Electric power system equipment operation risk intelligent identification and early warning system

Abstract

The invention discloses a power system equipment operation risk intelligent identification and early warning system. The system comprises an environmental parameter acquisition module, a passive ultraviolet fluorescence detection module, an active ultraviolet laser excitation module, a background fluorescence analysis and dynamic threshold calculation module, an abnormal signal confirmation module, an intelligent determination module and a closed loop verification and alarm module. Environment characteristic indexes are calculated by monitoring temperature, humidity and air pressure in real time; collecting and analyzing a natural ultraviolet fluorescence signal, actively emitting ultraviolet laser to excite environmental fluorescence response, and adaptively calculating a dynamic anomaly judgment threshold value; abnormal signals are accurately confirmed by utilizing multi-dimensional feature combination and feature space analysis, and corona discharge and line discharge are distinguished through intelligent judgment; and finally, carrying out automatic secondary verification positioning on the visible light image, and automatically generating a discharge early warning signal. The system has the characteristics of high sensitivity, strong interference resistance and high intelligent degree, and the discharge detection reliability is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical equipment condition monitoring and fault diagnosis, and particularly relates to an intelligent risk identification and early warning system for the operation of power system equipment. Background Art

[0002] In a power system, the long-term stable operation of electrical equipment is an important foundation for ensuring the safety of the power grid. However, as the operation time of electrical equipment increases, the natural aging of insulating materials, structural defects of equipment, or changes in external environmental factors, local discharge phenomena may gradually occur in the equipment. If such discharge phenomena cannot be detected and processed in time at an early stage, they may further develop into insulation breakdown, short-circuit faults, or even serious accidents. Therefore, it is of great significance to detect and accurately judge the early discharge phenomena of electrical equipment as early as possible to ensure equipment safety and reduce operation risks.

[0003] At present, the monitoring methods for electrical equipment discharge phenomena mainly include various technologies such as ultrasonic detection, infrared thermal imaging detection, high-frequency pulse current detection, and ultraviolet fluorescence detection. Among them, the ultraviolet fluorescence detection technology has gradually attracted attention and has been applied to a certain extent due to its non-contact detection, high sensitivity, and long-range detection capabilities. However, in practical applications, the existing ultraviolet fluorescence detection systems still have deficiencies in terms of sensitivity and reliability. Especially in outdoor complex environmental conditions, they are easily interfered, resulting in false alarms or missed alarms in monitoring results. In addition, the degree of automation of the existing technology in discharge signal type identification and automatic verification and alarm is relatively low, and the overall detection accuracy and on-site adaptability have not yet met the actual requirements for the safe and efficient operation and maintenance of the power system. Therefore, it is necessary to further improve the reliability, accuracy, and intelligence level of the early discharge monitoring system for electrical equipment. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent risk identification and early warning system for the operation of power system equipment, which has the advantages of effectively ensuring the long-term stable and safe operation of electrical equipment and reducing the operation risks of the equipment.

[0005] The above technical object of the present invention is achieved through the following technical solutions:

[0006] An intelligent risk identification and early warning system for the operation of power system equipment, comprising:

[0007] An environmental parameter acquisition module, configured to collect temperature, humidity, and air pressure data in the monitoring area in real time, and calculate and output an environmental characteristic index;

[0008] The passive ultraviolet fluorescence detection module is used to collect natural ultraviolet fluorescence signal data in the monitoring area in real time, and based on the stored historical anomaly judgment threshold, quickly and preliminarily judge the natural ultraviolet fluorescence signal data for anomalies. When an anomaly is judged to exist, an abnormal fluorescence trigger signal is output;

[0009] The active ultraviolet laser excitation module is connected to the passive ultraviolet fluorescence detection module and is used to start after receiving the abnormal fluorescence trigger signal. According to the fluorescence characteristics of the abnormal signal, it adaptively adjusts the intensity, frequency and time-domain coding mode of the laser pulse, and emits laser pulses to the environmental area outside the abnormal position to stimulate gas molecules in the area to generate fluorescence response signals;

[0010] The background fluorescence analysis and dynamic threshold calculation module is connected to the active ultraviolet laser excitation module and the environmental feature analysis module, and is used to collect the environmental fluorescence response signals after laser excitation, calculate the dynamic anomaly judgment threshold according to the environmental fluorescence response signals and environmental feature indices, and provide it to the passive ultraviolet fluorescence detection module and the abnormal signal confirmation module. The passive ultraviolet fluorescence detection module updates the stored historical anomaly judgment threshold according to the dynamic anomaly judgment threshold to obtain a new historical anomaly judgment threshold;

[0011] The abnormal signal confirmation module is connected to the passive ultraviolet fluorescence detection module and the background fluorescence analysis and dynamic threshold calculation module, and is used to perform multi-dimensional feature combination and analyze whether there is a real anomaly in the natural ultraviolet fluorescence signal data initially judged as abnormal output by the passive ultraviolet fluorescence detection module in combination with the dynamic anomaly judgment threshold. If it is judged that there is a real anomaly, a real abnormal signal is output;

[0012] The intelligent judgment module is connected to the abnormal signal confirmation module and is used to perform multi-period integration and adaptive analysis judgment on the real abnormal ultraviolet fluorescence signal, distinguish corona discharge and real line discharge, and output a line discharge confirmation signal;

[0013] The closed-loop verification and alarm module is connected to the intelligent judgment module and is used to perform secondary visible light verification and positioning on the abnormal position after receiving the line discharge confirmation signal, and generate a discharge warning signal.

[0014] By adopting the above technical solutions, through the real-time monitoring of environmental parameters and active laser excitation technology, high-precision discrimination between natural ultraviolet fluorescence signals and background fluorescence is achieved, effectively improving the sensitivity and accuracy of early identification of abnormal discharges. At the same time, based on multi-dimensional feature analysis of dynamic thresholds, the risks of misjudgment and missed judgment are reduced, the recognition reliability of real discharge phenomena is enhanced, automatic closed-loop verification and alarm of discharge abnormal states are realized, and the automation and intelligent level of system operation are significantly improved.

[0015] Further setting: The passive ultraviolet fluorescence detection module includes:

[0016] An ultraviolet detector array unit for real-time acquisition of natural ultraviolet fluorescence signal data in the monitoring area;

[0017] A fluorescence signal feature extraction unit for extracting features from the natural ultraviolet fluorescence signal data to obtain feature data, specifically including fluorescence intensity features, fluorescence frequency features, fluorescence spectrum features, and fluorescence frequency features;

[0018] An initial anomaly judgment unit for receiving the extracted feature data and comparing it with the historical anomaly judgment threshold stored in the threshold storage unit to determine whether there is an anomaly;

[0019] A marking trigger unit for recording and outputting the spatial position information of the ultraviolet fluorescence where an anomaly is determined, and outputting an abnormal fluorescence trigger signal;

[0020] A threshold storage unit for updating the stored historical anomaly judgment threshold according to the dynamic anomaly judgment threshold provided by the background fluorescence analysis and dynamic threshold calculation module.

[0021] By adopting the above technical solutions, using an ultraviolet detector array and a feature extraction method, comprehensively extract the intensity, frequency, and spectrum features of natural ultraviolet fluorescence signals, and accurately capture weak fluorescence signals in the initial stage of discharge. Through the initially abnormal screening with the dynamically updated historical anomaly judgment threshold, the sensitivity and real-time performance of anomaly detection are improved.

[0022] Further settings: The active ultraviolet laser excitation module includes:

[0023] A pulsed laser unit for emitting pulsed ultraviolet laser, with continuously adjustable pulse power and width;

[0024] A fluorescence signal feature analysis unit for receiving the feature data of the passive ultraviolet fluorescence detection module and calculating an anomaly level score based on the feature data;

[0025] A time-domain coding control unit adaptively sets the power, frequency, and time-domain coding mode of the laser pulse according to the anomaly level score, where:

[0026] When the anomaly level score is within the first level score range, a simple periodic coding with low power and low frequency is adopted;

[0027] When the anomaly level score is within the second level score range, a periodic pulse train coding with medium power and medium frequency is adopted;

[0028] When the anomaly level score is within the third level score range, a complex non-periodic multi-level coding with high power and high frequency is adopted;

[0029] The servo optical pan-tilt unit is used to receive the spatial position information output by the passive ultraviolet fluorescence detection module, and adjust the emission direction of the laser pulse in real time according to the spatial position information, so as to ensure that the emitted pulsed ultraviolet laser avoids the occurrence area of abnormal fluorescence signals and excites the background fluorescence response signals in the monitoring area outside the abnormal position.

[0030] By adopting the above technical solution, an active ultraviolet laser excitation method with adaptive time-domain coding is adopted, and the laser power, frequency and coding mode are automatically adjusted according to the abnormal signal level, flexibly adapting to the actual on-site situation, accurately exciting the background fluorescence signal, further improving the accurate recognition ability of real discharge anomalies, effectively reducing the interference effect, and enhancing the adaptability of the detection system to environmental changes.

[0031] Further setting: The background fluorescence analysis and dynamic threshold calculation module includes:

[0032] The fluorescence response acquisition unit collects the environmental background fluorescence response signals after the active laser pulse excitation in real time to obtain the actual fluorescence characteristics;

[0033] The expected characteristic calculation unit calculates the expected background fluorescence characteristics according to the pre-established fluorescence characteristic database and environmental characteristic index;

[0034] The characteristic deviation calculation unit calculates the relative deviation between the actual fluorescence response signal and the expected fluorescence characteristics to obtain the fluorescence characteristic deviation;

[0035] The dynamic anomaly threshold calculation unit calculates the dynamic anomaly judgment threshold according to the fluorescence characteristic deviation;

[0036] The dynamic threshold output unit outputs the dynamic anomaly thresholds of each characteristic to the passive ultraviolet fluorescence detection module and the abnormal signal confirmation module in real time, for updating the historical anomaly thresholds and multi-dimensional characteristic anomaly analysis.

[0037] By adopting the above technical solution, through the dynamic threshold calculation method, the deviation between the background fluorescence response signal and the expected characteristics is analyzed in real time, and the anomaly judgment threshold is adaptively updated to ensure that the anomaly judgment threshold always accurately reflects the on-site real-time environmental conditions, significantly improving the accuracy and robustness of the system anomaly judgment, and effectively reducing the false alarm rate and missed alarm rate.

[0038] Further setting: The multi-dimensional characteristic combination is combined with the dynamic anomaly judgment threshold to analyze and judge whether there is a real anomaly. If it is determined that there is a real anomaly, the real anomaly signal is output, which specifically includes the following steps:

[0039] Obtain the characteristic data of the natural ultraviolet fluorescence signal data initially judged to have an anomaly;

[0040] Construct a real-time characteristic vector and a dynamic threshold characteristic vector according to the characteristic data and the dynamic anomaly judgment threshold;

[0041] Calculate the Euclidean distance between the real-time feature vector and the dynamic threshold feature vector;

[0042] By comparing the Euclidean distance with a preset distance threshold, determine whether the abnormal signal is truly abnormal. If the Euclidean distance is greater than the preset distance threshold, it is judged as truly abnormal and the real abnormal signal is output.

[0043] By adopting the above technical solution, using the feature space distance calculation method between the real-time feature vector and the dynamic threshold feature vector, combined with the multi-dimensional dynamic anomaly judgment threshold, the precise re-judgment of the preliminary abnormal signal is realized, effectively distinguishing the real anomaly and the environmental interference signal, further improving the reliability and accuracy of anomaly recognition, and ensuring the real existence of abnormal discharge.

[0044] Further set: The intelligent judgment module includes:

[0045] A multi-cycle integral analysis unit that performs integral statistics on the natural ultraviolet fluorescence signal with anomalies for multiple power frequency cycles to obtain an integral value;

[0046] A feature parameter statistics unit that calculates the average value and standard deviation of the integral value set;

[0047] A pulse periodicity feature extraction unit that calculates the phase concentration degree of the abnormal natural ultraviolet fluorescence signal within the power frequency cycle;

[0048] An adaptive intelligent judgment unit that forms a feature vector based on the integral average value, integral standard deviation, and phase concentration degree features, and uses a machine learning algorithm for classification judgment to distinguish corona discharge and line discharge;

[0049] A discharge confirmation signal output unit that, if the adaptive intelligent judgment unit determines it as line discharge, outputs a line discharge confirmation signal to the closed-loop verification and alarm module in real time.

[0050] By adopting the above technical solution, through multi-cycle integration, statistical analysis, and pulse periodicity feature extraction, combined with the machine learning classification algorithm, corona discharge and line discharge are accurately distinguished, the recognition ability of different discharge types is improved, enabling on-site personnel to take corresponding treatment measures for different discharge types, reducing the complexity and cost of equipment maintenance, and improving the operation and maintenance efficiency.

[0051] Further set: The visible light secondary verification and positioning of the abnormal position to generate a discharge warning signal specifically includes the following steps:

[0052] Take a visible light image of the abnormal area in real time;

[0053] According to the spatial position information of the abnormal area, determine the image position of the abnormal area through the camera calibration parameters;

[0054] Extract the gray - scale contrast features and spark detection features of the abnormal area;

[0055] Use a machine - learning model to automatically classify the abnormal image features and output the category determination probability of the abnormal image;

[0056] Set an alarm probability threshold. When the classification probability is higher than the threshold, automatically output a discharge warning signal.

[0057] By adopting the above - mentioned technical solution, an automated secondary verification method based on spatial positioning and image processing is used to analyze the visible - light image of the abnormal area in real - time, automatically confirm the discharge abnormal position, generate a reliable alarm warning signal, realize the efficient closed - loop verification of abnormal discharge faults and on - site automatic alarm, and greatly improve the on - site operation and maintenance automation level and operation safety.

[0058] In summary, the present invention has the following beneficial effects: Through active ultraviolet laser excitation and precise multi - dimensional feature analysis technology, accurate identification of early weak discharge signals is realized, effectively improving the sensitivity of early abnormal detection and significantly advancing the timeliness of fault warning. Fully considering the real - time impact of environmental parameter changes, the adaptive correction of dynamic thresholds and feature space distance analysis technology are adopted to effectively reduce the false - alarm impact of environmental background fluorescence signals and other interference factors, and improve the accuracy and reliability of monitoring. By using the multi - cycle integration, pulse periodicity analysis and adaptive machine - learning classification technology of the intelligent determination module, efficient and accurate discrimination between corona discharge and line discharge is realized, providing clear guidance for fault handling in on - site operation and maintenance, and reducing the maintenance cost and difficulty. Through the visible - light image automatic positioning and machine - learning classification confirmation technology of the closed - loop verification and alarm module, automatic secondary verification, positioning and alarm of abnormal discharge phenomena are realized, significantly reducing the need for manual intervention and improving the on - site operation and maintenance automation level. In summary, the overall technical solution provided by the present invention is outstanding in improving the sensitivity, accuracy, intelligence and automation level of early detection of electrical equipment discharge, can effectively ensure the long - term stable and safe operation of electrical equipment, reduce the equipment operation risk, and has significant economic and social benefits. Description of the Drawings

[0059] Figure 1 It is the overall flow block diagram of the embodiment. Detailed Embodiment

[0060] The following further details the present invention with reference to the accompanying drawings.

[0061] Embodiment:

[0062] As Figure 1 shown, the intelligent risk identification and warning system for power system equipment includes:

[0063] An environmental parameter acquisition module, which is used to collect temperature, humidity and air pressure data in the monitoring area in real time and calculate and output an environmental characteristic index; this index is comprehensively calculated from the real-time temperature T(t), humidity H(t) and air pressure P(t) in the monitoring area, and the specific formula is as follows:

[0064] E idx (t) = α·T(t) + β·H(t) + γ·P(t)

[0065] Among them, α, β, and γ are environmental characteristic weight coefficients calibrated through experiments.

[0066] A passive ultraviolet fluorescence detection module, which is used to collect natural ultraviolet fluorescence signal data in the monitoring area in real time, and based on the stored historical anomaly judgment threshold, quickly and preliminarily judge the natural ultraviolet fluorescence signal data for anomalies, and output an abnormal fluorescence trigger signal when it is judged that there is an anomaly;

[0067] An active ultraviolet laser excitation module, connected to the passive ultraviolet fluorescence detection module, which is used to start after receiving the abnormal fluorescence trigger signal, adaptively adjust the intensity, frequency and time-domain coding mode of the laser pulse according to the fluorescence characteristics of the abnormal signal, and emit laser pulses to the environmental area outside the abnormal position to excite gas molecules in the area to generate fluorescence response signals;

[0068] A background fluorescence analysis and dynamic threshold calculation module, connected to the active ultraviolet laser excitation module and the environmental characteristic analysis module, which is used to collect the environmental fluorescence response signal after laser excitation, calculate the dynamic anomaly judgment threshold based on the environmental fluorescence response signal and the environmental characteristic index, and provide it to the passive ultraviolet fluorescence detection module and the abnormal signal confirmation module. The passive ultraviolet fluorescence detection module updates the stored historical anomaly judgment threshold according to the dynamic anomaly judgment threshold to obtain a new historical anomaly judgment threshold;

[0069] An abnormal signal confirmation module, connected to the passive ultraviolet fluorescence detection module and the background fluorescence analysis and dynamic threshold calculation module, which is used to perform multi-dimensional feature combination and combine the dynamic anomaly judgment threshold analysis on the natural ultraviolet fluorescence signal data initially determined to be abnormal output by the passive ultraviolet fluorescence detection module to determine whether there is a real anomaly. If it is determined that there is a real anomaly, a real anomaly signal is output;

[0070] An intelligent judgment module, connected to the abnormal signal confirmation module, which is used to perform multi-period integration and adaptive analysis judgment on the real abnormal ultraviolet fluorescence signal, distinguish corona discharge and real line discharge, and output a line discharge confirmation signal;

[0071] The closed-loop verification and alarm module, connected to the intelligent determination module, is used to perform secondary visible light verification and positioning on the abnormal location after receiving the line discharge confirmation signal, and generate a discharge warning signal.

[0072] Through the real-time monitoring of environmental parameters and the active laser excitation technology, the high-precision distinction between natural ultraviolet fluorescence signals and background fluorescence is realized, effectively improving the sensitivity and accuracy of early identification of abnormal discharges. At the same time, based on the multi-dimensional feature analysis of dynamic thresholds, the risks of false judgment and missed judgment are reduced, the recognition reliability of real discharge phenomena is enhanced, the automatic closed-loop verification and alarm of discharge abnormal states are realized, and the automation and intelligence levels of system operation are significantly improved.

[0073] Among them, the passive ultraviolet fluorescence detection module specifically includes:

[0074] The ultraviolet detector array unit is used to collect the data of natural ultraviolet fluorescence signals in the monitoring area in real time. The detection band of the ultraviolet detector array unit is 240nm to 280nm, the response sensitivity is better than 10pW / cm2, and the signal is collected at a sampling frequency of not less than 1000Hz;

[0075] The phase-locked amplification circuit unit, connected to the ultraviolet detector array unit, is used to perform phase-sensitive detection and phase-locked amplification on the weak ultraviolet fluorescence signal output by the narrowband filtering unit. The phase-locked amplification circuit is used to perform fine filtering and amplification processing on the ultraviolet fluorescence pulse signal in a specific frequency range (such as 500Hz to 2kHz), and improve the detection sensitivity and signal-to-noise ratio of the signal in real time. A high-speed analog-to-digital converter (ADC) is used to perform high-speed sampling and quantization on the filtered fluorescence pulse signal in real time; the sampling frequency is preferably greater than 10MHz to completely capture the time-domain waveform characteristics of the fluorescence pulse, including pulse amplitude, duration, rising edge and falling edge characteristics.

[0076] The fluorescence signal feature extraction unit is used to extract feature data from the natural ultraviolet fluorescence signal data, specifically including fluorescence intensity features, fluorescence frequency features, fluorescence spectrum features and fluorescence frequency features;

[0077] The preliminary abnormal judgment unit is used to receive the extracted feature data and compare it with the historical abnormal judgment threshold stored in the threshold storage unit to judge whether there is an abnormality; directly compare the fluorescence intensity feature with the historical abnormal judgment threshold. When the real-time fluorescence intensity exceeds the historical abnormal judgment threshold, it is preliminarily determined that there is an abnormality, or when the difference between the fluorescence frequency feature, fluorescence spectrum feature, and fluorescence pulse count feature and the historical abnormal judgment threshold exceeds the preset range, it is preliminarily determined that there is an abnormality. When at least one feature parameter meets the above preliminary judgment conditions, it is judged that there is an abnormality. The historical abnormal judgment threshold is initially determined by test or empirical data and is corrected in real time by the subsequent background fluorescence analysis module.

[0078] A marker trigger unit, which is used to record and output the spatial position information of the ultraviolet fluorescence judged to be abnormal, and output an abnormal fluorescence trigger signal; when it is judged as an abnormal fluorescence signal, it uses the built-in spatial position information unit (such as a rotary encoder, a pitch angle encoder, a laser rangefinder) to record the three-dimensional spatial coordinate information of the fluorescence abnormal position in real time, ensuring that the subsequent active excitation module can accurately avoid this abnormal area; at the same time, it outputs an abnormal fluorescence trigger signal to the subsequent active excitation module to start the active laser excitation and background fluorescence response detection process.

[0079] A threshold storage unit, which is used to update the stored historical abnormal judgment threshold according to the dynamic abnormal judgment threshold provided by the background fluorescence analysis and dynamic threshold calculation module. It adaptively updates the stored historical abnormal judgment threshold according to the preset update strategy (such as moving average, exponential weighted average), ensuring the real-time performance and reliability of the abnormal judgment benchmark.

[0080] An ultraviolet detector array and a feature extraction method are adopted to comprehensively extract the intensity, frequency and spectrum characteristics of the natural ultraviolet fluorescence signal, and accurately capture the weak fluorescence signal at the initial stage of discharge. Through the preliminary abnormal screening with the dynamically updated historical abnormal judgment threshold, the sensitivity and real-time performance of abnormal detection are improved.

[0081] The active ultraviolet laser excitation module includes:

[0082] A pulsed laser unit, which is used to emit pulsed ultraviolet laser, and the pulse power and width can be continuously adjusted; a pulsed ultraviolet laser with a wavelength in the range of 240nm - 280nm is used as the excitation light source, and the pulse power of this laser is continuously adjustable in the range of 0.1mJ - 10mJ, the pulse frequency is continuously adjustable in the range of 10Hz - 10kHz, and the pulse width is continuously adjustable within 1ns - 100ns to meet the requirements of different excitation conditions;

[0083] A fluorescence signal feature analysis unit, which is used to receive the feature data of the passive ultraviolet fluorescence detection module and calculate the abnormal level score according to the feature data; it normalizes the fluorescence pulse count feature (number of pulses per unit time), fluorescence intensity feature (average intensity of pulse amplitude), fluorescence frequency feature (frequency of pulse appearance), and fluorescence spectrum feature (normalized intensity of characteristic spectral line) to obtain the corresponding normalized features. Next, a weighted scoring method is used to calculate the abnormal level score P, and the specific formula is:

[0084] P = w C ·C n + w F ·F n + w f ·f n + w S ·Sn

[0085] w C and w F and wf, w S are the weight coefficients of each characteristic parameter. Generally, the selected weight range is: w c ≈0.3, w F ≈0.4, w f ≈0.2, w S ≈0.1. The specific weight value can be adjusted according to experimental calibration and needs to satisfy that the sum of conditions is 1. C n and F n and f n and S n are the characteristic values after normalization of the fluorescence pulse count characteristic, fluorescence intensity characteristic, fluorescence frequency characteristic, and fluorescence spectrum characteristic respectively.

[0086] After calculating the abnormal level score P and combining the statistical analysis of experimental data, the abnormal level score is divided into the following three level ranges:

[0087] The first level (low): 1.0 ≤ P < 1.5, and the characteristic manifestation is that the fluorescence intensity is relatively low (F n <1.2), and the pulse count is relatively small (C n <1.1).

[0088] The second level (medium): 1.5 ≤ P < 2.5, and the characteristic manifestation is that the fluorescence intensity is medium (1.2 ≤ F n <1.8), and the pulse count is moderate (1.1 ≤ C n <1.5).

[0089] The third level (high): P ≥ 2.5, and the characteristic manifestation is that the fluorescence intensity is relatively high (F n ≥1.8), and the pulse count is frequent (C n ≥1.5).

[0090] The specific numerical settings within the above level ranges can be fine-tuned according to the actual on-site working conditions or the statistical results of experimental data to meet the actual monitoring requirements.

[0091] The time-domain coding control unit adaptively sets the power, frequency, and time-domain coding mode of the laser pulse according to the abnormal level score, where:

[0092] When the abnormal level score is within the first-level score range, the laser unit emits simple periodic laser pulses with low power (0.1 - 1 mJ) and low frequency (10 - 100 Hz);

[0093] When the abnormality level score is within the second level score range, the laser unit emits a periodic pulse train laser with medium power (1 to 5 mJ) and medium frequency (100 Hz to 1 kHz);

[0094] When the abnormality level score is within the third level score range, the laser unit emits high-power (5-10mJ), high-frequency (1kHz-10kHz) complex non-periodic multi-level time-domain coded laser pulses to stimulate the ambient gas molecules to produce stronger and more stable fluorescence response signals;

[0095] The servo optical pan-tilt unit is used to receive the spatial position information output by the passive ultraviolet fluorescence detection module, and adjust the emission direction of the laser pulse in real time according to the spatial position information to ensure that the emitted pulsed ultraviolet laser avoids the area where the abnormal fluorescence signal occurs, and stimulates the monitoring area outside the abnormal position to generate a background fluorescence response signal. The servo pan-tilt unit includes a pitch and azimuth dual-axis servo motor, a controller and an encoder, which can accurately point the laser emission direction to the area outside the abnormal position in real time, prevent the active laser from overlapping the real abnormal area, and ensure the effectiveness of the environmental background fluorescence excitation.

[0096] Adopting active ultraviolet laser excitation method with adaptive time domain coding, the laser power, frequency and coding mode are automatically adjusted according to the abnormal signal level, which can flexibly adapt to the actual situation on site, accurately stimulate the background fluorescence signal, further improve the ability to accurately identify real discharge anomalies, effectively reduce the influence of interference, and enhance the adaptability of the detection system to environmental changes.

[0097] The background fluorescence analysis and dynamic threshold calculation module includes:

[0098] The fluorescence response acquisition unit collects the environmental background fluorescence response signal after active laser pulse excitation in real time to obtain the actual fluorescence characteristics; and receives the background fluorescence response signal generated by the gas molecules (mainly N2, O2, etc.) in the environmental area after the active ultraviolet laser excitation module emits the laser pulse in real time.

[0099] The expected feature calculation unit calculates the expected background fluorescence features according to the pre-established fluorescence feature database and environmental feature index; including fluorescence intensity, pulse count, frequency and spectrum distribution, and the calculation formula is:

[0100]

[0101] X exp (t) is the characteristic expected value (including intensity, pulse count, frequency, and spectrum characteristics), X base is the base eigenvalue, is the benchmark environmental characteristic index, k X is the characteristic sensitivity coefficient.

[0102] A feature deviation calculation unit calculates the relative deviation between the actual fluorescence response signal and the expected fluorescence feature to obtain the fluorescence feature deviation.

[0103] A dynamic anomaly threshold calculation unit weights the relative deviations of fluorescence intensity, pulse count, frequency, and spectral features respectively according to the fluorescence feature deviation, and calculates a multi-dimensional dynamic anomaly judgment threshold. The calculation formula is:

[0104] Th dyn，X (t) = Th base，X ·[1 + ω X ·ΔX(t)]

[0105] Th dyn，X (t): Real-time dynamic anomaly threshold (corresponding to intensity, count, frequency, and spectrum respectively); Th base，X : Base threshold; ω X : Feature weight coefficient.

[0106] A dynamic threshold output unit outputs the dynamic anomaly thresholds of each feature to the passive ultraviolet fluorescence detection module and the anomaly signal confirmation module in real time, for updating the historical anomaly threshold and multi-dimensional feature anomaly analysis.

[0107] Among them, the reference environmental feature index is obtained through multiple background fluorescence signal measurement experiments under typical environmental conditions set as standard references. The specific process is as follows: Select standard environmental conditions, such as the environmental temperature is 25°C, the relative humidity is 50%, and the air pressure is 101325 Pa (standard atmospheric pressure); calculate by continuously measuring environmental parameters under this condition. Define a reference index as a reference base point to facilitate the rapid interpolation calculation of features under other conditions, improve the real-time performance and unity of the calculation, and facilitate the feature comparison of different environments. The base threshold is also determined by statistical methods through multiple laser active excitation experiments under the above standard reference environmental conditions, and statistically analyzing the fluctuation range and stable value of each feature.

[0108] Through the dynamic threshold calculation method, the deviation between the background fluorescence response signal and the expected feature is analyzed in real time, and the anomaly judgment threshold is adaptively updated to ensure that the anomaly judgment threshold always accurately reflects the on-site real-time environmental conditions, significantly improving the accuracy and robustness of the system's anomaly judgment, and effectively reducing the false alarm rate and missed alarm rate.

[0109] The above-mentioned multi-dimensional feature combination combined with the dynamic anomaly judgment threshold is used to analyze and judge whether there is a real anomaly. If it is determined that there is a real anomaly, a real anomaly signal is output. The specific steps are as follows:

[0110] Obtain the feature data of the natural ultraviolet fluorescence signal data initially judged to have an anomaly, including fluorescence intensity feature, fluorescence pulse count feature, fluorescence frequency feature, and fluorescence spectrum feature;

[0111] Construct a real-time feature vector and a dynamic threshold feature vector based on the feature data and the dynamic anomaly judgment threshold;

[0112] The real-time feature vector is:

[0113]

[0114] The dynamic threshold feature vector is:

[0115] X Th (t) = [1, 1, 1,..., 1]

[0116] Each feature in the above vectors is normalized. Based on the dynamic threshold, it is convenient to compare in a unified feature space.

[0117] Calculate the Euclidean distance between the real-time feature vector and the dynamic threshold feature vector; the calculation formula is:

[0118]

[0119] D(t): The Euclidean distance between the real-time feature vector and the dynamic threshold vector, X s，i (t): The real-time normalized feature value, X Th，i (t): The value of the i-th dimension in the dynamic threshold vector, always 1, N: The number of feature dimensions.

[0120] By comparing the Euclidean distance with the preset distance threshold, determine whether the abnormal signal is truly abnormal. When the Euclidean distance is greater than the preset distance threshold, it is judged as truly abnormal and the true abnormal signal is output. When the Euclidean distance is less than or equal to the preset distance threshold, it is determined as environmental interference and false alarm, and the signal is ignored.

[0121] Using the feature space distance calculation method between the real-time feature vector and the dynamic threshold feature vector, combined with the multi-dimensional dynamic anomaly judgment threshold, realize the precise re-judgment of the preliminary abnormal signal, effectively distinguish the true anomaly and the environmental interference signal, further improve the reliability and accuracy of anomaly recognition, and ensure the true existence of abnormal discharge.

[0122] The intelligent judgment module includes:

[0123] A multi-cycle integral analysis unit that performs integral statistics on the natural ultraviolet fluorescence signal with anomalies for multiple power frequency cycles to obtain an integral value, and performs integral statistical analysis on the true abnormal ultraviolet fluorescence signal output by the abnormal signal confirmation module for multiple power frequency cycles (each cycle T0 = 20ms). The specific method is as follows: Suppose the pulse amplitude of the ultraviolet fluorescence signal measured in real time in the j-th cycle is F j (t), and the integral value of the j-th cycle is obtained by integration.

[0124] The partial discharge signal changes periodically with the power frequency voltage. Multi-cycle integral statistics are beneficial to capturing the persistence and stability of the signal (refer to the IEC60270 partial discharge detection standard).

[0125] The characteristic parameter statistical unit statistically analyzes the set of integral values {Ij} within multiple consecutive power frequency cycles (usually 5 - 10 cycles), and calculates the average integral value and the standard deviation of the integral value. The multi-cycle statistical characteristics are used to distinguish corona discharge (with obvious periodicity and a small standard deviation) from line discharge (with strong persistence and a large standard deviation).

[0126] The pulse periodicity feature extraction unit calculates the phase concentration of abnormal natural ultraviolet fluorescence signals within the power frequency cycle, which is obtained by using the phase vector method. Corona discharge usually only appears near the voltage peak phase (with high concentration), while line discharge has no obvious phase concentration feature (with low concentration).

[0127] The adaptive intelligent decision-making unit synthesizes the aforementioned statistical characteristics (average integral value, integral standard deviation, pulse phase concentration) to form a feature vector Y, and performs adaptive classification based on machine learning algorithms (such as support vector machine SVM or BP neural network), and defines the decision function:

[0128] f(Y) = W·Y + b

[0129] Where W is the weight vector, which is optimized and determined through historical experimental training data, and b is the classification threshold, which is optimized and determined through historical experimental data. If f(Y) ≥ 0, it is determined as line discharge; if f(Y) < 0, it is determined as corona discharge. The machine learning model can automatically explore the internal relationships between multi-dimensional features, improving the accuracy and reliability of discharge type discrimination (refer to IEC TS62478).

[0130] The discharge confirmation signal output unit, if the adaptive intelligent decision-making unit determines it as line discharge, will output the line discharge confirmation signal to the closed-loop verification and alarm module in real time. If it is determined as corona discharge, the confirmation signal will not be output temporarily and will only be recorded in the system log. Line discharge (such as arc) has a greater impact on equipment safety and is given priority for processing and alarming. Corona discharge has a smaller impact and can be processed later (in line with the power system operation and maintenance standard DL / T1432).

[0131] By adopting the above technical solutions, through multi-cycle integration, statistical analysis, and pulse periodicity feature extraction, combined with machine learning classification algorithms, corona discharge and line discharge can be accurately distinguished, enhancing the recognition ability of different discharge types, enabling on-site personnel to take corresponding treatment measures for different discharge types, reducing the complexity and cost of equipment maintenance, and improving the operation and maintenance efficiency.

[0132] The visible light secondary verification and positioning of the abnormal location to generate a discharge warning signal specifically includes the following steps:

[0133] Take visible light images of abnormal areas in real time. Discharge phenomena are often accompanied by visible light signs (such as arc sparks). Through visible light confirmation, the abnormal location can be visually verified (refer to the IEC TS 60034 standard, Electrical equipment condition monitoring guidelines). Although it is difficult to detect in the initial stage, on the basis of locating the abnormal situation, the accuracy of determining the location can be increased;

[0134] According to the spatial position information of the abnormal area, determine the image position of the abnormal area through camera calibration parameters, and use computer vision methods to accurately map the spatial position to the image, which is convenient for subsequent automated image analysis processing (refer to "Principles of Computer Vision and Image Processing");

[0135] Extract the gray contrast feature and spark detection feature of the abnormal area. The spark detection feature uses the threshold method to detect the proportion of local high-brightness pixels in the image area;

[0136] Use a machine learning model to automatically classify the abnormal image features and output the class determination probability of the abnormal image. A pre-trained abnormal image classification model (such as a random forest classifier or a neural network).

[0137] Set an alarm probability threshold. When the classification probability is higher than the threshold, automatically output a discharge warning signal.

[0138] Adopt an automated secondary verification method based on spatial positioning and image processing, analyze the visible light image of the abnormal area in real time, automatically confirm the discharge abnormal position, generate a reliable alarm warning signal, realize the efficient closed-loop verification of abnormal discharge faults and on-site automated alarm, and greatly improve the on-site operation and maintenance automation level and operation safety.

[0139] In summary, the present invention has the following beneficial effects: Through active ultraviolet laser excitation and precise multi-dimensional feature analysis technology, accurate identification of early weak discharge signals is achieved, effectively improving the sensitivity of early anomaly detection and significantly advancing the timeliness of fault warning. Fully considering the real-time impact of environmental parameter changes, the adaptive correction of dynamic thresholds and feature space distance analysis technology are adopted to effectively reduce the false alarm impact of environmental background fluorescence signals and other interference factors, enhancing the accuracy and reliability of monitoring. By applying the multi-cycle integration, pulse periodicity analysis, and adaptive machine learning classification technology of the intelligent determination module, efficient and accurate differentiation between corona discharge and line discharge is realized, providing clear guidance for fault handling in on-site operation and maintenance, and reducing the maintenance cost and difficulty. Through the visible light image automatic positioning and machine learning classification confirmation technology of the closed-loop verification and alarm module, automatic secondary verification, positioning, and alarm of abnormal discharge phenomena are achieved, significantly reducing the need for manual intervention and improving the level of on-site operation and maintenance automation. In summary, the overall technical solution provided by the present invention is outstanding in terms of improving the sensitivity, accuracy, intelligence, and automation of early detection of electrical equipment discharge, can effectively ensure the long-term stable and safe operation of electrical equipment, reduce the operation risk of the equipment, and has significant economic and social benefits.

[0140] The above-described embodiments do not constitute a limitation on the protection scope of the technical solution. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the above embodiments shall be included in the protection scope of the technical solution.

Claims

1. An intelligent risk identification and early warning system for power system equipment, characterized in that Including: An environmental parameter acquisition module, which is used to collect temperature, humidity and air pressure data in the monitoring area in real time, and calculate and output an environmental characteristic index; A passive ultraviolet fluorescence detection module, which is used to collect natural ultraviolet fluorescence signal data in the monitoring area in real time, and based on the stored historical anomaly judgment threshold, quickly and preliminarily judge the natural ultraviolet fluorescence signal data for anomalies. When it is judged that there is an anomaly, an abnormal fluorescence trigger signal is output; An active ultraviolet laser excitation module, connected to the passive ultraviolet fluorescence detection module, which is used to start after receiving the abnormal fluorescence trigger signal, adaptively adjust the intensity, frequency and time-domain coding mode of the laser pulse according to the fluorescence characteristics of the abnormal signal, and emit laser pulses to the environmental area outside the abnormal position to excite gas molecules in the area to generate a fluorescence response signal; A background fluorescence analysis and dynamic threshold calculation module, connected to the active ultraviolet laser excitation module and the environmental characteristic analysis module, which is used to collect the environmental fluorescence response signal after laser excitation, calculate a dynamic anomaly judgment threshold based on the environmental fluorescence response signal and the environmental characteristic index, and provide it to the passive ultraviolet fluorescence detection module and the abnormal signal confirmation module. The passive ultraviolet fluorescence detection module updates the stored historical anomaly judgment threshold according to the dynamic anomaly judgment threshold to obtain a new historical anomaly judgment threshold; An abnormal signal confirmation module, connected to the passive ultraviolet fluorescence detection module and the background fluorescence analysis and dynamic threshold calculation module, which is used to perform multi-dimensional feature combination and combined with the dynamic anomaly judgment threshold analysis on the natural ultraviolet fluorescence signal data initially determined to be abnormal output by the passive ultraviolet fluorescence detection module to determine whether there is a real anomaly. If it is determined that there is a real anomaly, a real anomaly signal is output; An intelligent judgment module, connected to the abnormal signal confirmation module, which is used to perform multi-period integration and adaptive analysis and judgment on the real abnormal ultraviolet fluorescence signal, distinguish corona discharge and real line discharge, and output a line discharge confirmation signal; A closed-loop verification and alarm module, connected to the intelligent judgment module, which is used to perform secondary visible light verification and positioning on the abnormal position after receiving the line discharge confirmation signal, and generate a discharge warning signal.

2. The intelligent identification and early warning system for the operation risk of power system equipment according to claim 1, characterized in that The passive ultraviolet fluorescence detection module includes: An ultraviolet detector array unit, which is used to collect natural ultraviolet fluorescence signal data in the monitoring area in real time; A fluorescence signal feature extraction unit, which is used to extract features from the natural ultraviolet fluorescence signal data to obtain feature data, specifically including fluorescence intensity feature, fluorescence frequency feature, fluorescence spectrum feature and fluorescence frequency feature; An abnormal preliminary judgment unit, which is used to receive the extracted feature data and compare it with the historical anomaly judgment threshold stored in the threshold storage unit to judge whether there is an anomaly; A marking trigger unit, which is used to record and output the spatial position information of the ultraviolet fluorescence judged to be abnormal, and output an abnormal fluorescence trigger signal; A threshold storage unit, which is used to update the stored historical anomaly judgment threshold according to the dynamic anomaly judgment threshold provided by the background fluorescence analysis and dynamic threshold calculation module.

3. The early-stage ultraviolet fluorescence remote detection system for electrical equipment discharge according to claim 2, wherein, The active ultraviolet laser excitation module includes: A pulsed laser unit, which is used to emit pulsed ultraviolet laser, and the pulse power and width can be continuously adjusted; A fluorescence signal characteristic analysis unit, used to receive characteristic data of the passive ultraviolet fluorescence detection module and calculate an abnormality grade score based on the characteristic data; The time domain coding control unit adaptively sets the power, frequency and time domain coding mode of the laser pulse according to the abnormality level score, wherein: When the abnormal level score is within the first level score range, a simple periodic encoding with low power and low frequency is used; When the abnormality level score is within the second level score range, medium power and medium frequency periodic pulse train encoding is used; When the abnormality level score is within the third level score range, high-power, high-frequency complex non-periodic multi-level coding is used; The servo optical pan-tilt unit is used to receive the spatial position information output by the passive ultraviolet fluorescence detection module, and adjust the emission direction of the laser pulse in real time according to the spatial position information to ensure that the emitted pulsed ultraviolet laser avoids the occurrence area of the abnormal fluorescence signal and stimulates the monitoring area outside the abnormal position to generate a background fluorescence response signal.

4. The intelligent risk identification and early warning system for the operation of power system equipment according to claim 1, wherein The background fluorescence analysis and dynamic threshold calculation module includes: The fluorescence response acquisition unit collects the environmental background fluorescence response signal after active laser pulse excitation in real time to obtain the actual fluorescence characteristics; An expected feature calculation unit calculates the expected feature of background fluorescence according to a pre-established fluorescence feature database and an environmental feature index; A characteristic deviation calculation unit calculates the relative deviation between the actual fluorescence response signal and the expected fluorescence characteristic to obtain the fluorescence characteristic deviation; A dynamic abnormality threshold calculation unit, which calculates a dynamic abnormality judgment threshold according to a fluorescence characteristic deviation; The dynamic threshold output unit outputs the dynamic anomaly threshold of each feature to the passive ultraviolet fluorescence detection module and the anomaly signal confirmation module in real time for updating the historical anomaly threshold and multi-dimensional feature anomaly analysis.

5. The intelligent risk identification and early warning system for the operation of power system equipment according to claim 2, wherein, The method of combining multi-dimensional features with dynamic abnormality judgment threshold analysis to determine whether an abnormality actually exists, and outputting a real abnormality signal if an abnormality actually exists specifically includes the following steps: Acquiring characteristic data of natural ultraviolet fluorescence signal data that is preliminarily determined to be abnormal; Constructing real-time feature vectors and dynamic threshold feature vectors based on feature data and dynamic anomaly judgment thresholds; Calculate the Euclidean distance between the real-time feature vector and the dynamic threshold feature vector; By comparing the Euclidean distance with the preset distance threshold, it is determined whether the abnormal signal is a real abnormality. If the Euclidean distance is greater than the preset distance threshold, it is determined that there is a real abnormality and a real abnormality signal is output.

6. The intelligent risk identification and early warning system for the operation of power system equipment according to claim 1, wherein The intelligent determination module comprises: The multi-cycle integral analysis unit performs integral statistics of multiple power frequency cycles on the natural ultraviolet fluorescence signal with abnormalities to obtain the integral value; The characteristic parameter statistics unit calculates the mean value and standard deviation of the integral value set; The pulse periodicity feature extraction unit calculates the phase concentration of the abnormal natural ultraviolet fluorescence signal within the power frequency period; The adaptive intelligent judgment unit forms a feature vector based on the integral mean value, integral standard deviation and phase concentration characteristics, and uses a machine learning algorithm for classification and judgment to distinguish between corona discharge and line discharge; The discharge confirmation signal output unit outputs a line discharge confirmation signal to the closed-loop verification and alarm module in real time if the adaptive intelligent determination unit determines that the line is discharged.

7. The intelligent risk identification and early warning system for power system equipment operation according to claim 1, wherein The visible light secondary verification and positioning of the abnormal position to generate a discharge warning signal specifically includes the following steps: Real-time capture the visible light image of the abnormal area; According to the spatial position information of the abnormal area, determine the image position of the abnormal area through the camera calibration parameters; Extract the gray contrast feature and the spark detection feature of the abnormal area; Use a machine learning model to automatically classify the abnormal image features and output the category determination probability of the abnormal image; Set an alarm probability threshold. When the classification probability is higher than the threshold, automatically output a discharge warning signal.

Images