An online monitoring system and detection method for hydrogen leakage in thermal power plant generators

Abstract

This invention relates to the field of power equipment safety monitoring, and discloses an online monitoring system and detection method for hydrogen leakage from a thermal power plant generator. The system includes a distributed solid-state palladium alloy sensor array, a multi-channel impedance harmonic signal conditioning unit, an edge computing processing terminal, and a host computer monitoring platform. The multi-channel impedance harmonic signal conditioning unit applies a composite AC excitation signal to the distributed solid-state palladium alloy sensor array. The edge computing processing terminal receives digital voltage sampling data, calculates the complex impedance spectrum characteristic parameters and nonlinear harmonic distortion coefficients, performs pure hydrogen locking discrimination and surface contamination correction, constructs a concentration gradient vector field, and calculates the spatial coordinates of the leakage source. By analyzing the nonlinear harmonic response to eliminate environmental interference, using complex impedance spectrum separation technology to eliminate the influence of surface contamination, and combining the inverse flowline tracing algorithm to invert the leakage location, the system solves the problems of high false alarm rate and inability to locate the leak in generator hydrogen monitoring.

Description

Technical Field

[0001] This invention relates to the field of power equipment safety monitoring, specifically to an online monitoring system and detection method for hydrogen leakage from generators in thermal power plants. Background Technology

[0002] Hydrogen-cooled generators utilize the excellent thermal conductivity of hydrogen for cooling. However, the flammable and explosive nature of hydrogen makes leak monitoring a critical aspect of ensuring the safe operation of power plants. Current generator hydrogen leak monitoring primarily relies on thermal conductivity and catalytic combustion gas sensors. These sensors operate based on changes in resistance and heat generated when a gas-sensitive material comes into contact with the gas being measured. The internal operating environment of a generator is complex, including electromagnetic interference, temperature fluctuations, and mechanical vibrations.

[0003] In existing technologies, traditional sensors face a serious problem of cross-sensitivity. The cooling gas inside the generator contains trace amounts of oil mist, water vapor, and volatiles from insulating materials. These non-target gas components adsorb onto the surface of the gas-sensitive material, causing non-specific changes in conductivity. This makes it difficult for the monitoring system to separate the hydrogen signal from the ambient background noise, easily leading to false alarms and missed alarms.

[0004] During long-term operation, suspended particles and oil sludge inside the generator gradually deposit on the sensor surface, forming a contamination layer. This contamination layer hinders the contact between gas molecules and the sensitive membrane, altering the sensor's fundamental impedance characteristics and causing decreased sensitivity and zero-point drift. Existing detection methods lack mechanisms for assessing and correcting the sensor's surface condition, making it impossible to distinguish online whether the measurement drift stems from changes in hydrogen concentration or sensor aging and contamination. This leads to a decrease in the reliability of measurement data over time, increasing the workload of maintenance and calibration.

[0005] Existing monitoring systems typically employ single-point or sparsely distributed multi-point configurations, only capable of acquiring gas concentration values ​​at specific locations. Due to the complex internal airflow structure and rapid air circulation speed of the generator, leaked hydrogen quickly diffuses and fills the stator frame space. Single concentration threshold determination methods cannot reflect the spatial distribution of the concentration field and cannot trace the specific location of the leak source. Maintenance personnel must manually inspect sealing tiles, coolers, and pipe joints one by one after shutting down and replacing the gas, resulting in long troubleshooting cycles and impacting unit utilization. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an online monitoring system and detection method for hydrogen leakage in thermal power plant generators. It solves the problems of high false alarm rates caused by interference from non-target gases, measurement drift caused by oil mist contamination of sensors, and difficulty in accurately locating the spatial position of the leakage source in existing generator hydrogen monitoring technologies.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an online monitoring system and detection method for hydrogen leakage from a thermal power plant generator.

[0008] This invention provides an online monitoring system and detection method for hydrogen leakage in thermal power plant generators. The invention primarily addresses the technical problems of low sensitivity, high false alarm rate, difficulty in accurate location, and sensor drift due to contamination in the monitoring of hydrogen leakage inside generators.

[0009] The system provided by this invention includes a distributed solid-state palladium alloy sensor array, a multi-channel impedance harmonic signal conditioning unit, an edge computing processing terminal, and a host computer monitoring platform. The distributed solid-state palladium alloy sensor array consists of multiple solid-state palladium alloy sensor nodes physically installed in the monitoring area inside the generator. The solid-state palladium alloy sensor nodes are manufactured using microelectromechanical systems (MEMS) technology, and their structure includes a thermally insulating suspended film layer deposited on a silicon substrate, a micro-hot plate heating electrode, an electrically insulating isolation layer, interdigitated electrode pairs, and a gas-sensitive palladium alloy nanofilm. The gas-sensitive palladium alloy nanofilm is made of palladium-silver alloy material and has a thickness at the mesoscale. It utilizes the nonlinear electron scattering effect generated when hydrogen enters the crystal lattice to convert changes in hydrogen concentration into changes in electrical impedance characteristics.

[0010] In terms of signal excitation and acquisition, this invention utilizes a multi-channel impedance harmonic signal conditioning unit to apply a composite AC excitation signal to each solid-state palladium alloy sensor node. The composite AC excitation signal consists of a fundamental frequency large-amplitude carrier signal superimposed with a wideband scanning small signal. The current amplitude of the fundamental frequency large-amplitude carrier signal is set to be greater than the critical threshold for exciting the nonlinear polarization effect of the palladium alloy thin film, used to induce a nonlinear response; the wideband scanning small signal contains multiple scanning frequency points distributed on a logarithmic scale, used to obtain a wideband impedance spectrum. A high-bandwidth transimpedance amplifier is connected to the sensor node using a four-wire Kelvin connection mode, and a synchronous sampling analog-to-digital converter acquires the response signal containing high-order harmonic components.

[0011] In terms of signal processing and recognition, this invention utilizes an edge computing terminal to perform a Fast Fourier Transform (FFT) to acquire frequency domain data, and calculates complex impedance spectrum characteristic parameters and nonlinear harmonic distortion coefficients based on the frequency domain data. The system employs pure hydrogen-locked discrimination logic to extract the spectral amplitude magnitudes of the fundamental frequency large-amplitude carrier signal, second harmonic component, and third harmonic component from the frequency domain data, and calculates a normalized nonlinear distortion response index. When the nonlinear distortion response index meets preset conditions, the system confirms that the detected signal originates from a lattice phase transition caused by hydrogen, thereby eliminating false alarms from other interfering gases.

[0012] This invention addresses contamination resistance and self-healing by performing surface contamination correction calculations based on complex impedance spectroscopy. The system separates the complex impedance vector into real and imaginary characteristic components, and calculates the additional equivalent resistance caused by the surface contamination layer using the imaginary characteristic component at a high-frequency reference point. The system then uses this additional equivalent resistance to perform a subtraction correction operation on the real characteristic component at a low-frequency detection point, obtaining the bulk resistance of the gas-sensitive thin film after removing the influence of surface contamination. When the additional equivalent resistance exceeds the cleaning trigger threshold, the intelligent self-healing control unit controls the sensor surface temperature to rise to the target cleaning temperature, utilizing thermal desorption to remove surface contaminants.

[0013] This invention introduces an inverse streamline tracing algorithm for leak source localization. The system uses an inverse distance weighted interpolation algorithm to map spatially discrete concentration data into a continuous three-dimensional concentration scalar field, and performs gradient operations on this scalar field to generate a concentration gradient vector field. The system then performs inverse streamline tracing within this concentration gradient vector field, using sensor node locations exceeding a preset safety threshold as seed points, and iteratively searches along the local gradient direction until a convergence criterion is met. The convergence endpoint of the search path is then calculated as the spatial coordinates of the leak source.

[0014] This invention also combines real-time hydrogen concentration values ​​and concentration change rates to construct a comprehensive hazard index, and performs graded interlocking control based on the range of the comprehensive hazard index, including triggering the exhaust fan, shutting off the hydrogen supply, starting nitrogen purging, and sending a shutdown interlocking command, thus realizing closed-loop management from monitoring to protection.

[0015] Compared with the prior art, the present invention has the following beneficial technical effects:

[0016] 1. This invention applies a composite AC excitation signal through a multi-channel impedance harmonic signal conditioning unit, and uses a fundamental frequency large amplitude carrier signal to excite the nonlinear electron scattering effect of palladium alloy nanofilm. The nonlinear harmonic distortion coefficient is calculated based on the second and third harmonic components and pure hydrogen lock-in discrimination is performed. The unique lattice phase transition characteristics of hydrogen are used to distinguish interfering gases, thus avoiding the influence of the generator's internal environment temperature and humidity and non-target gases on the hydrogen concentration detection results.

[0017] 2. This invention analyzes the characteristic parameters of the complex impedance spectrum through edge computing processing terminal, separates the high-frequency imaginary part characteristic components of the complex impedance vector to quantify the additional equivalent resistance generated by the surface contamination layer, obtains the body resistance of the gas-sensitive thin film by subtraction correction operation, and combines the heating desorption mechanism of the intelligent self-healing control unit to maintain the consistency of sensor measurement data in the environment where there is oil mist and dust inside the generator, thereby reducing zero drift caused by sensor surface contamination.

[0018] 3. This invention collects spatial discrete concentration data through a distributed solid-state palladium alloy sensor array, constructs a continuous three-dimensional concentration scalar field using an edge computing processing terminal and generates a concentration gradient vector field, and uses an inverse flow line tracing algorithm to iteratively search for the spatial coordinates of the leakage source along the local gradient direction. Based on the distribution trend of the concentration field, the location of the leakage point is inverted, thus realizing the spatial positioning of the hydrogen leakage source inside the generator. Attached Figure Description

[0019] Figure 1 This is a structural block diagram of the online hydrogen leakage monitoring system for thermal power plant generators according to the present invention;

[0020] Figure 2 This is a schematic diagram of the signal processing and pure hydrogen locking process of the present invention;

[0021] Figure 3 This is a flowchart of the leakage source localization algorithm of the present invention;

[0022] Figure 4 This is a schematic diagram of the system cleaning and interlocking control logic of the present invention.

[0023] Among them, 100, Online monitoring system for hydrogen leakage in generators of thermal power plants; 10, Solid-state palladium alloy sensor array; 10a, Stator core air duct monitoring subset; 10b, End sealing area monitoring subset; 10c, Cooler inlet and outlet monitoring subset; 11, Solid-state palladium alloy sensor node; 111, Silicon substrate; 111a, Back cavity; 112, Thermally insulating suspended film layer; 113, Micro-hot plate heating electrode; 114, Electrically insulating isolation layer; 115, Interdigitated electrode pair; 115a, Connection signal injection electrode; 115b, Signal extraction electrode; 116, Gas-sensitive palladium alloy nanofilm; 20, Multi-channel impedance harmonic signal conditioning unit; 210, Digital waveform synthesis module; 211, Phase accumulator unit; 212, Waveform memory unit; 213, High-speed digital-to-analog converter unit; 220 1. Voltage-to-current conversion drive module; 230. Differential voltage extraction module; 240. Analog signal preprocessing module; 241. Programmable gain amplifier unit; 242. Anti-aliasing filter unit; 21. Direct digital frequency synthesis excitation source; 22. High bandwidth transimpedance amplifier; 23. Synchronous sampling analog-to-digital converter; 30. Edge computing processing terminal; 31. Digital signal processor; 31a. Spectrum analysis module; 31b. Harmonic feature calculation module; 31c. Logic discrimination module; 32. Vector feature extraction module; 33. Pollution correction calculation module; 40. Host computer monitoring platform; 41. Vector field construction module; 42. Anticurrent line tracing module; 50. Intelligent self-healing control unit; 51. Pulse heating drive module; 52. Logic arbitration module; 53. Relay output interface. Detailed Implementation

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

[0025] like Figure 1 As shown, the present invention provides an online monitoring system 100 for hydrogen leakage from a generator in a thermal power plant. The system 100 includes: a distributed solid-state palladium alloy sensor array 10, a multi-channel impedance harmonic signal conditioning unit 20, an edge computing processing terminal 30, and a host computer monitoring platform 40.

[0026] The distributed solid-state palladium alloy sensor array 10 consists of multiple solid-state palladium alloy sensor nodes 11. The solid-state palladium alloy sensor nodes 11 are physically installed in the generator stator cooling airflow, end cover sealing area, and hydrogen cooler inlet and outlet areas. The solid-state palladium alloy sensor nodes 11 are configured to convert changes in hydrogen concentration into changes in electrical impedance characteristics. The solid-state palladium alloy sensor nodes 11 internally contain a nanoscale solid-state palladium alloy thin film deposited on a microelectromechanical system (MEMS) hotplate.

[0027] The multi-channel impedance harmonic signal conditioning unit 20 is electrically connected to the distributed solid-state palladium alloy sensor array 10. The multi-channel impedance harmonic signal conditioning unit 20 includes a direct digital frequency synthesis excitation source 21, a high-bandwidth transimpedance amplifier 22, and a synchronous sampling analog-to-digital converter 23.

[0028] The direct digital frequency synthesis excitation source 21 is configured to apply a composite AC excitation signal to each solid-state palladium alloy sensor node 11. The composite AC excitation signal consists of a fundamental frequency large-amplitude carrier signal superimposed with a wideband scanning small signal. The composite AC excitation signal generated by the direct digital frequency synthesis excitation source 21... The following functional relationship is satisfied:

[0029] ;

[0030] In the formula, The time for generating the direct digital frequency synthesis excitation source 21 The intensity of the composite AC excitation current at any given moment; The current amplitude of the fundamental frequency carrier is set to be greater than the critical threshold for exciting the nonlinear polarization effect of the palladium alloy thin film; This refers to the frequency value of the baseband carrier. This represents the total number of scan frequency points. To scan the current amplitude of small signals; For the frequency scan sequence, the first Frequency values ​​at each frequency point; For the first The initial phase angle of each scan frequency component.

[0031] A high-bandwidth transimpedance amplifier 22 is connected to the output of the solid-state palladium alloy sensor node 11. The high-bandwidth transimpedance amplifier 22 is configured to receive the response current signal of the solid-state palladium alloy sensor node 11 under the action of a composite AC excitation signal, and convert the response current signal into a voltage response signal.

[0032] The synchronous sampling analog-to-digital converter 23 is connected to the output of the high-bandwidth transimpedance amplifier 22. The synchronous sampling analog-to-digital converter 23 is configured to discretize and sample the voltage response signal to generate digital voltage sample data. The sampling frequency of the synchronous sampling analog-to-digital converter 23 satisfies the Nyquist sampling theorem and can cover the highest frequency component in the composite AC excitation signal as well as the third harmonic component of the fundamental carrier.

[0033] The edge computing processing terminal 30 is connected to the multi-channel impedance harmonic signal conditioning unit 20 via a high-speed data bus. The edge computing processing terminal 30 integrates a digital signal processor 31. The digital signal processor 31 is configured to receive digital voltage sampling data and perform fast Fourier transform operations.

[0034] The edge computing processing terminal 30 is configured to calculate the complex impedance spectrum characteristic parameters and the nonlinear harmonic distortion coefficient based on frequency domain data. The edge computing processing terminal 30 stores a pure hydrogen locking discrimination logic algorithm and a gradient vector localization algorithm. The edge computing processing terminal 30 is configured to invert the hydrogen concentration value based on the calculated complex impedance spectrum characteristic parameters and the nonlinear harmonic distortion coefficient, and to calculate the spatial coordinates of the leakage source.

[0035] The host computer monitoring platform 40 is connected to the edge computing processing terminal 30 via an industrial Ethernet communication network. The host computer monitoring platform 40 is configured to receive and display hydrogen concentration data, spatial coordinates of the leak source, and system alarm status. The host computer monitoring platform 40 is equipped with a distributed control system communication interface 41. The distributed control system communication interface 41 is configured to send alarm signals and shutdown interlock commands to the power plant's distributed control system.

[0036] The system 100 also includes a power management module 50. The power management module 50 provides stable DC operating voltages for the distributed solid-state palladium alloy sensor array 10, the multi-channel impedance harmonic signal conditioning unit 20, and the edge computing processing terminal 30.

[0037] In the online hydrogen leakage monitoring system 100 for thermal power plant generators provided by the present invention, the spatial distribution structure of the distributed solid-state palladium alloy sensor array 10 is configured to match the hydrodynamic characteristics of the generator in order to support subsequent gradient vector positioning calculations.

[0038] The distributed solid-state palladium alloy sensor array 10 contains a number of Solid-state palladium alloy sensor node 11. To accurately describe the spatial location of each solid-state palladium alloy sensor node 11, the system establishes a coordinate system with the generator rotor's central axis as... The radial direction of the shaft and generator stator is axis (corresponding to the Cartesian coordinate system) ), the circumferential tangential direction of the generator is The axes are converted from a three-dimensional cylindrical coordinate system to a unified three-dimensional rectangular coordinate system. The position vector of each solid-state palladium alloy sensor node 11 in three-dimensional space is... Defined as:

[0039] ;

[0040] In the formula, For the first The spatial position vector of a solid palladium alloy sensor node 11; For the first The coordinates of a solid palladium alloy sensor node 11 in the horizontal direction perpendicular to the ground plane; For the first The coordinates of a solid palladium alloy sensor node 11 in the vertical direction perpendicular to the ground plane; For the first The coordinate values ​​of a solid palladium alloy sensor node 11 along the generator axis; This represents the total number of nodes 11 in the solid-state palladium alloy sensor.

[0041] The distributed solid-state palladium alloy sensor array 10 is divided into three monitoring subsets according to the internal airflow structure of the generator: stator core airflow monitoring subset 10a, end sealing area monitoring subset 10b, and cooler inlet and outlet monitoring subset 10c.

[0042] The stator core air duct monitoring subset 10a is distributed at the outlet of the radial ventilation groove of the generator stator frame. The solid palladium alloy sensor node 11 in the stator core air duct monitoring subset 10a is along the generator axial direction (…). The nodes (in the axial direction) are arranged at equal intervals. The axial spacing between two adjacent solid palladium alloy sensor nodes 11 The value is set to one-tenth of the effective length of the stator core. The stator core air duct monitoring subset 10a is configured to detect abnormal hydrogen concentrations caused by stator bar insulation damage or cooling water pipe leaks.

[0043] The end-sealing area monitoring subset 10b is distributed within the end cover spaces on both sides of the generator, at the turbine and exciter ends. Within each end cover space, solid-state palladium alloy sensor nodes 11 are uniformly arranged circumferentially, with a central angle interval of 90 degrees between adjacent nodes. The end-sealing area monitoring subset 10b primarily monitors the hydrogen-side sealing tiles and the outgoing bushing area. The end-sealing area monitoring subset 10b is configured to detect hydrogen leakage diffusion caused by sealing oil system failure.

[0044] The cooler inlet and outlet monitoring subset 10c is installed at the inlet of the hot air zone and the outlet of the cold air zone of the generator's built-in hydrogen cooler. The cooler inlet and outlet monitoring subset 10c is arranged in pairs.

[0045] All solid-state palladium alloy sensor nodes 11 are fixed to the low-velocity region of the generator housing inner wall using dedicated explosion-proof brackets. The spatial coordinates of each solid-state palladium alloy sensor node 11 are as follows. During the system installation and debugging phase, the data is precisely measured and stored in the configuration database of the edge computing processing terminal 30.

[0046] When hydrogen leaks inside the generator, the hydrogen plume generated at the leak point forms a specific concentration distribution field under the action of the fan head inside the generator. The distributed solid-state palladium alloy sensor array 10 discretizes the continuous concentration field into a spatial concentration dataset through the aforementioned spatially discretized sampling points. :

[0047] ;

[0048] In the formula, This is a dataset of the overall concentration distribution. For the position vector The solid palladium alloy sensor node 11 at this location detects the temperature and pressure corrected hydrogen volume concentration value. This spatial concentration dataset... The data is transmitted as input to the edge computing processing terminal 30 for subsequent calculation of the concentration gradient vector and inversion of the leakage source location.

[0049] The multi-channel impedance harmonic signal conditioning unit 20 provided by the present invention is configured to achieve high-precision complex impedance measurement and nonlinear harmonic capture. Its internal circuit structure is interconnected with the solid palladium alloy sensor node 11 using a four-wire Kelvin connection mode.

[0050] The multi-channel impedance harmonic signal conditioning unit 20 includes, in its physical architecture: a digital waveform synthesis module 210, a voltage-to-current conversion drive module 220, a differential voltage extraction module 230, and an analog signal preprocessing module 240.

[0051] The digital waveform synthesis module 210 is connected to the edge computing processing terminal 30. The digital waveform synthesis module 210 includes a phase accumulator unit 211, a waveform memory unit 212, and a high-speed digital-to-analog converter unit 213.

[0052] A voltage-to-current conversion drive module 220 is connected between the high-speed digital-to-analog converter unit 213 and the current injection terminal of the solid-state palladium alloy sensor node 11. The voltage-to-current conversion drive module 220 employs a modified Howland current pump circuit topology and is configured to convert analog voltage control signals... Converted into a constant composite AC excitation signal The transfer function of the voltage-to-current conversion drive module 220 satisfies the following relationship:

[0053] ;

[0054] In the formula, The composite AC excitation current injected into the sensor; The transconductance gain coefficient of the voltage-to-current conversion drive module 220 is expressed in Siemens (S). This is the control voltage signal output by the high-speed digital-to-analog converter unit 213. This module ensures [the signal's integrity] through a negative feedback mechanism. The waveform does not change with the load impedance of node 11 of the solid palladium alloy sensor.

[0055] The differential voltage extraction module 230 is connected to the voltage sensing terminal of the solid-state palladium alloy sensor node 11. The differential voltage extraction module 230 includes a high input impedance instrumentation amplifier unit 231. The differential voltage extraction module 230 outputs the raw response voltage signal. With sensor impedance The relationship is:

[0056] ;

[0057] In the formula, This represents the total voltage response across the sensor. This represents the linear impedance magnitude of solid-state palladium alloy sensor node 11 at the current frequency. The excitation angular frequency; The impedance phase angle; The high-order harmonic voltage component is generated by the nonlinear effect of the palladium-hydrogen system.

[0058] The analog signal preprocessing module 240 is connected between the differential voltage extraction module 230 and the synchronous sampling analog-to-digital converter 23. The analog signal preprocessing module 240 includes a programmable gain amplifier unit 241 and an anti-aliasing filter unit 242.

[0059] Programmable gain amplifier unit 241 is configured to automatically adjust signal chain gain based on the current signal amplitude. To match the dynamic range of the synchronous sampling analog-to-digital converter 23.

[0060] Anti-aliasing filter unit 242 is configured to filter out noise components above the Nyquist frequency. The cutoff frequency of anti-aliasing filter unit 242 is... Set as baseband carrier frequency More than ten times, to ensure the third harmonic component 3 The signal passes without attenuation. The final conditioned voltage signal is input to the synchronous sampling analog-to-digital converter 23. satisfy:

[0061] ;

[0062] In the formula, The analog input voltage of the analog-to-digital converter; This is the linear gain of the programmable gain amplifier unit 241; For the anti-aliasing filter unit 242 at frequency The amplitude-frequency response coefficient at that point.

[0063] The synchronous sampling analog-to-digital converter 23 receives trigger signals from the same clock source and... Digital quantization is performed to generate digital voltage sampling data, which is then transmitted to the edge computing processing terminal 30 for spectrum analysis.

[0064] The solid-state palladium alloy sensor node 11 provided by this invention is manufactured based on microelectromechanical systems (MEMS) technology, and its internal core sensing unit adopts a multilayer thin film stacked structure. Structurally, the solid-state palladium alloy sensor node 11 includes: a silicon substrate 111, a thermally insulating suspended film layer 112, a micro-hot plate heating electrode 113, an electrically insulating isolation layer 114, interdigitated electrode pairs 115, and a gas-sensitive palladium alloy nanofilm 116.

[0065] A thermally insulating suspended film layer 112 is deposited on the upper surface of the silicon substrate 111 and covers the opening of the back cavity 111a. The thermally insulating suspended film layer 112 is composed of a composite dielectric film formed by alternating deposition of silicon nitride and silicon dioxide. The thermally insulating suspended film layer 112 is configured to support the functional layer above while blocking heat conduction to the silicon substrate 111, ensuring the thermal field concentration in the heated area.

[0066] The hot plate heating electrode 113 is located on the upper surface of the thermally insulating suspended film layer 112. The hot plate heating electrode 113 is deposited from platinum metal using photolithography and sputtering processes, exhibiting a serpentine, meandering wiring structure. Both ends of the hot plate heating electrode 113 are connected to the power management module 50 via wire bonding. The hot plate heating electrode 113 is configured to generate Joule heating upon receiving a pulsed current, rapidly heating and cleaning or modulating the temperature of the gas-sensitive palladium alloy nanofilm 116 above it. The hot plate heating electrode 113 integrates temperature sensing functionality, and its temperature coefficient of resistance... Used for real-time feedback of the temperature value of the heating zone :

[0067] ;

[0068] In the formula, The current absolute temperature of the heating electrode 113 of the micro-hot plate; For temperature The resistance value of the micro-hot plate obtained by measurement below; To be at the reference temperature The nominal resistance value below; is the linear temperature coefficient of resistance of platinum material.

[0069] Interdigitated electrode pairs 115 are deposited on the upper surface of the electrically insulating layer 114. Each interdigitated electrode pair 115 consists of a pair of interlaced gold electrode comb-like arrays. The interdigitated electrode pair 115 includes a signal injection electrode 115a and a signal extraction electrode 115b, both of which are connected to the multi-channel impedance harmonic signal conditioning unit 20. The geometric parameters of the interdigitated electrode pair 115 determine the sensor's reference capacitance and effective detection area.

[0070] A gas-sensitive palladium alloy nanofilm 116 was uniformly deposited and coated over the interdigitated electrode pair 115 and its gaps using a magnetron sputtering process. The gas-sensitive palladium alloy nanofilm 116 is made of a palladium-silver alloy, with the atomic percentage of silver set at 23% to 25% to suppress the volume expansion and fragmentation of pure palladium material during the hydrogen absorption phase transition. The thickness of the gas-sensitive palladium alloy nanofilm 116 is [not specified in the original text]. The thickness is controlled between 10 and 100 nanometers, falling within the mesoscopic scale. This thickness range allows the gas-sensitive palladium alloy nanofilm 116 to exhibit a significant nonlinear electron scattering effect when a lattice phase transition occurs due to hydrogen adsorption. The gas-sensitive palladium alloy nanofilm 116 serves as a variable impedance element connecting the signal injection electrode 115a and the signal extraction electrode 115b, and its complex impedance... It includes real and imaginary components determined by grain boundaries and surface effects.

[0071] The solid-state palladium alloy sensor node 11 provided by this invention possesses specific thermoelectric coupling characteristics, which are jointly determined by the thermal conduction mechanism and electrical response mechanism between the micro-hot plate heating electrode 113 and the gas-sensitive palladium alloy nanofilm 116. The solid-state palladium alloy sensor node 11 is configured to switch between a constant-temperature detection mode and a pulse cleaning mode. Its thermal response speed and the uniformity of the temperature field directly determine the system's detection signal-to-noise ratio and anti-contamination capability.

[0072] Thermal coupling is achieved between the micro-hot plate heating electrode 113 and the gas-sensitive palladium alloy nanofilm 116 via an electrically insulating layer 114. In the thermodynamic model, the solid palladium alloy sensor node 11 is equivalent to a device with thermal resistance... With heat capacity This constitutes a first-order thermal inertial system. When the power management module 50 injects heating power into the micro-hotplate heating electrode 113... At that time, the steady-state temperature rise on the surface of the gas-sensitive palladium alloy nanofilm 116 The following heat balance equation is satisfied:

[0073] ;

[0074] In the formula, Joule heat power generated by heating electrode 113 of the micro-hot plate; The driving current for injecting the micro-hot plate heating electrode 113; The resistance value of the heating electrode 113 of the micro-hot plate at the operating temperature; The total thermal conductivity from the heated area to the environment is given by the value of the thermal resistance. The reciprocal of; The operating temperature of the 116 gas-sensitive palladium alloy nanofilm surface; The ambient background temperature inside the generator compartment; This represents the power component lost through thermal radiation.

[0075] The thermoelectric coupling characteristics of solid palladium alloy sensor node 11 also include transient thermal response characteristics. This characteristic is determined by the system's thermal time constant. Definition. Thermal time constant. The time required for the sensor to heat from ambient temperature to the pulse cleaning temperature is determined by the following formula:

[0076] ;

[0077] In the formula, The thermal time constant; Equivalent thermal resistance; This is the equivalent heat capacity of the sensor; The effective density of the multilayer film stacked structure; The effective volume of the heated region; To achieve an effective specific heat capacity, in this embodiment, the back cavity 111a is designed with thermal insulation. The value is set to less than 20 milliseconds, which enables the gas-sensitive palladium alloy nanofilm 116 to respond to fast thermally modulated signals with frequencies higher than 10 Hz.

[0078] The conductivity of the gas-sensitive palladium alloy nanofilm 116 is modulated by both temperature and hydrogen concentration. To isolate the temperature effect in signal processing, the system utilizes the inherent temperature coefficient of resistance of the gas-sensitive palladium alloy nanofilm 116 for reference definition. The reference resistance of the gas-sensitive palladium alloy nanofilm 116 in a hydrogen-free environment is... With temperature The changes follow the following linear relationship:

[0079] ;

[0080] In the formula, For temperature Pure palladium thin film resistor; To be at the calibration temperature The initial resistance below; Here is the temperature coefficient of resistance of the palladium-silver alloy thin film. This thermoelectric coupling equation is stored in the edge computing processing terminal 30 and used to subtract the resistance drift component caused by temperature fluctuations when calculating the resistance change caused by hydrogen.

[0081] In the online monitoring method for hydrogen leakage in thermal power plant generators provided by this invention, the direct digital frequency synthesis excitation source 21 is configured to construct a specific composite AC excitation signal. The composite AC excitation signal It is designed to simultaneously include a large-signal carrier component for exciting nonlinear effects and a small-signal scanning component for probing the linear impedance spectrum.

[0082] The direct digital frequency synthesis excitation source 21 generates discrete time-sampled data through a digital synthesis algorithm, and outputs it to the solid-state palladium alloy sensor node 11 via a digital-to-analog converter and a voltage-to-current converter circuit. Composite AC excitation signal. The time-domain mathematical model is defined as:

[0083] ;

[0084] In the formula, In time The total excitation current value injected into the solid palladium alloy sensor node 11 at all times; To represent the current amplitude of the baseband carrier signal; To represent a fixed frequency of the baseband carrier signal; It is a time variable; This represents the total number of superimposed scan frequency points; The amplitude of a single-frequency component of a wideband scanning small signal; For the first Frequency values ​​of each scan frequency point; For the first The initial phase angle of each scan frequency component.

[0085] In the above mathematical model, the current amplitude of the fundamental frequency carrier signal... It is set as a high-energy-level excitation parameter. The values ​​are configured such that the voltage drop across the gas-sensitive palladium alloy nanofilm 116 can cross the linear region of its current-voltage characteristic curve. Specifically, The generated electric field strength is sufficient to drive the nonlinear migration and rearrangement of adsorbed hydrogen atoms within the palladium lattice, thereby inducing the harmonic distortion required for subsequent steps. The fixed frequency of the fundamental carrier signal... It is usually set in the low-frequency band, with a value ranging from 100 Hz to 1000 Hz, as the reference frequency for harmonic analysis.

[0086] Amplitude of single-frequency component of wideband scanning small signal It is set as a low-energy perturbation parameter. and The proportional relationship satisfies This proportional limitation ensures that the scanning signal does not cause significant nonlinear effects, thus guaranteeing the applicability of the superposition principle in small-signal analysis. The total number of scanning frequency points. Determined based on the frequency resolution requirements of the impedance spectrum. Frequency value of each scan frequency point In frequency range The inner distribution is based on a logarithmic scale, where 10 Hz It is 100 kHz.

[0087] Direct digital frequency synthesis excitation source 21 generates composite AC excitation signal At that time, a phase-truncation accumulation technique is used. This is applied to each frequency component. (include and all Phase accumulator unit 211 is based on the system sampling clock. Calculate phase increment :

[0088] ;

[0089] In the formula, For frequency The phase step value of the component within each clock cycle; This refers to the bit width of the phase accumulator; This is the system's master clock frequency. The digital waveform synthesis module 210 superimposes the instantaneous amplitude values ​​of all components in the digital domain to generate a composite waveform data stream. This data stream is then converted into an analog current signal to drive the solid-state palladium alloy sensor array 10.

[0090] By applying this composite AC excitation signal The system can simultaneously acquire the nonlinear distortion response of the solid palladium alloy sensor node 11 to a large signal carrier and the complex impedance spectrum response to a wideband small signal within a single measurement cycle, providing an independent data source for subsequent pure hydrogen locking and concentration inversion.

[0091] This invention utilizes the unique nonlinear current-voltage characteristics of the solid palladium alloy sensor node 11 under hydrogen absorption conditions to physically distinguish hydrogen from non-hydrogen interference. The gas-sensitive palladium alloy nanofilm 116 exhibits non-ohmic conductivity that dynamically varies with hydrogen concentration when in contact with hydrogen.

[0092] A large amplitude fundamental frequency carrier current applied by the direct digital frequency synthesis excitation source 21 Under the influence of hydrogen absorption, the gas-sensitive palladium alloy nanofilm 116 no longer obeys linear Ohm's law. Its instantaneous response voltage... Described as relating to excitation current Nonlinear polynomial functions:

[0093] ;

[0094] In the formula, This represents the total voltage response across node 11 of the solid palladium alloy sensor. The input is the composite AC excitation current; This is the linear reference impedance of the sensor at zero hydrogen concentration; The concentration of hydrogen gas entering the crystal lattice; The second-order nonlinear distortion coefficient is positively correlated with hydrogen concentration, and its physical source is hydrogen-induced lattice asymmetry distortion. The third nonlinear distortion coefficient is positively correlated with hydrogen concentration, and its physical origin is the nonlinear scattering mechanism of electron-phonon hydrogen atoms under high current density.

[0095] The above formula shows that a phase transition occurs if and only if hydrogen gas enters the palladium lattice and triggers the phase transition. and Only then will it be a non-zero value. After substituting the fundamental frequency component into the above equation and expanding it, the response voltage is... The frequency of the lieutenant general is The second harmonic component and frequency are The third harmonic component. The amplitude of these higher harmonic components directly maps to the hydrogen atom occupancy rate within the crystal lattice.

[0096] In contrast, interfering substances such as oil mist, water vapor, or dust present in the generator nacelle adhere only to the physical surface of the gas-sensitive palladium alloy nanofilm 116 and cannot penetrate the intercrystalline lattice. These surface contamination layers are electrically equivalent to a parallel or series network of linear resistors or capacitors. The response voltage for surface interfering substances... Following the principle of linear superposition:

[0097] ;

[0098] In the formula, This refers to the voltage drop caused by surface interference. This is the equivalent linear impedance of the surface contamination layer. Because... It is a constant (at a single frequency). spectral components and They are completely identical and will not generate new higher harmonic frequency components.

[0099] Based on the aforementioned physical mechanism, this system can achieve exclusive identification of hydrogen by detecting the presence of high-order harmonic components in the response signal. Regardless of the thickness of the oil film or ash deposits on the surface, as long as they do not alter the palladium lattice structure, no harmonic signals will be generated, thus fundamentally eliminating false alarms caused by the complex internal environment of the generator affecting hydrogen leakage monitoring.

[0100] See attached document Figure 2 The edge computing processing terminal 30 provided by this invention integrates a digital signal processor 31, which is configured to execute a pure hydrogen lock-in algorithm based on frequency domain analysis. In order to achieve accurate signal separation and logical judgment, the digital signal processor 31 is logically divided into a spectrum analysis module 31a, a harmonic feature calculation module 31b, and a logical discrimination module 31c.

[0101] The spectrum analysis module 31a is configured to receive discrete-time digital voltage sampling data output by the synchronous sampling analog-to-digital converter 23. To suppress spectral leakage, the spectral analysis module 31a first... Apply the Blackman-Harris window function for processing, then execute. Point Fast Fourier Transform.

[0102] The harmonic characteristic calculation module 31b is connected to the spectrum analysis module 31a. The harmonic characteristic calculation module 31b is configured to sample data from the frequency domain. The system extracts amplitude information at specific frequency points based on the sampling rate. With the number of transformation points Positioning base frequency Corresponding index Second harmonic Corresponding index and third harmonic Corresponding index The harmonic characteristic calculation module 31b calculates the nonlinear distortion response index based on the extracted amplitude. Its calculation formula is defined as:

[0103] ;

[0104] In the formula, is the normalized nonlinear distortion response exponent, and is a dimensionless value; This represents the magnitude of the spectral amplitude of the fundamental frequency carrier signal. The magnitude of the spectral amplitude of the second harmonic component; This represents the magnitude of the spectral amplitude of the third harmonic component. This formula eliminates the influence of excitation signal intensity fluctuations on the discrimination logic by normalizing the higher harmonic energy to the fundamental frequency energy.

[0105] The logic discrimination module 31c is connected to the harmonic characteristic calculation module 31b. The logic discrimination module 31c has a preset pure hydrogen determination threshold. The threshold This setting is based on the system noise floor level under clean air conditions, and is typically three times the root mean square value of the system noise floor. The logic discrimination module 31c is configured to execute binary comparison logic to generate a pure hydrogen lock flag. :

[0106] ;

[0107] In the formula, This is a pure hydrogen lock flag; a value of 1 indicates that the lattice phase transition response is indeed caused by hydrogen; a value of 0 indicates that the signal source is linear interference (such as oil or water vapor) or system noise.

[0108] If and only if When the value is 1, the edge computing processing terminal 30 will then initiate the subsequent concentration inversion algorithm. If... If the value is 0, the system marks the current measurement data as non-hydrogen interference, does not perform concentration calculations, and maintains a zero output. The logic discrimination module 31c is also equipped with a continuous confirmation mechanism, requiring confirmation from three consecutive measurement cycles. A valid hydrogen leak alarm signal is triggered only when all values ​​are 1, thus filtering out occasional electromagnetic pulse interference. Through the above logic, the system achieves exclusive locking of hydrogen signals in complex electromagnetic and chemical environments.

[0109] The edge computing processing terminal 30 provided by the present invention integrates a vector feature extraction module 32, which is configured to analyze the complex impedance spectrum characteristics of the solid palladium alloy sensor node 11 based on the response data under the composite AC excitation signal.

[0110] The vector feature extraction module 32 establishes a data connection with the synchronous sampling analog-to-digital converter 23 and the direct digital frequency synthesis excitation source 21. During each measurement cycle, the vector feature extraction module 32 synchronously receives time-domain digital voltage sampling data. With the corresponding time-domain digital current excitation sampling data To obtain impedance information at specific frequency points, the vector feature extraction module 32 performs discrete Fourier transforms on the two sampled datasets, mapping the time-domain signal to the frequency-domain complex plane. This is done for a preset frequency set. Each frequency point Calculate its corresponding complex impedance value. .

[0111] Obtain the complex impedance vector Subsequently, the vector feature extraction module 32 performs an orthogonal decomposition operation, separating the complex impedance vector into real and imaginary characteristic components. The real characteristic components correspond to the sum of the bulk resistance and contact resistance of the gas-sensitive palladium alloy nanofilm 116, while the imaginary characteristic components correspond to the reactance generated by the grain boundary capacitance and the surface double-layer capacitance. The decomposition process follows the following mathematical relationship:

[0112] ;

[0113]

[0114] ;

[0115] In the formula, In frequency The real part of the impedance (resistance component) below. In frequency The imaginary part of the impedance (reactance component) below. This represents the magnitude of the complex impedance. This is the phase difference between voltage and current, i.e., the impedance phase angle.

[0116] The vector feature extraction module 32, through the above calculations, expands the single physical detection signal into a multi-dimensional frequency domain feature space. For signals containing... The system constructs a scanning process at discrete frequency points with a length of [missing information]. frequency domain feature vector :

[0117] ;

[0118] In the formula, The input feature vector used for subsequent concentration inversion; This is a matrix transpose operation.

[0119] This frequency domain feature separation step establishes the data foundation for the dual-channel inversion. Among them, The component is mainly controlled by the electron scattering effect caused by hydrogen atoms entering the palladium lattice, which manifests as a change in resistance related to hydrogen concentration; The components are mainly controlled by the adsorption state on the film surface and the integrity of the microstructure, exhibiting frequency-dependent capacitive changes. By separating the real and imaginary parts, the system can distinguish between the bulk impedance drift caused by changes in hydrogen concentration and the interfacial impedance drift caused by environmental humidity or surface contamination, thus providing orthogonalized characteristic data for subsequent accurate concentration calculations.

[0120] The edge computing processing terminal 30 provided by the present invention integrates a pollution correction calculation module 33. This module is configured to compensate the real resistance value using the imaginary part characteristic component of the complex impedance spectrum in order to eliminate the interference of non-hydrogen deposits on the sensor surface on concentration inversion.

[0121] The pollution correction calculation module 33 establishes a data connection with the vector feature extraction module 32 and receives feature components containing real parts. With imaginary characteristic components The frequency domain eigenvector. In the actual operating environment of the generator compartment, oil mist and dust will gradually deposit on the surface of the solid palladium alloy sensor node 11, forming a contamination film with dielectric properties. Electrically, this contamination film manifests as a parasitic complex impedance network connected in parallel with the gas-sensitive palladium alloy nanofilm 116.

[0122] The contamination correction calculation module 33 performs corrections based on the following physical model: the gas-sensitive palladium alloy nanofilm 116 mainly exhibits pure resistive characteristics at low frequencies, with its resistance value varying with hydrogen concentration; while the surface contamination film mainly exhibits capacitive reactive characteristics, with its reactance value varying with the thickness of the contamination layer and the dielectric constant. Therefore, the imaginary impedance component in the high-frequency band directly reflects the degree of surface contamination and is independent of hydrogen concentration.

[0123] The pollution correction calculation module 33 first selects high-frequency reference points from the eigenvectors. The imaginary impedance at (set to 100 kHz) Calculate the additional equivalent resistance caused by surface contamination. The calculation process uses a second-order polynomial regression model:

[0124] ;

[0125] In the formula, This is a spurious resistance increment introduced by the surface contamination layer; In the reference frequency The absolute value of the imaginary part of the impedance measured below; This is a linear correction coefficient, and its unit is dimensionless. This is the second-order nonlinearity correction coefficient, and its unit is Siemens (S). This is the reference deviation constant. The above coefficients... , and These are fixed parameters obtained by pre-calibrating the sensor in a standard oil mist environment and stored in the non-volatile memory of the edge computing processing terminal 30.

[0126] The additional equivalent resistance is calculated. Then, the pollution correction calculation module 33 selects low-frequency detection points. Real impedance at (set to 1 kHz) Using the original measured value, a subtraction correction operation is performed to obtain the true palladium film resistance. :

[0127] ;

[0128] In the formula, The corrected bulk resistance of the gas-sensitive thin film, which is determined solely by hydrogen concentration and temperature; In order to detect frequency The following is the total real impedance value output by the vector feature extraction module 32.

[0129] The pollution correction calculation module 33 will calculate the... The output is then sent to the subsequent concentration mapping unit. Through this surface contamination correction model, the system orthogonally decouples the low-frequency real part from the high-frequency imaginary part, ensuring that even when the sensor surface is covered with a non-conductive oil film or dust layer, it can still accurately obtain the true resistance value reflecting the palladium lattice state, thus avoiding false alarms caused by environmental contamination.

[0130] See attached document Figure 3 The generator hydrogen leakage monitoring system provided by the present invention includes a leakage source location unit 40, which integrates a vector field construction module 41. The vector field construction module 41 is configured to map the discrete concentration data collected by the solid palladium alloy sensor array 10 into a continuous three-dimensional spatial concentration scalar field, and further convert it into a gradient vector field indicating the direction of the leakage source.

[0131] The vector field construction module 41 first receives data from... Real-time monitoring data of each solid palladium alloy sensor node 11. The three-dimensional spatial coordinate position of each solid palladium alloy sensor node 11 within the generator nacelle is pre-calibrated and fixed. Define the first... The spatial coordinates of the sensors are At the current moment The detected hydrogen concentration value is The spatial discrete dataset obtained by the vector field construction module 41 Represented as:

[0132] ;

[0133] In the formula, It is a spatially discrete dataset containing both location and concentration information; This represents the total number of sensors; For the first The position vectors of each sensor node in the Cartesian coordinate system; For the first The actual hydrogen concentration value output by each sensor node after surface contamination correction.

[0134] To infer the concentration distribution in unsampled areas between sensor nodes, the vector field construction module 41 uses an inverse distance weighted interpolation algorithm to construct a continuous three-dimensional concentration scalar field. For the monitoring area any point in space Its estimated concentration The calculation formula is as follows:

[0135] ;

[0136] ;

[0137] In the formula, For a point in space The interpolated concentration scalar at the location; For calculation points With the sensor nodes The Euclidean distance between them; In this embodiment, the distance weight decay power is used as the exponent. The value is set to 2 to reflect the physical property that the concentration decreases inversely with the square of the distance during gas diffusion; symbol This indicates the number of valid sensor nodes participating in the interpolation calculation.

[0138] In obtaining a continuous concentration scalar field Subsequently, the vector field construction module 41 performs gradient calculations on the scalar field to generate a concentration gradient vector field describing the gas diffusion trend. Each vector in the gradient vector field points in the direction of the fastest increase in concentration, i.e., towards the potential source of leakage. Gradient vector The calculation formula is defined as follows:

[0139]

[0140] In the formula, For a point in space The three-dimensional concentration gradient vector at the location For Hamiltonian operators; , , To represent the concentration scalar field in axis, axis, Partial derivatives along the axial direction; respectively along axis, axis, The unit direction vector along the axis.

[0141] The vector field construction module 41, through the above calculations, transforms the originally discrete single-point monitoring data into vector field data with direction indication function. This gradient vector field... It contains not only information on the concentration magnitude but also information on the direction of the concentration change rate. When a hydrogen leak occurs, the concentration field inside the generator compartment will form an extreme region centered on the leak point. The gradient vector streamlines output by the vector field construction module 41 will converge towards this extreme region, thus providing a topological basis for subsequent leak source coordinate inversion.

[0142] The leak source location unit 40 provided by this invention integrates a reverse flow tracing module 42. This reverse flow tracing module 42 is configured to use a numerical iteration method to perform a reverse path search from a low concentration region to a high concentration region in the concentration gradient vector field generated by the vector field construction module 41, thereby locking the spatial coordinates of the hydrogen leak source.

[0143] The reverse flow tracking module 42 establishes a data communication connection with the vector field construction module 41 to receive three-dimensional concentration gradient vector field data. The reverse flow tracing module 42 first performs the seed point selection step. The system traverses the spatial discrete dataset. Select concentration measurement value Exceeding the preset safety threshold The former The locations of the solid palladium alloy sensor nodes are used as the initial set of iterative seed points. .

[0144] For each initial seed point The reverse flow tracing module 42 performs discretized gradient climbing iterative calculations. In the... In the next iteration step, the current position Next position The update follows a fixed-step displacement rule along the local gradient direction. The position update formula is defined as:

[0145] ;

[0146] In the formula, For the first The tracking path is in the first The three-dimensional spatial coordinate vector after the next iteration; For the first The tracking path is in the first The three-dimensional spatial coordinate vector at the next iteration; The iteration step size is the distance traveled in the direction of movement during each iteration; To be at the current location The gradient vector at that point indicates the direction of path evolution; For vectors The modulus (or norm / size) is used to normalize the vector.

[0147] The reverse flow tracing module 42 has an iteration termination condition. For each tracing path, iteration stops when either the position convergence criterion or the gradient magnitude criterion is met. The position convergence criterion is defined as the position change between two consecutive iterations being less than a preset position tolerance. :

[0148] ;

[0149] In the formula, Let be the displacement magnitude of the position vector; This represents the minimum convergence distance allowed by the system. When the iteration terminates, the coordinates obtained from the last calculation are recorded as the [number]th [point]. The convergence endpoint of the streamlines To eliminate potential local extremum traps or noise biases that may exist in single-path search, the reverse flow tracing module 42 employs a multi-point weighted centroid clustering algorithm to calculate the final leakage source coordinates. The algorithm applies to all convergence endpoints. Spatial clustering is calculated using the following formula:

[0150] ;

[0151] ;

[0152] In the formula, The calculated three-dimensional coordinates of the hydrogen leak source inside the generator compartment are estimated. The total number of valid tracing streamlines involved in the calculation; For the first Coordinates of the converging endpoints of the streamlines; For the first The weighting coefficients of each convergence endpoint; For the convergence endpoint The scalar value of hydrogen concentration obtained by interpolation.

[0153] Using the aforementioned reverse streamline tracing algorithm, the system can automatically draw three-dimensional streamline clusters pointing towards high-concentration regions. The reverse streamline tracing module 42 outputs... The location of the physical leak point is directly mapped into the digital twin model of the generator set and marked as a bright spot on the central monitoring screen, achieving automated and precise positioning without human intervention.

[0154] See attached document Figure 4 The system provided by the present invention includes an intelligent self-healing control unit 50, which integrates a pulse heating drive module 51. The pulse heating drive module 51 is configured to apply a high-current pulse of a specific waveform to the micro-hot plate assembly 113 in the solid palladium alloy sensor node 11 to remove non-hydrogen contaminants adsorbed on the surface of the gas-sensitive palladium alloy nanofilm 116 through the Joule heating effect.

[0155] The intelligent self-healing control unit 50 establishes a communication connection with the pollution correction calculation module 33. When the pollution correction calculation module 33 calculates the surface additional equivalent resistance... Exceeding the preset cleaning trigger threshold At this time, the intelligent self-healing control unit 50 sends an interrupt command to the direct digital frequency synthesis excitation source 21 to suspend the normal measurement cycle and activate the pulse heating drive module 51 to enter the self-healing cleaning mode.

[0156] The pulse heating drive module 51 uses a stepped heating strategy to generate drive current sampling data. To prevent physical delamination of the gas-sensitive palladium alloy nanofilm 116 due to thermal shock. During the heating phase, the electrical power applied to the micro-hot plate assembly 113 is converted into heat energy, raising the sensor surface temperature. The temperature rapidly rises to the target desorption temperature. The dynamic response of the sensor surface temperature follows the following thermodynamic equation:

[0157] ;

[0158] In the formula, In time The instantaneous surface temperature of the sensor at any given moment; The ambient reference temperature inside the generator compartment; The instantaneous drive current amplitude output by the pulse heating drive module 51; This is the heating resistance value of the micro-hot plate assembly 113; The thermal conductivity of the sensor structure is expressed in watts per Kelvin (W / K). The thermal time constant of the system is determined by the specific heat capacity and mass of the sensor material.

[0159] The pulse heating drive module 51 adjusts the pulse heating drive module 51. The amplitude and duty cycle will Precisely control the target cleaning temperature In this embodiment, The temperature was set to 350 degrees Celsius. At this temperature, the kinetic energy gained by oil and water molecules adsorbed on the palladium film surface is sufficient to overcome van der Waals forces or chemical bond energies, resulting in thermal desorption. The desorption rate of pollutants... Following the Arrhenius equation:

[0160] ;

[0161] In the formula, is the desorption rate constant of surface contaminants; It is the pre-exponential factor (frequency factor), which is related to the vibrational frequency of the adsorbed molecules; The activation energy for the desorption of pollutants; Boltzmann's constant; The absolute temperature of the target for controlled cleaning.

[0162] This formula shows that by raising the temperature to... Desorption rate It increases exponentially, thus achieving rapid removal of the contaminant layer within a short pulse.

[0163] After the cleaning pulse ends, the pulse heating drive module 51 executes a controlled cooling program. Drive current Instead of cutting off immediately, the temperature decreases at a linear slope, causing the sensor surface temperature to drop. The temperature gradually returns to its operating temperature. This process aims to release thermal stress and prevent microcracks from forming in the gas-sensitive membrane due to rapid contraction.

[0164] After the cleaning cycle is completed, the intelligent self-healing control unit 50 resets the measurement circuit, and the system automatically performs a reference impedance calibration. If the calibration is successful... If the temperature drops below the cleaning termination threshold, the system resumes normal monitoring; if it remains above the threshold, the system will increment... The amplitude is adjusted and the pulse thermodynamic cleaning process described above is repeated until the sensor performance is restored or the maximum number of safe cleaning cycles is reached. Through this mechanism, the system maintains the long-term sensitivity and reliability of the sensor array in an oil mist environment without the need for manual maintenance.

[0165] The intelligent self-healing control unit 50 provided by this invention integrates a logic arbitration module 52 and a relay output interface 53. The logic arbitration module 52 is configured to receive real-time hydrogen concentration data processed by a concentration inversion algorithm. It also combines the concentration change rate to perform multi-level safety interlock judgments, so as to trigger graded protection actions in the event of a leakage accident.

[0166] The logic arbitration module 52 first performs a differential operation on the real-time concentration sampling data to calculate the current concentration rise rate. The calculation process uses the finite difference method, and its mathematical expression is as follows:

[0167] ;

[0168] In the formula, For at any time The rate of change in hydrogen concentration, expressed in parts per million per second (ppm / s); For a moment Real-time concentration value; This represents the concentration value at the previous sampling time. This represents the time interval for system sampling.

[0169] To comprehensively assess leakage risk, logic arbitration module 52 constructs a comprehensive hazard index. This index depends not only on the current absolute concentration but also on the rate of increase in concentration, thus enabling early warning of rapid deflagration risks. The comprehensive hazard index is calculated using the following formula:

[0170]

[0171] In the formula, Dimensionless overall risk index; This is the explosive concentration value of hydrogen (40,000 ppm). The reference value for the limiting temperature rise rate defined for the system; This represents the weighting coefficient for the concentration term; This is the weighting coefficient for the rate of change term. In this embodiment, and The sum is 1.

[0172] The logic arbitration module 52 has three preset, incremental security thresholds: warning threshold. Alarm threshold and the trip threshold The logic arbitration module 52 will calculate the... The value is compared in real time with the above threshold, and the corresponding physical control strategy is executed through the relay output interface 53.

[0173] when Greater than and less than At that time, the system determined it to be a minor level one leak. The logic arbitration module 52 outputs the first control signal through the relay output interface 53 to close the exhaust fan circuit in the generator compartment and perform forced ventilation dilution.

[0174] when Greater than and less than At that time, the system determined it to be a level two serious leak. The logic arbitration module 52 outputs a second control signal through the relay output interface 53, driving the electromagnetic shut-off valve on the hydrogen supply pipeline to close, and simultaneously opening the nitrogen purging valve to replace the inside of the generator casing with inert gas.

[0175] when At this time, the system determines that the situation is a Level 3 emergency. The logic arbitration module 52 outputs a third control signal through the relay output interface 53. This signal is directly connected to the main control protection circuit (DCS system) of the generator set, triggering the generator emergency shutdown procedure.

[0176] Furthermore, the logic arbitration module 52 is also equipped with fault shielding logic. If the pure hydrogen locking flag is output by the aforementioned harmonic characteristic calculation module 31b... If the value is 0, or if the pollution correction calculation module 33 reports that the sensor is in a cleaning and maintenance state, the logic arbitration module 52 will forcibly lock the sensor. If the value is 0, even if the sensor outputs a high-level signal, no interlocking action will be triggered. This logic ensures that the system only responds to genuine hydrogen leak events, preventing false trips caused by electromagnetic interference or equipment maintenance.

Claims

1. An online monitoring system for hydrogen leakage from a thermal power plant generator, characterized in that, The system includes: A distributed solid-state palladium alloy sensor array consists of multiple solid-state palladium alloy sensor nodes physically installed in the monitoring area inside the generator. The solid-state palladium alloy sensor nodes convert changes in hydrogen concentration into changes in electrical impedance characteristics. A multi-channel impedance harmonic signal conditioning unit is electrically connected to the distributed solid-state palladium alloy sensor array, applies a composite AC excitation signal to each of the solid-state palladium alloy sensor nodes, receives a response current signal, and converts the response current signal into digital voltage sampling data. The composite AC excitation signal is composed of a fundamental frequency large amplitude carrier signal superimposed with a wideband scanning small signal. An edge computing processing terminal is connected to the multi-channel impedance harmonic signal conditioning unit, receives the digital voltage sampling data, performs a fast Fourier transform operation to obtain frequency domain data, calculates complex impedance spectrum characteristic parameters and nonlinear harmonic distortion coefficients based on the frequency domain data, and inverts the hydrogen concentration value based on the complex impedance spectrum characteristic parameters and nonlinear harmonic distortion coefficients to calculate the spatial coordinates of the leakage source. The host computer monitoring platform communicates with the edge computing processing terminal to receive and display the hydrogen concentration value and the spatial coordinates of the leakage source.

2. The online monitoring system for hydrogen leakage in a thermal power plant generator according to claim 1, characterized in that, The solid palladium alloy sensor node is manufactured using microelectromechanical systems (MEMS) technology. The structure of the solid palladium alloy sensor node includes a silicon substrate, a thermally insulating suspended film layer deposited on the silicon substrate, a micro hot plate heating electrode located on the thermally insulating suspended film layer, an electrically insulating isolation layer covering the micro hot plate heating electrode, interdigitated electrode pairs deposited on the electrically insulating isolation layer, and a gas-sensitive palladium alloy nanofilm covering the interdigitated electrode pairs. The gas-sensitive palladium alloy nanofilm is made of palladium-silver alloy material. The thickness of the gas-sensitive palladium alloy nanofilm is at the mesoscopic scale. Under the electric field driven by the fundamental frequency large amplitude carrier signal, a nonlinear electron scattering effect is generated when hydrogen enters the crystal lattice.

3. The online monitoring system for hydrogen leakage in a thermal power plant generator according to claim 1, characterized in that, The multi-channel impedance harmonic signal conditioning unit includes a direct digital frequency synthesis excitation source, a high-bandwidth transimpedance amplifier, and a synchronous sampling analog-to-digital converter. The direct digital frequency synthesis excitation source generates the composite AC excitation signal using phase cutoff accumulation technology, wherein the current amplitude of the fundamental frequency large amplitude carrier signal is set to be greater than the critical threshold for exciting the nonlinear polarization effect of the palladium alloy thin film, and the wideband scanning small signal contains multiple scanning frequency points distributed on a logarithmic scale. The high-bandwidth transimpedance amplifier is connected to the solid-state palladium alloy sensor node using a four-wire Kelvin connection mode, converting the response current signal into a voltage response signal; The sampling frequency of the synchronous sampling analog-to-digital converter covers the highest frequency component in the composite AC excitation signal and the third harmonic component of the fundamental frequency large amplitude carrier signal.

4. The online monitoring system for hydrogen leakage in a thermal power plant generator according to claim 1, characterized in that, The edge computing processing terminal integrates a digital signal processor, which executes pure hydrogen lock-in discrimination logic. The pure hydrogen lock-in discrimination logic includes: extracting the spectral amplitude magnitude values ​​of the fundamental frequency large amplitude carrier signal, the second harmonic component, and the third harmonic component from the frequency domain data; Calculate the normalized nonlinear distortion response index, which is related to the energy of the second harmonic component and the third harmonic component; When the nonlinear distortion response index is greater than the preset pure hydrogen determination threshold, a valid pure hydrogen lock flag is generated to confirm that the detected signal originates from a lattice phase transition caused by hydrogen.

5. The online monitoring system for hydrogen leakage in a thermal power plant generator according to claim 1, characterized in that, The edge computing processing terminal performs surface contamination correction calculations based on complex impedance spectroscopy; The edge computing processing terminal separates the complex impedance vector of the solid palladium alloy sensor node into real and imaginary characteristic components; The edge computing processing terminal selects the imaginary feature component at the high-frequency reference point to calculate the additional equivalent resistance caused by the surface contamination layer. It then uses the additional equivalent resistance to perform a subtraction correction operation on the real feature component at the low-frequency detection point to obtain the gas-sensitive film body resistance after removing the influence of surface contamination. The gas-sensitive film body resistance is then used to invert the hydrogen concentration value.

6. The online monitoring system for hydrogen leakage in a thermal power plant generator according to claim 1, characterized in that, The system also includes a leak source location unit, which performs the following operations: The spatially discrete concentration data collected by the distributed solid-state palladium alloy sensor array is mapped into a continuous three-dimensional concentration scalar field using an inverse distance weighted interpolation algorithm. Gradient calculations are performed on the three-dimensional concentration scalar field to generate a concentration gradient vector field indicating the gas diffusion trend; In the concentration gradient vector field, reverse flow tracing is performed. The location of the solid palladium alloy sensor node that exceeds the preset safety threshold is used as the seed point. Iterative search is performed along the local gradient direction until at least one of the location convergence criterion and gradient magnitude criterion is satisfied. The convergence endpoint of the search path is calculated as the spatial coordinates of the leakage source.

7. The online monitoring system for hydrogen leakage in a thermal power plant generator according to claim 5, characterized in that, The system also includes an intelligent self-healing control unit; When the additional equivalent resistance calculated by the edge computing processing terminal exceeds the cleaning trigger threshold, the intelligent self-healing control unit pauses the measurement cycle, applies a step-heating driving current to the solid palladium alloy sensor node, raises the sensor surface temperature to the cleaning target temperature and maintains it for a predetermined time, removes surface contaminants by thermal desorption, and controls the temperature to gradually drop after cleaning.

8. The online monitoring system for hydrogen leakage in a thermal power plant generator according to claim 1, characterized in that, The edge computing processing terminal calculates the real-time rate of change of the hydrogen concentration value and constructs a comprehensive hazard index by combining the real-time concentration value. The system performs tiered interlocking control based on the magnitude of the comprehensive hazard index: when the comprehensive hazard index is in the first range, the exhaust fan is triggered; when the comprehensive hazard index is in the second range, the hydrogen supply is shut off and nitrogen purging is initiated; when the comprehensive hazard index is in the third range, a shutdown interlocking command is sent.

9. The online monitoring system for hydrogen leakage in a thermal power plant generator according to claim 1, characterized in that, The distributed solid-state palladium alloy sensor array is divided into a stator core air duct monitoring subset, an end sealing area monitoring subset, and a cooler inlet and outlet monitoring subset according to the generator's internal airflow structure. The stator core air duct monitoring subset is distributed at the radial ventilation groove outlet of the generator stator frame and is arranged at equal intervals along the axial direction; the end sealing area monitoring subset is distributed in the end cover space on both sides of the turbine end and exciter end of the generator and is evenly arranged in the circumferential direction; the cooler inlet and outlet monitoring subset is arranged at the hot air zone inlet and cold air zone outlet of the hydrogen cooler.

10. A detection method for an online monitoring system for hydrogen leakage in a thermal power plant generator, characterized in that, The detection method, applicable to an online hydrogen leakage monitoring system for a thermal power plant generator as described in any one of claims 1-9, includes the following steps: S1. Control the multi-channel impedance harmonic signal conditioning unit to generate a composite AC excitation signal and inject it into each solid palladium alloy sensor node in the distributed solid palladium alloy sensor array. The composite AC excitation signal consists of a fundamental frequency large amplitude carrier signal and a wideband scanning small signal. S2. The response signal of each solid palladium alloy sensor node is synchronously acquired through the multi-channel impedance harmonic signal conditioning unit, the response signal is converted into digital voltage sampling data and transmitted to the edge computing processing terminal. S3. The edge computing processing terminal is used to perform frequency domain transformation on the digital voltage sampling data to obtain frequency domain data. The fundamental frequency large amplitude carrier signal, second harmonic component and third harmonic component are extracted from the frequency domain data to calculate the nonlinear harmonic distortion coefficient. The complex impedance spectrum characteristic parameters are analyzed from the frequency domain data. The real part characteristic components are corrected according to the imaginary part characteristic components in the complex impedance spectrum characteristic parameters. Pure hydrogen lock-in discrimination is performed according to the nonlinear harmonic distortion coefficient. S4. When hydrogen leakage is confirmed, the hydrogen concentration value at each solid palladium alloy sensor node is inverted using the corrected complex impedance spectrum characteristic parameters to construct a three-dimensional concentration gradient vector field inside the generator. The spatial coordinates of the leakage source are calculated using the inverse flow line tracing algorithm, and the hydrogen concentration value and the spatial coordinates of the leakage source are sent to the host computer monitoring platform for display.

Images