Intelligent control method and system for condensation module of transformer moisture absorber

Abstract

The application belongs to the technical field of transformers, and particularly relates to an intelligent control method and system for a condensing module of a transformer moisture absorber, which comprises the following steps: collecting original air humidity data, preprocessing the original air humidity data to obtain an extreme point sequence; determining a local amplitude oscillation index of a current extreme point; constructing a local extreme point sequence of the current extreme point, obtaining a trend sequence of the current extreme point, and determining a noise dominant factor of the current extreme point; obtaining an optimized sliding step length of the current extreme point; extracting a real change trend of the original air humidity data, and intelligently controlling the condensing module. The application analyzes the numerical characteristics of the extreme points and the numerical change characteristics between the extreme points, adapts the sliding step length of the current extreme point, overcomes the limitations of the fixed sliding step length, filters out the noise, and improves the real-time performance and accuracy of the intelligent control of the condensing module of the transformer moisture absorber.

Description

Technical Field

[0001] This invention relates to the field of transformer technology. More specifically, this invention relates to an intelligent control method and system for a transformer dehumidifier condensation module. Background Technology

[0002] As the core hub of the power transmission and distribution system, the dielectric strength and chemical stability of the insulating oil inside the power transformer are crucial to ensuring the safe operation of the power grid. During long-term operation, the transformer is affected by copper and iron losses caused by load current fluctuations and the temperature difference between day and night. The volume of the insulating oil will undergo continuous thermal expansion and contraction. In order to balance the pressure difference between the inside and outside of the oil tank, the transformer oil conservator is connected to the outside atmosphere through a humidifier to carry out frequent gas exchange, which is the transformer's breathing function. If the air entering the transformer contains a large amount of moisture, it will become damp, which will reduce the withstand voltage of the insulating oil and even cause a short circuit accident.

[0003] Therefore, modern high-performance dehumidifiers have gradually incorporated condensation dehumidification technology, attempting to force water vapor in the air to condense into liquid water and discharge it before the air passes through the silica gel tank, thus achieving maintenance-free active dehumidification. However, transformers are typically deployed in outdoor substations, operating in extremely harsh and complex environments. Humidity sensors installed at the dehumidifier's air inlet not only face normal humidity fluctuations but also constant interference from sudden environmental noises such as strong winds and heavy rain. These interferences cause the collected humidity data sequence to exhibit non-stationary and non-linear oscillations. Traditional control strategies, often based on simple threshold judgments or mean filtering, struggle to extract the true trend of environmental humidity changes from data heavily contaminated by noise. A local mean decomposition algorithm for signal time-frequency analysis adaptively decomposes complex signals into several product function PF components, effectively handling non-stationary signals and extracting the true breathing trend from noisy humidity data.

[0004] When iteratively decomposing non-stationary humidity data, the core steps of the local mean decomposition algorithm rely on constructing a local mean function and an envelope estimation function. Both of these require moving average processing of the signal's local extrema. In existing local mean decomposition algorithms, the step size of the moving average is usually set to the span between two adjacent extrema. This fixed and passive step size setting fails to perceive the signal's distribution density characteristics in the time domain. When outdoor environments experience heavy rainfall or high-frequency electromagnetic interference, the data collected by humidity sensors will be mixed with a large amount of high-frequency, large-amplitude impulse noise. This noise will cause extremely dense pseudo-extrema to emerge in the data waveform, drastically shortening the time interval between adjacent extrema. If the span between adjacent extrema is still used as the moving step size, the moving step size will become extremely small, causing the moving average operation to lose its intended smoothing and denoising capabilities. Instead, it will extremely accurately depict the noise profile, resulting in high-frequency noise being incorrectly retained in the decomposed low-frequency envelope, leading to severe mode mixing. Ultimately, these decomposition results containing false high humidity characteristics will misjudge the current ambient humidity, causing the condenser module to start frequently when dehumidification is not needed, or to shut down due to interference when dehumidification is needed. This not only causes huge energy waste, but also causes the semiconductor cooling chip to fail prematurely due to frequent thermal cycling stress.

[0005] Among existing related technologies, Chinese patent document CN104979078B, entitled "A Maintenance-Free Dehumidifier and Its Control Method," discloses a control system involving intelligent collaboration between a transformer and a dehumidifier. In the data processing stage, after receiving raw data from various sensors such as air humidity, the control module uses the traditional local mean decomposition method to remove noise from the raw data. Subsequently, it performs data feature extraction and transformer condensation and moisture absorption status assessment. When denoising air humidity data, this patent application uses a standard local mean decomposition denoising algorithm. Its sliding step size is usually a fixed parameter set based on global noise statistical characteristics. Its drawback is that it does not consider the non-stationary, non-Gaussian noise interference caused by sudden micro-airflow and strong electromagnetic effects on sensors under actual outdoor extreme temperature differences or high-voltage electromagnetic environments. When faced with complex transformer operating conditions, the fixed sliding step size is prone to misjudging sudden changes in real air humidity as noise filtering or misretaining strong environmental extreme noise as real signals, resulting in insufficient adaptive capability. Summary of the Invention

[0006] To address the technical problem that existing local mean decomposition algorithms in transformer desiccant condenser module control often misinterpret environmental noise interference as humidity change trends, leading to over-decomposition or signal distortion and failing to accurately extract the true trend, this invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides an intelligent control method for a transformer dehumidifier condensing module, comprising: collecting raw air humidity data, preprocessing it to obtain an extreme point sequence; determining the local amplitude oscillation index of the current extreme point based on the average absolute value of the difference between the current extreme point and its two adjacent extreme points and the second-order difference; constructing a local extreme point sequence of the current extreme point, obtaining a trend sequence of the current extreme point based on the local extreme point sequence, determining the noise dominance factor of the current extreme point based on the correlation between the local extreme point sequence and the trend sequence, combined with the sampling time interval between the current extreme point and its two adjacent extreme points; correcting the original sliding step size based on the noise dominance factor to obtain an optimized sliding step size for the current extreme point; using the optimized sliding step size to perform local mean decomposition on the raw air humidity data, extracting the true trend of change of the raw air humidity data, and intelligently controlling the condensing module based on the comparison result of the true trend of change with a preset threshold.

[0008] This invention highlights the amplitude characteristics of real humidity changes by calculating the local amplitude oscillation index and utilizing the difference between the average absolute value of the difference between extreme points and the second-order difference. It also identifies noise interference by calculating the noise-dominant factor, enhances the anti-interference capability of the local mean decomposition algorithm in complex environments by adaptively correcting the original sliding step size, and achieves quantitative evaluation of the transformer desiccant condenser module by extracting the real trend of original air humidity data based on the optimized sliding step size and performing intelligent control based on the comparison with a preset threshold. This accurately identifies humidity anomalies caused by environmental humidity, improving the accuracy and robustness of control.

[0009] Preferably, the acquisition of raw air humidity data includes: using a humidity sensor inside the dust cover of the air inlet of the transformer dehumidifier to acquire raw air humidity data.

[0010] Preferably, the preprocessing to obtain the extreme point sequence includes: for the collected raw air humidity data, constructing a data buffer of a preset length in real time, and performing extreme point search on all data points in the data buffer to obtain the extreme point sequence.

[0011] Preferably, the local amplitude oscillation index satisfies the expression: In the formula, This is the index of the current extreme point. For the first The local amplitude oscillation index at the current extreme point For the first The extreme point and the first The difference between the extreme points For the first The extreme point and the first The difference between the extreme points For the first The extreme point and the first The difference between the extreme points.

[0012] This invention achieves the evaluation of the local amplitude oscillation index by constructing a positive correlation function between the average absolute value of the difference between extreme points and the second-order difference term. The average absolute value of the difference between extreme points reflects the average level of amplitude change between extreme points, while the maximum value of the second-order difference term amplifies the response to noise. This results in extreme points that belong to the actual humidity change being calculated with a smaller local amplitude oscillation index, while extreme points affected by noise are calculated with a larger local amplitude oscillation index, thus providing a reliable basis for the calculation of the noise-dominant factor.

[0013] Preferably, the method for obtaining the trend sequence is as follows: select several extreme points with the smallest sampling time interval with the current extreme point, and construct a local extreme point sequence of the current extreme point with the current extreme point; perform linear fitting on all data points in the data buffer to obtain a trend function; substitute the sampling time of all extreme points in the local extreme point sequence into the trend function, and extract the trend values ​​corresponding to all extreme points in the local extreme point sequence to form a trend sequence.

[0014] Preferably, the noise dominance factor satisfies the expression: In the formula, This is the index of the current extreme point. For the first The noise-dominant factor at each extreme point For the first The index of local amplitude fluctuations at extreme points For the first The extreme point and the first The sampling time interval between each extreme point For the first The extreme point and the first The sampling time interval between each extreme point For the first A sequence of local extreme points of extreme points For the first Trend sequence of extreme points For covariance calculation, This is for variance calculation.

[0015] This invention constructs a composite function comprising a covariance term, a variance product ratio term, and a sampling time interval term to evaluate the noise dominance factor. The covariance term and the variance product ratio term reflect the correlation between the local extreme point sequence and the trend sequence, while the sampling time interval term enhances the interval contribution of the noise region. This results in extreme points affected by noise having a larger noise dominance factor, while extreme points belonging to the actual humidity change maintain a lower dominance factor, thereby effectively distinguishing between humidity trends and background noise.

[0016] Preferably, the optimized sliding step size satisfies the expression: In the formula, This is the index of the current extreme point. For the first The optimized sliding step size for each extreme point. For the first The original sliding step size for each extreme point This is the step gain coefficient. For the first The noise-dominant factor at each extreme point.

[0017] This invention achieves adaptive sliding step size by dynamically adjusting the original sliding step size. At extreme points with strong noise interference, the sliding step size is automatically increased to suppress misclassification, while at extreme points with significant humidity changes, a smaller step size is maintained to preserve the true response, thus ensuring the accuracy of extracting the true trend of humidity data.

[0018] Preferably, the extraction of the true trend of the original air humidity data includes: using a local mean decomposition algorithm combined with an optimized sliding step size to decompose the original air humidity data, obtaining several PF components and a residual component, discarding the first few high-frequency PF components, and superimposing the remaining PF components and the residual component to reconstruct the true trend of the original air humidity data.

[0019] Preferably, the intelligent control of the condensing module includes: immediately starting the condensing module to rated power in response to the actual trend value of the humidity data being greater than a preset condensing start threshold; and immediately shutting down the condensing module in response to the actual trend value of the humidity data not being greater than the preset condensing start threshold.

[0020] This invention achieves intelligent control of the condensing module by comparing the actual trend value of humidity data with a preset condensing start threshold. The actual trend value of humidity data reflects the moisture absorption state of the transformer and the humidity level of the environment, providing a reliable benchmark for the precise adjustment of the condensing module. This effectively prevents the risk of transformer insulation damage or internal corrosion caused by abnormal humidity, and ensures the long-term stable operation of high-voltage equipment.

[0021] Secondly, the present invention provides an intelligent control system for a transformer desiccant condensing module, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned intelligent control method for a transformer desiccant condensing module is implemented.

[0022] By adopting the above technical solution, a computer program is generated from the intelligent control method of the transformer desiccant condensation module and stored in the memory so that it can be loaded and executed by the processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0023] The beneficial effects of this invention are as follows: This invention solves the technical problem that traditional local mean decomposition algorithms easily misjudge environmental noise as humidity changes in the transformer dehumidifier environment by introducing an adaptive sliding step size correction mechanism based on multi-feature fusion, which leads to control distortion.

[0024] This invention establishes the intrinsic mapping relationship between the local amplitude oscillation index and the fluctuation affected by noise by analyzing the average absolute value of the difference between extreme points and the second difference. On this basis, it further determines the noise-dominant factor and tightly couples the index reflecting amplitude changes with the index reflecting sequence correlation, thereby realizing the identification of noise.

[0025] This invention adaptively corrects the original sliding step size for each extreme point, achieving accurate matching between the sliding step size and local data features. For extreme points affected by noise, the step size is automatically increased to prevent misclassification; for extreme points representing genuine humidity changes, the step size is maintained to preserve completeness. Ultimately, this invention improves the accuracy and anti-interference capability of humidity data processing in transformer dehumidifiers, enhances the robustness of condenser module control, and enables timely and accurate identification of humidity anomalies, providing a solid technical guarantee for transformer insulation performance maintenance and safe equipment operation. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating an intelligent control method for a transformer desiccant condensation module according to the present invention; Figure 2 This is a comparison diagram showing the effects of performing decomposition using the original sliding step size in the existing technology and the optimized sliding step size in this invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] This invention discloses an intelligent control method for a transformer dehumidifier condensation module, referring to... Figure 1 This includes steps S001 to S005, specifically: S001: Collect raw air humidity data, perform preprocessing, and obtain the extreme point sequence.

[0030] Specifically, a humidity sensor is used inside the dust cover of the air inlet of the transformer dehumidifier to collect raw air humidity data at a preset sampling frequency. For the collected raw air humidity data, a data buffer of a preset length is constructed in real time, and extreme point search is performed on all data points in the data buffer to obtain an extreme point sequence. In this embodiment, the sampling frequency is set to 10Hz and the length of the data buffer is set to 60 seconds. In other embodiments, the implementer can set the sampling frequency and the length of the data buffer according to the actual implementation situation.

[0031] S002: Determine the local amplitude oscillation index of the current extreme point based on the average of the absolute values ​​of the differences between the current extreme point and its two adjacent extreme points, as well as the second-order difference.

[0032] It should be noted that existing local mean decomposition algorithms, by pre-setting a fixed sliding step size when constructing the local mean function and envelope estimation function, cannot adaptively adjust the smoothing intensity according to the density of signal fluctuations when dealing with humidity data containing high-frequency burst noise. This causes the fitted envelope to over-fit noise peaks, resulting in distortion and an inability to accurately extract the true humidity change trend. Consequently, this severely affects the control accuracy of the condenser module. Based on signal time-frequency analysis theory and the physical inertial characteristics of transformer breathing, the change in the true humidity signal at extreme points should be continuous and gradual. However, noise interference can cause drastic amplitude fluctuations between adjacent extreme points, manifested in the waveform as sharp bends with large slope changes. Therefore, this invention determines the local amplitude oscillation index of the current extreme point to characterize the intensity of signal waveform oscillation at the current extreme point.

[0033] Specifically, the local amplitude oscillation index satisfies the expression: ; In the formula, This is the index of the current extreme point. For the first The index of local amplitude fluctuations at extreme points For the first The extreme point and the first The difference between the extreme points For the first The extreme point and the first The difference between the extreme points For the first The extreme point and the first The difference between the extreme points.

[0034] In the formula, The larger it is, the more likely it is to be the first The greater the average change in value between the first extreme point and its two adjacent extreme points, the more significant the change in the value between the first extreme point and its two adjacent extreme points. The more abrupt the value of the extreme point, the more it indicates that the... The greater the likelihood that an extreme point is affected by noise, the more likely the _th extreme point is to be affected by noise. The larger the local amplitude oscillation index at each extreme point, the greater the oscillation index. The larger it is, the more likely it is to be the first The more abrupt the waveform change at the extreme point, the more it indicates that the... The greater the probability that the extreme point is affected by noise, the greater the credibility. The larger the local amplitude oscillation index at each extreme point, the greater the oscillation index.

[0035] S003: Construct a local extreme point sequence for the current extreme point. Based on the local extreme point sequence, obtain the trend sequence for the current extreme point. According to the correlation between the local extreme point sequence and the trend sequence, and combined with the sampling time interval between the current extreme point and its two adjacent extreme points, determine the noise dominance factor of the current extreme point.

[0036] It should be noted that the violent air intake behavior caused by the cooling and contraction of insulating oil due to a sharp drop in transformer load, and the noise impact from raindrop interference in severe weather, both exhibit drastic amplitude jumps. This makes it difficult to accurately distinguish between the real signal and noise based solely on amplitude differences, leading to over-smoothing of the real breathing signal or incomplete noise removal. Based on the principles of physical inertia and signal randomness theory, the noise generated by raindrop impacts on the sensor is a high-frequency transient interference, with its extreme points distributed extremely densely along the time axis and exhibiting chaotic local trends. In contrast, the actual air intake behavior of the transformer is constrained by the inertia of airflow, resulting in relatively gradual humidity changes. This leads to a relatively sparse distribution of extreme points, with the direction of local fluctuations consistently closely matching the overall trend. Therefore, this invention determines the noise dominance factor at the current extreme point to characterize the probability that the current extreme point is dominated by high-frequency environmental noise.

[0037] Specifically, the 10 extreme points with the smallest sampling time interval from the current extreme point are selected, and a local extreme point sequence is constructed with the current extreme point. Linear fitting is performed on all data points in the data buffer to obtain a trend function. The sampling times of all extreme points in the local extreme point sequence are substituted into the trend function to obtain the trend values ​​corresponding to all extreme points in the local extreme point sequence, thus forming a trend sequence. In this embodiment, the number of extreme points with the smallest sampling time interval from the current extreme point is set to 10. In other embodiments, the implementer can set it according to the actual implementation situation.

[0038] Specifically, the noise-dominant factor satisfies the expression: ; In the formula, This is the index of the current extreme point. For the first The noise-dominant factor at each extreme point For the first The index of local amplitude fluctuations at extreme points For the first The extreme point and the first The sampling time interval between each extreme point For the first The extreme point and the first The sampling time interval between each extreme point For the first A sequence of local extreme points of extreme points For the first Trend sequence of extreme points For covariance calculation, This is for variance calculation.

[0039] In the formula, The larger it is, the more likely it is to be the first The greater the likelihood that an extreme point is affected by noise, the more likely the _th extreme point is to be affected by noise. The larger the noise dominance factor at each extreme point, the greater the noise dominance factor. The smaller the value, the better. The denser the distribution of extreme points at the nth extreme point, the more it indicates that the nth extreme point is... The closer the humidity change characteristics at the extreme points match the humidity change characteristics under noise influence, the more likely it is that the... The greater the likelihood that an extreme point is affected by noise, the more likely the _th extreme point is to be affected by noise. The larger the noise dominance factor at each extreme point, the greater the noise dominance factor. In essence, it is the first The absolute value of the Pearson correlation coefficient between the local extreme point sequence and the trend sequence of the nth extreme point reflects the nth extreme point. The similarity between the local extreme point sequence and the trend sequence of the nth extreme point is such that the smaller the value, the better. The more chaotic the local fluctuations at the extreme point, the more it indicates that the... The more the local fluctuations at the extreme point conform to the noise characteristics, the more the _th_ extreme point is considered to be _th_ extreme point. The greater the likelihood that an extreme point is affected by noise, the more likely the _th extreme point is to be affected by noise. The larger the noise dominance factor at each extreme point, the greater the noise dominance factor.

[0040] S004: The optimized sliding step size for the current extreme point is obtained by correcting the original sliding step size based on the noise dominance factor.

[0041] It should be noted that after obtaining the noise dominance factor of the current extreme point, this invention will calculate the optimized sliding step size of the current extreme point based on the noise dominance factor. The traditional local mean decomposition algorithm has a relatively fixed way of setting the sliding step size when constructing the local mean function and the envelope estimation function. This invention, however, uses the noise dominance factor to calculate the optimized sliding step size of the current extreme point, so that the local mean decomposition algorithm can filter out noise more accurately.

[0042] Specifically, the optimized sliding step size satisfies the expression: ; In the formula, This is the index of the current extreme point. For the first The optimized sliding step size for each extreme point. For the first The original sliding step size for each extreme point is taken as: and The larger of the two values, For the first The noise-dominant factor at each extreme point The step gain coefficient is set to 2.5 in this embodiment. In other embodiments, the implementer can set it according to the actual implementation situation. For example, when the transformer is in an extremely harsh environment and the ambient noise interference is large, the step gain coefficient can be appropriately increased to enhance the algorithm's ability to smoothly suppress high-frequency pulse noise and prevent the condensation module from malfunctioning. When the sensitivity requirement for capturing the weak breathing signal of the transformer is high, the step gain coefficient can be appropriately reduced to prevent the loss of real signal details due to excessive step stretching, thereby improving the accuracy of restoring the humidity change trend.

[0043] In the formula, The larger it is, the more likely it is to be the first The greater the probability that the extreme point is located in a strong noise pollution area, the more likely the th extreme point is to be located in a strong noise pollution area. The larger the sliding step size after optimization of each extreme point, the more likely noise will be completely filtered out.

[0044] S005: The local mean decomposition of the original air humidity data is performed using the optimized sliding step size to extract the true trend of the original air humidity data, and the condensation module is intelligently controlled based on the comparison between the true trend and the preset threshold.

[0045] Specifically, intelligent control of the condensation module includes: The original air humidity data is decomposed using a local mean decomposition algorithm with preset parameters and an optimized sliding step size to obtain several PF components and one residual component. The first two high-frequency PF components are discarded, and the remaining PF components are superimposed with the residual component to reconstruct the true trend of the original air humidity data. In this embodiment, the convergence threshold for terminating the sliding average iteration in the local mean decomposition algorithm is set to 0.001, the maximum number of filtering iterations is set to 15, and the number of high-frequency PF components discarded before reconstruction is set to 2. In other embodiments, implementers can set the convergence threshold for terminating the sliding average iteration, the maximum number of filtering iterations, and the number of high-frequency PF components discarded before reconstruction according to the actual implementation situation. If the actual trend value of the original air humidity data is greater than the preset condensation start threshold, the transformer is absorbing moisture and the environment is humid, so the condensation module is immediately started to the rated power; if the actual trend value of the original air humidity data is not greater than the preset condensation start threshold, the transformer stops absorbing moisture and the environment is dry, so the condensation module is immediately shut down.

[0046] like Figure 2 As shown in the figure, this diagram compares the effectiveness of the existing technology using the original sliding step size and the optimized sliding step size of this invention in extracting humidity change trends under environmental noise interference. The horizontal axis represents sampling time, the vertical axis represents humidity value, and the shaded area in the background indicates the time period of noise interference. Within the noise interference range, the curve corresponding to the existing technology exhibits severe oscillations, incorrectly following the noise waveform and deviating significantly from the true trend, because the fixed step size cannot adapt to high-density noise extreme points. In contrast, the curve corresponding to this invention benefits from the dynamic correction mechanism of the noise-dominant factor on the sliding step size, automatically increasing the sliding step size when noise interference is detected, significantly enhancing the smoothing effect. The results show that the curve corresponding to this invention remains stable even in strong noise regions, accurately coinciding with the true humidity trend. This proves that this invention can effectively isolate environmental interference, accurately restore the transformer's true humidification behavior, and provide a reliable basis for the precise control of the condensation module.

[0047] This invention also discloses an intelligent control system for a transformer desiccant condensing module, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent control method for a transformer desiccant condensing module according to the present invention is implemented.

[0048] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. An intelligent control method for a transformer dehumidifier condensation module, characterized in that, include: Raw air humidity data is collected, preprocessed, and a sequence of extreme points is obtained. The local amplitude oscillation index of the current extreme point is determined based on the average of the absolute values ​​of the differences between the current extreme point and its two adjacent extreme points, as well as the second difference. Construct a local extreme point sequence for the current extreme point, obtain a trend sequence for the current extreme point based on the local extreme point sequence, and determine the noise dominance factor for the current extreme point based on the correlation between the local extreme point sequence and the trend sequence, combined with the sampling time interval between the current extreme point and its two adjacent extreme points. The optimized sliding step size for the current extreme point is obtained by correcting the original sliding step size based on the noise dominance factor. The optimized sliding step size is used to perform local mean decomposition on the original air humidity data to extract the true trend of change in the original air humidity data, and the condensation module is intelligently controlled based on the comparison result of the true trend of change with the preset threshold.

2. The intelligent control method for a transformer dehumidifier condensation module according to claim 1, characterized in that, The process of collecting raw air humidity data includes: using a humidity sensor inside the dust cover of the air inlet of the transformer dehumidifier to collect raw air humidity data.

3. The intelligent control method for a transformer dehumidifier condensation module according to claim 1, characterized in that, The preprocessing to obtain the extreme point sequence includes: for the collected raw air humidity data, constructing a data buffer of a preset length in real time, and performing extreme point search on all data points in the data buffer to obtain the extreme point sequence.

4. The intelligent control method for a transformer dehumidifier condensation module according to claim 1, characterized in that, The local amplitude oscillation index satisfies the expression: ； In the formula, This is the index of the current extreme point. For the first The local amplitude oscillation index at the current extreme point For the first The extreme point and the first The difference between the extreme points For the first The extreme point and the first The difference between the extreme points For the first The extreme point and the first The difference between the extreme points.

5. The intelligent control method for a transformer dehumidifier condensation module according to claim 3, characterized in that, The method for obtaining the trend sequence is as follows: select several extreme points with the smallest sampling time interval with the current extreme point, and construct a local extreme point sequence of the current extreme point with the current extreme point; perform linear fitting on all data points in the data buffer to obtain the trend function; substitute the sampling time of all extreme points in the local extreme point sequence into the trend function, extract the trend value corresponding to all extreme points in the local extreme point sequence, and form the trend sequence.

6. The intelligent control method for a transformer dehumidifier condensation module according to claim 1, characterized in that, The noise dominance factor satisfies the expression: ； In the formula, This is the index of the current extreme point. For the first The noise-dominant factor at each extreme point For the first The index of local amplitude fluctuations at extreme points For the first The extreme point and the first The sampling time interval between each extreme point For the first The extreme point and the first The sampling time interval between each extreme point For the first A sequence of local extreme points of extreme points For the first Trend sequence of extreme points For covariance calculation, This is for variance calculation.

7. The intelligent control method for a transformer dehumidifier condensation module according to claim 1, characterized in that, The optimized sliding step size satisfies the expression: ； In the formula, This is the index of the current extreme point. For the first The optimized sliding step size for each extreme point. For the first The original sliding step size for each extreme point This is the step gain coefficient. For the first The noise-dominant factor at each extreme point.

8. The intelligent control method for a transformer dehumidifier condensation module according to claim 1, characterized in that, The extraction of the true trend of change in the original air humidity data includes: using the local mean decomposition algorithm combined with the optimized sliding step size to decompose the original air humidity data, obtaining several PF components and one residual component, discarding the first few high-frequency PF components, and superimposing the remaining PF components and the residual component to reconstruct the true trend of change in the original air humidity data.

9. The intelligent control method for a transformer dehumidifier condensation module according to claim 1, characterized in that, The intelligent control of the condensing module includes: immediately starting the condensing module to rated power when the actual change trend value of the original air humidity data is greater than the preset condensing start threshold; and immediately shutting down the condensing module when the actual change trend value of the original air humidity data is not greater than the preset condensing start threshold.

10. An intelligent control system for a transformer dehumidifier condensation module, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement an intelligent control method for a transformer desiccant condensing module according to any one of claims 1-9.

Images