels at a single sampling time, arranged in a fixed order.

[0110] The baseline sensitivity matrix is ​​a matrix-based calibration result of the response of each sensing channel to changes in the concentration of each target gas under baseline operating conditions. It is a set of matrix element values ​​obtained by calibration using 5 to 8 standard concentration gradients, with 3 to 5 repeated experiments for each gradient.

[0111] The timing transmission correction matrix is ​​a matrix parameter used to describe the influence of magnetocaloric timing deviation on the equivalent sensitivity of each sensing channel. Preferably, it is a set of matrix element values ​​obtained based on gas arrival delay tests from 0 to 60 seconds and fitted under multiple fault conditions. It should at least cover typical timing patterns such as early gas arrival, delayed arrival, and partial overlap.

[0112] The temperature correction matrix is ​​a matrix parameter used to describe the impact of average housing temperature difference changes on the sensitivity of each sensing channel. Preferably, it is a set of matrix element values ​​obtained from multi-temperature point calibration fitting between 20°C and 80°C. It should at least cover the temperature drift range under common equipment operating environments and local overheating conditions.

[0113] The hydrogen concentration estimate for the previous cycle is the hydrogen concentration estimate obtained from the previous round of multi-component concentration inversion, used to characterize the continued impact of hydrogen on the cross-interference of the sensor array in the current cycle.

[0114] The hydrogen interference correction matrix is ​​a matrix parameter used to describe the impact of hydrogen concentration changes on the sensitivity and cross-interference level of each sensing channel, and is used to suppress the misleading effect of hydrogen on the inversion results of other target gases. Preferably, it is a set of matrix element values ​​obtained by calibration fitting based on hydrogen interference ranging from 0 μL / L to 500 μL / L. It should at least cover common hydrogen interference levels under conditions ranging from slight discharge to significant decomposition.

[0115] The integrated sensitivity model is a dynamic sensitivity model obtained by superimposing the baseline sensitivity matrix with the time-series transmission correction term, temperature correction term, and hydrogen interference correction term. It is used to more realistically reflect the response relationship of the sensing array to each target gas under the current time-series state, thermal state, and interference state.

[0116] The observation variance of each channel is a statistic used to characterize the observation noise and fluctuation of each sensing channel.

[0117] The regularization parameter is a constraint strength parameter calculated based on the sum of the observation variances of each channel, the magnitude of the sensor array response vector, and the numerical stability term.

[0118] The observation weight matrix is ​​a diagonal matrix composed of the inverses of the observation variances of each channel. It is used to give greater weight to channels with less noise in the inversion, while reducing the weight of channels with more noise, thereby improving the robustness of multi-component gas concentration inversion.

[0119] The identity matrix is ​​a standard matrix with one diagonal element and zero all other elements. It is used to provide a reference matrix of the same order as the integrated sensitivity model during the regularization inversion process. Before performing the regularization inversion operation, the sensor array response vector and the integrated sensitivity model must be divided by their respective set full-scale reference thresholds to convert them into pure numerical matrices. This ensures that the regularization parameters and all matrix addition operations in the inversion process are performed in a dimensionless space, eliminating the constraint bias caused by different concentration units.

[0120] The multi-component gas concentration values ​​are estimated results of the concentration of each target gas obtained by integrating the sensitivity model, the observation weight matrix, and the regularized inversion structure, and are used to describe the composition of the current fault gas components.

[0121] The total concentration is the sum of the concentration values ​​of various gases and their components. It is used to normalize the concentration values ​​of each target gas and form a concentration weight that reflects the proportion of each component.

[0122] Concentration weight is a proportional value obtained by dividing the concentration values ​​of a target gas by the sum of its concentrations and the numerical stability term. It is used to represent the relative proportion of the target gas in the current gas mixture and to update the equivalent diffusion coefficient.

[0123] The diffusion coefficients of each target gas are parameters representing the molecular diffusion capacity of various target gases under given temperature and pressure conditions. They are used to obtain the equivalent diffusion coefficients for the next period by weighting them according to concentration. Preferably, the set of diffusion coefficients is obtained by converting them based on a standard gas property database and a working temperature and pressure correction formula. For specific scenarios where trace fault characteristic gases inside the equipment diffuse freely in the vast majority of background insulating gases, the concentration-weighted arithmetic mean of the diffusion coefficients of each individual trace component in the background gas is used as an engineering simplification approximation to calculate the equivalent diffusion coefficient of the trace gas mixture.

[0124] The gas category index is a discrete number used to distinguish different target gas categories.

[0125] The channel index is a discrete number used to distinguish different observation channels in the sensor array.

[0126] In detail, the channel composition and preprocessing method of the sensor array response vector are as follows: the sensor array includes at least 2 to 8 sensor channels with different response characteristics to the target decomposition gas; before entering the vector, the output of each channel is subjected to zero-point correction, initial temperature drift compensation, outlier removal and amplitude normalization in sequence; if a channel is saturated or distorted in the current period, the channel is marked as invalid and given a minimal weight in the observation weight matrix; for example, a 4-channel array can be composed of a hydrogen sensitive channel, a sulfide sensitive channel, an oxidizing gas sensitive channel and a reference channel, and the sensor array response vector is formed in a fixed order.

[0127] In detail, the calibration method for the reference sensitivity matrix and the timing transmission correction matrix, temperature correction matrix, and hydrogen interference correction matrix is ​​as follows: First, the target gases are graded and calibrated at different concentrations under a standard gas platform to obtain the reference sensitivity matrix; then, the timing transmission correction matrix is ​​calibrated by manually setting different arrival delay conditions; then, the temperature correction matrix is ​​calibrated under different shell temperature differences; finally, the hydrogen interference correction matrix is ​​calibrated in a mixed gas containing different hydrogen background concentrations; each matrix element is preferably obtained through least squares fitting or piecewise linear fitting; for example, the reference matrix can be established by fixing the temperature and hydrogen background first, and then the time delay, temperature difference, and hydrogen interference can be introduced item by item to calibrate the corresponding correction matrices in layers.

[0128] In detail, the estimation window and update method for the observation variance of each channel are as follows: For each channel, the response residuals or repeated sampling results of the most recent 5 to 50 valid periods are saved, and the observation variance of each channel is updated using the sliding window variance calculation method; when the periodic samples are insufficient, the calibration stage variance is used as the initial value; if a channel experiences sudden noise, exponential smoothing update can be enabled to avoid drastic changes in variance; for example, for a channel, the residual sequence of the most recent 20 periods is saved, and the oldest data can be deleted, the latest data can be added, and the variance can be recalculated when a new period arrives.

[0129] In detail, the constraint handling method for excessively large or small regularization parameters is as follows: set upper and lower limits for the calculated regularization parameters; the lower limit is preferably set to 0.0001, and the upper limit is preferably set to 10; when the original calculated value is less than the lower limit, it is processed according to the lower limit to prevent the inversion from being too sensitive; when the original calculated value is greater than the upper limit, it is processed according to the upper limit to prevent excessive smoothing from compressing the concentration estimate; for example, when the magnitude of the response vector of a certain period is abnormally large, causing the regularization parameter to be close to 0, it is still used in the inversion operation according to the lower limit value.

[0130] In detail, the physical constraint method for negative values ​​of multi-component gas concentrations is as follows: If the inversion result of a target gas is negative, first check whether there is drift or model mismatch in the channel group; in the output stage, the negative value is truncated to 0, and the absolute value before truncation is included in the model residual statistics; if the negative value occurs repeatedly more than a preset number of times, the comprehensive sensitivity model is recalibrated or the weight of the relevant matrix elements of the gas is reduced; for example, if the sulfur dioxide concentration obtained by inversion in a certain period is negative 3, the final output is 0, and a negative value truncation event is recorded.

[0131] In detail, the handling method for the sum of concentrations approaching zero during concentration weight calculation is as follows: when the sum of concentrations is lower than the minimum concentration threshold, the equivalent diffusion coefficient is no longer updated using the current concentration weight, but the equivalent diffusion coefficient of the previous effective period is directly used; if the system is in the first period and there are no historical values, then the empirical initial value is used; for example, if the concentration of all gases in a certain period is close to zero, it indicates that the observation information for this period is insufficient, and at this time it is not appropriate to use the noise-dominated proportion to update the equivalent diffusion coefficient.

[0132] Preferably, the cumulative value of the gas effective observability is calculated to update the correction coefficient, and the gas effective observability for the next period is corrected using the updated correction coefficient, including: The overall gas strength is obtained by calculating the modulus of the concentration values ​​of the multi-component gas: The cumulative value of the effective observability of the gas is obtained by integrating the effective observability of the gas: The regression vector is constructed by multiplying the accumulated value, the absolute value of the magnetocaloric timing deviation corresponding to the previous cycle, and the average shell temperature difference by the accumulated value. The updated correction coefficients and the current covariance matrix are derived recursively from the regression vector, the overall gas intensity, and the previous covariance matrix: Multiply the updated correction coefficient components by the absolute values ​​of the magnetocaloric time series deviations and the average shell temperature difference corresponding to the first and next cycles, respectively, and sum them. Then multiply by the base observability to obtain the corrected gas effective observability for the next cycle: in, The overall strength of the gas; This is the cumulative value of the effective observability of the gas; For the regression vector, The average shell temperature difference; For recursive gain, For the previous covariance matrix, This is the previous correction factor. The updated correction coefficients, This is the current covariance matrix; For the corrected gas effective observability of the next period, Based on observability, , , The updated correction coefficient components, This represents the magnetocaloric timing deviation for the next cycle. This represents the average shell temperature difference for the next cycle.

[0133] The overall gas intensity is a comprehensive intensity index obtained by taking the modulus of the concentration values ​​of multiple components of gas. It is used to characterize the strength of the overall response of all target gases at the current monitoring time.

[0134] The cumulative value of gas effective observability is the cumulative amount obtained by integrating the gas effective observability over time over a complete period, and is used to characterize the total amount of overall observable resources in the current period.

[0135] The regression vector is a vector composed of the cumulative value of the effective observability of the gas and the product of the cumulative value with the absolute value of the magnetocaloric time series deviation and the average shell temperature difference of the previous cycle.

[0136] The previous covariance matrix is ​​the uncertainty matrix saved by the recursive algorithm at the previous update time.

[0137] The recursive gain is an online adjustment gain calculated based on the previous covariance matrix and the current regression vector. It is used to determine the strength of the influence of the new observation data on the correction coefficient update. The larger the recursive gain, the stronger the influence of the new data on the correction result.

[0138] The previous correction coefficient is a vector of correction coefficients that has been updated and saved in the previous period, and is used as the initial reference for the recursive update in the current period.

[0139] The updated correction coefficients are a new correction coefficient vector obtained by recursively using the current regression vector, the overall gas intensity, and the previous covariance matrix. It is used to reflect the system's latest correction capability after absorbing the current period's observation information.

[0140] The current covariance matrix is ​​the new uncertainty matrix obtained after this recursive update. It is used in the next period as the previous covariance matrix to participate in the recursion, reflecting the latest parameter uncertainty state.

[0141] The updated correction coefficient components are the components in the updated correction coefficient that correspond to the constant term, the magnetocaloric timing deviation term, and the average shell temperature difference term, respectively, and are used to make partial corrections to the basic observability for the next cycle.

[0142] The baseline observability is the base value of the gas effective observability for the next cycle calculated by the aforementioned effective arrival time model before the latest correction factor is added.

[0143] The corrected gas observability for the next cycle is the result of multiplying the base observability by the correction factor composed of the updated correction coefficient components. Before substituting into the regression vector for recursive calculation, the cumulative value, the absolute value of the deviation, and the average shell temperature difference must be divided by their respective standard reference values ​​to ensure that the resulting regression vector is a pure numerical vector. This ensures that the correction coefficient components obtained through recursive calculation and the resulting correction factor are dimensionless proportions, guaranteeing that the physical meaning and units of the gas observability before and after correction remain consistent.

[0144] In detail, the initialization method of the previous covariance matrix and the previous correction coefficients is as follows: before the start of the first cycle, the previous correction coefficients are initialized as a vector of 1, 0, 0, so that the basic observability works in an uncorrected state; the previous covariance matrix is ​​initialized as a large diagonal matrix to indicate that the system has a large uncertainty about the initial correction coefficients; for example, the diagonal of the previous covariance matrix can be initialized to 100 or 1000, while the off-diagonal elements are set to 0.

[0145] In detail, the numerical divergence suppression method in the recursive update is as follows: a maximum allowable value is set for the recursive gain, symmetry and positive definiteness corrections are periodically performed on the current covariance matrix, and the updated correction coefficients are limited after each update; when the observed residuals increase abnormally, the update of this cycle can be skipped, and only the previous correction coefficients and the previous covariance matrix are retained; for example, if the overall gas intensity changes abnormally due to a sensor failure in a certain cycle, the system can directly freeze the recursive update to prevent the correction coefficients from diverging all at once.

[0146] In detail, the upper and lower limits of the gas effective observability for the next cycle after correction are constrained as follows: upper and lower limits are set for both the correction factor and the gas effective observability for the next cycle after final correction; the correction factor is preferentially limited to between 0.2 and 5; the lower limit of the final observability is 0, and the upper limit can be set to 2 to 5 times the historical maximum value of the basic observability; for example, if the correction factor is too large and the correction factor is calculated to be 8, it is still processed according to the upper limit of 5 to avoid the subsequent sampling time being dominated by extreme values.

[0147] In detail, the rules for inheriting correction coefficients for the first period and abnormal periods are as follows: the first period uses initial correction coefficients and does not perform historical inheritance; starting from the second period, if the observation data of this period is valid, it is updated normally and the updated correction coefficients are inherited to the next period; if this period is determined to be an abnormal period, the previous correction coefficients and the previous covariance matrix are directly inherited and no update is performed; for example, when the temperature sensor is disconnected, the gas concentration is all zero, or the matrix inversion fails, it can be regarded as an abnormal period.

[0148] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. An adaptive environmental energy harvesting method based on SF6 online monitoring, applied to a dual-source system with co-loop and disparate locations, comprising a magnetic field energy harvesting unit, a thermoelectric energy harvesting unit, a gas sensing unit, and an energy storage supercapacitor, characterized in that... include: The energy cycle is divided according to the energy storage natural valley value of the supercapacitor and dual-source energy is collected simultaneously. Extract the magnetocaloric timing deviation of the dual-source energy within the energy cycle; The reference arrival time is obtained, and the effective arrival time is obtained by correcting the reference arrival time using the magnetocaloric timing deviation, thereby constructing the effective observability of the gas; The dual-source collection weight is generated based on the magnetocaloric timing deviation, and the dual-source collection weight is converted into an equivalent input load. The instantaneous stored energy is obtained, and the instantaneous stored energy is multiplied by the effective observability of the gas to obtain a comprehensive evaluation index. Based on the comprehensive evaluation index, the earliest moment that meets the total energy requirement is found as the unique sampling moment. At the unique sampling time, the gas sensing unit is controlled to perform monitoring, and the concentration values ​​of multi-component gases are inverted using a comprehensive sensitivity model that incorporates the magnetocaloric timing deviation. The cumulative value of the gas effective observability is calculated to update the correction coefficient, and the gas effective observability for the next period is corrected using the updated correction coefficient.

2. The adaptive environmental energy harvesting method based on SF6 online monitoring according to claim 1, characterized in that, The energy cycle is divided according to the natural valley value of the energy storage supercapacitor, and dual-source energy is collected simultaneously, including: The voltage and current at the magnetic field energy harvesting terminal are simultaneously collected and multiplied to obtain the instantaneous power of the magnetic field; the voltage and current at the thermoelectric energy harvesting terminal are simultaneously collected and multiplied to obtain the instantaneous power of the thermoelectric field. The instantaneous stored energy is calculated by collecting the terminal voltage of the energy storage supercapacitor and combining it with the capacitance value of the energy storage supercapacitor. The moment when the first derivative of the instantaneous stored energy with respect to time is equal to zero and the second derivative with respect to time is greater than zero is taken as the valley moment; The time interval between two adjacent valley values ​​is defined as the energy cycle; Within the energy cycle, the instantaneous power of the magnetic field and the instantaneous power of the thermoelectric field are integrated over time to obtain the dual-source energy, which includes magnetic field harvesting energy and thermoelectric harvesting energy.

3. The adaptive environmental energy harvesting method based on SF6 online monitoring according to claim 2, characterized in that, Extracting the magnetocaloric timing deviation of the dual-source energy within the energy cycle includes: Calculate the mean and standard deviation of the instantaneous magnetic field power within the energy cycle, as well as the mean and standard deviation of the instantaneous thermoelectric power. The instantaneous magnetic field power is standardized using the mean magnetic field power, the standard deviation of the magnetic field power, and the numerical stability term to obtain the normalized magnetic field power; The normalized thermoelectric power is obtained by standardizing the instantaneous thermoelectric power using the mean thermoelectric power, the standard deviation of thermoelectric power, and the numerical stability term. The product of the normalized magnetic field power and the time derivative of the normalized thermoelectric power is subtracted from the product of the normalized thermoelectric power and the time derivative of the normalized magnetic field power. The product is then integrated over time within the energy cycle, and half of the result is taken to obtain the directed area of ​​the trajectory. The arc length of the trajectory is obtained by taking the square root of the sum of the square of the time derivative of the normalized magnetic field power and the square of the time derivative of the normalized thermoelectric power, and then integrating over time within the energy period. The magnetocaloric timing deviation is obtained by dividing twice the directional area of ​​the trajectory by the square of the arc length of the trajectory and the sum of the numerical stability term.

4. The adaptive environmental energy harvesting method based on SF6 online monitoring according to claim 3, characterized in that, Obtain the reference arrival time, correct the reference arrival time using the magnetocaloric timing deviation to obtain the effective arrival time, and then construct the effective observability of the gas, including: The average shell temperature difference is obtained by taking the difference between the shell temperature of the target air chamber and the ambient temperature; The equivalent velocity of natural convection is calculated by combining gravitational acceleration, gas thermal expansion coefficient, gas kinematic viscosity, gas thermal diffusivity, the average shell temperature difference, and the equivalent convection characteristic length. The equivalent diffusion rate is obtained by dividing the equivalent diffusion coefficient of the period by the equivalent transmission path length. The reference arrival time is obtained by dividing the equivalent transmission path length by the sum of the natural convection equivalent velocity, the diffusion equivalent velocity, and the numerical stability term; The absolute value integral of the difference between the normalized magnetic field power time derivative and the normalized thermoelectric power time derivative is calculated and divided by the sum of the trajectory arc length and the numerical stability term to obtain the asynchronous intensity correction term. The effective arrival time is obtained by multiplying the reference arrival time by the sum of the products of the asynchronous intensity correction term and the absolute value of the magnetocaloric timing deviation. An exponential decay function is constructed using the effective arrival time, and the exponential decay function is integrated with the instantaneous thermoelectric power to obtain the effective observability of the gas.

5. The adaptive environmental energy harvesting method based on SF6 online monitoring according to claim 4, characterized in that, Generate dual-source collection weights based on the magnetocaloric timing deviation, and convert the dual-source collection weights into an equivalent input load, including: Divide the magnetic field harvesting energy by the sum of the magnetic field harvesting energy, the thermoelectric harvesting energy, and the numerical stability term, and then divide by the sum of the products of the asynchronous intensity correction term and the absolute value of the magnetocaloric timing deviation to obtain the initial magnetic field weight. Divide the thermoelectric harvested energy by the sum of its value, the magnetic field harvested energy, and the numerical stability term, and then multiply by the sum of the products of the asynchronous intensity correction term and the absolute value of the magnetocaloric timing deviation to obtain the initial thermoelectric weight. Divide the initial magnetic field weight and the initial thermoelectric weight by the sum of the two respectively to obtain the magnetic field collection weight and thermoelectric collection weight that constitute the dual-source collection weight; The square of the open-circuit voltage of the corresponding energy source in the previous cycle is divided by four times the product of the maximum available power of the corresponding energy source in the previous cycle and the corresponding dual-source collection weight, and the sum of the numerical stability terms, to convert it into the equivalent input load of the energy source.

6. The adaptive environmental energy harvesting method based on SF6 online monitoring according to claim 5, characterized in that, The instantaneous stored energy is obtained, and the instantaneous stored energy is multiplied by the effective observability of the gas to obtain a comprehensive evaluation index. Based on the comprehensive evaluation index, the earliest moment that meets the total energy requirement is selected as the unique sampling moment, including: The total energy required is obtained by multiplying the preheating power consumption, sampling power consumption, processing power consumption and transmission power consumption by their respective durations, summing them, and then adding the product of standby power consumption and the effective arrival time. The comprehensive evaluation index is obtained by multiplying the instantaneous energy storage by the effective observability of the gas; The points in the next cycle where the first derivative of the comprehensive evaluation index with respect to time is equal to zero and the second derivative is less than zero constitute the set of local maxima. The net collected power is obtained by adding the product of the magnetic field power conversion efficiency, the magnetic field collection weight, and the instantaneous magnetic field power to the product of the thermoelectric power conversion efficiency, the thermoelectric collection weight, and the instantaneous thermoelectric power, and then subtracting the standby power consumption. In the set of local maxima, the earliest time in which the integral value of the net collected power from the valley time at the start of the next cycle to that time is greater than or equal to the total required energy is selected as the unique sampling time.

7. The adaptive environmental energy harvesting method based on SF6 online monitoring according to claim 6, characterized in that, At the unique sampling time, the gas sensing unit is controlled to perform monitoring, and the concentration values ​​of multiple components are inverted using a comprehensive sensitivity model that incorporates the magnetocaloric timing deviation, including: The sensor array response vector is acquired at the unique sampling time. The comprehensive sensitivity model is obtained by adding the product of the reference sensitivity matrix, the magnetocaloric timing deviation and the timing transmission correction matrix, the average shell temperature difference and the temperature correction matrix, and the hydrogen estimated concentration of the previous cycle and the hydrogen interference correction matrix. The sum of the observation variances of each channel is divided by the sum of the square of the magnitude of the response vector of the sensor array and the numerical stability term to obtain the regularization parameter; a diagonal observation weight matrix is ​​constructed from the reciprocals of the observation variances of each channel. The transpose of the integrated sensitivity model, the observation weight matrix, and the integrated sensitivity model are multiplied together, and the product of the transpose, the observation weight matrix, and the identity matrix is ​​added to obtain the inverse matrix. The inverse matrix is ​​then multiplied by the transpose of the integrated sensitivity model, the observation weight matrix, and the sensor array response vector in sequence to obtain the multi-component gas concentration value. The concentration values ​​of the multi-component gases of each type of gas are divided by the sum of the total concentration and the sum of the numerical stability terms to obtain the corresponding concentration weights. The diffusion coefficients of each target gas are multiplied by their corresponding concentration weights and then summed to update the equivalent diffusion coefficients for the next period.

8. The adaptive environmental energy harvesting method based on SF6 online monitoring according to claim 7, characterized in that, The cumulative value of the gas effective observability is calculated to update the correction coefficients. The updated correction coefficients are then used to correct and derive the gas effective observability for the next period, including: The overall gas strength is obtained by calculating the modulus of the concentration values ​​of the multi-component gas. The cumulative value of the effective observability of the gas is obtained by integrating the effective observability of the gas. The product of the cumulative value, the absolute value of the magnetocaloric time series deviation corresponding to the previous cycle, and the average shell temperature difference with the cumulative value constitutes a regression vector; The updated correction coefficients and the current covariance matrix are derived recursively from the regression vector, the overall gas intensity, and the previous covariance matrix. The updated correction coefficient components are multiplied by the absolute values ​​of the magnetocaloric time series deviations and the average shell temperature difference corresponding to the first and next cycles, respectively, and summed. Then, multiplied by the basic observability, the corrected gas effective observability for the next cycle is obtained.

Images