CN112580993B - Power grid equipment fault probability analysis method - Google Patents

Abstract

The invention relates to a power grid equipment fault probability analysis method, and belongs to the technical field of power regulation and control. The method comprises the following steps: taking the health state of the equipment as a center, and carrying out multi-dimensional data acquisition and structural analysis; cross identification and cleaning of data; respectively establishing an equipment health state evaluation model in each dimension; setting weight coefficients of all dimension factors based on expert experience; automatic optimization adjustment of multidimensional data weight coefficients based on historical data; analyzing and calculating the equipment tripping probability by combining historical data; and handling based on different tripping probabilities. The method can effectively calculate and analyze the fault probability of the power grid equipment, carries out corresponding processing, and is easy to popularize and apply.

Description

A kind of power grid equipment failure probability analysis method

technical field

The invention belongs to the technical field of power regulation, and in particular relates to a failure probability analysis method for power grid equipment.

Background technique

With the continuous expansion of the scale of the power grid, the influx of AC and DC transmission hybrid and distributed power sources, the wiring mode and operation mode of the power grid are becoming more and more complicated. On the other hand, with the promotion of the integrated mode of large-scale operation and regulation, hundreds of substations and related lines are centrally regulated by a small number of operating personnel, and the work pressure and mental pressure of personnel are increasing day by day.

In the unattended mode of the substation, the regulators rely on remote signaling, telemetry and other information to monitor the operation status of the power grid equipment; due to the large number of equipment under their jurisdiction, they receive a large amount of various kinds of information every day, which makes the operators tired. . How to rely on technical means to dig out the failure risk of power grid equipment from various types of information has become the focus and difficulty of the work of regulators at all levels. In the process of power grid operation and management, remote signaling alarms, telemetry out-of-limits, online monitoring of power transmission and transformation, on-site inspections, and test data may all feedback equipment problems and deterioration trends. Analysis, mining equipment health status. However, the following deficiencies generally exist:

1. Among the various types of data used to analyze equipment status, many are unstructured data, and there are a certain amount of incorrect data, duplicate data, missing data, and conflicting data, which affect the quality of equipment status analysis.

2. There are many dimensions that affect the operation status of equipment, and the same dimension is divided into different sub-factors and severity levels. In the process of evaluating the status of various equipment, it is difficult to form a unified judgment system between each dimension for synthesis. Evaluate.

3. Past judgments rely too much on expert experience to make judgments. There are problems in the number of samples, adaptability to different operating modes, and adaptability to new features of the current power grid, which may easily lead to judgment deviations.

Therefore, how to overcome the shortcomings of the existing technology is an urgent problem to be solved in the current field of power regulation technology.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the deficiencies of the prior art, and provide a power grid equipment failure probability analysis method combining expert experience and historical data. The causal relationship model is built, analyzed and perfected. On this basis, various types of information are cross-identified, and the data is deduplicated, completed, and corrected. Based on the corrected data, the related theories of the AHP are used to build equipment health Status analysis model; based on historical equipment health assessment results and records of unplanned power outages, the equipment health assessment model and failure probability model are adaptively adjusted.

For achieving the above object, the technical scheme adopted in the present invention is as follows:

A method for analyzing the failure probability of power grid equipment, comprising the following steps:

In step (1), multi-dimensional data collection and structured analysis are carried out with the health status of the equipment as the center;

Step (2), cross-identification and cleaning of data;

In step (3), an equipment health status assessment model is established for each dimension;

Step (4), the initial setting of the weight coefficients of each dimension factor based on expert experience;

Step (5), based on the automatic optimization adjustment of the multidimensional data weight coefficient of historical data;

Step (6), analyze and calculate the equipment tripping probability in combination with historical data;

Step (7), dealing with different trip probabilities.

Further, preferably, the data collected in step (1) includes four dimensions: equipment defects, historical operating conditions, operating years and family hidden dangers;

The said family hidden danger is the average number of the corresponding type of defects in the model of the manufacturer;

Analyze the collected data separately.

Further, preferably, the specific method of step (2) is: cleaning the data by building a cross-identification model between multi-dimensional data; the details are as follows:

(2.1) Identification of telemetry data quality: based on the normal range and change speed limit of telemetry, identify the number of hops and errors in telemetry; based on the historical data of telemetry values, extract the unchanged telemetry; Perform power balance verification on active and reactive power on each side of transformers, lines, and bus bars to identify abnormal data; clean up hop counts, wrong counts, unchanged telemetry data, and abnormal data in telemetry;

(2.2) Cross-identification of remote signaling and telemetry data: identify the operation mode of each interval through remote signaling and equipment topology; establish a feature point model for telemetry in operating intervals and a feature point model for telemetry in non-operating intervals; through remote signaling and telemetry corresponding intervals State feature points, perform cross-identification, clean up false remote signaling alarms, extract a list of abnormal telemetry data, and provide them to the staff for inspection and processing;

The feature point model of the operating interval telemetry is that the active power or reactive power is not 0 in the operating interval except for the case of empty charging of the line; the feature point model of the non-operating interval telemetry is that the active power, reactive power and current of the non-operating interval are 0.

Further, preferably, the specific method of step (3) is: using the data cleaned in step (2), for each dimension factor, establish an equipment health status assessment model respectively, and carry out the assessment of the equipment health status, specifically:

(3.1) For defect records: According to the type of defective equipment and the classification of defects, establish a standard table of deduction points for defects, and deduct points for the defects existing in the equipment; set a general deduction standard according to the level of defects, deduct 41 points for critical defects, and deduct 41 points for major defects. 11 points are deducted for defects, and 3 points are deducted for general defects;

The final deduction value of the defect factor is recorded as S ₁ ;

(3.2) For historical operating conditions: the historical operating conditions are divided into short-circuit impact accumulation, line heavy overload, main transformer heavy overload, and oil temperature exceeding the limit;

①For the accumulation of short-circuit shocks, 11 points will be deducted for each shock, and the trip will be accumulated once. If the reclosing is unsuccessful, the points will be deducted according to the interval between the successful reclosing and the tripping again. The points will be deducted once every 1 second, and 11 points will be deducted each time;

②For the heavy overload of the line and the heavy overload of the main transformer, the calculation formula of the corresponding deduction for the damage to the equipment is:

The deductions are:

Among them, L _A is the average value of the current limit violation, L _M is the maximum value of the current limit violation, and T(L) is the duration of the current limit violation;

③ For the oil temperature exceeding the limit, considering the exceeding temperature and the accumulated time, the calculation formula for the deduction corresponding to the damage to the main transformer is:

The deduction value is: (e ^(t-75)×0.1 -1)×T(t)

Among them, t is the maximum oil temperature during the oil temperature overrun period, and T(t) is the duration of this overrun;

Short-circuit impact, overload of line or main transformer, and oil temperature exceeding the limit, the deductions calculated by the three models are accumulated in units of equipment, as the overall deduction of historical operating conditions of the equipment;

The final deduction value of the historical operating condition factor is recorded as S ₂ ;

(3.3) For the operating life factor, let the standard life of the equipment be Y, and the current equipment has been running for y years;

For equipment operating for more than 90% of its standard life, the deductions are:

No points will be deducted for equipment whose operating time does not exceed 90% of the standard life;

The final deduction value of the operating years factor is recorded as S ₃ ;

(3.4) For family hidden dangers, the deduction is as follows:

Among them, F is a certain type of defect, P(F) is the average probability of such defects in the same equipment type and voltage level, P _D (F) is the average probability of such defects in a manufacturer's model; S ₁ (F ) is the deduction value corresponding to defect F in the defect deduction standard table.

The final deduction value of the family hidden danger factor is recorded as S ₄ .

Further, preferably, the specific method of step (4) is: specifying the initial coefficients of each factor by experts, that is, specifying a situation, in the case of a certain defect, heavy overload of equipment, equipment operating years, and family hidden dangers, various The proportion of factors affecting the health of equipment, take the average value of multiple experts, and then calculate the effective coefficient through the calculation of various factors and the final average proportion of various factors to calculate the effective coefficient;

In the process of parameter initial setting, the analytic hierarchy process is used to complete, and the specific method is as follows:

(4.1) Case combination extraction of different factors: For defect records, historical operating conditions, operating years, and family hidden dangers, typical samples are selected and combined to obtain combined cases of multiple factors;

(4.2) Experts assess the relationship of importance of each factor: select 3 or more experts to compare and assess the weights of different factors in the selected case combination for equipment tripping, and input them as parameters of the evaluation matrix;

(4.3) Calculate the weight of the contribution of each factor to the equipment tripping judgment by the analytic hierarchy process, which is respectively recorded as L _n ; wherein, n=1, 2, 3, 4, corresponding to four types of factors respectively;

(4.4) Adjust the coefficients of the evaluation model established in step (3) in turn: for the four types of factors, through the model calculation, the obtained deduction points are S _n , n=1, 2, 3, 4, corresponding to four class factor; combined with the results calculated in (5.3), the calculation formula of the coefficient R _n calculated according to this group of case combinations is:

The coefficient R _n is the result calculated for this case combination;

m是案例组合的序号，取值从1到M，M是案例组合的数量；缺陷记录对应的系数，经专家评定之后对应的系数计算公式为：(4.5) Set the initial coefficient according to the combination of multiple groups of cases: for defect records, each group of cases will analyze a coefficient R ₁ , and record this sequence as

m is the serial number of the case combination, ranging from 1 to M, where M is the number of case combinations; the coefficient corresponding to the defect record, and the corresponding coefficient calculation formula after expert evaluation is:

This parameter will be used as the initial value for parameter adjustment;

For the other three factors, corresponding calculations are performed in turn to obtain the initial values of all factors.

Further, preferably, step (5) is based on the initial setting of the weight coefficients of different dimension factors according to expert experience, by adjusting the weight coefficients, and combining historical data to perform consistency verification on different weight combinations to obtain the optimal coefficient. combination;

Specific steps are as follows:

(5.1) Extract the equipment that has experienced unplanned power outages within the scope of analysis;

的25％为步长，以

为有效区间，遍历四个系数的不同组合，提取可选的系数组合列表，采用步骤(5.3)至步骤(5.5)进行最优系数的粗调；(5.2) For each initial coefficient, n=1, 2, 3, and 4, which correspond to four types of factors, respectively.

25% of the step size to

For the effective interval, traverse the different combinations of the four coefficients, extract a list of optional coefficient combinations, and use steps (5.3) to (5.5) to perform rough adjustment of the optimal coefficients;

(5.3) Select a historical time interval with complete defects and operating conditions, for each coefficient combination in the coefficient combination list, take days as the unit, according to the evaluation method of step (3), take the selected coefficient combination as the coefficient, Calculate the deductions for each device in the power grid;

(5.4) Detect the consistency of each group of coefficients: for each device, sort the deductions from large to small for each device; traverse the sequence of deductions for each day from front to back. If an unplanned power outage occurs, but the deduction value of n unplanned outages is less than the current deduction value, n is added to the inconsistency; the total number of inconsistencies is recorded as U _C ;

(5.5) Selection of the optimal effective coefficient: For the inconsistency quantity U _C calculated by the coefficient combination (R ₁ , R ₂ , R ₃ , R ₄ ), calculate the inconsistency coefficient according to the following formula:

以0.01为步长，以

为有效区间，遍历四个系数的不同组合，提取可选的系数组合列表，按照步骤(5.3)至步骤(5.5)的方法，进行最优系数的细调；(5.6) Select two sets of coefficient combinations with smaller inconsistent coefficients in the coarse adjustment process, and for each coefficient in each set of coefficient combinations

with a step size of 0.01, with

For the valid interval, traverse the different combinations of the four coefficients, extract a list of optional coefficient combinations, and perform fine-tuning of the optimal coefficients according to the methods from steps (5.3) to (5.5);

Take the parameter combination with the smallest inconsistent coefficient in the fine tuning result as the optimal coefficient combination, denoted as

Further, it is preferred that step (6) is to comprehensively evaluate the equipment failure probability in combination with the time nodes of equipment fault tripping and unplanned power outages in history and the equipment health assessment result;

The specific method is as follows: the calculation of the deduction of each equipment is carried out with the optimal combination of coefficients obtained in step (5) as the effective coefficient, and the total deduction of the equipment is as follows:

Calculate separately for each type of equipment, count trips, unplanned maintenance, planned maintenance, and the corresponding effective number of faults;

In the process of calculating the number of effective faults, trips, unplanned maintenance, and planned maintenance are counted according to 1, 0.5, and 0.25 as coefficients, namely:

The number of valid faults = the number of trips + the number of other unplanned maintenance × 0.5 + the number of planned maintenance × 0.25

First, calculate the failure probability of the equipment when there is no deduction;

Then remove the influence of the probability of failure without deduction: the effective number of equipment trips due to health problems, and calculate the total effective number of failures minus the expected value of equipment failure without deduction:

The effective quantity affected by the deduction = the effective quantity - the probability of equipment failure without deductions × the total number of equipment operating days;

If the calculated value is negative, take 0;

The deduction corresponds to the calculation of the probability of failure:

Further, preferably, step (7) is based on different tripping probabilities of the equipment, and adopts different coping solutions for power failure maintenance, load transfer, and formulating plans. The specific methods are:

For equipment with a trip probability greater than 10%, if it has a power outage condition, it needs to be outaged for maintenance;

For equipment with a failure probability greater than 5% and less than or equal to 10%, try to transfer the load carried to the equipment in a relatively healthy operating state for power supply in advance;

For equipment with a failure probability greater than 0.1% and less than or equal to 5%, make a plan for equipment failure and strengthen equipment inspection;

Equipment with a failure probability of less than or equal to 0.1% does not need to take special measures.

The structured analysis described in the present invention is divided into two aspects: one is to associate with equipment, to standardize the analysis of the text (that is, to unify the writing of numbers and equipment types in the text), through the factory station, voltage level , equipment type, equipment name keywords for correlation matching; the second is to classify equipment defects according to equipment type and defect type, and establish a keyword model for text matching (maintain the keywords corresponding to each classification, through the relationship of or, and, not Combining keywords), the defects are corresponding to the corresponding defect classification.

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

The invention has a reasonable design, and proposes a method for realizing the overall analysis of the health state and trip probability of the equipment by comprehensively analyzing the multi-dimensional data related to the equipment. The rules of historical data are automatically improved, and abnormal data is eliminated and completed. Based on different dimensions, a device state evaluation model is established to evaluate the health state of the device. Consistency comparison equations are established between dimensions. The coefficients are initially set based on expert experience, and automatic learning and parameter adjustment are performed based on historical data. The health status of the equipment is qualitatively and quantitatively evaluated by combining various factors.

Compared with the previous equipment state assessment methods, the method described in the present invention can establish a set of consistency comparison equations between different dimensions, which is more in line with the real probability of equipment problems in history on the basis of meeting the expectations of expert experience.

Description of drawings

FIG. 1 is a flow chart of a method for analyzing the failure probability of power grid equipment according to the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the embodiments.

Those skilled in the art will understand that the following examples are only for illustrating the present invention and should not be construed as limiting the scope of the present invention. If no specific technology or condition is indicated in the examples, the technology or condition described in the literature in the field or the product specification is used. If the materials or equipment used are not marked with the manufacturer, they are all conventional products that can be obtained through purchase.

As shown in Figure 1, a power grid equipment failure probability analysis method combining expert experience and historical data includes the following steps:

In step (1), multi-dimensional data collection and structured analysis are carried out with the health status of the equipment as the center;

Step (2), cross-identification and cleaning of data;

In step (3), an equipment health status assessment model is established for each dimension;

Step (4), the initial setting of the weight coefficients of each dimension factor based on expert experience;

Step (5), based on the automatic optimization adjustment of the multidimensional data weight coefficient of historical data;

Step (6), analyze and calculate the equipment tripping probability in combination with historical data;

Step (7), dealing with different trip probabilities.

The concrete realization method of described step (1) is:

The health status of the device mainly includes the following aspects:

1) Equipment defects: including manually recorded defects (from the defect processing process or log), and defects analyzed from the alarm signal (from the remote signaling alarm signal, excluding feedback operation, maintenance and debugging and other normal operation management processes issued). After the signal is received, the information reflecting the abnormal operation of transformers, switches and secondary equipment is extracted); according to the type of defects, the defects are classified; defects from different sources should be deduplicated;

2) Historical operating conditions: extract out-of-limit information (main transformer oil temperature out-of-limit, line and main transformer load out-of-limit, bus voltage out-of-limit) from remote signaling data, extract active power, reactive power, current, Voltage and main transformer oil temperature measurement data, obtain the waveform information of voltage and current at the time of fault from the fault recorder; extract;

3) Operating life: According to the standard life corresponding to the type of equipment, according to whether the actual operating time exceeds 90% of the standard life, it is judged whether the equipment has entered the old age; the equipment entering the old age shall be deducted according to the relationship between the current operating time and the standard life. Minute;

4) Family hidden dangers: Extract the average number of various types of defects in the equipment, and the number of corresponding types of defects in a certain manufacturer + model. For the manufacturer models whose defect probability exceeds the average defect probability by more than 5 times, it is included in the family hidden danger.

Analyze the collected data separately. Structural parsing is divided into two aspects: one is to associate with equipment. This part is to standardize the parsing of the text (unification of the writing of numbers and equipment types in the text), through the factory station, voltage level, equipment type, The equipment name keyword is correlated and matched; the second is to classify the equipment defects according to the equipment type and defect type, and establish a keyword model for text matching (maintain the keywords corresponding to each classification, and analyze the keywords through the or, and, and non-relationships. combination) to map the defects to the corresponding defect classifications.

The specific implementation method of the step (2) is: cleaning the data by building a cross-identification model between multi-dimensional data. The method of building and cleaning the model includes the following:

1) Identification of telemetry data quality: based on the normal range and change speed limit of telemetry, identify the number of hops and errors in telemetry; based on the historical data of telemetry values, extract the unchanged telemetry; , The active power and reactive power of each side of the line and bus are checked for power balance, and abnormal data is identified;

2) Cross-identification of remote signaling and telemetry data: identify the operation mode of each interval through remote signaling + equipment topology; establish the feature point model of the telemetry of the operating interval (except for the case of empty charging of the line, the active or reactive power is not used in the operating interval. 0), the feature point model of non-operational interval telemetry (active power, reactive power, and current of non-operational interval are 0); cross-identification is performed through the interval state characteristic points corresponding to remote signaling and telemetry, and the false remote signaling alarms are cleaned up The list of abnormal telemetry data is extracted and provided to the staff for inspection and processing.

The specific implementation method of the step (3) is to analyze the health state of the equipment from the aspects of defect records, historical operating conditions, operating years, and family hidden dangers. For each type of factor, a point deduction system is established to evaluate the health status of the equipment.

1) Defect record: including manually recorded defects and defects analyzed from the alarm signal. Filter out maintenance and debugging, operation-related, and AVC-related signals, extract valid alarms, and obtain corresponding equipment defects. According to the type of defective equipment and the classification of defects, establish a standard table of deduction points for defects, and deduct points for the defects of the equipment; General defects deduct 3 points), and the deduction standard for some types of defects can be adjusted according to the actual situation.

The process of creating the deduction standard table is as follows:

①Set general deduction standards according to the level of defects (41 points for critical defects, 11 points for major defects, and 3 points for general defects);

②According to the actual situation, each locality shall make special settings for some defect types with a large gap between the severity of defects and the general deduction standard of defect grade, and set the deduction value separately;

③In the actual use process, first check the deduction value set separately for the defect type; if the corresponding deduction value is not set, select the corresponding general deduction standard according to the defect level.

2) Historical operating conditions: subdivided into short-circuit shock accumulation, line heavy overload, main transformer heavy overload, and oil temperature exceeding the limit. Evaluation models are built in several dimensions to evaluate the health status of the equipment.

The cumulative short-circuit impact considers the impact caused by equipment tripping in history, and considers the number of impacts and the duration of the impact. Points will be deducted according to the following rules: 11 points will be deducted for each impact, and the trip will be accumulated once. Points will be deducted at intervals of time, 11 points will be deducted every 1 second, and the above deductions will be used as the accumulated deductions for short-circuit shocks.

Considering the equipment load rate and accumulated time for heavy overload of equipment, the calculation formula of the corresponding deduction for equipment damage is as follows:

Among them, L _A is the average value of the current limit violation, _LM is the maximum value of the current limit violation, and T(L) is the duration of the current limit violation.

For the oil temperature exceeding the limit, considering the exceeding temperature and the accumulated time, the calculation formula of the deduction corresponding to the damage to the main transformer is as follows:

(e ^{(t-75)×0.1-1} )×T(t)

Among them, t is the highest oil temperature during the oil temperature overrun period, and T(t) is the duration of this overrun.

Short-circuit impact, line or main transformer overload, and oil temperature exceeding the limit, the deductions calculated by the three models are accumulated in units of equipment, as the overall deduction of the historical operating conditions of the equipment.

(注：运行年限不超过标准寿命90％的设备，不进行扣分)。3) Operating years: The standard life of the equipment is Y, and the current equipment operating time is y years. For equipment operating for more than 90% of its standard life, the deductions are:

(Note: No points will be deducted for equipment whose operating life does not exceed 90% of the standard life).

4) Family hidden danger: Take the manufacturer + model as the standard, study the total average probability of various defects and the average probability of each manufacturer's model. Calculate the defect rate of a device by dividing the number of defects by the total time the device has been running (in days). Calculate the situation of all equipment separately, as well as the situation of sub-manufacturer + model. For a certain manufacturer and model, if the corresponding defect rate is more than 5 times the defect rate of the same defect of the same type of equipment, then for the equipment of the relevant manufacturer and model, the deduction formula is:

In the formula, F is a certain type of defect, P(F) is the average probability of such defects in the same equipment type and voltage level, P _D (F) is the average probability of a manufacturer's model for this defect; S ₁ (F ) is the deduction value corresponding to defect F in the defect deduction standard table.

The above four types of models should be weighted according to certain coefficients in actual calculation, and the method of coefficient setting will be described later.

The specific implementation method of the step (4) is: through experts specifying the initial coefficients of each factor (specifying a situation, when a certain defect occurs, the equipment is overloaded, the equipment running years, and the hidden dangers of the family, various factors affect the health of the equipment. The proportion of influence, take the average of multiple experts, and then calculate the score through various factors, and calculate the effective coefficient corresponding to the final average proportion of various factors). In the process of initial setting of this parameter, the analytic hierarchy process is used to complete, and the specific steps are as follows:

1) Case combination extraction of different factors: For defect records, historical operating conditions, operating years, and family hidden dangers, typical samples are selected and combined to obtain combined cases of multiple factors. For example, for the main transformer, it is possible to extract the abnormal oil temperature alarm (defect record), the load rate of 98% running for 24 hours (historical operating conditions), the running time reaching the standard operating years (operating years), and the main transformer of the same manufacturer has a cold zone. The probability of full stop alarm is 10 times the average probability of the main transformer of the same voltage level (family hidden danger), as a case combination.

2) Expert evaluation of the importance of each factor: select 3 or more experts to compare and evaluate the weights of different factors in the selected case combination for equipment tripping, and input them as parameters of the evaluation matrix.

3) Calculate the weight of each factor's contribution to the equipment tripping judgment by the AHP, and denote it as L _n (n=1, 2, 3, 4, corresponding to four types of factors respectively).

4) Adjust the coefficients of the evaluation model established in step (3) in turn: for the four types of factors, through the model calculation, the obtained deduction points are S _n (n=1, 2, 3, 4, respectively corresponding to the four types factor); combined with the results calculated in (5.3), the calculation formula of the coefficient R _n calculated according to this group of case combinations is:

Note that the above calculation results are calculated for one case combination.

(m是案例组合的序号，取值从1到M，M是案例组合的数量)。缺陷记录对应的系数，经专家评定之后对应的系数计算公式为：5) Set the initial coefficient according to the combination of multiple groups of cases: for defect records, a coefficient R ₁ will be analyzed for each group of cases, and this sequence is recorded as

(m is the serial number of the case combination, ranging from 1 to M, where M is the number of case combinations). The coefficient corresponding to the defect record, after the expert evaluation, the corresponding coefficient calculation formula is:

This parameter will be used as the initial value of parameter adjustment, and further automatic adjustment needs to be carried out in combination with historical data. The corresponding calculations are performed for the four factors in turn to obtain the initial values of all factors.

The specific implementation method of the step (5) is: to automatically adjust the coefficients of each type to make the evaluation system more reasonable, and the steps are as follows:

1) Extract the equipment that has experienced unplanned power outages within the scope of analysis. The scope of analysis here is to take a period of time when the defect records, remote signaling, and telemetry information are complete, and to take data of more than one year. The time should not be too far away from the present. For example, the data from January 1 last year to the current can be taken.

(n从1到4，代表四类因素)，以各个系数的

的25％为步长，以

为有效区间，遍历四个系数的不同组合(比如四个系数分别取

)，提取可选的系数组合列表，进行最优系数的粗调，具体过程如步骤下面3)至5)所示；2) For each initial coefficient

(n is from 1 to 4, representing four types of factors), with the

25% of the step size to

For the valid interval, traverse different combinations of the four coefficients (for example, the four coefficients are taken separately

), extract the optional coefficient combination list, carry out the rough adjustment of the optimal coefficient, and the concrete process is as shown in steps 3) to 5) below;

3) Select a historical time interval with complete defects and operating conditions, and for each coefficient combination in the coefficient combination list, take days as the unit, according to the evaluation method of step (3), take the selected coefficient combination as the coefficient, The deduction of each device in the power grid is calculated;

4) Detect the consistency of each group of coefficients: for each device, sort from large to small according to the daily deduction value; traverse the sequence of daily deduction values from front to back, if there is no occurrence of a certain day For unplanned power outages, but the deduction value for n unplanned outages is less than the current deduction value, n will be added to the inconsistency. The total number of inconsistencies, denoted as U _C .

5) Selection of the optimal effective coefficient: for the inconsistency quantity U _C calculated by the coefficient combination R _n , the inconsistency coefficient is calculated according to the following formula:

以0.01为步长，以

为有效区间，遍历四个系数的不同组合，提取可选的系数组合列表，按照前面3)至5)的方法，进行最优系数的细调。Select two sets of coefficient combinations with small inconsistent coefficients in the coarse adjustment process, and for each coefficient in each set of coefficient combinations

with a step size of 0.01, with

For the effective interval, traverse different combinations of the four coefficients, extract a list of optional coefficient combinations, and perform fine adjustment of the optimal coefficients according to the methods 3) to 5) above.

Take the parameter combination with the smallest inconsistent coefficient in the fine tuning result as the optimal coefficient combination, denoted as

The specific implementation method of the step (6) is: the calculation of the deduction of each device is calculated with the optimal coefficient combination obtained in the step (5) as the effective coefficient, and the total deduction of the device is:

For each type of equipment, it is calculated separately, and the number of effective faults corresponding to trips, unplanned maintenance, and planned maintenance is counted.

In the process of calculating the number of effective faults, trips, unplanned maintenance, and planned maintenance are counted according to 1, 0.5, and 0.25 as coefficients, namely:

The number of valid faults = the number of trips + the number of other unplanned maintenance × 0.5 + the number of planned maintenance × 0.25

First, calculate the failure probability of the equipment when there is no deduction (so-called with or without deduction, according to the calculation process of step (3), the deduction values of each part are superimposed, and the day when the total deduction is 0 is recorded as no deduction);

Then remove the influence of the probability of failure when no points are deducted: the effective number of equipment trips increased by the equipment due to health problems (here corresponds to the deduction of equipment points), the total effective number of faults is subtracted from the expected value of equipment failure when no points are deducted. Do the calculation (there is also a certain probability that the device will fail when there is no health problem (that is, without any deductions)):

The effective quantity affected by the deduction of points = the effective quantity - the probability of equipment failure without deductions × the total equipment operation day

Takes 0 if the computed value is negative.

The deduction corresponds to the calculation of the probability of failure:

The following example is used to illustrate: Suppose there are 100 main transformers, 40 of them have been running for 300 days, and the other 60 have been running for 500 days. The cumulative running time is: 40×300+60×500=42,000 units·days.

The deduction of points during the period and the corresponding situation of failure and maintenance are as follows:

100台主变中，有扣分的累计天数为5000台·天；期间扣分以天为单位进行积分，得到的总扣分为20000分；在设备存在扣分的期间，相应设备发生故障、非计划停电、计划停电的次数分别为5次、10次、20次。

Among the 100 main transformers, the cumulative number of days with points deduction is 5,000 units·days; during the period, the points are deducted in units of days, and the total deduction points obtained are 20,000 points; during the period of equipment deduction, the corresponding equipment fails, The number of unplanned power outages and planned outages are 5, 10, and 20, respectively.

100台主变中，无扣分的累计天数为42000-5000＝37000台·天；在设备不存在扣分的期间，相应设备发生故障、非计划停电、计划停电的次数分别为6次、12次、24次。

Among the 100 main transformers, the accumulated days without points deduction is 42000-5000=37000 units·days; during the period when the equipment does not have points deduction, the number of corresponding equipment failures, unplanned power outages, and planned power outages are 6 times and 12 times respectively. times, 24 times.

but:

No deductions (on the day) effective quantity = 6×1+12×0.5+24×0.25=18

Deducted points (on the day) effective quantity = 5 × 1 + 10 × 0.5 + 20 × 0.25 = 15

Total number of valid faults = 18 + 15 = 33

Deductions affect the effective failure probability = 33-0.000486 × 42000 ≈ 33-25.3 = 7.7

If the deduction of a main transformer is 10 points on a certain day, the corresponding failure probability is:

The specific implementation method of the step (7) is as follows: according to the failure probability level of the power equipment, different measures are taken to deal with it.

For equipment with a failure probability greater than 10%, if power failure conditions exist, power failure is required for maintenance; for equipment with a failure probability greater than 5%, and equipment with a failure probability of less than or equal to 10%, try to transfer the load (especially the important load) to the operating state in advance. Power supply to healthy equipment; equipment with a failure probability greater than 0.1%, and equipment with a failure probability of less than or equal to 5%, make a plan for equipment failure, strengthen equipment inspection, and take the next step according to the development trend and speed of equipment health status. Measures; no special measures are required for equipment with a failure probability of less than or equal to 0.1%.

The above specific ratios are adjusted according to the actual load level of the power grid in different places.

Applications

The following takes the analysis process of the power grid in a certain area as an example to illustrate. The data range analyzed was from January 1, 2017 to January 1, 2020, a total of 3 years. The data before July 1, 2019 is used as learning data, and the relevant parameters are adjusted adaptively; the data from July 1, 2019 to January 1, 2020 is used as test data to verify the effectiveness of the algorithm. The analysis process is mainly based on the analysis of the main variable related data.

1) Taking the health status of the equipment as the center, carry out multi-dimensional data collection and structured analysis

Access device information, topology information, remote signaling, and telemetry information from the D5000 system. Remote signaling information includes accident, anomaly, displacement, notification information, and over-limit information is obtained separately. Telemetry information 5 minutes a section. The equipment account information is supplemented by the docking in the PMS.

Connect with OMS to obtain maintenance, defect records, and trip records.

External environment, obtain wildfire information (recorded information includes affected lines and towers).

Perform structural analysis on maintenance, defect records, and trip records in OMS data, associate relevant data with equipment models in D5000 through keyword matching, and classify defect information according to defect types.

2) Cross-identification and cleaning of data

After structured analysis of the data, through cross-identification, the main problems found are shown in Table 1:

Table 1

3) Establish equipment health status assessment models for each dimension

According to the method described above, an equipment health status assessment model is established in four dimensions: equipment defects, historical operating conditions, operating years, and family hidden dangers.

4) Initial setting of the weight coefficients of each dimension factor based on expert experience;

Using the analytic hierarchy process, the proportion of each factor is analyzed. The following takes the main variation as an example to describe the initialization process of the corresponding weights of each factor. The specific content and evaluation results are shown in Table 2:

Table 2

5) Automatic optimization and adjustment of multidimensional data weight coefficients based on historical data

Taking the main transformer as an example, the tripping and unplanned power outages of the main transformer during the period from January 1, 2017 to July 1, 2019 are analyzed, and the relevant parameters of the failure probability are calculated in combination with the deduction of points during the corresponding period. During this period, a total of 36 unplanned outages (including fault trips) occurred.

By automatically adjusting the coefficients of the factors of the above four types of equipment (respectively, defect records, historical operating conditions, operating years, and family hidden dangers), the initial step size is set to 25% of each factor. After finding the roughly optimal combination of coefficients, use 0.01 as the step size to fine-tune the coefficients of each factor. The adjusted results are shown in Table 3:

table 3

factor poor optimal combination Precise and optimal combination Defect record 0.80 0.81 Short-circuit impact accumulation 0.59 0.59 Operating years 0.51 0.52 family hidden danger 0.62 0.62 Inconsistency probability 3.91% 2.32%

6) Analyze and calculate the equipment tripping probability in combination with historical data;

By analyzing the data from July 1, 2019 to January 1, 2020, there were a total of 9 unplanned power outages that occurred during this period.

After calculation, the probability of inconsistency during this period is 2.69%. The overall calculation results are as expected.

7) Real-time analysis of equipment tripping probability

From May 1, 2020, online analysis of the failure probability of equipment in the power grid was carried out; in May, 2 cases of failure probability greater than 10% were found (one was that the line was affected by the wildfire, and the other was that the main alarm signal was analyzed from the alarm signal. There are 9 cases of defects); the failure probability is between 0.1% and 5%, mainly due to the influence of defect records and historical operating conditions.

8) Follow-up treatment based on different trip probabilities

For the case where the failure probability is greater than 10%, take power outage maintenance; for the case where the failure probability is between 0.1% and 5%, check the corresponding plan when the equipment fails, and strengthen the inspection of the equipment.

The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above-mentioned embodiments, and the descriptions in the above-mentioned embodiments and the description are only to illustrate the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have Various changes and modifications fall within the scope of the claimed invention. The claimed scope of the present invention is defined by the appended claims and their equivalents.

Claims (7)

1. a power grid equipment failure probability analysis method, is characterized in that, comprises the following steps: In step (1), multi-dimensional data collection and structured analysis are carried out with the health status of the equipment as the center; Step (2), cross-identification and cleaning of data; In step (3), an equipment health status assessment model is established for each dimension; Step (4), the initial setting of the weight coefficients of each dimension factor based on expert experience; Step (5), based on the automatic optimization adjustment of the multidimensional data weight coefficient of historical data; Step (6), analyze and calculate the equipment tripping probability in combination with historical data; Step (7), dealing with different tripping probabilities; The specific method of step (4) is: specify the initial coefficient of each factor by experts, that is, specify a situation, in the case of a certain defect, heavy equipment overload, equipment operating years, and family hidden dangers, the impact of various factors on equipment health. Proportion, take the average value of multiple experts, and then calculate the effective coefficient through the calculation of the scores of various factors and the final average proportion of various factors to calculate the effective coefficient; In the process of parameter initial setting, the analytic hierarchy process is used to complete, and the specific method is as follows: (4.1) Case combination extraction of different factors: For defect records, historical operating conditions, operating years, and family hidden dangers, typical samples are selected and combined to obtain combined cases of multiple factors; (4.2) Experts assess the relationship of importance of each factor: select 3 or more experts to compare and assess the weights of different factors in the selected case combination for equipment tripping, and input them as parameters of the evaluation matrix; (4.3) Calculate the weight of the contribution of each factor to the equipment tripping judgment by the AHP, denoted as L _n ; wherein, n=1, 2, 3, 4, corresponding to four types of factors respectively; (4.4) Adjust the coefficients of the evaluation model established in step (3) in turn: for the four types of factors, through the model calculation, the obtained deduction points are S _n , n=1, 2, 3, 4, corresponding to the four types respectively factor; combined with the results calculated in (4.3), the calculation formula of the coefficient R _n calculated according to this group of case combinations is:

The coefficient R _n is the result calculated for this set of case combinations; (4.5) Set the initial coefficient according to the combination of multiple groups of cases: for defect records, each group of cases will analyze a coefficient R ₁ , and record this sequence as

m is the serial number of the case combination, ranging from 1 to M, where M is the number of case combinations; the coefficient corresponding to the defect record, and the corresponding coefficient calculation formula after expert evaluation is:

This parameter will be used as the initial value for parameter adjustment; For the other three factors, corresponding calculations are performed in turn to obtain the initial values of all factors. 2. The power grid equipment failure probability analysis method according to claim 1, wherein the data collected in step (1) includes four dimensions of equipment defects, historical operating conditions, operating years and family hidden dangers; The said family hidden danger is the average number of corresponding types of defects in a certain model of a certain manufacturer; Analyze the collected data separately. 3. The power grid equipment failure probability analysis method according to claim 1, wherein the specific method of step (2) is: by building a cross-identification model between multidimensional data, the data is cleaned; the details are as follows: (2.1) Identification of telemetry data quality: based on the normal range and change speed limit of telemetry, identify the number of hops and errors in telemetry; based on the historical data of telemetry values, extract the unchanged telemetry; Perform power balance verification on active and reactive power on each side of transformers, lines, and bus bars to identify abnormal data; clean up hop counts, wrong counts, unchanged telemetry data, and abnormal data in telemetry; (2.2) Cross-identification of remote signaling and telemetry data: identify the operation mode of each interval through remote signaling and equipment topology; establish a feature point model for telemetry in operating intervals and a feature point model for telemetry in non-operating intervals; through remote signaling and telemetry corresponding intervals State feature points, perform cross-identification, clean up false remote signaling alarms, extract a list of abnormal telemetry data, and provide them to the staff for inspection and processing; The feature point model of the operating interval telemetry is that the active power or reactive power is not 0 in the operating interval except for the case of empty charging of the line; the feature point model of the non-operating interval telemetry is that the active power, reactive power and current of the non-operating interval are 0. 4. The power grid equipment failure probability analysis method according to claim 1, wherein the specific method of step (3) is: using the data cleaned by step (2), for each dimension factor, establish equipment health The state evaluation model is used to evaluate the health state of the equipment, specifically: (3.1) For defect records: According to the type of defective equipment and the classification of defects, establish a standard table of deduction points for defects, and deduct points for the defects existing in the equipment; set a general deduction standard according to the level of defects, deduct 41 points for critical defects, and deduct 41 points for major defects. 11 points are deducted for defects, and 3 points are deducted for general defects; The final deduction value of the defect factor is recorded as S ₁ ; (3.2) For historical operating conditions: the historical operating conditions are divided into short-circuit impact accumulation, line heavy overload, main transformer heavy overload, and oil temperature exceeding the limit; ①For the accumulation of short-circuit shocks, 11 points will be deducted for each shock, and the trip will be accumulated once. If the reclosing is unsuccessful, the points will be deducted according to the interval between the successful reclosing and the tripping again. The points will be deducted once every 1 second, and 11 points will be deducted each time; ②For the heavy overload of the line and the heavy overload of the main transformer, the calculation formula of the corresponding deduction for the damage to the equipment is: The deductions are:

Among them, L _A is the average value of the current limit violation, L _M is the maximum value of the current limit violation, and T(L) is the duration of the current limit violation; ③ For the oil temperature exceeding the limit, considering the exceeding temperature and the accumulated time, the calculation formula for the deduction corresponding to the damage to the main transformer is: The deduction value is: (e ^(t-75)×0.1 -1)×T(t) Among them, t is the maximum oil temperature during the oil temperature overrun period, and T(t) is the duration of this overrun; Short-circuit impact, overload of line or main transformer, and oil temperature exceeding the limit, the deductions calculated by the three models are accumulated in units of equipment, as the overall deduction of historical operating conditions of the equipment; The final deduction value of the historical operating condition factor is recorded as S ₂ ; (3.3) For the operating life factor, let the standard life of the equipment be Y, and the current equipment has been running for y years; For equipment operating for more than 90% of its standard life, the deductions are:

No points will be deducted for equipment whose operating time does not exceed 90% of the standard life; The final deduction value of the operating years factor is recorded as S ₃ ; (3.4) For family hidden dangers, the deduction is as follows:

Among them, F is a certain type of defect, P(F) is the average probability of such defects in the same equipment type and voltage level, P _D (F) is the average probability of such defects in a manufacturer's model; S ₁ (F ) is the deduction value corresponding to defect F in the defect deduction standard table; The final deduction value of the family hidden danger factor is recorded as S ₄ . 5. The power grid equipment failure probability analysis method according to claim 1, wherein step (5) is based on the initial setting of weight coefficients of different dimension factors according to expert experience, by adjusting the weight coefficients, combined with historical data Consistency check is performed on different weight combinations to obtain the optimal coefficient combination; Specific steps are as follows: (5.1) Extract the equipment that has experienced unplanned power outages within the scope of analysis; (5.2) For each initial coefficient

Corresponding to the four types of factors, the coefficients of each

25% of the step size to

For the effective interval, traverse the different combinations of the four coefficients, extract a list of optional coefficient combinations, and use steps (5.3) to (5.5) to perform rough adjustment of the optimal coefficients; (5.3) Select a historical time interval with complete defects and operating conditions, for each coefficient combination in the coefficient combination list, take days as the unit, according to the evaluation method of step (3), take the selected coefficient combination as the coefficient, Calculate the deductions for each device in the power grid; (5.4) Detect the consistency of each group of coefficients respectively: for each device, sort from large to small according to the daily deduction | value; traverse the sequence of daily deduction values from front to back, if a day If there is no unplanned power outage, but there are n unplanned power outages with a deduction value smaller than the current deduction value, add n to the inconsistency; the total number of inconsistencies is recorded as U _C ; (5.5) Selection of the optimal effective coefficient: For the inconsistency quantity U _C calculated by the coefficient combination (R ₁ , R ₂ , R ₃ , R ₄ ), calculate the inconsistency coefficient according to the following formula:

(5.6) Select two sets of coefficient combinations with smaller inconsistent coefficients in the coarse adjustment process, and for each coefficient in each set of coefficient combinations

with a step size of 0.01, with

For the valid interval, traverse the different combinations of the four coefficients, extract a list of optional coefficient combinations, and perform fine-tuning of the optimal coefficients according to the methods from steps (5.3) to (5.5); Take the parameter combination with the smallest inconsistent coefficient in the fine tuning result as the optimal coefficient combination, denoted as

6. The power grid equipment failure probability analysis method according to claim 5 is characterized in that, step (6) is to combine the time node of equipment failure tripping and unplanned power outage in the history, combined with equipment health assessment result, to equipment failure probability conduct a comprehensive assessment; The specific method is as follows: the calculation of the deduction of each equipment is carried out with the optimal combination of coefficients obtained in step (5) as the effective coefficient, and the total deduction of the equipment is as follows:

Calculate separately for each type of equipment, count trips, unplanned maintenance, planned maintenance, and the corresponding effective number of faults; In the process of calculating the number of effective faults, trips, unplanned maintenance, and planned maintenance are counted according to 1, 0.5, and 0.25 as coefficients, namely: The number of valid faults = the number of trips + the number of other unplanned maintenance × 0.5 + the number of planned maintenance × 0.25 First, calculate the failure probability of the equipment when there is no deduction;

Then remove the influence of the probability of failure without deduction: the effective number of equipment trips due to health problems, and calculate the total effective number of failures minus the expected value of equipment failure without deduction: The effective quantity affected by the deduction = the effective quantity - the probability of equipment failure without deductions × the total number of equipment operating days; If the calculated value is negative, take 0; The deduction corresponds to the calculation of the probability of failure:

7. The power grid equipment failure probability analysis method according to claim 1, is characterized in that, step (7) is based on the different tripping probability of equipment, adopts different coping treatment plans of outage maintenance, load transfer, and formulating plans, and the concrete method is as follows: : For equipment with a trip probability greater than 10%, if it has a power outage condition, it needs to be repaired by a power outage; For equipment with a failure probability greater than 5% and less than or equal to 10%, try to transfer the load carried to the equipment in a relatively healthy operating state for power supply in advance; For equipment with a failure probability greater than 0.1% and less than or equal to 5%, make a plan for equipment failure and strengthen equipment inspection; Equipment with a failure probability of less than or equal to 0.1% does not need to take special measures.

Images