A method for state monitoring and fault early warning of dry-type transformers on offshore platforms

Abstract

The present invention relates to the technical field of monitoring dry-type transformers on offshore platforms, and more specifically, to a method for monitoring the status and providing fault warnings for dry-type transformers on offshore platforms. The method includes collecting historical noise information of sensors, obtaining the proportion of conventional noise positions of each sensor noise and the signal-to-noise ratio variation trend; and calculating the aging influence and concentrated aging influence of each sensor noise in combination with the influence weight of each sensor noise. The present invention obtains the influence weight of each sensor noise in actual collection by the proportion of conventional noise positions and the signal-to-noise ratio variation trend, and uses the influence weight to feedback the degree of influence of each sensor noise on sensor aging. By calculating the aging influence and concentrated aging influence of each sensor noise, and matching the aging influence interval, the method is used to evaluate the aging degree of sensors in the current offshore platform dry-type transformer, thereby ensuring the accuracy of the prediction results and reducing the monitoring work and difficulty.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring dry-type transformers on offshore platforms, and in particular to a method for monitoring the state and providing fault early warning for dry-type transformers on offshore platforms. Background Art

[0002] In order to ensure the normal operation of offshore platform dry-type transformers and obtain various operating parameters, the existing offshore platform dry-type transformer system contains multiple types of sensors, which can not only cooperate with the offshore platform dry-type transformer to achieve normal operation, but also provide real-time feedback of monitoring data.

[0003] Due to the harsh environment of dry-type transformers on offshore platforms and the different locations of various sensors, the degree to which they are affected by the environment varies. For example, seawater corrosion, mechanical vibration caused by waves, and the influence of bad weather will all affect the sensors, affecting not only their monitoring accuracy but also their service life. Therefore, sensor monitoring work in dry-type transformers on offshore platforms is essential.

[0004] There are generally two types of monitoring directions for existing sensors. One is direct monitoring, which involves manual on-site monitoring on offshore platforms and monitoring and processing through various sensor instruments. This processing method is cumbersome and labor-intensive, and is difficult to monitor for independent offshore platforms. The other is indirect monitoring, which involves monitoring the signals fed back by the sensors and analyzing them to determine the aging status of the sensors, thereby providing early warning of impending failures. Different signals contain different information, and predictions often require continuous information segmentation. However, the accuracy of the final prediction results is low.

[0005] Among the signals fed back by the sensors, the signal noise contains the most information and is easily distinguished. Different noise types, locations, and signal-to-noise ratios will cover different fault information. If the information covered by the noise can be analyzed and used as the basis for fault prediction, the monitoring workload can be reduced and the difficulty of monitoring can be lowered.

[0006] In order to address the above problems, there is an urgent need for a method to monitor the sensor status and provide fault warning in offshore platform dry-type transformers based on the noise information differences of the sensors. Summary of the Invention

[0007] The object of the present invention is to provide a method for state monitoring and fault early warning of dry-type transformers on offshore platforms, so as to solve the problems raised in the above-mentioned background technology.

[0008] To achieve the above objectives, a method for state monitoring and fault early warning of dry-type transformers on offshore platforms is provided, comprising the following steps:

[0009] S1. Monitor the sensor signal of the offshore platform dryer and obtain the noise position in the sensor signal. and the signal-to-noise ratio change ;

[0010] S2. Collect historical sensor noise information and obtain the proportion of regular noise positions of each sensor noise and the signal-to-noise ratio variation trend;

[0011] S3, according to the proportion of conventional noise positions As well as the signal-to-noise ratio change trend, calculate the influence weight of each sensor noise in actual acquisition ;

[0012] S4. Combine the influence weights of each sensor noise , calculate the aging effect of each sensor noise and the concentrated aging effect ;

[0013] S5. Establish the aging impact interval , different aging impact ranges Corresponding to different degrees of aging;

[0014] S6. Impact on concentrated aging Aging impact interval Match, obtain the aging degree of the current sensor, mark the direct impact on sensor noise, and warn of sensor failure based on the aging degree.

[0015] As a further improvement of this technical solution, the noise position in the sensor signal is obtained in S1 The method comprises the following steps:

[0016] S1.1. Divide the signal time domain time axis and use sliding window variance to locate the noise segment;

[0017] S1.2. Set the window width in the time domain signal, marked as W;

[0018] S1.3, use the window variance algorithm to calculate the local variance near the time point t, marked as ;

[0019] S1.4. Define baseline variance ,when When , it is determined to be a noise interval, and the time range of the current noise interval is obtained as the position of the current noise in the current signal.

[0020] As a further improvement of this technical solution, the window variance algorithm in S1.3 is as follows:

[0021] ;

[0022] ;

[0023] in The signal at time The sampling value of is the average value of the data in the window, W is the window width in the time domain signal, is the local variance around time point t.

[0024] As a further improvement of this technical solution, the signal-to-noise ratio variation in the sensor signal is obtained in S1 The method comprises the following steps:

[0025] S1.5. Get the signal power of the current signal and noise power ;

[0026] S1.6, use the signal-to-noise ratio logarithmic algorithm to calculate the signal-to-noise ratio ;

[0027] S1.7. Combine the interval time T of the collected signal to obtain the change in signal-to-noise ratio .

[0028] As a further improvement of this technical solution, the signal-to-noise ratio logarithm algorithm in S1.6 is as follows:

[0029] ;

[0030] ;

[0031] is the signal power of the current signal, is the noise power, is the signal-to-noise ratio, as well as is the signal-to-noise ratio of two adjacent signals collected at an interval of T, where T is the interval time.

[0032] As a further improvement of this technical solution, the ratio of the conventional noise position of each sensor noise is obtained in S2 The method comprises the following steps:

[0033] S2.1. Obtain the position of the same noise in the signal per unit time;

[0034] S2.2. Calculate the probability of the same noise occurring at different positions in the signal within the same batch;

[0035] S2.3. Establish probability thresholds;

[0036] When the occurrence probability ≥ probability threshold, the noise position corresponding to the current occurrence probability is the regular noise position, and the occurrence probability is the actual proportion ;

[0037] When the occurrence probability is less than the probability threshold, the noise position corresponding to the current occurrence probability is the abnormal position and is not used as a reference.

[0038] As a further improvement of the present technical solution, the signal-to-noise ratio change trend in S2 includes the range of increase and the degree of range influence;

[0039] The range of the increase is the range of the signal-to-noise ratio change;

[0040] The influence degree of the range is the change in the signal-to-noise ratio of the monitoring results on both sides of the adjacent 's changing trend.

[0041] As a further improvement of this technical solution, the influence weight of each sensor noise in the actual acquisition is calculated in S3 The method comprises the following steps:

[0042] S3.1. Proportion of passing through conventional noise locations And the signal-to-noise ratio change trend prediction, formulate the comparison order of the current monitoring noise, specifically the increase range - range impact degree - the proportion of conventional noise position , as a comparison item;

[0043] S3.2. Assign initial weights to each comparison item. The values of the initial weights are different and are defined based on the impact of each comparison item on the corresponding noise.

[0044] S3.3. After completing the comparison of the three comparison items, select the initial weight of the sensor noise of the successfully compared items and perform the influence weight renew;

[0045] When the sensor noises of the three successfully compared items are all different, the initial weights of the corresponding comparison items are obtained and combined and allocated;

[0046] When the sensor noise of two successfully compared items is the same, the initial weights of the corresponding comparison items are obtained and multiplied to update the influence weight of the current sensor noise. , the remaining one comparison item that is successfully matched and has different noise from the above-mentioned sensor is still assigned an initial weight in a combined manner;

[0047] When the sensor noise of three successfully compared items is the same, the initial weights of the corresponding comparison items are obtained and multiplied to update the influence weight of the current sensor noise. .

[0048] As a further improvement of the present technical solution, the S4 sensor noises include thermal noise, 1 / f noise, popcorn noise and drift noise;

[0049] Thermal noise is the noise caused by the thermal motion of carriers inside the conductor;

[0050] 1 / f noise is the noise caused by carrier mobility fluctuations due to defects in the sensor material;

[0051] Popcorn noise is the noise caused by carrier capture at semiconductor lattice defects in the sensor;

[0052] Drift noise is the noise caused by the breaking / oxidation of chemical bonds in the sensitive material of the sensor.

[0053] As a further improvement of this technical solution, the concentrated aging influence is calculated in S4 The output signal optimization model algorithm is adopted, and its algorithm formula is as follows:

[0054] ;

[0055] in is the thermal noise term, and its specific calculation formula is as follows:

[0056] .

[0057] To obey the standard normal distribution A random process, that is, a Gaussian white noise with a mean of 0 and a variance of 1, , which is the Boltzmann constant, is the absolute temperature, is the equivalent noise resistance, is the system bandwidth;

[0058] in is the 1 / f noise term, and its specific calculation formula is as follows:

[0059] ;

[0060] in is the time constant, is the power law index and has a range of , is the normalization constant, f is the frequency, and the variables are replaced by ,Right now ;

[0061] in is the popcorn item, and its specific calculation formula is as follows:

[0062] ;

[0063] in, represents the total number of pulses at time t, is the time when the i-th pulse occurs, is the amplitude of the ith pulse, is a unit step function;

[0064] in is the drift term, and its specific calculation formula is as follows:

[0065] ;

[0066] in is the linear aging coefficient, Assign a value to the initial transient drift, is the aging stabilization time constant;

[0067] 、 、 as well as The influence weights of thermal noise, 1 / f noise, popcorn noise and drift noise are respectively , when comparing sensor noise not involved in the project, the corresponding impact weight =1, the corresponding impact weight of the sensor noise involved is the updated weight.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] In this offshore platform dry-type transformer condition monitoring and fault warning method, the influence weight of each sensor noise in actual acquisition is obtained through the proportion of conventional noise positions and the signal-to-noise ratio change trend. The influence degree of each sensor noise on sensor aging is fed back through the influence weight. The aging influence amount and the concentrated aging influence amount of each sensor noise are calculated, and the aging influence amount interval is matched to evaluate the aging degree of the sensors in the current offshore platform dry-type transformer. This not only ensures the accuracy of the prediction results, but also reduces the monitoring work and difficulty. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 is a flow chart of the overall method of the present invention;

[0071] Figure 2 A diagram showing the steps of a method for obtaining a noise position in a sensor signal according to the present invention;

[0072] Figure 3 A diagram showing the steps of a method for obtaining a change in the signal-to-noise ratio in a sensor signal according to the present invention;

[0073] Figure 4 A diagram showing the steps of a method for obtaining the proportion of the normal noise position of each sensor noise according to the present invention;

[0074] Figure 5 This is a step diagram of the method for calculating the influence weight of each sensor noise in actual acquisition according to the present invention. DETAILED DESCRIPTION

[0075] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0076] See also Figure 1 As shown, a method for monitoring sensor status and providing fault warning in an offshore platform dry-type transformer based on noise information differences of the sensor is provided, comprising the following steps:

[0077] S1. Monitor the sensor signal of the offshore platform dryer and obtain the noise position in the sensor signal. and the signal-to-noise ratio change ;

[0078] S2. Collect historical sensor noise information and obtain the proportion of regular noise positions of each sensor noise and the signal-to-noise ratio variation trend;

[0079] S3, according to the proportion of conventional noise positions As well as the signal-to-noise ratio change trend, calculate the influence weight of each sensor noise in actual acquisition ;

[0080] S4. Combine the influence weights of each sensor noise , calculate the aging effect of each sensor noise and the concentrated aging effect ;

[0081] S5. Establish the aging impact interval , different aging impact ranges Corresponding to different degrees of aging;

[0082] S6. Impact on concentrated aging Aging impact interval Match, obtain the aging degree of the current sensor, mark the direct impact on sensor noise, and warn of sensor failure based on the aging degree.

[0083] The specific contents are as follows:

[0084] Since the noise types carried in the sensor feedback signal are diverse, the corresponding locations of different sensor noises are different, and the signal-to-noise ratio changes differently with the aging trend of the sensor, the corresponding changes will also be different. Therefore, the above differences can be used as features to predict sensor noise;

[0085] First, it is necessary to monitor the sensor signal of the offshore platform dryer and obtain the noise position in the sensor signal and the signal-to-noise ratio change , since the noise exists in the signal in a segmented form, such as Figure 2 As shown, the specific method of locating the noise position is as follows:

[0086] First, the signal time domain axis is divided, and the sliding window variance is used to locate the noise segment. Since noise usually manifests as random fluctuations of the signal, the time domain position of the noise can be located by monitoring the mutation of the local variance. The first step is to set the window width in the time domain signal, marked as W (take 5-10 times the signal period, and the monitoring personnel preset it according to the actual situation), calculate the local variance near the time point t, marked as , and the local variance Reflects the degree of fluctuation of the data in the window, local variance The larger the value, the greater the signal fluctuation in the window, and the more likely there is noise. On the contrary, the local variance The smaller the value, the more stable the signal. ,in The signal at time The sampling value of is the average value of the data in the window, and , the calculation results are based on 3 Criteria, defining baseline variance ,when When , it is determined to be a noise interval, and the time range of the current noise interval is obtained as the position of the current noise in the current signal.

[0087] Signal-to-noise ratio change As another distinguishing item of noise in the present invention, Figure 3 As shown, during the calculation process, the signal power of the current signal is first obtained and noise power , using the signal-to-noise ratio logarithm algorithm to calculate the signal-to-noise ratio ,Right now , combined with the interval time T of the collected signal, obtain the change in signal-to-noise ratio ,in as well as is the signal-to-noise ratio of two adjacent signals collected at an interval of time T.

[0088] In the actual monitoring process, the noise of each sensor often presents a noise position Regularity changes and signal-to-noise ratio changes Regular changes, such as the signal-to-noise ratio variation in thermal noise / month, so in order to predict the current sensor aging noise, it is necessary to collect the sensor's historical noise information in advance and obtain the proportion of the regular noise position of each sensor noise. And the signal-to-noise ratio change trend, such as Figure 4 As shown, the proportion of regular noise positions is obtained Here’s how:

[0089] First, we need to obtain the position of the same noise in the signal within a unit time. As can be seen from the above content, it is the time range, and calculate the probability of the same noise in different positions in the signal under the same batch. The same batch here is the monitoring time range, for example, the calculation is performed in the signal data obtained within a month, and a probability threshold is set. When the probability of occurrence ≥ the probability threshold, the noise position corresponding to the current probability of occurrence is the regular noise position, and the probability of occurrence is the actual proportion ,On the contrary, when the occurrence probability is less than the probability threshold, the noise position corresponding to the current ,occurrence probability is the abnormal position and is not used as a reference;

[0090] For the signal-to-noise ratio change trend, it specifically includes the range of increase and the degree of influence of the range, such as the change in the signal-to-noise ratio in thermal noise / month, that is, every month, the signal-to-noise ratio monitored in the current month Compared to the previous month, the lowest , when the signal-to-noise ratio change of the noise displayed by two adjacent monitoring results is 3 / month and / month, first of all, the two monitoring results are consistent with the signal-to-noise ratio change in thermal noise The range of increase, at the same time, as time changes, the corresponding range of influence becomes increasingly higher, that is, the signal-to-noise ratio change The signal-to-noise ratio changes in the above example are getting bigger and bigger. It also meets the scope of influence. At this time, relative to the trend of signal-to-noise ratio change, the current monitoring results are completely consistent with the final impact weight. It will also increase.

[0091] In order to further refine the final prediction results, the present invention introduces the influence weight As a parameter to predict sensor noise, such as Figure 5 The specific calculation method is as follows:

[0092] First, the ratio of the regular noise position And the signal-to-noise ratio change trend prediction, formulate the comparison order of the current monitoring noise, specifically the increase range - range impact degree - the proportion of conventional noise position , as the comparison item, and the initial weight division of each comparison item is different. The value of each initial weight is defined according to the impact of a single comparison item on the corresponding noise. For example, for noise with a single noise position distribution, the proportion of its corresponding regular noise position is The initial weight of will be higher than the other two comparison items;

[0093] During the comparison process, first calculate the signal-to-noise ratio change of the current monitoring noise according to the above steps. First, determine which sensor noise increase range is consistent with the target, and obtain the initial weight of the matching comparison item corresponding to the matching sensor noise. Then, continue the comparison of the next comparison item in the order of the comparison items, which is the comparison of the range impact degree.

[0094] When the signal-to-noise ratio of the current monitoring noise changes If the impact degree of the matching range belongs to the same sensor noise as the previous comparison item, then the current monitoring noise is the impact weight of the sensor noise. Will be updated to and The product of and is the initial weight of different comparison items for the same sensor noise;

[0095] The signal-to-noise ratio change of the current monitoring noise If the impact degree of the matched range does not belong to the same sensor noise as the previous comparison item, then the initial weight of the impact degree of the other sensor noise on the matched range is obtained. For example, the monitoring data obtained in the same batch show three sets of signal-to-noise ratio changes. , respectively 、 as well as , during the first comparison project matching process, 、 as well as Both belong to the range of thermal noise increase, and the corresponding initial weights are obtained. At the same time, the range of thermal noise influence is greater than RdB / month, and the range of drift noise influence is greater than PdB / month. After calculation, when - >RdB / month, and - >RdB / month, then the signal-to-noise ratio change of the monitoring noise at this time According to the degree of influence of the range in thermal noise, obtain the corresponding initial weight, calculate the product of the two initial weights, and update the influence weight ,when - >PdB / month, and - >PdB / month, then the signal-to-noise ratio change of the monitoring noise at this time According to the range influence degree of drift noise, the corresponding initial weight is obtained, and the influence weight of the monitoring noise is obtained at this time. It is expressed as a combination of the initial weight of the first term that corresponds to the range of the increase in sensor noise and the initial weight of the second term that corresponds to the range of the impact of the sensor noise;

[0096] Finally, the proportion of regular noise positions Matching, after the match is successful, the initial weight of the comparison item is the ratio of the corresponding regular noise position When the sensor noise corresponding to the noise position matched by the current monitoring noise position coincides with the sensor noise corresponding to the previous item or two comparison items, the influence weight is calculated in the form of product. Update, otherwise, influence weights in a combined manner renew.

[0097] Completion Impact Weight After the calculation work, it is necessary to combine the influence weights of each sensor noise , calculate the aging effect of each sensor noise and the concentrated aging effect , wherein the sensor noise involved in the present invention includes thermal noise, 1 / f noise, popcorn noise and drift noise;

[0098] Thermal noise is the noise caused by the thermal motion of carriers inside the conductor;

[0099] 1 / f noise is the noise caused by carrier mobility fluctuations due to defects in the sensor material;

[0100] Popcorn noise is the noise caused by carrier capture at semiconductor lattice defects in the sensor;

[0101] Drift noise is the noise caused by the breaking / oxidation of chemical bonds in the sensitive material of the sensor.

[0102] The specific calculation adopts the output signal optimization model algorithm, and its algorithm formula is as follows:

[0103] ;

[0104] in is the thermal noise term, and its specific calculation formula is as follows:

[0105] .

[0106] is a random process that obeys the standard normal distribution N(0,1), that is, Gaussian white noise with mean 0 and variance 1. , which is the Boltzmann constant, is the absolute temperature, is the equivalent noise resistance, is the system bandwidth;

[0107] in is the 1 / f noise term, and its specific calculation formula is as follows:

[0108] ;

[0109] in is the time constant, is the power law index and has a range of , is the normalization constant, f is the frequency, and the variables are replaced by ,Right now ;

[0110] in is the popcorn item, and its specific calculation formula is as follows:

[0111] ;

[0112] in, represents the total number of pulses at time t, is the time when the i-th pulse occurs, is the amplitude of the ith pulse, is a unit step function;

[0113] in is the drift term, and its specific calculation formula is as follows:

[0114] ;

[0115] in is the linear aging coefficient, Assign a value to the initial transient drift, is the aging stabilization time constant;

[0116] Further 、 、 as well as The influence weights of thermal noise, 1 / f noise, popcorn noise and drift noise are respectively , when comparing sensor noise not involved in the project, the corresponding impact weight =1, the corresponding influence weight of the sensor noise involved is the updated weight. When all three comparison items belong to thermal noise, the aging effect is = , corresponding to the influence weight of thermal noise That is, the product of the initial weights of each comparison item. It is worth noting that for the noise position, the initial weight after the product needs to be multiplied by the corresponding proportion to obtain the final updated influence weight , when the three comparison items belong to 1 / f noise, popcorn noise and drift noise respectively, , 、 as well as They correspond to the initial weights of the matched comparison items respectively, and the value of a single sensor noise is the single aging influence.

[0117] At the same time, in order to further improve the prediction efficiency, the intervals of various aging influence quantities are formulated in advance. , different aging impact ranges Corresponding to different aging degrees, and with the concentrated aging effect Perform numerical matching to obtain the influence interval of successful matching , obtain the aging degree corresponding to the current monitoring result, where the sensor noise with the highest single aging impact is marked as directly affecting the sensor noise, and warn of sensor failure based on the aging degree. It is worth noting that in the aging impact interval During the formulation process, it is necessary to determine it based on actual conditions, such as the model of the sensor to be tested and the environment in which the offshore platform dry-type transformer is located.

[0118] The present invention obtains the influence weight of each sensor noise in actual acquisition through the proportion of conventional noise positions and the signal-to-noise ratio change trend, and uses the influence weight to feedback the degree of influence of each sensor noise on sensor aging. The aging influence amount and the concentrated aging influence amount of each sensor noise are calculated, and the aging influence amount interval is matched to evaluate the aging degree of sensors in the current offshore platform dry-type transformer. This not only ensures the accuracy of the prediction results, but also reduces the monitoring work and reduces the difficulty of monitoring.

[0119] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the state and early warning of faults of dry-type transformers on offshore platforms, characterized in that: The steps include: S1. Monitor the sensor signal of the offshore platform dryer and obtain the noise position in the sensor signal. and the signal-to-noise ratio change ; The noise position in the sensor signal is obtained in S1 The method comprises the following steps: S1.

1. Divide the signal time domain time axis and use sliding window variance to locate the noise segment; S1.

2. Set the window width in the time domain signal, marked as W; S1.3, use the window variance algorithm to calculate the local variance near the time point t, marked as ; S1.

4. Define baseline variance ,when When , it is determined to be a noise interval, and the time range of the current noise interval is obtained as the position of the current noise in the current signal; S2. Collect historical sensor noise information and obtain the proportion of regular noise positions of each sensor noise and the signal-to-noise ratio variation trend; The ratio of the normal noise position of each sensor noise obtained in S2 The method comprises the following steps: S2.

1. Obtain the position of the same noise in the signal per unit time; S2.

2. Calculate the probability of the same noise occurring at different positions in the signal within the same batch; S2.

3. Establish probability thresholds; When the occurrence probability ≥ probability threshold, the noise position corresponding to the current occurrence probability is the regular noise position, and the occurrence probability is the actual proportion ; When the occurrence probability is less than the probability threshold, the noise position corresponding to the current occurrence probability is the abnormal position and is not used as a reference; S3, according to the proportion of conventional noise positions As well as the signal-to-noise ratio change trend, calculate the influence weight of each sensor noise in actual acquisition ; The calculation in S3 of the influence weight of each sensor noise in the actual acquisition The method comprises the following steps: S3.

1. Proportion of passing through conventional noise locations And the signal-to-noise ratio change trend prediction, formulate the comparison order of the current monitoring noise, specifically the increase range - range impact degree - the proportion of conventional noise position , as a comparison item; S3.

2. Assign initial weights to each comparison item. The values of the initial weights are different and are defined based on the impact of each comparison item on the corresponding noise. S3.

3. After completing the comparison of the three comparison items, select the initial weight of the sensor noise of the successfully compared items and perform the influence weight renew; When the sensor noises of the three successfully compared items are all different, the initial weights of the corresponding comparison items are obtained and combined and allocated; When the sensor noise of two successfully compared items is the same, the initial weights of the corresponding comparison items are obtained and multiplied to update the influence weight of the current sensor noise. , the remaining one comparison item that is successfully matched and has different noise from the above-mentioned sensor is still assigned an initial weight in a combined manner; When the sensor noise of three successfully compared items is the same, the initial weights of the corresponding comparison items are obtained and multiplied to update the influence weight of the current sensor noise. ; S4. Combine the influence weights of each sensor noise , calculate the aging effect of each sensor noise and the concentrated aging effect ; The S4 sensor noise includes thermal noise, 1 / f noise, popcorn noise and drift noise; Thermal noise is the noise caused by the thermal motion of carriers inside the conductor; 1 / f noise is the noise caused by carrier mobility fluctuations due to defects in the sensor material; Popcorn noise is the noise caused by carrier capture at semiconductor lattice defects in the sensor; Drift noise is the noise caused by the breakage / oxidation of chemical bonds in the sensitive material of the sensor; Calculation of concentrated aging influence in S4 The output signal optimization model algorithm is adopted, and its algorithm formula is as follows: ； in is the thermal noise term, and its specific calculation formula is as follows: ； To obey the standard normal distribution A random process, that is, a Gaussian white noise with a mean of 0 and a variance of 1, , which is the Boltzmann constant, is the absolute temperature, is the equivalent noise resistance, is the system bandwidth; in is the 1 / f noise term, and its specific calculation formula is as follows: ； in is the time constant, is the power law index and has a range of , is the normalization constant, f is the frequency, and the variables are replaced by ,Right now ; in is the popcorn item, and its specific calculation formula is as follows: ； in, represents the total number of pulses at time t, is the time when the i-th pulse occurs, is the amplitude of the ith pulse, is a unit step function; in is the drift term, and its specific calculation formula is as follows: ； in is the linear aging coefficient, Assign a value to the initial transient drift, is the aging stabilization time constant; 、 、 as well as The influence weights of thermal noise, 1 / f noise, popcorn noise and drift noise are respectively , when comparing sensor noise not involved in the project, the corresponding impact weight =1, the corresponding impact weight of the sensor noise involved is the updated weight; S5. Establish the aging impact interval , different aging impact ranges Corresponding to different degrees of aging; S6. Impact on concentrated aging Aging impact interval Match, obtain the aging degree of the current sensor, mark the direct impact on sensor noise, and warn of sensor failure based on the aging degree.

2. The offshore platform dry-type transformer condition monitoring and fault early warning method according to claim 1 is characterized by: The window variance algorithm in S1.3 is as follows: ； ； in The signal at time The sampling value of is the average value of the data in the window, W is the window width in the time domain signal, is the local variance around time point t.

3. The offshore platform dry-type transformer condition monitoring and fault early warning method according to claim 1 is characterized by: The S1 obtains the signal-to-noise ratio change in the sensor signal The method comprises the following steps: S1.

5. Get the signal power of the current signal and noise power ; S1.6, use the signal-to-noise ratio logarithmic algorithm to calculate the signal-to-noise ratio ; S1.

7. Combine the interval time T of the collected signal to obtain the change in signal-to-noise ratio .

4. The offshore platform dry-type transformer condition monitoring and fault early warning method according to claim 3 is characterized by: The signal-to-noise ratio logarithm algorithm in S1.6 is as follows: ； ； is the signal power of the current signal, is the noise power, is the signal-to-noise ratio, as well as is the signal-to-noise ratio of two adjacent signals collected at an interval of T, where T is the interval time.

5. The offshore platform dry-type transformer condition monitoring and fault early warning method according to claim 1 is characterized by: The signal-to-noise ratio change trend in S2 includes the range of increase and the degree of range impact; The range of the increase is the range of the signal-to-noise ratio change; The influence degree of the range is the change in the signal-to-noise ratio of the monitoring results on both sides of the adjacent 's changing trend.

Images