Multi-sensor fusion electrical equipment control state monitoring method and system

Abstract

The invention discloses a multi-sensor fusion electrical equipment control state monitoring method and system, and relates to the technical field of data processing, and the method comprises the steps: obtaining a first monitoring time period, presetting a data processing time window, obtaining a second monitoring time period, and obtaining a state data sequence; obtaining the ith pre-data weight and the ith trend coefficient corresponding to the jth sensor; the ith pre-data weight of the jth sensor is corrected according to the ith trend coefficient of each sensor, and the ith target data weight of the jth sensor is formed; and obtaining the state weight of the jth sensor according to the plurality of target data weights of the jth sensor, and obtaining the real-time state index of the electrical equipment according to the real-time state data collected by each sensor and the state weight of each sensor. The method has the advantages of reliability optimization, multi-dimensional false alarm suppression and trend collaborative verification.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a multi-sensor fusion electrical equipment control status monitoring method and system. Background Art

[0002] In the field of electrical equipment status monitoring, although multi-sensor fusion technology has been widely used in the collaborative perception of parameters such as temperature, vibration, and current, the existing multi-sensor fusion electrical equipment control status monitoring process under complex working conditions has defects.

[0003] First, the physical characteristics of sensor reliability decaying over time are not considered, such as the cumulative effect of measurement deviations caused by aging in sensors that have been in operation for a long time. At the same time, the differences in data stability between different sensors under abnormal working conditions may be homogenized. Dynamic correlations between multiple physical quantities, such as a sudden increase in current accompanied by a temperature rise trend, are not considered. Existing methods do not establish such a cross-dimensional trend consistency verification mechanism and cannot introduce conflicting information through a weight redistribution mechanism. Second, the sensor's attenuation error continuously affects decision-making, and the impact of sensor maintenance timeliness on data reliability is not considered. At the same time, data anomaly determination relies too much on single-point threshold comparison. When some sensors experience boundary fluctuations (such as temperature oscillations near the safety threshold), the existing static correlation model cannot effectively identify trend contradictions between multiple sensors, resulting in an increase in the false alarm rate. Therefore, new monitoring methods are urgently needed to overcome the technical problems of existing monitoring processes in sensor reliability optimization and false alarm suppression. Summary of the Invention

[0004] In view of the defects in the prior art, the present invention provides a multi-sensor fusion method and system for monitoring the control status of electrical equipment.

[0005] A multi-sensor fusion electrical equipment control status monitoring method includes: obtaining a first monitoring time period of the electrical equipment, and presetting a data processing time window, and dividing the first monitoring time period into multiple second monitoring time periods according to the preset data processing time window, and obtaining a state data sequence collected by each sensor in each second monitoring time period; obtaining an i-th pre-data weight and an i-th trend coefficient corresponding to the j-th sensor according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period; correcting the i-th pre-data weight of the j-th sensor according to the i-th trend coefficient of each sensor, and forming an i-th target data weight of the j-th sensor; obtaining a state weight of the j-th sensor according to the multiple target data weights of the j-th sensor, and obtaining a real-time state indicator of the electrical equipment according to the real-time state data collected by each sensor and the state weight of each sensor.

[0006] Optionally, obtaining the i-th pre-data weight corresponding to the j-th sensor based on the status data sequence collected by the j-th sensor in the i-th second monitoring time period includes: obtaining the last maintenance time point corresponding to the j-th sensor, and obtaining the proxy time point of the i-th second monitoring time period; obtaining the time attenuation factor of the j-th sensor in the i-th second monitoring time period based on the last maintenance time point and the proxy time point of the j-th sensor; obtaining the i-th pre-data weight corresponding to the j-th sensor based on the status data sequence and the time attenuation factor collected by the j-th sensor in the i-th second monitoring time period.

[0007] Optionally, the i-th pre-data weight corresponding to the j-th sensor is obtained according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period as follows: Among them, W pij is the i-th pre-data weight corresponding to the j-th sensor, m is the number of sensors, λ is the time decay factor, n ij is the number of data in the state data sequence collected by the j-th sensor in the i-th second monitoring time period, is the k+1th data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, is the kth data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, X jmax is the upper limit of the acquisition range of the jth sensor, X jmin is the lower limit of the acquisition range of the jth sensor.

[0008] Optionally, the i-th trend coefficient corresponding to the j-th sensor is obtained according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period and is expressed as:

[0009] Among them, C ij is the i-th trend coefficient corresponding to the j-th sensor, n ij is the number of data in the state data sequence collected by the j-th sensor in the i-th second monitoring time period, is the k+1th data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, is the kth data in the state data sequence collected by the jth sensor in the i-th second monitoring time period.

[0010] Optionally, the i-th predicted data weight of the j-th sensor is modified according to the i-th trend coefficient of each sensor as follows: Among them, W tijis the weight of the i-th target data corresponding to the j-th sensor, W pij is the i-th pre-data weight corresponding to the j-th sensor, W aj is the weight adjustment value corresponding to the jth sensor, m is the number of sensors, C ij is the i-th trend coefficient corresponding to the j-th sensor, C iv is the i-th trend coefficient corresponding to the v-th sensor.

[0011] Optionally, obtaining the real-time status indicators of electrical equipment based on the real-time status data collected by each sensor and the status weight of each sensor includes: obtaining the data security threshold of each sensor; obtaining the real-time status indicators of electrical equipment based on the real-time status data collected by each sensor, the status weight of each sensor and the data security threshold.

[0012] Optionally, the real-time status index of the electrical equipment is obtained based on the real-time status data collected by each sensor and the status weight of each sensor, and is expressed as:

[0013] Among them, I S is the real-time status indicator of the electrical equipment, m is the number of sensors, W j is the state weight of the jth sensor, W c is the state weight of the cth sensor, X nj is the real-time status data collected by the jth sensor, X jth is the data security threshold of the j-th sensor.

[0014] A multi-sensor fusion electrical equipment control status monitoring system is also provided, and the system includes: a data acquisition module, which is used to obtain a first monitoring time period of the electrical equipment, and preset a data processing time window, and divide the first monitoring time period into multiple second monitoring time periods according to the preset data processing time window, and obtain the status data sequence collected by each sensor in each second monitoring time period; a first data processing module, which is used to obtain the i-th pre-data weight and i-th trend coefficient corresponding to the j-th sensor based on the status data sequence collected by the j-th sensor in the i-th second monitoring time period; a second data processing module, which is used to correct the i-th pre-data weight of the j-th sensor according to the i-th trend coefficient of each sensor, and form the i-th target data weight of the j-th sensor; a third data processing module, which is used to obtain the status weight of the j-th sensor based on the multiple target data weights of the j-th sensor, and obtain the real-time status index of the electrical equipment according to the real-time status data collected by each sensor and the status weight of each sensor.

[0015] Optionally, the first data processing module is also used to: obtain the last maintenance time point corresponding to the jth sensor, and obtain the proxy time point of the i-th second monitoring time period; obtain the time attenuation factor of the j-th sensor in the i-th second monitoring time period based on the last maintenance time point and the proxy time point of the j-th sensor; obtain the i-th pre-data weight corresponding to the j-th sensor based on the state data sequence and time attenuation factor collected by the j-th sensor in the i-th second monitoring time period.

[0016] Optionally, the third data processing module is further used to: obtain the data security threshold of each sensor; and obtain the real-time status index of the electrical equipment based on the real-time status data collected by each sensor, the status weight of each sensor and the data security threshold.

[0017] The beneficial effects of the present invention are embodied in:

[0018] In the entire multi-sensor fusion electrical equipment control status monitoring, first of all, the data preprocessing mechanism based on time window segmentation realizes the refined capture of equipment operation characteristics, and takes into account response capture and data processing by adjusting the window length, ensuring the extraction of effective multi-sensor data under different working conditions (such as equipment start and stop, load mutation, etc.); further, a dual evaluation system of time decay and trend consistency verification is introduced, and a dynamic correlation model across physical quantities is constructed on the basis of quantifying the reliability of individual sensors. It can not only dynamically correct the data credibility of aging sensors according to maintenance timeliness, but also identify contradictory signals in the sensor group through trend coefficient matching, such as abnormal working conditions where current surges but there is no temperature rise response, so as to To a certain extent, it can distinguish sensor failures from real equipment abnormalities; further, the dynamic weight correction mechanism retains the sensor's own data quality assessment results through the coupling calculation of pre-data weights and trend verification increments, and strengthens the collaborative decision-making ability of multi-source data through group trend consistency verification, so that the influence of single sensor drift errors is effectively suppressed, while the multi-parameter abnormalities caused by real faults are significantly amplified through the weight superposition effect; further, the status indicator design completely eliminates the risk of underreporting caused by the mutual offset of data in the existing threshold comparison method, and by aggregating the exceeding data and allocating the contribution according to the weight, the indicator value strictly reflects the degree of deviation from the equipment health status and identifies the urgency of sudden failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly describes the drawings required for the specific embodiments or the description of the prior art. Similar elements or parts are generally identified by similar reference numerals throughout the drawings. Elements or parts in the drawings are not necessarily drawn to scale.

[0020] Figure 1A schematic diagram of the steps of a multi-sensor fusion electrical equipment control state monitoring method according to one embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of a portion of the steps in S2 of the multi-sensor fusion electrical equipment control status monitoring method of the present invention;

[0022] Figure 3 This is a schematic diagram of part of step S4 in the multi-sensor fusion electrical equipment control status monitoring method of the present invention. DETAILED DESCRIPTION

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0024] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. In addition, the terms "first," "second," etc. are used only to distinguish the descriptions and are not to be understood as indicating or implying relative importance.

[0026] like Figure 1 As shown, a multi-sensor fusion electrical equipment control state monitoring method is provided, comprising:

[0027] S1. Obtain a first monitoring time period of the electrical equipment, preset a data processing time window, divide the first monitoring time period into multiple second monitoring time periods according to the preset data processing time window, and obtain a state data sequence collected by each sensor in each second monitoring time period;

[0028] S2. Obtaining the i-th pre-data weight and the i-th trend coefficient corresponding to the j-th sensor according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period;

[0029] S3. Modify the i-th predicted data weight of the j-th sensor according to the i-th trend coefficient of each sensor, and form the i-th target data weight of the j-th sensor;

[0030] S4. Obtain the state weight of the jth sensor according to the weights of multiple target data of the jth sensor, and obtain the real-time state index of the electrical equipment according to the real-time state data collected by each sensor and the state weight of each sensor.

[0031] In this embodiment, it should be noted that, in S1, a refined analysis of the electrical equipment status data is achieved through dynamic segmentation of time windows. First, the continuous first monitoring time period is divided into a plurality of second monitoring time periods of equal length according to the preset data processing time window, and the data in each sub-window will be independently subjected to feature extraction and correlation verification. This segmentation method can effectively capture the transient response characteristics of the equipment under different working conditions, such as parameter transitions in the equipment startup phase, periodic fluctuations during steady-state operation, or abnormal disturbances caused by sudden loads. By dividing the time window, the trend fuzziness problem caused by the homogenization of long-period data can be avoided, and the basis for local consistency analysis is provided for the subsequent dynamic adjustment of sensor weights. For example, for the monitoring of high-voltage circuit breakers, the 24-hour continuous operation data can be divided into 15-minute windows, and the coordinated change patterns of parameters such as temperature, current, and mechanical vibration in each window are analyzed separately.

[0032] Furthermore, the division of time windows needs to be adaptively adjusted in combination with the actual operating characteristics of the equipment and the sampling frequency of the sensor. Taking the gearbox monitoring of wind turbines as an example, under conditions of severe wind speed fluctuations, the length of the second monitoring period can be automatically shortened to 30 seconds to capture the instantaneous correlation between gear meshing vibration and oil temperature changes with higher time resolution; while in periods of stable wind speeds, the window is extended to 5 minutes to reduce data processing overhead. Within each time window, the raw signals of multi-source sensors are synchronously collected and a state data sequence aligned by timestamp is formed to ensure the spatiotemporal alignment basis for subsequent trend consistency analysis. For example, in transformer winding temperature monitoring, the top oil temperature, hot spot temperature and cooler flow data within the same window will be synchronously extracted to identify abnormal conditions of local overheating.

[0033] In S2, dual evaluation realizes the quantification of sensor data quality and trend characteristics. First, for the sensor data sequence in each time window, a time attenuation factor is constructed in combination with the maintenance timeliness. This factor increases nonlinearly with the extension of the sensor service life, effectively reflecting the cumulative effect of measurement deviations caused by aging. For example, in the pitch monitoring of wind turbines, the yaw vibration sensor that has been maintained recently will obtain a smaller attenuation factor, while the blade strain sensor that has not been calibrated for a long time will be assigned a larger attenuation value, ensuring that historical maintenance records affect the data credibility assessment in real time. At the same time, by calculating the normalized ratio of the fluctuation amplitude of adjacent data points in the window to the sensor range, the degree of abnormal data fluctuation is quantified - when the current sensor experiences a sharp jump under a short-term overload condition, its pre-data weight will be significantly reduced to avoid excessive impact of transient interference on the overall assessment.

[0034] Furthermore, trend coefficient extraction captures the overall direction of change in the data sequence and establishes a basic criterion for cross-sensor trend association. For example, in transformer oil temperature monitoring, if the top oil temperature sensor within a window shows a continuous upward trend, while the bottom oil temperature sensor shows a downward trend, this contradictory trend pattern will be identified. When a sudden increase in current is positively correlated with an upward trend in oil temperature, the consistency of the trend coefficient will increase the weight of the relevant sensor; conversely, if the cooling water pump flow sensor fails to detect an increasing flow trend when the temperature rises, its contradictory signal will be marked by the opposite sign of the trend coefficient, triggering a weight correction mechanism. This dual assessment not only takes into account the degradation of individual sensor reliability, but also establishes a verification framework for the dynamic association of multiple physical quantities.

[0035] In S3, a dynamic weight correction mechanism is used to identify and redistribute weights when trends conflict across multiple sensors. First, cross-sensor data correlation is assessed based on the sign matching of trend coefficients. When a sensor's trend direction conflicts with that of the majority of sensors, the sensor's anomaly weight is amplified through weighted analysis. For example, in high-voltage switchgear monitoring, if a phase current sensor shows a sustained upward trend while the corresponding busbar temperature sensor shows a downward trend, this violation of the physical law of the positive correlation between current and temperature is identified as a trend conflict. The target weight of the current sensor is automatically increased, allowing subsequent analysis to focus more on this anomalous data and facilitating the identification of localized overheating hazards caused by abnormal contact resistance. Furthermore, when multiple sensors demonstrate synergistic trends (e.g., increased vibration accompanied by rising oil temperature), their weights are positively enhanced through consistency verification, increasing the decision weight of normal operating data.

[0036] Furthermore, the correction process employs a relative weight adjustment strategy, preserving the sensor's own reliability assessment results (pre-data weights) while also incorporating incremental adjustments based on group trend verification. For example, in motor bearing monitoring, if a vibration sensor detects a shock pulse but the temperature sensor shows no abnormal temperature rise, the target weight of the vibration sensor is dynamically increased by calculating the matching degree between the vibration sensor's trend coefficient and the trends of other sensors, such as temperature and speed. This mechanism ensures that isolated outliers caused by single sensor failures or environmental interference do not excessively influence the overall judgment. Meanwhile, multi-parameter trend anomalies caused by true equipment failures are significantly amplified through the weight stacking effect, effectively distinguishing random fluctuations from true faults. For example, when a cooling fan failure causes an abnormal increase in transformer oil temperature, the negative correlation between the oil temperature trend and the fan speed trend triggers a significant adjustment to the oil temperature sensor's weight, giving the temperature rise data a higher weight in the comprehensive assessment.

[0037] In S4, a weighted aggregation mechanism enables precise monitoring of equipment status. First, safety thresholds for each sensor are dynamically set based on equipment design parameters and historical operating data. For example, the transformer top oil temperature threshold is determined by combining the insulation material's heat resistance rating with an ambient temperature compensation model, while the gearbox vibration threshold is adaptively adjusted based on the spectral characteristics under different speed conditions. During the real-time evaluation phase, when a sensor's monitored value exceeds the safety threshold, the magnitude of the excess is weighted according to the sensor's status weight. Sensors with recent maintenance and consistent trends are more reliable, and their excess data contributes significantly to the overall indicator. When calculating the status indicator, only positive deviations exceeding the safety threshold are collected. By normalizing the real-time excess magnitude of each sensor with its status weight and then performing a weighted accumulation, a non-negative status indicator is generated. For example, when monitoring partial discharge at a cable terminal, if an ultrasonic sensor detects a discharge signal but does not reach the threshold, its data is not included in the indicator calculation. However, if a UHF sensor detects a pulse current exceeding the threshold, its high weight directly drives the status indicator up significantly, effectively reflecting the severity of insulation degradation.

[0038] Furthermore, the status indicators are set up so that they form a strict positive correlation mapping relationship with the health status of the equipment. When the status indicator value exceeds the zero threshold, it means that an abnormality has been detected in the control status of the monitored electrical equipment. For example, in the monitoring of generator slip rings, if the carbon brush temperature sensor shows a slight excess value but the vibration sensor value is normal, the status indicator will only reflect the value corresponding to the temperature anomaly; when the spark sensor and the vibration sensor simultaneously detect violent discharge and mechanical impact, the excess data of the two will produce a superposition effect after weight amplification, causing the status indicator to grow rapidly, accurately representing the urgency of the ring fire fault. This mechanism not only avoids the defect of multi-sensor data offsetting each other in traditional methods, but also can identify sudden faults through the rate of indicator increase - for example, when the dielectric loss of the transformer bushing exceeds the standard, the speed at which the indicator value jumps from 0 to the critical value can reflect the acceleration process of insulating oil cracking.

[0039] In summary, in the entire multi-sensor fusion electrical equipment control status monitoring, first, the data preprocessing mechanism based on time window segmentation realizes the refined capture of equipment operation characteristics, and takes into account response capture and data processing by adjusting the window length, ensuring the extraction of effective multi-sensor data under different working conditions (such as equipment start and stop, load mutation, etc.); further, a dual evaluation system of time decay and trend consistency verification is introduced, and a dynamic correlation model across physical quantities is constructed on the basis of quantifying the reliability of individual sensors. It can not only dynamically correct the data credibility of aging sensors according to maintenance timeliness, but also identify contradictory signals in the sensor group through trend coefficient matching, such as abnormal working conditions where the current surges but there is no temperature rise response, so as to To a certain extent, it can distinguish sensor failures from real equipment anomalies. Furthermore, the dynamic weight correction mechanism retains the sensor's own data quality assessment results through the coupled calculation of pre-data weights and trend verification increments, and strengthens the collaborative decision-making ability of multi-source data through group trend consistency verification, so that the impact of single sensor drift errors is effectively suppressed, while the multi-parameter anomalies caused by real failures are significantly amplified through the weight superposition effect. Furthermore, the status indicator design completely eliminates the risk of underreporting caused by the mutual offset of data in the existing threshold comparison method. By aggregating excess data and allocating contributions based on weights, the indicator value strictly reflects the degree of deviation from the equipment's health status and identifies the urgency of sudden failures. In summary, the entire solution performs particularly well in false alarm suppression. Its trend contradiction detection function can effectively filter out isolated abnormal signals caused by environmental interference or sensor boundary fluctuations. At the same time, the maintenance of the timeliness compensation mechanism greatly reduces the probability of misjudgment of sensors that have not been calibrated in time.

[0040] like Figure 2 As shown, in one embodiment, obtaining the i-th pre-data weight corresponding to the j-th sensor according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period in S2 includes:

[0041] S21. Obtain the last maintenance time point corresponding to the j-th sensor, and obtain the proxy time point of the i-th second monitoring time period;

[0042] S22. Obtaining a time attenuation factor of the jth sensor in the i-th second monitoring time period according to the last maintenance time point and the proxy time point of the jth sensor;

[0043] S23 . Obtain an i th pre-data weight corresponding to the j th sensor according to the state data sequence collected by the j th sensor in the i th second monitoring time period and the time attenuation factor.

[0044] In this embodiment, it should be noted that in S21, a sensor reliability baseline is established through maintenance time tracking. The timestamp of the last maintenance of each sensor is recorded, and the time decay effect is calculated in combination with the time midpoint (proxy time point) of the current monitoring window. For example, in the wind turbine pitch monitoring, the maintenance timestamp of the recently replaced gearbox vibration sensor is closer to the current monitoring window, and its time decay effect is weaker; while the generator winding temperature sensor that has not been calibrated for a long time, its maintenance time is several months away from the current window, and its potential aging risk will be automatically identified. This mechanism converts operation and maintenance records into quantitative parameters, providing an objective basis for subsequent weight calculations.

[0045] In S22, a time-attenuation factor is used to dynamically compensate for sensor drift errors. For example, λ = 2-exp[-(proxy time point - last maintenance time point)], where λ ranges between 1 and 2. The time-attenuation factor varies nonlinearly with the maintenance interval. For example, in high-voltage circuit breaker contact temperature monitoring, a newly calibrated sensor has an attenuation factor close to a minimum of 1 within the first monitoring window. Over time, when the sensor enters the 10th monitoring window, the attenuation factor increases as the maintenance interval lengthens. This, to a certain extent, addresses the error accumulation problem caused by sensor aging in existing methods.

[0046] In S23, the timeliness characteristics and data quality are integrated to achieve dynamic pre-weight allocation. When calculating the pre-weight, the dual effects of the time decay factor and the real-time data fluctuation characteristics are simultaneously considered. For example, in transformer oil chromatography monitoring, although a hydrogen sensor that has not been maintained in time shows stable data, its basic weight will be compressed due to the large time decay factor. On the other hand, even if another acetylene sensor that has been recently calibrated experiences brief fluctuations, it can still maintain a high weight due to its small time decay factor and the fluctuation amplitude is within a reasonable range. This mechanism is particularly critical in the monitoring of relay cabinets in nuclear power plants. It can tolerate occasional electromagnetic interference fluctuations of sensors and implement preventive weight reduction for sensors that have not been maintained for a long time.

[0047] In one embodiment, in S2, the i-th pre-data weight corresponding to the j-th sensor is obtained according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period, and is expressed as:

[0048] in,

[0049] W pij is the i-th pre-data weight corresponding to the j-th sensor, m is the number of sensors, λ is the time decay factor, n ij is the number of data in the state data sequence collected by the j-th sensor in the i-th second monitoring time period, is the k+1th data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, is the kth data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, X jmax is the upper limit of the acquisition range of the jth sensor, X jmin is the lower limit of the acquisition range of the jth sensor.

[0050] In this embodiment, it should be noted that It is used for the basic weight distribution, which is used to distribute the initial weights equally and ensure that all sensors have the same reference.

[0051] Further, is the normalized fluctuation amplitude; where the numerator Used to calculate the sum of the absolute differences of adjacent data points and quantify the volatility of the sequence. In the denominator, (n ij -1) can be seen as converting the sum of fluctuations into the average value of the total number of adjacent differences, and on the other hand, it can be seen as converting the denominator (X jmax -X jmin ) is increased by a multiple of the total number of adjacent differences to average out the effects of different dimensions. The normalized result is that the fluctuation amplitude is compressed to [0, 1] to achieve cross-sensor comparability.

[0052] Furthermore, the time decay factor λ nonlinearly amplifies or reduces the impact of fluctuations. That is, the longer the maintenance period, the larger λ, the stronger the exponential decay effect, and the more significant the weight reduction. At the same time, exp{-λ·normalized fluctuation amplitude} is used to map normalized fluctuations to weight decay, ensuring that the greater the fluctuation and the longer the maintenance, the lower the weight.

[0053] In summary, sensor aging compensation is achieved, directly linked to maintenance timeliness. The longer the maintenance interval, the larger the λ value. For example, if the maintenance interval of a current sensor exceeds 3 months, λ increases from 1.2 to 1.8, and the weight is reduced by 13% under the same data fluctuation, suppressing the accumulation of aging errors. Furthermore, the normalized fluctuation calculation avoids homogenization processing while achieving cross-sensor comparability. For example, the same percentage fluctuation of a vibration sensor (range 0-10mm / s) and a current sensor (0-100A) has an equivalent weight penalty. Therefore, by quantifying the time correlation of fluctuations and normalizing them across sensors, the entire expression transforms abstract issues such as sensor reliability degradation, range differences, and maintenance timeliness into computable pre-data weights. This enables accurate identification and suppression of abnormal signals before data fusion, laying the foundation for subsequent trend consistency verification.

[0054] In one embodiment, in S2, the i-th trend coefficient corresponding to the j-th sensor is obtained according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period, and is expressed as:

[0055] in,

[0056] C ij is the i-th trend coefficient corresponding to the j-th sensor, n ij is the number of data in the state data sequence collected by the j-th sensor in the i-th second monitoring time period, is the k+1th data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, is the kth data in the state data sequence collected by the jth sensor in the i-th second monitoring time period.

[0057] In this embodiment, it should be noted that For adjacent difference calculation, the difference of all adjacent data points of the sensor in the time window is calculated to quantify the instantaneous change direction and intensity; for example, if the data sequence is [25, 26, 27, 28, 29°C], the adjacent difference is +1°C (26-25), +1°C (27-26), etc. Further, all adjacent differences are accumulated to obtain the trend intensity. The sum of positive values ​​indicates an overall upward trend, and the sum of negative values ​​indicates a downward trend; for example, if the difference is [1, 1, 1, 1], the sum is 4, indicating a continuous increase. Further, Mapping is achieved, that is, the cumulative sum is mapped to discrete trend directions: 1 means the sum is greater than 0, indicating an overall upward trend; -1 means the sum is less than 0, indicating an overall downward trend; 0 means the sum is equal to 0, indicating no trend (rarely occurs).

[0058] In summary, this method enables cross-dimensional trend correlation verification, ignoring the specific magnitude of change and focusing solely on the trend direction. This allows for horizontal comparison of sensors with different dimensions (e.g., current A and temperature °C). For example, if a sudden increase in current (trend 1) is accompanied by a temperature rise (trend 1), it is considered a physical correlation; if a temperature drop (trend -1) occurs, a conflict flag is triggered. Furthermore, even if a sensor exhibits a fixed aging bias (e.g., a temperature sensor is 5°C higher overall), as long as the trend is correct (e.g., a continuous increase), the trend coefficient remains valid. For example, the data sequence [30, 31, 32, 33, 34°C] for an aged sensor still correctly outputs trend 1. Furthermore, this method suppresses misjudgments of random fluctuations, as short-term fluctuations are offset by trends. For example, if the adjacent differences in the data sequence [25, 26, 25, 26, 27°C] are [1, -1, 1, 1], and the cumulative sum is 2, trend 1 is still output, preventing oscillations from being misjudged as no trend. In summary, the trend coefficient calculated by the entire expression is the basis for solving multi-sensor trend contradiction detection and weight optimization, and provides a trend analysis framework for equipment status monitoring.

[0059] In one embodiment, the modification of the i-th pre-data weight of the j-th sensor according to the i-th trend coefficient of each sensor in S3 is expressed as:

[0060] in,

[0061] W tij is the weight of the i-th target data corresponding to the j-th sensor, W pij is the i-th pre-data weight corresponding to the j-th sensor, W aj is the weight adjustment value corresponding to the jth sensor, m is the number of sensors, C ij is the i-th trend coefficient corresponding to the j-th sensor, C iv is the i-th trend coefficient corresponding to the v-th sensor.

[0062] In this embodiment, it should be noted that sgn(C ij ·C iv ) is the trend matching calculation, which is used to calculate the sign (1, 0, or -1) of the trend product of the jth sensor and all other sensors (v≠j); for example, if the trend of sensor j is 1 and the trends of other sensors are -1, then sgn(C ij ·C iv )=sgn(-1)=-1. Further, Only records accumulated are positive matches That is, more sensors have the same trend, and the majority of sensors support trend j. Negative matches are filtered by max[0, negative number]. If all sensors have the opposite trend to j, the accumulated sum is negative and the correction term is reset to zero.

[0063] Furthermore, the denominator m represents the total number of sensors, which is used to ensure the comparability and fairness of weight adjustments for sensors of different sizes through normalization. At the same time, m standardizes the correction amplitude and accumulates the sum. The maximum theoretical value is m-1 (when all other sensors are consistent with j), the minimum value is -(m-1) (all opposite), after filtering it is 0, and finally divided by m, the correction term is compressed to the interval [0, 1] to avoid imbalance in weight adjustment caused by differences in the number of sensors; for example: when m=4, the maximum correction term is W aj , when m=10, the maximum value of the correction term is still W aj At the same time, the contribution of each sensor's trend to the correction term is evenly distributed across the total number m, reflecting a democratic decision-making mechanism where "the minority obeys the majority." For example, in a system of 10 sensors, if 5 support trend j, its correction term contribution is 0.56; whereas, among 6 sensors, if 3 support it, the contribution is 0.6, highlighting the stronger influence of majority opinion in a small sensor system.

[0064] In summary, majority consistency verification is achieved. Only when the majority of sensors support trend j is its weight increased, avoiding false increases caused by accidental consistency in a few sensors. For example, if the current sensor shows an upward trend and more than half of the temperature and vibration sensors also show an upward trend, it is determined to be a true overload and its weight is increased. Furthermore, isolated anomalies are suppressed. If the trend of sensor j conflicts with the majority (such as a sudden increase in current but no change in temperature), the cumulative sum is less than or equal to 0, and the weight is not adjusted, avoiding false alarms. Random fluctuation immunity is achieved. Short-term trend conflicts require an overall positive cumulative sum to trigger correction, and are not disrupted by accidental fluctuations. Furthermore, the reliability of group decision-making is improved. Through majority sensor trend consistency verification, it is ensured that alarms are only triggered when multiple physically related parameters are abnormal. For example, abnormal transformer oil temperature must be accompanied by a downward trend in fan speed to avoid false positives in a single sensor.

[0065] like Figure 3 As shown, in one embodiment, obtaining the real-time status index of the electrical device according to the real-time status data collected by each sensor and the status weight of each sensor in S4 includes:

[0066] S41, obtaining the data security threshold of each sensor;

[0067] S42. Acquire a real-time status indicator of the electrical equipment according to the real-time status data collected by each sensor, the status weight of each sensor, and the data security threshold.

[0068] In this embodiment, it should be noted that in S41, a dynamic safety threshold system is constructed by integrating the physical characteristics of the equipment with the operation history, breaking through the limitations of the existing fixed threshold. The setting of the safety threshold is generally based on the static threshold setting of the equipment design parameters. For example, the technical source can be the International Electrotechnical Commission (IEC) standard or the equipment manufacturer's technical manual, and the fixed threshold is set according to the equipment nameplate parameters (such as rated current and temperature limit). For example: Transformer top oil temperature threshold: IEC 60076-7 stipulates that the top oil temperature of the natural oil circulation transformer shall not exceed 105°C, and the forced oil circulation transformer is 95°C; Motor vibration threshold: ISO 10816-3 standard sets the effective value of vibration velocity for different power motors (such as 1.8mm / s is normal and 4.5mm / s is an alarm). The setting of the safety threshold can also be based on the distribution of historical data, and a dynamic threshold is established for the historical normal operating condition data (such as mean ± standard deviation), such as setting the temperature threshold to 2 times the standard deviation of the historical mean; in specific use, real-time environmental variables (such as the impact of ambient temperature and humidity on heat dissipation efficiency) can be artificially introduced to achieve dynamic adjustment.

[0069] In S42, the magnitude of each sensor's exceeding the standard is first normalized using a safety threshold to eliminate dimensional differences, and then multiplied by the proportion of its state weight to the total weight to reflect differences in data credibility. For example, in high-voltage circuit breaker monitoring, if the weight of a current sensor is high due to recent calibration, its exceeding standard data will dominate the indicator calculation; while a temperature sensor that has not been maintained for a long time will have its contribution suppressed even if it is slightly exceeding the standard due to its low weight. This design performs well in gas turbine detonation monitoring: when multiple vibration sensors simultaneously detect high-frequency shocks with high weights, the indicator value increases rapidly; while isolated exceeding standard signals generated by a single sensor due to electromagnetic interference will not trigger false alarms due to their low weight ratio. This mechanism uses physical correlation verification and credibility weighting to enable status indicators to accurately represent the degree of equipment health degradation.

[0070] In one embodiment, the real-time status index of the electrical device is obtained in S4 based on the real-time status data collected by each sensor and the status weight of each sensor, which is expressed as:

[0071] in,

[0072] I S is the real-time status indicator of the electrical equipment, m is the number of sensors, W j is the state weight of the jth sensor, W c is the state weight of the cth sensor, X nj is the real-time status data collected by the jth sensor, X jth is the data security threshold of the j-th sensor.

[0073] In this embodiment, it should be noted that To normalize the weight, the state weight W of a single sensor is j Convert it into a proportion of the total weight to eliminate the impact of sensor quantity differences on decision-making and ensure that high-reliability sensors occupy a higher proportion in decision-making. In order to normalize the excess range, the real-time status data X nj and data security threshold X jth The difference is converted into a percentage to eliminate the incomparability of different dimensions (such as temperature °C, current A, vibration mm / s). Furthermore, max (0, normalized value) is a non-negative cutoff, and only the data exceeding the standard (X nj >X jth , abnormal data) is calculated. Data that does not exceed the standard or is below the threshold is not included in the calculation. The purpose is to filter out normal signals that do not exceed the limit. Further, the normalized weight of each sensor is multiplied by the excess amplitude, and the real-time status index I is accumulated. S ,The larger the real-time status indicator is, the more the device health status deviates from normal and the ,anomaly is driven by high-confidence sensor data.

[0074] In summary, through the weight W j Dynamically reflects sensor reliability (such as maintenance timeliness and data volatility). For example, sensors that have not been maintained for a long time have a low weight, and their exceeded-standard data contributes little to the indicator, thus suppressing aging errors. Furthermore, false alarm suppression is achieved. Due to the low weight of a single sensor false alarm, its impact on the overall indicator is limited. For example, a vibration sensor falsely reports a 20% exceeded standard, but its weight is only 0.05, and its contribution is 0.05*0.2=0.01. If a sensor trend conflicts with the majority (such as a sudden increase in current but no change in temperature), its weight in S3 is suppressed, further reducing the probability of false alarms. Furthermore, collaborative verification of multiple physical quantities is achieved. The indicator value only increases significantly when multiple sensors exceed the standard simultaneously and the trend is consistent.

[0075] It should also be noted that I S By quantifying the excess range and reliability of sensor data, the deviation degree of the equipment health status is directly reflected. S The numerical range of can define the following electrical equipment control status: I S When it is 0, the electrical equipment control state is normal, all sensor data are within the safety threshold, and the equipment operates normally; S ≤T1 (Level 1 threshold), the electrical equipment control state is slightly abnormal, a single or a small number of sensors exceed the standard, and attention should be paid to potential risks; T1 S ≤T2 (Level 2 threshold), the electrical equipment control state is moderately abnormal, multiple sensors exceed the standard, and emergency intervention is required; T2 S ​​​, the electrical equipment control status is severely abnormal, a large number of sensors exceed the specified value, and emergency shutdown maintenance is required. The Level 1 and Level 2 thresholds are dynamically adjusted based on the equipment type, historical data, and risk level, such as equipment design specifications (such as IEC standards) or historical failure data statistics.

[0076] A multi-sensor fusion electrical equipment control status monitoring system is also provided, the system comprising:

[0077] a data acquisition module, configured to acquire a first monitoring time period of the electrical equipment, preset a data processing time window, divide the first monitoring time period into a plurality of second monitoring time periods according to the preset data processing time window, and acquire a sequence of status data collected by each sensor in each second monitoring time period;

[0078] A first data processing module is configured to obtain an i-th pre-data weight and an i-th trend coefficient corresponding to the j-th sensor based on a state data sequence collected by the j-th sensor during the i-th second monitoring time period;

[0079] The second data processing module is used to obtain and modify the i-th predicted data weight of the j-th sensor according to the i-th trend coefficient of each sensor, and form the i-th target data weight of the j-th sensor;

[0080] The third data processing module is used to obtain the state weight of the jth sensor based on multiple target data weights of the jth sensor, and obtain the real-time state index of the electrical equipment based on the real-time state data collected by each sensor and the state weight of each sensor.

[0081] In one embodiment, the first data processing module is also used to: obtain the last maintenance time point corresponding to the jth sensor, and obtain the proxy time point of the i-th second monitoring time period; obtain the time attenuation factor of the j-th sensor in the i-th second monitoring time period based on the last maintenance time point and the proxy time point of the j-th sensor; obtain the i-th pre-data weight corresponding to the j-th sensor based on the state data sequence and time attenuation factor collected by the j-th sensor in the i-th second monitoring time period.

[0082] In one embodiment, the third data processing module is further used to: obtain the data security threshold of each sensor; and obtain the real-time status index of the electrical equipment based on the real-time status data collected by each sensor, the status weight of each sensor and the data security threshold.

[0083] In this embodiment, it should be noted that, regarding the above-mentioned multi-sensor fusion electrical equipment control status monitoring system, the specific method of performing operations has been described in detail in the implementation of the multi-sensor fusion electrical equipment control status monitoring method, and will not be elaborated here.

[0084] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the scope of protection of the present disclosure.

[0085] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0086] In addition, the various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A multi-sensor fusion electrical equipment control status monitoring method, characterized in that: include: Obtaining a first monitoring time period of the electrical equipment, and presetting a data processing time window, dividing the first monitoring time period into a plurality of second monitoring time periods according to the preset data processing time window, and obtaining a sequence of status data collected by each sensor in each second monitoring time period; Obtaining the i-th pre-data weight and the i-th trend coefficient corresponding to the j-th sensor according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period; The i-th predicted data weight of the j-th sensor is modified according to the i-th trend coefficient of each sensor, and the i-th target data weight of the j-th sensor is formed; The state weight of the jth sensor is obtained according to the weights of multiple target data of the jth sensor, and the real-time state index of the electrical equipment is obtained according to the real-time state data collected by each sensor and the state weight of each sensor.

2. The multi-sensor fusion electrical equipment control status monitoring method according to claim 1, characterized in that: The step of obtaining the i-th pre-data weight corresponding to the j-th sensor according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period includes: Get the last maintenance time point corresponding to the jth sensor, and get the proxy time point of the i-th second monitoring time period; Obtain the time attenuation factor of the jth sensor in the i-th second monitoring period according to the last maintenance time point and the proxy time point of the jth sensor; The i-th pre-data weight corresponding to the j-th sensor is obtained according to the state data sequence and the time attenuation factor collected by the j-th sensor in the i-th second monitoring time period.

3. The multi-sensor fusion electrical equipment control status monitoring method according to claim 1, characterized in that: The i-th pre-data weight corresponding to the j-th sensor is obtained according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period as follows: in, W pij is the i-th pre-data weight corresponding to the j-th sensor, m is the number of sensors, λ is the time decay factor, n ij is the number of data in the state data sequence collected by the j-th sensor in the i-th second monitoring time period, is the k+1th data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, is the kth data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, X jmax is the upper limit of the acquisition range of the jth sensor, X jmin is the lower limit of the acquisition range of the jth sensor.

4. The multi-sensor fusion electrical equipment control status monitoring method according to claim 1, characterized in that: The i-th trend coefficient corresponding to the j-th sensor is obtained according to the state data sequence collected by the j-th sensor in the i-th second monitoring time period as follows: in, C ij is the i-th trend coefficient corresponding to the j-th sensor, n ij is the number of data in the state data sequence collected by the j-th sensor in the i-th second monitoring time period, is the k+1th data in the state data sequence collected by the jth sensor in the i-th second monitoring time period, is the kth data in the state data sequence collected by the jth sensor in the i-th second monitoring time period.

5. The multi-sensor fusion electrical equipment control status monitoring method according to claim 1, characterized in that: The correction of the i-th pre-data weight of the j-th sensor according to the i-th trend coefficient of each sensor is expressed as: in, W tij is the weight of the i-th target data corresponding to the j-th sensor, W pij is the i-th pre-data weight corresponding to the j-th sensor, W aj is the weight adjustment value corresponding to the jth sensor, m is the number of sensors, C ij is the i-th trend coefficient corresponding to the j-th sensor, C iv is the i-th trend coefficient corresponding to the v-th sensor.

6. The multi-sensor fusion electrical equipment control status monitoring method according to claim 1, characterized in that: The step of obtaining the real-time status index of the electrical equipment based on the real-time status data collected by each sensor and the status weight of each sensor includes: Obtain the data security threshold of each sensor; The real-time status indicators of electrical equipment are obtained based on the real-time status data collected by each sensor, the status weight of each sensor and the data security threshold.

7. The multi-sensor fusion electrical equipment control status monitoring method according to claim 1, characterized in that: The real-time status index of the electrical equipment obtained based on the real-time status data collected by each sensor and the status weight of each sensor is expressed as: in, I S is the real-time status indicator of the electrical equipment, m is the number of sensors, W j is the state weight of the jth sensor, W c is the state weight of the cth sensor, X nj is the real-time status data collected by the jth sensor, X jth is the data security threshold of the j-th sensor.

8. A multi-sensor fusion electrical equipment control status monitoring system, characterized in that: The system comprises: a data acquisition module, configured to acquire a first monitoring time period of the electrical equipment, preset a data processing time window, divide the first monitoring time period into a plurality of second monitoring time periods according to the preset data processing time window, and acquire a sequence of status data collected by each sensor in each second monitoring time period; A first data processing module is configured to obtain an i-th pre-data weight and an i-th trend coefficient corresponding to the j-th sensor based on a state data sequence collected by the j-th sensor during the i-th second monitoring time period; The second data processing module is used to obtain and modify the i-th predicted data weight of the j-th sensor according to the i-th trend coefficient of each sensor, and form the i-th target data weight of the j-th sensor; The third data processing module is used to obtain the state weight of the jth sensor based on multiple target data weights of the jth sensor, and obtain the real-time state index of the electrical equipment based on the real-time state data collected by each sensor and the state weight of each sensor.

9. The multi-sensor fusion electrical equipment control status monitoring system according to claim 8, characterized in that: The first data processing module is further configured to: Get the last maintenance time point corresponding to the jth sensor, and get the proxy time point of the i-th second monitoring time period; Obtain the time attenuation factor of the jth sensor in the i-th second monitoring period according to the last maintenance time point and the proxy time point of the jth sensor; The i-th pre-data weight corresponding to the j-th sensor is obtained according to the state data sequence and the time attenuation factor collected by the j-th sensor in the i-th second monitoring time period.

10. The multi-sensor fusion electrical equipment control status monitoring system according to claim 9, characterized in that: The third data processing module is further configured to: Obtain the data security threshold of each sensor; The real-time status indicators of electrical equipment are obtained based on the real-time status data collected by each sensor, the status weight of each sensor and the data security threshold.

Images