Electric locomotive transformer fault monitoring device and method based on gas photoacoustic spectroscopy

Abstract

The present invention discloses a device and method for monitoring transformer faults of electric locomotives based on gas photoacoustic spectroscopy, which relates to the field of electric power technology. The steps of the method include: installing a speed sensor and a vibration sensor on the transformer itself to monitor the transformer vibration impact value during the operation of the electric locomotive; obtaining a water compensation coefficient by installing a temperature sensor and a pressure sensor on the inner wall of the photoacoustic pool; setting a gas concentration monitoring strategy, and adjusting the parameters of the gas concentration monitoring strategy in real time by combining the transformer vibration impact value and the water compensation coefficient; establishing a transformer fault prediction model to predict the type of transformer faults of electric locomotives. The present invention solves the noise impact caused by the vibration caused by the operation of the electric locomotive and the impact of water vapor on the photoacoustic signal during the monitoring of the transformer faults of the electric locomotive.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to a device and method for monitoring transformer faults of an electric locomotive based on gas photoacoustic spectroscopy. Background Art

[0002] With the development of modern industry and the rapid growth of the national economy, the requirements for the stability and reliability of power supply are increasing day by day. However, during the long-term high-load operation of electric locomotive transformers, due to the effects of various factors such as heat, electricity, machinery and chemistry, the internal insulation components will gradually age and even produce defects, leading to partial discharge and the generation of harmful gases. These fault gases will dissolve in the transformer oil to varying degrees, and their concentration and type can reflect the possible fault types of the transformer. Therefore, the monitoring and analysis of dissolved gases in transformer oil has become an important means to judge the operating status of the transformer.

[0003] Traditional transformer fault monitoring methods mainly include offline analysis and online monitoring. Although the offline analysis method is accurate, it has the disadvantages of low efficiency, time-consuming and labor-intensive, and failure to detect hidden dangers in a timely manner. The most widely used online monitoring method is gas chromatography, which requires regular replacement of carrier gas, has a complex system and high maintenance costs, and the measurement accuracy needs to be improved.

[0004] In this context, gas photoacoustic spectroscopy (PAS) technology, as a new online monitoring technology for dissolved gases in transformer oil, has gradually attracted the attention of researchers due to its advantages of high measurement accuracy, short detection time, good long-term stability, no need for carrier gas, simple structure, and simultaneous online monitoring of multiple gases. Photoacoustic spectroscopy technology is based on the principle of photoacoustic effect. It detects gas concentration by measuring the sound signal generated after the gas absorbs light energy. It has significant advantages such as zero background, high sensitivity, wide detection dynamic range, and signal intensity is independent of light intensity.

[0005] For example, an existing Chinese patent with publication number CN114609046A discloses a high-precision photoacoustic spectroscopy monitoring method for gas in transformer oil, and the method includes the following specific steps: performing spectral monitoring on the gas in the transformer oil through a spectral monitoring module, obtaining gas deviation values ​​and spectral deviation values ​​of different dissolved gases in the transformer oil and sending them to a status diagnosis module; diagnosing the working status of the transformer oil using the status diagnosis module, and obtaining status diagnosis values ​​of different dissolved gases in the transformer oil by combining the gas deviation value and the spectral deviation value and sending them to a health assessment module; and the health assessment module assesses the health status of the transformer oil.

[0006] However, this design does not take into account that the background noise in the environment may interfere with the acquisition of the acoustic spectrum signal. In practical applications, the vibration environment generated by the electric locomotive during operation poses a challenge to the online monitoring of gas photoacoustic spectroscopy technology. The vibration of the electric locomotive may not only cause physical displacement or deformation of the monitoring equipment, affecting the stability and accuracy of the photoacoustic signal, but may also interfere with the analysis of the photoacoustic signal by introducing additional noise signals. This vibration interference not only increases the complexity and difficulty of the monitoring technology, but may also reduce the reliability and effectiveness of the monitoring results.

[0007] Secondly, the absorption characteristics of water vapor to light are significantly different from those of target gases (such as hydrogen, methane, acetylene, etc.) dissolved in transformer oil. In photoacoustic spectroscopy measurement, if the concentration of water vapor is high, its absorption spectrum may overlap or interfere with the absorption spectrum of the target gas, resulting in deviations in the measurement results. At the same time, water vapor may also affect the temperature and pressure distribution in the photoacoustic cell. The generation of the photoacoustic effect depends on the thermoelastic effect of gas molecules, and the presence of water vapor will change the thermophysical properties of oil and gas, thereby affecting the intensity and frequency characteristics of the photoacoustic signal. This effect is particularly significant in environments with large temperature changes, which may lead to a decrease in the stability and reliability of the monitoring system.

[0008] Therefore, how to effectively deal with the impact of tram vibration and water vapor on the fault monitoring technology of electric locomotive transformers and explore effective solutions is of great significance for improving the accuracy and stability of the monitoring system and ensuring the safe operation of electric locomotive transformers. To this end, the present invention provides an electric locomotive transformer fault monitoring device and method based on gas photoacoustic spectroscopy. Summary of the invention

[0009] The object of the present invention is to provide an electric locomotive transformer fault monitoring device and method based on gas photoacoustic spectroscopy to solve the existing problems raised in the above background technology.

[0010] To achieve the above object, the present invention provides the following technical solution: an electric locomotive transformer fault monitoring device based on gas photoacoustic spectroscopy, comprising:

[0011] Photoacoustic spectroscopy detection device, used to monitor transformer faults in electric locomotives;

[0012] The oil-gas separation device is arranged on the transformer and is used to extract insulating oil from the transformer and perform oil-gas separation to output the gas to the photoacoustic spectroscopy detection device;

[0013] The photoacoustic spectroscopy detection device comprises:

[0014] NEMS light source, used to generate light pulses;

[0015] A photoacoustic cell, used to provide a closed environment and receive light pulses generated by the NEMS light source;

[0016] A filter wheel, the filter wheel comprising a filter disposed on the filter wheel, the filter wheel being used to receive light pulses outputted by the photoacoustic cell and transmit light with a wavelength corresponding to the filter through the filter;

[0017] Photoacoustic detector, used to receive light transmitted through the filter and analyze the fault of the electric locomotive transformer.

[0018] A further improvement of the present invention is that a speed sensor and a vibration sensor are provided in the transformer, and a vibration impact monitoring strategy is configured to monitor the speed of the electric locomotive during operation and the vibration impact value of the transformer at the corresponding speed;

[0019] A temperature sensor and a pressure sensor are installed on the inner wall of the photoacoustic cell, and a water compensation coefficient monitoring strategy is configured to calculate the water compensation coefficient through the water compensation coefficient monitoring strategy;

[0020] The photoacoustic detector is also provided with a gas concentration monitoring strategy, and the gas concentration monitoring strategy includes adjusting the gas concentration monitoring strategy parameters in real time in combination with the transformer vibration impact value and the water compensation coefficient;

[0021] Based on the gas concentration monitoring strategy, a transformer fault prediction model is established to predict the fault type of the electric locomotive transformer.

[0022] A further improvement of the present invention is that the filter wheel includes filters, and the filters include methane filters, ethylene filters and hydrogen filters; the methane filters, ethylene filters and hydrogen filters are used to obtain methane sound wave signals, ethylene sound wave signals and hydrogen sound wave signals.

[0023] On the other hand, the present invention provides a method for monitoring transformer faults in electric locomotives based on gas photoacoustic spectroscopy, comprising the following steps:

[0024] S1. Set up a vibration impact monitoring strategy to monitor the vibration impact value of the transformer during the operation of the electric locomotive;

[0025] S2. Setting a water compensation coefficient monitoring strategy, and obtaining the water compensation coefficient by combining the humidity data and pressure data in the photoacoustic cell;

[0026] S3, setting a gas concentration monitoring strategy, and adjusting the gas concentration monitoring strategy parameters in real time by combining the transformer vibration impact value and the water compensation coefficient;

[0027] S4. Establish a transformer fault prediction model to predict the fault type of electric locomotive transformer.

[0028] The present invention is further improved in that the calculation of the transformer vibration impact value includes the following steps: obtaining the average speed of the electric locomotive within a time T through the speed sensor ; Obtain the transformer surface vibration displacement data with an interval of time t within time T through the vibration sensor Based on the difference of the transformer surface vibration displacement data at the current time node and the previous time node, the current vibration acceleration of the transformer is obtained by secondary derivation , get the average value of transformer surface vibration acceleration within time T ; The transformer vibration impact value is obtained by weighted summing the mean value of the transformer surface vibration acceleration and the average speed of the electric locomotive.

[0029] The present invention is further improved in that the water compensation coefficient monitoring strategy includes calculating the absolute humidity of the gas sample in the photoacoustic cell through the relative humidity of the gas sample in the photoacoustic cell, the relative humidity represents the ratio of the mass of water vapor in the photoacoustic cell to the water vapor saturated state at the same temperature when calculating the absolute humidity, and the absolute humidity represents the total mass of water vapor in the photoacoustic cell; by experimentally establishing a relative humidity data set in the photoacoustic cell at different temperatures and a corresponding saturated water vapor pressure data set on the inner wall of the photoacoustic cell, the relative humidity under the current temperature conditions is extracted. Saturated water vapor pressure on the inner wall of the photoacoustic cell The absolute humidity water vapor pressure is calculated by multiplying the product of the saturated water vapor pressure on the inner wall of the photoacoustic cell under the current temperature condition and the relative humidity to a percentage. The current pressure on the inner wall of the photoacoustic cell is obtained by the pressure sensor. The absolute humidity water vapor pressure is calculated by the ratio of the absolute humidity water vapor pressure to the difference between the current pressure on the inner wall of the photoacoustic cell and the absolute humidity water vapor pressure. Multiply to get the current absolute humidity in the photoacoustic pool .

[0030] A further improvement of the present invention is that the water compensation coefficient monitoring strategy also includes obtaining a methane acoustic signal sequence, an ethylene acoustic signal sequence, a hydrogen acoustic signal sequence with an interval of time t within a time T obtained through the methane filter, the ethylene filter and the hydrogen filter, and an absolute humidity sequence in the current photoacoustic cell, and linearly fitting the methane acoustic signal sequence, the ethylene acoustic signal sequence, the hydrogen acoustic signal sequence and the absolute humidity sequence in the current photoacoustic cell to obtain an absolute humidity calibration curve in the current photoacoustic cell, and calculating the slope of the curve, which is recorded as the water compensation coefficient.

[0031] A further improvement of the present invention is that the gas concentration monitoring strategy includes obtaining the Allan deviation of the methane acoustic wave signal sequence to obtain a methane concentration Allan deviation sequence, traversing the methane concentration Allan deviation sequence, calculating the absolute value of the difference between two adjacent data, setting a difference noise impact threshold, and when it is monitored that the absolute value of the difference between three consecutive adjacent data in the methane concentration Allan deviation sequence is greater than the difference noise impact threshold, the absolute value of the difference between the two adjacent data at this time is collected and included in the difference noise impact sequence, and the ratio of the number of data in the difference noise impact sequence to the number of methane acoustic wave signal sequences is obtained as the first methane noise impact coefficient, and the ratio of the mean of the difference noise impact sequence to the difference noise impact threshold is calculated as the second methane noise impact coefficient.

[0032] A further improvement of the present invention is that the gas concentration monitoring strategy also includes setting a vibration influence standard value, obtaining a ratio of the transformer vibration influence value to the vibration influence standard value and recording it as a vibration influence coefficient, calculating the mean of the methane acoustic wave signal sequence, and then calculating the difference between the mean and the product of the water compensation coefficient and the absolute humidity in the current photoacoustic pool, and the methane gas concentration is obtained by multiplying the difference with the first noise influence coefficient, the second noise influence coefficient and the vibration influence coefficient.

[0033] A further improvement of the present invention is that the gas concentration monitoring strategy also includes obtaining the first noise influence coefficient of ethylene, the second noise influence coefficient of ethylene, the first noise influence coefficient of hydrogen, the second noise influence coefficient of hydrogen in the same manner as above, and calculating the ethylene gas concentration and the hydrogen gas concentration.

[0034] A further improvement of the present invention is that the process of establishing the transformer fault prediction model includes establishing a fault gas database, the fault gas database includes different methane gas concentrations, ethylene gas concentrations and hydrogen gas concentration ranges and fault types, the fault types include transformer safety, oil overheating, spark discharge in the oil and water ingress, the fault gas database also includes the ratio of each gas, and the fault gas database is used as model training data and the current concentration and ratio of each gas are used as model input to judge the transformer fault type.

[0035] A further improvement of the present invention is that the transformer fault prediction model includes a first-layer fault type judgment algorithm and a second-layer fault type judgment algorithm. First, the current concentration and ratio of each gas are input into the first-layer fault type judgment algorithm, and a first-layer transformer fault type is output. Finally, the current concentration and ratio of each gas are input into the second-layer fault type judgment algorithm, and a second-layer transformer fault type is output. If the first-layer transformer fault type is the same as the second-layer transformer fault type, the fault is used as the output of the transformer fault prediction model. If the first-layer transformer fault type is different from the second-layer transformer fault type, return to S1 and calculate the fault type of the next time period T.

[0036] A further improvement of the present invention is that the first-layer fault type judgment algorithm judges the transformer fault type through the three-ratio method, establishes a fault type coding database, encodes the ratio of methane to ethylene as A, the ratio of methane to hydrogen as B, and the ratio of ethylene to hydrogen as C, sets the coding intervals of code A, code B, and code C, obtains a coding table of four fault types, compares the current concentration and ratio of each gas with the coding table of the four fault types, and obtains a layer of transformer fault types.

[0037] A further improvement of the present invention is that the second-layer fault type judgment algorithm is based on a random forest model, uses historical data in a historical fault gas database as training data, uses the current concentration and ratio of each gas as input data, and outputs the second-layer transformer fault type.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The present invention firstly considers the noise influence of the gas concentration measurement accuracy in the gas concentration monitoring strategy, and secondly considers the noise influence of the sudden vibration on the measurement accuracy during the operation of the electric locomotive, so as to adjust the gas concentration monitoring strategy parameters, thereby reducing the noise influence caused by the vibration caused by the collision of the electric locomotive during the monitoring of the electric locomotive transformer fault;

[0040] 2. Secondly, the water compensation coefficient is calculated through the data of the temperature sensor and the pressure sensor, and the gas concentration monitoring strategy is adjusted in real time, which improves the accuracy and reliability of monitoring;

[0041] 3. By using the established fault gas database and double-layer fault type judgment algorithm, the intelligent prediction of the fault type of the electric locomotive transformer is realized, and the accuracy and robustness of the diagnosis are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of a photoacoustic spectroscopy detection device in a fault monitoring device for electric locomotive transformer based on gas photoacoustic spectroscopy of the present invention;

[0043] Figure 2 A schematic diagram of a built-in strategy of a transformer fault monitoring device for electric locomotives based on gas photoacoustic spectroscopy according to the present invention;

[0044] Figure 3 This is a flow chart of a method for monitoring transformer faults in electric locomotives based on gas photoacoustic spectroscopy according to the present invention;

[0045] Figure 4 This is a flow chart of the gas concentration monitoring strategy of the present invention. DETAILED DESCRIPTION

[0046] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0047] The term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0048] Example 1

[0049] This embodiment discloses a transformer fault monitoring device for electric locomotive based on gas photoacoustic spectroscopy, comprising:

[0050] Photoacoustic spectroscopy detection device, used to monitor transformer faults in electric locomotives;

[0051] The oil-gas separation device is arranged on the transformer and is used to extract insulating oil from the transformer and perform oil-gas separation to output the gas to the photoacoustic spectroscopy detection device;

[0052] The above devices together constitute a dual-stage absorption-enhanced photoacoustic spectroscopy, which enables the device to achieve the measurement of flowing gas that is prohibitive for ordinary photoacoustic spectroscopy. The dual-stage absorption-enhanced structure isolates the gas sample from the photoacoustic detector, and the total volume of the entire gas chamber is less than 3mL;

[0053] The oil-gas separation device comprises a pump and a valve, which are used to extract the insulating oil in the oil tank through a sampling valve on the oil tank.

[0054] Figure 1 The present invention shows a schematic diagram of a photoacoustic spectrum detection device in a transformer fault monitoring device for electric locomotives based on gas photoacoustic spectroscopy, wherein the photoacoustic spectrum detection device comprises:

[0055] NEMS light source is used to generate light pulses. NEMS technology allows the light source and other optical components (such as filters, beam guides, etc.) to be highly integrated on a chip, which makes the monitoring device smaller and lighter, suitable for installation on mobile electric locomotives.

[0056] A photoacoustic cell, used to provide a closed environment and receive light pulses generated by the NEMS light source;

[0057] A filter wheel, the filter wheel includes a filter arranged on the filter wheel, the filter wheel is used to receive light pulses output by the photoacoustic cell, and transmit light with a wavelength corresponding to the filter through the filter; corresponding to the NEMS light source, the filter wheel carries three filters for selectively transmitting light within the wavelength range corresponding to the three filters, the filters include a methane filter, an ethylene filter and a hydrogen filter; the methane filter, the ethylene filter and the hydrogen filter are used to obtain methane acoustic wave signals, ethylene acoustic wave signals and hydrogen acoustic wave signals;

[0058] Photoacoustic detector, used to receive light transmitted through the filter and analyze the fault of the electric locomotive transformer;

[0059] Figure 2 A schematic diagram of a built-in strategy of a transformer fault monitoring device for electric locomotives based on gas photoacoustic spectroscopy is shown in the present invention;

[0060] The transformer is provided with a speed sensor and a vibration sensor, and is configured with a vibration impact monitoring strategy for monitoring the speed of the electric locomotive during operation and the vibration impact value of the transformer at the corresponding speed;

[0061] A temperature sensor and a pressure sensor are installed on the inner wall of the photoacoustic cell, and a water compensation coefficient monitoring strategy is configured to calculate the water compensation coefficient through the water compensation coefficient monitoring strategy;

[0062] The photoacoustic detector is also provided with a gas concentration monitoring strategy, and the gas concentration monitoring strategy includes adjusting the gas concentration monitoring strategy parameters in real time in combination with the transformer vibration impact value and the water compensation coefficient;

[0063] Based on the gas concentration monitoring strategy, a transformer fault prediction model is established to predict the fault type of the electric locomotive transformer.

[0064] Example 2

[0065] Figure 3 The present invention shows a flow chart of a method for monitoring transformer faults in electric locomotives based on gas photoacoustic spectroscopy, which includes the following steps:

[0066] S1. Install a speed sensor and a vibration sensor on the transformer itself to monitor the transformer vibration impact value during the operation of the electric locomotive; the calculation of the transformer vibration impact value includes the following steps: obtaining the average speed of the electric locomotive within a time T through the speed sensor , the collection interval is t; the vibration displacement data of the transformer surface at an interval of t within time T is obtained by the vibration sensor , based on the difference of transformer surface vibration displacement data at the current time node and the previous time node , the transformer surface vibration displacement data difference is taken for the second derivative to obtain the transformer current vibration acceleration , get the average value of transformer surface vibration acceleration within time T ; The mean value of the transformer surface vibration acceleration and the average speed of the electric locomotive The weighted summation is performed to obtain the transformer vibration impact value, and the calculation formula is: ,in, represents the average speed weight of the electric locomotive, Represents the mean weight of the transformer surface vibration acceleration.

[0067] S2. A water compensation coefficient is obtained by installing a temperature sensor and a pressure sensor on the inner wall of the photoacoustic cell; the water compensation coefficient monitoring strategy includes calculating the absolute humidity of the gas sample in the photoacoustic cell by the relative humidity of the gas sample in the photoacoustic cell, wherein the relative humidity represents the ratio of the mass of water vapor in the photoacoustic cell to the mass of water vapor in the saturated state of water vapor at the same temperature when calculating the absolute humidity, and the absolute humidity represents the total mass of water vapor in the photoacoustic cell; a relative humidity data set in the photoacoustic cell at different temperatures and a corresponding saturated water vapor pressure data set on the inner wall of the photoacoustic cell are established by experiments;

[0068] When light shines on gas molecules, if the frequency of the light matches a certain electronic energy level transition of the molecule, the molecule will absorb the energy of the photon and transition from a lower energy level (ground state) to a higher energy level (excited state). The molecule in the excited state is unstable and will seek to return to the ground state to release excess energy. In the process of non-radiative de-excitation, the excess energy of the molecule is transferred to the surrounding medium in the form of heat energy. This process causes the local temperature to rise, forming a tiny heat source. Changes in temperature will cause changes in the density of the medium, because an increase in temperature will cause the medium to expand and the density to decrease. Changes in density will cause changes in pressure. When the medium expands, the local pressure decreases; when the medium cools and contracts, the pressure recovers. This periodic pressure change forms a sound wave. Therefore, the present invention can obtain the water compensation coefficient through a pressure sensor installed on the inner wall of the photoacoustic pool;

[0069] Extract relative humidity under current temperature conditions Saturated water vapor pressure on the inner wall of the photoacoustic cell The absolute humidity water vapor pressure is calculated by multiplying the saturated water vapor pressure on the inner wall of the photoacoustic cell under the current temperature conditions and the relative humidity to a percentage. , obtain the current pressure of the inner wall of the photoacoustic cell through the pressure sensor , then the absolute humidity water vapor pressure is calculated by the ratio of the absolute humidity water vapor pressure to the difference between the current photoacoustic cell inner wall pressure and the absolute humidity water vapor pressure. Multiply to get the current absolute humidity in the photoacoustic pool , is 1 Pascal.

[0070] Obtaining a methane acoustic wave signal sequence, an ethylene acoustic wave signal sequence, a hydrogen acoustic wave signal sequence, and an absolute humidity sequence in the current photoacoustic cell with an interval time of t within a time T obtained by the methane filter, the ethylene filter, and the hydrogen filter, and performing linear fitting on the methane acoustic wave signal sequence, the ethylene acoustic wave signal sequence, the hydrogen acoustic wave signal sequence, and the absolute humidity sequence in the current photoacoustic cell to obtain an absolute humidity calibration curve in the current photoacoustic cell, and calculating the slope of the curve, which is recorded as a water compensation coefficient .

[0071] S3, setting a gas concentration monitoring strategy, and adjusting the gas concentration monitoring strategy parameters in real time by combining the transformer vibration impact value and the water compensation coefficient; Figure 4 The flow chart of the gas concentration monitoring strategy of this embodiment is shown;

[0072] The gas concentration monitoring strategy includes obtaining the Allan deviation of the methane acoustic signal sequence , and obtain the Allan deviation series of methane concentration, where represents the i-th methane acoustic wave signal, , traverse the Allan deviation sequence of methane concentration, calculate the absolute value of the difference between two adjacent data, and set the difference noise impact threshold Since the vibration of the electric locomotive during operation is random and unpredictable, and has a uniform power distribution at all frequencies, it can be regarded as white noise. The characteristic of white noise is that it has no specific frequency preference, and its Allan variance tends to be stable after long-term averaging. Therefore, when the absolute value of the difference between three consecutive adjacent data in the Allan deviation sequence of methane concentration is monitored to be greater than the difference noise impact threshold, the absolute value of the difference between the two adjacent data at this time is collected and included in the difference noise impact sequence, and the number of data in the difference noise impact sequence is obtained. and the number of methane acoustic signals The ratio of methane is used as the first noise influence coefficient , calculate the mean of the difference noise affected sequence The ratio of the difference noise impact threshold is used as the methane second noise impact coefficient .

[0073] Setting the vibration influence standard value The ratio of the transformer vibration impact value to the vibration impact standard value is recorded as the vibration impact coefficient , calculate the mean of the methane acoustic wave signal sequence Then, calculate the difference between the product of the water compensation coefficient and the absolute humidity in the current photoacoustic cell. , then the methane gas concentration is obtained by multiplying the difference with the first noise influence coefficient, the second noise influence coefficient and the vibration influence coefficient. ;

[0074] The ethylene gas concentration is obtained through all the steps in S3 above. and hydrogen gas concentration ,in, represents the first noise impact coefficient of ethylene, represents the second noise impact coefficient of ethylene, Indicates the first noise influence coefficient of hydrogen, represents the second noise influence coefficient of hydrogen, represents the mean of the ethylene acoustic wave signal sequence, Represents the mean of the hydrogen acoustic wave signal sequence.

[0075] S4. Establish a transformer fault prediction model to predict the fault type of electric locomotive transformer.

[0076] The process of establishing the transformer fault prediction model includes establishing a fault gas database, which includes different methane gas concentrations, ethylene gas concentrations and hydrogen gas concentration ranges and fault types. The fault types include transformer safety, oil overheating, spark discharge in the oil and water ingress. The fault gas database also includes the ratio of each gas. The fault gas database is used as model training data, and the current concentration and ratio of each gas are used as model input to judge the transformer fault type.

[0077] The transformer fault prediction model includes a first-layer fault type judgment algorithm and a second-layer fault type judgment algorithm;

[0078] The first-layer fault type judgment algorithm judges the transformer fault type by the three-ratio method, establishes a fault type coding database, encodes the ratio of methane to ethylene as A, the ratio of methane to hydrogen as B, and the ratio of ethylene to hydrogen as C; sets the coding interval [0, ATcod1] of code A to be 0, (ATcod1, ATcod2] to be 1, (ATcod2, ATcod3] to be 2; sets the coding interval [0, BTcod1] of code B to be 0, (BTcod1, BTcod2] to be 1, (BTcod2, BTcod3] to be 2 cod3] takes the value of 2; the coding interval of code C [0, CTcod1] takes the value of 0, (CTcod1, CTcod2] takes the value of 1, (CTcod2, CTcod3] takes the value of 2; obtain the coding table of four fault types, compare the current concentration and ratio of each gas with the coding table of the four fault types, and obtain a layer of transformer fault types. For example, the current code of the gas inside the transformer is (1, 0, 0). By finding the fault type corresponding to (1, 0, 0) in the coding table of the four fault types as oil overheating, it is determined that the transformer fault type is oil overheating.

[0079] The second-layer fault type judgment algorithm is based on a random forest model, uses historical data in a historical fault gas database as training data, uses the current concentration and ratio of each gas as input data, and outputs a second-layer transformer fault type.

[0080] First, the current concentration and ratio of each gas is input into a first-layer fault type judgment algorithm, and a first-layer transformer fault type is output. Finally, the current concentration and ratio of each gas is input into a second-layer fault type judgment algorithm, and a second-layer transformer fault type is output. If the first-layer transformer fault type is the same as the second-layer transformer fault type, the fault is used as the output of the transformer fault prediction model. If the first-layer transformer fault type is different from the second-layer transformer fault type, return to S1 and calculate the fault type of the next time period T.

[0081] The threshold and weight settings can be set by the operator.

[0082] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0084] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0086] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. An electric locomotive transformer fault monitoring device based on gas photoacoustic spectroscopy, characterized in that: include: Photoacoustic spectroscopy detection device, used to monitor transformer faults in electric locomotives; The oil-gas separation device is arranged on the transformer and is used to extract insulating oil from the transformer and perform oil-gas separation to output the gas to the photoacoustic spectroscopy detection device; The photoacoustic spectroscopy detection device comprises: NEMS light source, used to generate light pulses; A photoacoustic cell, used to provide a closed environment and receive light pulses generated by the NEMS light source; A filter wheel, the filter wheel comprising a filter disposed on the filter wheel, the filter wheel being used to receive light pulses outputted by the photoacoustic cell and transmit light with a wavelength corresponding to the filter through the filter; Photoacoustic detector, used to receive light transmitted through the filter and analyze the fault of the electric locomotive transformer; The transformer is provided with a speed sensor and a vibration sensor, and is configured with a vibration impact monitoring strategy for monitoring the speed of the electric locomotive during operation and the vibration impact value of the transformer at the corresponding speed; A temperature sensor and a pressure sensor are installed on the inner wall of the photoacoustic cell, and a water compensation coefficient monitoring strategy is implemented to calculate the water compensation coefficient through the water compensation coefficient monitoring strategy; The photoacoustic detector also executes a gas concentration monitoring strategy, which includes adjusting the gas concentration monitoring strategy in real time in combination with the transformer vibration impact value and the water compensation coefficient; Based on the gas concentration monitoring strategy, a transformer fault prediction model is established to predict the fault type of the electric locomotive transformer; The filter wheel includes filters, and the filters include methane filters, ethylene filters, and hydrogen filters; the methane filters, ethylene filters, and hydrogen filters are used to obtain methane sound wave signals, ethylene sound wave signals, and hydrogen sound wave signals; The calculation of the transformer vibration impact value comprises the following steps: obtaining the average speed of the electric locomotive within a time T through the speed sensor ; Obtain the transformer surface vibration displacement data with an interval of time t within time T through the vibration sensor Based on the difference of the transformer surface vibration displacement data at the current time node and the previous time node, the current vibration acceleration of the transformer is obtained by secondary derivation , get the average value of transformer surface vibration acceleration within time T ; Performing a weighted summation on the mean value of the transformer surface vibration acceleration and the average speed of the electric locomotive to obtain the transformer vibration impact value; The water compensation coefficient monitoring strategy includes calculating the absolute humidity of the gas sample in the photoacoustic cell through the relative humidity of the gas sample in the photoacoustic cell, wherein the relative humidity represents the ratio of the mass of water vapor in the photoacoustic cell to the mass of water vapor in the saturated state of water vapor at the same temperature when calculating the absolute humidity, and the absolute humidity represents the total mass of water vapor in the photoacoustic cell; experimentally establishing a relative humidity data set in the photoacoustic cell at different temperatures and a corresponding saturated water vapor pressure data set on the inner wall of the photoacoustic cell, and extracting the relative humidity under the current temperature conditions. Saturated water vapor pressure on the inner wall of the photoacoustic cell The absolute humidity water vapor pressure is calculated by multiplying the product of the saturated water vapor pressure on the inner wall of the photoacoustic cell under the current temperature condition and the relative humidity to a percentage. The current pressure on the inner wall of the photoacoustic cell is obtained by the pressure sensor. The absolute humidity water vapor pressure is calculated by the ratio of the absolute humidity water vapor pressure to the difference between the current pressure on the inner wall of the photoacoustic cell and the absolute humidity water vapor pressure. Multiply to get the current absolute humidity in the photoacoustic pool ; The water compensation coefficient monitoring strategy also includes obtaining a methane acoustic wave signal sequence, an ethylene acoustic wave signal sequence, a hydrogen acoustic wave signal sequence, and an absolute humidity sequence in the current photoacoustic cell with an interval of time t within a time T obtained by the methane filter, the ethylene filter, and the hydrogen filter, and performing linear fitting on the methane acoustic wave signal sequence, the ethylene acoustic wave signal sequence, the hydrogen acoustic wave signal sequence, and the absolute humidity sequence in the current photoacoustic cell to obtain an absolute humidity calibration curve in the current photoacoustic cell, and calculating the slope of the curve, which is recorded as the water compensation coefficient; The gas concentration monitoring strategy includes obtaining the Allan deviation of the gas acoustic wave signal sequence to obtain the gas concentration Allan deviation sequence, traversing the gas concentration Allan deviation sequence, calculating the absolute value of the difference between two adjacent data, setting a difference noise impact threshold, and when it is monitored that the absolute value of the difference between three consecutive adjacent data in the gas concentration Allan deviation sequence is greater than the difference noise impact threshold, collecting the absolute value of the difference between the two adjacent data at this time and including it in the difference noise impact sequence, and obtaining the ratio of the number of data in the difference noise impact sequence to the number of gas acoustic wave signal sequences as the first gas noise impact coefficient, and calculating the ratio of the mean of the difference noise impact sequence to the difference noise impact threshold as the second gas noise impact coefficient; The gas concentration monitoring strategy also includes setting a vibration influence standard value, obtaining a ratio of the transformer vibration influence value to the vibration influence standard value and recording it as a vibration influence coefficient, calculating the mean of the gas acoustic wave signal sequence, and calculating the difference between the mean and the product of the water compensation coefficient and the absolute humidity in the current photoacoustic cell, and the gas concentration is obtained by multiplying the difference with the first noise influence coefficient, the second noise influence coefficient and the vibration influence coefficient; The gases in the gas concentration monitoring strategy include methane, ethylene and hydrogen. After obtaining the first noise influence coefficient of methane, the second noise influence coefficient of methane, the first noise influence coefficient of ethylene, the second noise influence coefficient of ethylene, the first noise influence coefficient of hydrogen, and the second noise influence coefficient of hydrogen in the above manner, the methane gas concentration, the ethylene gas concentration and the hydrogen gas concentration are calculated; The process of establishing the transformer fault prediction model includes establishing a fault gas database, which includes different methane gas concentrations, ethylene gas concentrations and hydrogen gas concentration ranges and fault types. The fault types include transformer safety, oil overheating, spark discharge in the oil and water ingress. The fault gas database also includes the ratio of each gas. The fault gas database is used as model training data, and the current concentration and ratio of each gas are used as model input to judge the transformer fault type.

2. A method for monitoring transformer faults of electric locomotives based on gas photoacoustic spectroscopy, which is performed based on the device for monitoring transformer faults of electric locomotives based on gas photoacoustic spectroscopy as claimed in claim 1, characterized in that: The following steps are involved: S1. Set up vibration impact monitoring strategy to monitor transformer vibration impact value during the operation of electric locomotive; S2. Setting a water compensation coefficient monitoring strategy, and obtaining the water compensation coefficient by combining the humidity data and pressure data in the photoacoustic cell; S3, setting a gas concentration monitoring strategy, and adjusting the gas concentration monitoring strategy parameters in real time by combining the transformer vibration impact value and the water compensation coefficient; S4. Establish a transformer fault prediction model to predict the fault type of electric locomotive transformer.

3. The electric locomotive transformer fault monitoring method based on gas photoacoustic spectroscopy according to claim 2 is characterized in that: The transformer fault prediction model includes a first-layer fault type judgment algorithm and a second-layer fault type judgment algorithm. First, the current concentration and ratio of each gas are input into the first-layer fault type judgment algorithm to output a first-layer transformer fault type. Finally, the current concentration and ratio of each gas are input into the second-layer fault type judgment algorithm to output a second-layer transformer fault type. If the first-layer transformer fault type is the same as the second-layer transformer fault type, the fault is used as the output of the transformer fault prediction model. If the first-layer transformer fault type is different from the second-layer transformer fault type, return to S1 to calculate the fault type of the next time period T.

4. The method for monitoring transformer faults in electric locomotives based on gas photoacoustic spectroscopy according to claim 3 is characterized in that: The first-layer fault type judgment algorithm judges the transformer fault type through the three-ratio method, establishes a fault type coding database, encodes the ratio of methane to ethylene as A, the ratio of methane to hydrogen as B, and the ratio of ethylene to hydrogen as C, sets the coding intervals of code A, code B, and code C, obtains a coding table of four fault types, compares the current concentration and ratio of each gas with the coding table of the four fault types, and obtains a layer of transformer fault types.

5. The electric locomotive transformer fault monitoring method based on gas photoacoustic spectroscopy according to claim 4 is characterized in that: The second-layer fault type judgment algorithm is based on a random forest model, uses historical data in a historical fault gas database as training data, uses the current concentration and ratio of each gas as input data, and outputs a second-layer transformer fault type.

Images