Electrical equipment intelligent fault prediction and diagnosis system based on Internet of Things

Abstract

The invention discloses an electrical equipment intelligent fault prediction and diagnosis system based on the Internet of Things, and relates to the technical field of electrical equipment fault detection. Multiple groups of historical data under various fault types of electrical equipment are collected, and contribution values corresponding to parameters are calculated; determining the weight corresponding to each parameter according to the contribution value, combining the actual sequence corresponding to each parameter by using the weight, calculating the similarity of the key I MF of the actual sequence and the fault corresponding sequence, predicting the fault and grade of the electrical equipment according to the similarity, and calculating the score of the ability of a diagnost to process different faults. And according to the ability score and the distance to the electrical equipment, a diagnostor is intelligently arranged. The fault diagnosis waiting time is shortened, and the diagnosis efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical equipment fault detection, and specifically to an intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things. Background Art

[0002] In the modern industrial production system, the stable operation of electrical equipment is extremely important. If a key electrical equipment suddenly fails, it will not only cause product quality problems, but also lead to delays in production plans and even may trigger safety accidents. With the rise of technologies such as big data and artificial intelligence, using these advanced technologies to achieve intelligent fault prediction and diagnosis of electrical equipment has become a popular research direction. However, the existing related technologies still have the following defects: in terms of fault prediction, the prediction accuracy is insufficient in the case of complex and changeable operating environments, and false alarms and missed alarms occur frequently; in terms of the deployment of diagnosticians, there is a lack of a scientific and reasonable evaluation and dispatch mechanism, and it is difficult to quickly match the most suitable diagnostician according to the type and severity of the fault, resulting in low fault handling efficiency and further expanding the losses caused by the fault. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention provides an intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things, which solves the problems of insufficient accuracy of electrical equipment fault prediction, unreasonable deployment of diagnosticians resulting in low fault handling efficiency and large losses.

[0004] To achieve the above objectives, the present invention is realized through the following technical solutions: An intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things, including:

[0005] A fault prediction module, which collects the parameter sequence of electrical equipment at a set time, integrates it into a sequence Q to be processed according to the corresponding weights, performs CEEMDAN decomposition on Q, calculates the energy proportion of each IMF, selects the IMF with the largest energy proportion as the key IMF of Q, calculates the similarity between the key IMF of Q and the key IMFs corresponding to different levels of faults, predicts the fault type and level of the electrical equipment through the similarity, and transmits the predicted fault type and level to the diagnostician intelligent scheduling module;

[0006] A diagnostician ability calculation module, which obtains the ability score of the diagnostician to handle faults through the number of repairs, repair time, and the normal duration of the equipment after repair, and transmits the ability score to the diagnostician intelligent scheduling module;

[0007] Diagnostician intelligent scheduling module. This module screens out diagnosticians who are good at handling the predicted fault types of electrical equipment from the diagnostician database to obtain the preliminary diagnostician set D1. Then, D1 is further screened according to the fault level to obtain the diagnostician set D2. The real-time distance di between each diagnostician in D2 and the faulty equipment is obtained. The comprehensive evaluation index Ri is calculated according to the formula Ri = ɑ * Sik - β * di, and Ri is sorted from large to small. The fault handling task is preferentially sent to the diagnostician with the highest Ri value. If the diagnostician receives the task within the specified time, the scheduling is completed. Otherwise, the system automatically sends the task information to the next diagnostician in the ranking until a diagnostician confirms the receipt of the task.

[0008] As a further solution of the present invention, before the fault prediction module, there also includes a data collection module, a parameter contribution calculation module, and a weight calculation module. The data collection module collects multiple groups of historical data under various fault types of electrical equipment and transmits the historical data to the parameter contribution calculation module. The parameter contribution calculation module classifies the collected historical data according to different fault types, labels each piece of data with a fault label and the specific fault level, determines the contribution value of each parameter corresponding to the fault according to the parameter fluctuation range, parameter change frequency, and causal analysis of the parameter and the fault, and transmits the contribution values of the parameters corresponding to different faults obtained to the weight determination module. Among them, the fault levels include light level, medium level, heavy level, and overweight level. The weight calculation module sorts the parameters of each fault type according to the contribution value, counts the number of the top 3 parameters in the sorting, takes the normalized parameter number as the weight occupied by each parameter, and transmits the weight occupied by each parameter to the fault prediction module.

[0009] As a further solution of the present invention, the specific steps to determine the contribution value corresponding to each parameter are as follows:

[0010] Calculate the absolute fluctuation range and relative fluctuation range of the fault according to the formulas △X = Xmax - Xmin and R = Xmax / Xmin, where Xmax and Xmin are respectively the maximum value and minimum value of the parameter in all fault data, and wabs and wrel are respectively the absolute fluctuation range weight and relative fluctuation range weight;

[0011] Set a threshold for each parameter, and count the number of times each parameter exceeds the set threshold as the change frequency f of the parameter;

[0012] Use expert judgment to score the causal relationship strength, and divide the causal relationship strength into 1 - 5 points, where 1 point means the causal relationship is very weak and 5 points means the causal relationship is very strong;

[0013] Calculate the contribution value of each parameter according to the formula Fi = w1 * Imi xi + w2 * (fi / max(f)) + w3 * ci, where Imi xi is the comprehensive fluctuation index of the i-th parameter, fi is its change frequency, ci is the causal relationship strength, and w1 + w2 + w3 = 1. w1, w2, and w3 can be adjusted according to specific circumstances.

[0014] As a further solution of the present invention, combine the absolute fluctuation range and the relative fluctuation range, and calculate the comprehensive fluctuation index according to the formula lmix = wabs * (△X / max(△X)) + wrel * ((R - 1) / max(R - 1)), where wabs and wrel are the absolute fluctuation range weight and the relative fluctuation range weight respectively, and wabs + wrel = 1.

[0015] As a further solution of the present invention, calculate the sequence Q to be processed according to the formula Q = wv * A1 + wi * B1 + wr * C1 + wt * D1 + wj * E1 + wn * F1 + wrr * G1, where wv, wi, wr, wt, wj, wn, and wrr are the weights of voltage, current, resistance, temperature, vibration, noise, and insulation resistance respectively, and A1, B1, C1, D1, E1, F1, and G1 are the actual sequences corresponding to voltage, current, resistance, temperature, vibration, noise, and insulation resistance respectively.

[0016] As a further solution of the present invention, calculate the similarity Sj between the real-time key IMF and the key IMFs of each fault. The similarity intervals corresponding to different faults in the light level, medium level, heavy level, and overweight level are [a1, a2), [a2, a3), [a3, a4), and [a4, a5] respectively. If Sj belongs to one of the above intervals, the corresponding fault and level of the electrical equipment can be predicted next. If Sj > a5, it is predicted that no fault will occur in the electrical equipment next.

[0017] As a further solution of the present invention, the specific steps for calculating the ability score of the diagnostician to handle each fault are as follows:

[0018] Calculate the repair times score according to the formula SA = (A / max(A)) * 10, where A is the repair times of the diagnostician for a certain fault, and max(A) is the maximum repair times of all diagnosticians for this fault;

[0019] Calculate the repair time score according to the formula SB = 10 - ((arg(B) + bo(B)) / max(arg(B) + bo(B))), where the sequence B = [b1, b2,..., bn], n is the repair times, and arg(B) and bo(B) are the average value and the fluctuation degree of B;

[0020] Calculate the normal duration score of the repaired equipment according to the formula SC = 5 + ((d iffarg - d iffextreme) / max(d iffarg - d iffextreme)) * 5, where d iffarg is the average difference and d iffextreme is the difference in the number of extreme points;

[0021] According to the formula S 总 = SA + SB + SC to calculate the ability score of each diagnostician to handle each type of fault.

[0022] As a further solution of the present invention, the specific steps for obtaining d iffarg and d iffextreme are as follows:

[0023] Collect the composition sequence C = [c1, c2,..., cn] of the normal duration after the diagnostician's repair;

[0024] Divide C into two equal parts, the front and the back. If n is odd, the first half has one more data than the second half. The first half sequence C1 = [c1, c2,..., cm], calculate the number of extreme points as extreme1 and the average value as arg1. The second half sequence C2 = [cm + 1,..., cn], calculate the number of extreme points as extreme2 and the average value as arg2;

[0025] Calculate the difference in the number of extreme points d iffextreme = extreme1 - extreme2, and calculate the average difference diffarg = arg2 - arg1.

[0026] As a further solution of the present invention, if Si k >= Tk, add the diagnostician i to D1, and select the diagnosticians corresponding to Si k >= Hq from D1 and add them to D2, where Si k is the ability score of each diagnostician i for the fault type k, Tk is the set threshold, and Hq is the lower limit of the diagnostician ability score corresponding to different levels.

[0027] As a further solution of the present invention, during the process of the diagnostician going to the fault site and handling the fault, the system continuously tracks the task progress, records the arrival time tstart of the diagnostician at the site, records the completion time tend of the fault handling, calculates the fault handling duration tdea l = tend - tstart, and at the same time, feeds back the task handling situation of the diagnostician to the diagnostician ability calculation module for real-time updating of the diagnostician's ability score.

[0028] The present invention provides an intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things. Compared with the prior art, it has the following beneficial effects:

[0029] (1) By collecting multi-parameter historical data and conducting comprehensive analysis, the present invention can comprehensively capture the variation characteristics of each parameter of electrical equipment in different fault states. The system can more accurately determine the contribution degree of each parameter to the fault based on these variation characteristics, and obtain the weight of each parameter using the contribution degree, thus preparing for the establishment of a more accurate fault prediction model;

[0030] (2) The present invention reasonably dispatches according to the proficiency of diagnosticians in different faults and the current distance, ensuring that the most suitable personnel can be quickly found for handling when a fault occurs, reducing the waiting time for fault diagnosis, and improving the diagnosis efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a system principle block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0033] As Figure 1 , the present invention provides an intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things, including:

[0034] A data collection module, the main function of which is to collect multiple groups of historical data of electrical equipment under various fault types, covering voltage, current, resistance, temperature, vibration, noise, and insulation resistance, collecting comprehensive and accurate data to prepare for determining effective weights for each parameter subsequently;

[0035] To collect this data, various sensors need to be installed on electrical equipment. Take the motors in a large factory as an example. Voltage sensors are installed to collect voltage. When the motor is operating normally, the voltage sensor can monitor in real time whether the input voltage is stable. If there is a sudden large fluctuation in the input voltage, it may indicate a power grid fault or a problem inside the motor. Current sensors are installed to collect current. The current sensor can monitor the working current in real time. When the motor load increases, the working current will increase accordingly. If the current value exceeds the normal range, it may indicate faults such as overload or short circuit in the motor. Thermistor sensors are installed to collect the resistance of the motor winding. It can monitor the change in the winding resistance in real time. An abnormal change in the winding resistance may mean problems such as short circuit, open circuit, or insulation damage in the winding. Thermocouple temperature sensors are installed on the stator core of the motor to collect temperature, which can monitor the temperature of the core in real time. When the motor operates at high load for a long time, the core temperature will rise. Piezoelectric vibration sensors are installed to collect vibration data. Piezoelectric vibration sensors are installed on the bearing housing of the motor. When the motor is operating normally, the vibration signal detected by the sensor is within the normal range. Once faults such as bearing wear or rotor imbalance occur, the vibration of the motor will intensify, and the vibration signal detected by the vibration sensor will change significantly. Noise sensors are installed to collect noise data. The sensor monitors the intensity of the noise in real time. Under normal circumstances, the noise of the motor is relatively stable and within a specific intensity range. When faults occur in the motor, such as bearing wear or rotor imbalance, the intensity of the noise will change significantly. Insulation resistance sensors are installed to collect the insulation resistance value between the motor winding and the shell. The sensor measures the insulation resistance regularly. Under normal circumstances, the insulation resistance value of the motor should be maintained at a relatively high level. If the insulation resistance value drops, it may mean that the insulation material is damp, aged, or damaged, posing safety hazards such as electric leakage and short circuit.

[0036] The parameter contribution calculation module classifies the collected historical data according to different fault types, and labels each piece of data with a clear fault label and specific fault level. For example, for motor fault data, it distinguishes whether it is an overheat fault, a winding short circuit fault, or a bearing fault, etc., and then gives a fault level classification for the specific fault, such as light level, medium level, heavy level, super heavy level, etc., so as to conduct in-depth analysis for specific fault types in the future.

[0037] For each fault type, calculate the parameter contribution based on the corresponding multiple historical data. The specific calculation method is as follows:

[0038] (1) Calculation of parameter fluctuation range: Calculate the absolute fluctuation range and relative fluctuation range of this fault according to the formulas △X = Xmax - Xmin and R = Xmax / Xmin. The absolute fluctuation range reflects the possible change span of the parameter during the fault occurrence. The relative fluctuation range takes into account the influence of the parameter's own magnitude on the change and can more intuitively reflect the relative degree of parameter change. Combine the two fluctuation ranges and calculate the comprehensive fluctuation index according to the formula lm_ix = wabs * (△X / max(△X)) + wrel * ((R - 1) / max(R - 1)), where Xmax and Xmin are the maximum and minimum values of this parameter in all fault data respectively, wabs and wrel are the weights of the absolute fluctuation range and relative fluctuation range respectively, and wabs + wrel = 1;

[0039] (2) Statistics of parameter change frequency: According to the characteristics of the parameter and practical experience, set a change threshold for each parameter. This threshold is used to judge whether the parameter change is significant. Traverse all fault data and count the number of times each parameter exceeds the set threshold as the change frequency f of this parameter. The change frequency reflects the frequency of significant changes of the parameter during the fault process. For example, for the temperature parameter, set that a temperature change exceeding 5°C is a significant change, and if the number of times exceeding 5°C in all fault data is 100, then 100 is the change frequency of this parameter;

[0040] (3) Analysis of the causal relationship between parameters and faults: Analyze in detail the change situations of each parameter before, during, and after the fault occurrence, and judge the logical causal relationship in time between the parameter change and the fault. For example, it is observed that the temperature of the motor winding continuously rises before the fault occurs, and then the motor has an overheating fault, indicating that the rising winding temperature may be one of the reasons for the motor overheating fault; if a certain parameter does not change, analyze whether the fault will still occur in the same way and degree. For example, if the temperature of the motor winding remains stable, although other parameters have certain changes, but the motor overheating fault does not occur, which further proves that there is a strong causal relationship between the winding temperature and the motor overheating fault. Expert judgment can be used to score the strength of the causal relationship, and the strength of the causal relationship is divided into 1 - 5 points, with 1 point indicating a very weak causal relationship and 5 points indicating a very strong causal relationship;

[0041] (4) Determine the parameter contribution factor and sort: Comprehensively consider the fluctuation range, change frequency, and causal relationship of the parameter, and calculate the contribution value of each parameter according to the formula Fi = w1 * Imix i + w2 * (fi / max(f)) + w3 * ci, where Imix iis the comprehensive fluctuation index of the i-th parameter, fi is its change frequency, ci is the causal relationship strength, and w1 + w2 + w3 = 1. w1, w2, and w3 can be adjusted according to specific situations and experience. For example, if more attention is paid to the change range of the parameter, w1 can be appropriately increased; if it is considered that the change frequency is more important, w2 can be appropriately increased; if more attention is paid to the causal relationship, w3 can be appropriately increased. The parameters are sorted in descending order according to the calculated contribution factor Fi. The larger Fi is, the higher the importance of the parameter to the fault.

[0042] Weight determination module, which statistically analyzes the sorting parameters of each fault type, statistically analyzes the parameters ranked in the top 3, and obtains the normalized values as the weights of each parameter.

[0043] For example, when there are faults in various electrical equipment in a power system, such as transformers, motors, switch cabinets, etc., the parameter sorting of all fault types is statistically analyzed. The statistical results show that among 10 fault types, the voltage parameter ranks in the top 3 for 4 times, the current parameter ranks in the top 3 for 3 times, the resistance parameter ranks in the top 3 for 2 times, the temperature parameter ranks in the top 3 for 15 times, the vibration parameter ranks in the top 3 for 2 times, the noise parameter ranks in the top 3 for 3 times, and the insulation resistance parameter ranks in the top 3 for 1 time. The total number of times all parameters rank in the top 3 is 30 times. Then the weight of the voltage parameter is 2 / 15, the weight of the current parameter is 1 / 10, the weight of the resistance parameter is 1 / 15, the weight of the temperature parameter is 1 / 2, the weight of the vibration parameter is 1 / 15, the weight of the noise parameter is 1 / 10, and the weight of the insulation resistance parameter is 1 / 30.

[0044] Fault prediction module, which collects the operating parameters of electrical equipment in real time at a set time interval TT, combines the previously determined parameter weights, integrates these parameters into the sequence Q to be processed, and then performs CEEMDAN decomposition on Q. CEEMDAN decomposition is an adaptive signal decomposition method that can decompose complex signals into multiple intrinsic mode functions IMFs with different frequency characteristics. For the decomposed IMF components, calculate the energy proportion of each IMF. The energy proportion reflects the energy contribution degree of the IMF in the original signal. Select the IMF with the largest energy proportion as the key IMF of sequence Q, and perform normalization processing on it, mapping the values of this IMF to the interval [0,1]. The purpose of this is to eliminate the influence of the data dimension, make the key IMF data of different times and different equipment comparable, and prepare for the subsequent fault prediction of electrical equipment.

[0045] For example, we are monitoring a large motor in a complex industrial environment, with a set time interval TT = 1 minute. At a certain moment, the following real-time data sequences are collected. For the sake of simplicity in calculation, only 3 data are listed in each sequence: voltage sequence A1 = [380, 382, 381], current sequence B1 = [50, 52, 51], resistance sequence C1 = [0.5, 0.52, 0.51], temperature sequence D1 = [60, 61, 60.5], vibration sequence E1 = [2.5, 2.6, 2.55], noise sequence F1 = [65, 66, 65.5], insulation resistance sequence G1 = [10, 10.2, 10.1]. After calculation by the weight determination module, the weights of each parameter are obtained as follows: voltage weight wv = 0.15, current weight wi = 0.2, resistance weight wr = 0.1, temperature weight wt = 0.2, vibration weight wj = 0.1, noise weight wn = 0.05, insulation resistance weight wrr = 0.2. According to the formula Q = wv*A1 + wi*B1 + wr*C1 + wt*D1 + wj*E1 + wn*F1 + wrr*G1, the sequence Q to be processed is calculated as Q = [84.55, 85.55, 85.05]. Subsequently, the sequence Q will be decomposed by CEEMDAN, the energy proportion of each IMF will be calculated, the IMF with the largest energy proportion will be found and normalized, and it will be used as the key IMF of the sequence Q for comparative analysis with historical fault data to determine whether the operation state of the motor is normal.

[0046] Calculate the similarity between the key IMF corresponding to the real-time sequence Q and the key IMFs corresponding to different levels of fault types, and judge the possible fault types of electrical equipment according to the similarity;

[0047] For example, in the historical key IMF library of a certain power company, the key IMFs corresponding to different levels of fault types such as minor faults and severe faults of transformers are stored. When a certain transformer is monitored in real time, the key IMF corresponding to the real-time sequence Q is obtained. Using the Euclidean distance calculation method, calculate the similarity Sj between the real-time key IMF and the key IMFs corresponding to each different level of fault type j in the historical key IMF library. The smaller Sj is, the higher the similarity;

[0048] If the calculated similarity Sj between the real-time key IMF and the key IMF corresponding to the minor fault of the transformer is 0.8, and the similarity interval of the minor fault of the transformer is [0.7, 0.9], then the system predicts that the transformer may have a minor fault. At this time, the system will send out a warning signal in time to call a diagnostician to check and maintain the transformer.

[0049] The reasons for selecting the IMF with the largest energy proportion as the key IMF of the sequence Q specifically include the following three points:

[0050] The IMF with the largest energy usually contains the most important information and energy distribution in the original signal. In the operation data of electrical equipment, it means that it can reflect the key features and change trends of the equipment operation state to the greatest extent. For example, in the motor fault monitoring, if the IMF with the largest energy reflects the frequency features related to vibration, then when there are abnormal changes in this IMF, it is very likely that there are faults in the mechanical structure of the motor, such as bearings, rotors, etc. By paying attention to this key IMF, the abnormal changes in the equipment operation state can be captured more directly and effectively;

[0051] The data collected during the operation of electrical equipment will inevitably be affected by various noise interferences. By selecting the IMF with the largest energy, the IMF components with smaller energy and possibly caused by noise can be filtered out to a certain extent, thereby improving the accuracy and reliability of data processing. For example, some high-frequency micro-fluctuations may be generated by sensor noise or environmental interference, and they account for a very small proportion in the total energy. By selecting the IMF with the largest energy, the influence of these noises can be minimized;

[0052] Selecting the IMF with the largest energy from multiple IMF components for subsequent analysis greatly simplifies the complexity of data analysis. In practical applications, in the face of a large amount of operation data and complex signal features, if each IMF is analyzed in detail, not only the calculation amount is huge, but also it is easy to get stuck in complex details and it is difficult to quickly and accurately judge the operation state of the equipment. By selecting the key IMF, the most important features can be focused on, potential fault hazards can be quickly discovered, and the efficiency of fault prediction can be improved.

[0053] Diagnostician ability calculation module, which comprehensively considers the repair times, repair time, and normal duration of the equipment after repair of the diagnostician for different types of faults, and calculates the proficiency score of the diagnostician for each type of fault;

[0054] For the repair times score, it is required that the more times of repairing a certain type of fault, the better. If the number of times of repairing a certain fault by the diagnostician is A, and the maximum number of times of repairing this fault by all diagnosticians is max(A), with a full score of 10 points, the score: SA=(A / max(A))*10;

[0055] For the repair time score, it is required that the shorter the time for each repair, the better. If the time sequence for each repair of a certain fault by the diagnostician is B = [b1, b2,..., bn], where n is the number of repairs, the average value is arg(B), and the degree of fluctuation is bo(B), with a full score of 10 points. Since it is required that both arg(B) and bo(B) should be smaller to get a higher score, so here a reverse calculation is adopted, and the score SB = 10 - ((arg(B)+bo(B)) / max(arg(B)+bo(B)));

[0056] For the normal duration of the equipment after maintenance, it is required that the normal duration of the equipment after each maintenance is as long as possible. If the sequence C = [c1, c2,..., cn] represents the normal duration of the equipment after the diagnostician's maintenance, the sequence C is evenly divided into two parts, the front and the back. If n is odd, the first half has one more data than the second half. The first half sequence C1 = [c1, c2,..., cm], calculate the number of its extreme points as extreme1 and the average value as arg1. The second half sequence C2 = [cm+1,..., cn], calculate the number of its extreme points as extreme2 and the average value as arg2. Calculate the difference in the number of extreme points diffextreme = extreme1 - extreme2, where extreme1 is always larger than extreme2. Calculate the difference in the average values diffarg = arg2 - arg1, where arg2 is always larger than arg1. The full score is 10 points, and the score SC = 5 + ((diffarg - diffextreme) / max(diffarg - diffextreme)) * 5. If diffarg - diffextreme is larger, it means that the number of extreme points decreases and the average value increases in the later stage, and the higher the score, that is, the better the performance;

[0057] According to the formula S 总 = SA + SB + SC to calculate the comprehensive score of this diagnostician for this type of fault.

[0058] Diagnostician intelligent scheduling module, which realizes the efficient and accurate scheduling of diagnosticians according to the fault prediction results, the professional capabilities and real-time location information of diagnosticians;

[0059] After the system predicts the fault type of the electrical equipment, select the diagnosticians who are good at handling this type of fault from the diagnostician database. The database stores the fault types that each diagnostician is good at and the corresponding ability scores. If there are u diagnosticians in the diagnostician database, the ability score of each diagnostician i for the fault type k is Sik. Set a screening threshold Tk, which is determined according to historical data and actual experience, and the thresholds for different fault types may be different. The screening condition is Sik >= Tk. The diagnostician i who meets this condition is initially selected to form the initial diagnostician set D1;

[0060] Faults of different levels require different professional capabilities of diagnosticians. D1 is screened again according to the fault level. The fault level can be divided into light, medium, heavy, and overweight. Different levels require diagnosticians with different diagnostic capabilities. Assuming the fault level is q, the lower limit of the ability score of diagnosticians corresponding to different levels is Hq, which is determined according to historical data and practical experience. For diagnostician i in set D1, if his ability score Sik >= Hq for the current fault type, he will be retained in the diagnostician set D2 after the second screening;

[0061] After determining set D2, obtain the real-time distance di between each diagnostician in the set and the faulty equipment, and calculate the comprehensive evaluation index Ri according to the formula Ri = ɑ * Sik - β * di, where ɑ and β are weight coefficients, 0 < ɑ, β < 1 and ɑ + β = 1. These two parameters need to be adjusted according to the actual situation to balance the importance of ability score and distance in the comprehensive evaluation. The larger ɑ is, the more attention is paid to the professional ability of the diagnostician. The larger β is, the more importance is attached to the distance factor. Sort the diagnosticians in set D2 in descending order according to the Ri value;

[0062] According to the sorting result, the fault handling task information is preferentially sent to the diagnostician with the highest Ri value. If this diagnostician confirms receiving the task within the specified time, the scheduling is completed. If this diagnostician is unable to undertake the task due to various reasons, such as being dealing with other urgent faults, unable to respond in time, etc., the system automatically sends the task information to the next diagnostician in the sorting, and so on, until a diagnostician confirms receiving the task;

[0063] During the process of the diagnostician going to the fault site and handling the fault, the system continuously tracks the task progress and records the time t when the diagnostician arrives at the site start and records the time t when the fault handling is completed end and calculates the fault handling duration t deal = t end - t start At the same time, record the handling result of the fault by the diagnostician, such as whether the fault is successfully eliminated, whether there are any remaining problems, etc.;

[0064] Feed back the task handling situation of the diagnostician to the diagnostician ability calculation module to update the ability score of the diagnostician. For example, if a diagnostician has a short handling duration and good handling results when dealing with a certain type of fault, his corresponding score may increase when calculating his ability score next time. On the contrary, if the handling duration is too long or the handling result is not ideal, the score may decrease. Through continuous feedback and optimization, the ability score of the diagnostician is made more accurate, thereby improving the overall performance of the diagnostician intelligent scheduling module.

[0065] Some of the data in the above formula are numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0066] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things, characterized in that, Including: A fault prediction module, which collects the parameter sequence of the electrical equipment at set time intervals, integrates it into a sequence Q to be processed according to corresponding weights, performs CEEMDAN decomposition on Q, calculates the energy proportion of each IMF, selects the IMF with the largest energy proportion as the key IMF of Q, calculates the similarity between the key IMF of Q and the key IMFs corresponding to different levels of faults, predicts the fault type and level of the electrical equipment through the similarity, and transmits the predicted fault type and level to the diagnostician intelligent scheduling module; A diagnostician ability calculation module, which obtains the ability score of the diagnostician to handle faults through the number of repairs, repair time, and the normal duration of the equipment after repair, and transmits the ability score to the diagnostician intelligent scheduling module; A diagnostician intelligent scheduling module, which screens out the diagnosticians who are good at handling this type of fault from the diagnostician database according to the predicted fault type of the electrical equipment to obtain a preliminary diagnostician set D1, further screens D1 according to the level of the fault to obtain a diagnostician set D2, obtains the real-time distance di between each diagnostician in D2 and the faulty equipment, calculates the comprehensive evaluation index Ri according to the formula Ri = ɑ * Sik - β * di, sorts Ri from large to small, and preferentially sends the fault handling task to the diagnostician with the highest Ri value. If this diagnostician receives the task within the specified time, the scheduling is completed; otherwise, the system automatically sends the task information to the diagnostician ranked next in the order until a diagnostician confirms receiving the task.

2. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 1, wherein Before the fault prediction module, there are also a data collection module, a parameter contribution calculation module, and a weight calculation module. The data collection module collects multiple groups of historical data under various fault types of the electrical equipment and transmits the historical data to the parameter contribution calculation module. The parameter contribution calculation module classifies the collected historical data according to different fault types, labels each piece of data with a fault label and the specific fault level, determines the contribution value of each parameter corresponding to this fault according to the parameter fluctuation range, parameter change frequency, and causal analysis of the parameter and the fault, and transmits the contribution values of the parameters corresponding to different faults obtained to the weight determination module. Among them, the fault levels include light level, medium level, heavy level, and extra-heavy level. The weight calculation module sorts the parameters of each fault type according to the contribution value, counts the number of the top 3 parameters in the sorting, takes the normalized parameter number as the weight of each parameter, and transmits the weight of each parameter to the fault prediction module.

3. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 2, characterized in that, The specific steps to determine the contribution value corresponding to each parameter are as follows: Calculate the absolute fluctuation range and relative fluctuation range of this fault according to the formulas △X = Xmax - Xmin and R = Xmax / Xmin, where Xmax and Xmin are respectively the maximum and minimum values of this parameter in all fault data, and wabs and wrel are respectively the absolute fluctuation range weight and relative fluctuation range weight; Set a threshold for each parameter and count the number of times each parameter exceeds the set threshold as the change frequency f of this parameter; Expert judgment is used to score the strength of the causal relationship, and the strength of the causal relationship is divided into 1-5 points. 1 point indicates a very weak causal relationship, and 5 points indicates a very strong causal relationship; Calculate the contribution value of each parameter according to the formula Fi = w1*Imixi + w2*(fi / max(f)) + w3*ci, where Imixi is the comprehensive fluctuation index of the i-th parameter, fi is its change frequency, ci is the strength of the causal relationship, and w1 + w2 + w3 = 1. w1, w2, and w3 can be adjusted according to specific circumstances.

4. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 3, characterized in that, Combine the absolute fluctuation range and the relative fluctuation range, and calculate the comprehensive fluctuation index according to the formula lmix = wabs*(△X / max(△X)) + wrel*((R - 1) / max(R - 1)), where wabs and wrel are the weights of the absolute fluctuation range and the relative fluctuation range respectively, and wabs + wrel = 1.

5. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 2, wherein Calculate the sequence Q to be processed according to the formula Q = wv*A1 + wi*B1 + wr*C1 + wt*D1 + wj*E1 + wn*F1 + wrr*G1, where wv, wi, wr, wt, wj, wn, and wrr are the weights of voltage, current, resistance, temperature, vibration, noise, and insulation resistance respectively, and A1, B1, C1, D1, E1, F1, and G1 are the actual sequences corresponding to voltage, current, resistance, temperature, vibration, noise, and insulation resistance respectively.

6. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 1, characterized in that, Calculate the similarity Sj between the real-time key IMF and the key IMF of each fault. The similarity intervals corresponding to different faults in the light, medium, heavy, and overweight levels are [a1, a2), [a2, a3), [a3, a4), and [a4, a5] respectively. If Sj belongs to one of the above intervals, the corresponding fault and level of the electrical equipment can be predicted next. If Sj > a5, it is predicted that no fault will occur in the electrical equipment next.

7. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 1, wherein The specific steps to calculate the score of the diagnostician's ability to handle each fault are as follows: Calculate the repair times score according to the formula SA = (A / max(A)) * 10, where A is the number of repairs of a certain fault by the diagnostician, and max(A) is the maximum number of repairs of this fault by all diagnosticians; Calculate the repair time score according to the formula SB = 10 - ((arg(B) + bo(B)) / max(arg(B) + bo(B))), where the sequence B = [b1, b2,..., bn], n is the number of repairs, and arg(B) and bo(B) are the average value and the degree of fluctuation of B; Calculate the score of the normal duration of the equipment after repair according to the formula SC = 5 + ((diffarg - diffextreme) / max(diffarg - diffextreme)) * 5, where diffarg is the difference in average values and diffextreme is the difference in the number of extreme points; According to the formula S 总 = SA + SB + SC, calculate the score of the diagnostician's ability to handle each type of fault.

8. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 7, wherein, The specific steps to obtain diffarg and diffextreme are as follows: Collect the sequence C = [c1, c2,..., cn] composed of the normal duration after the diagnostician's repair; Divide C into two equal parts, the front and the back. If n is odd, the front part has one more data than the back part. The front part sequence C1 = [c1, c2,..., cm], calculate the number of its extreme points as extreme1 and the average value as arg1. The back part sequence C2 = [cm+1,..., cn], calculate the number of its extreme points as extreme2 and the average value as arg2; Calculate the difference in the number of extreme points diffextreme = extreme1 - extreme2, and calculate the difference in the average value diffarg = arg2 - arg1.

9. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 1, characterized in that, If Sik >= Tk, add diagnostician i to D1, and then select the diagnosticians corresponding to Sik >= Hq from D1 and add them to D2, where Sik is the ability score of each diagnostician i for fault type k, Tk is the set threshold, and Hq is the lower limit of the ability scores of diagnosticians corresponding to different levels.

10. The intelligent fault prediction and diagnosis system for electrical equipment based on the Internet of Things according to claim 1, characterized in that, During the process of the diagnostician going to the fault site and handling the fault, the system continuously tracks the task progress, records the arrival time tstart of the diagnostician at the site, records the completion time tend of the fault handling, calculates the fault handling duration tdeal = tend - tstart, and at the same time, feeds back the task handling situation of the diagnostician to the diagnostician ability calculation module for real-time updating of the diagnostician's ability score.

Images