A fault early warning method based on oilfield network electrical equipment

Abstract

The application discloses a kind of based on oilfield network power equipment's fault early warning method, this method is first based on oilfield network power equipment fault early warning requirement, build suitable for oilfield network power probability type data training algorithm;Then the historical failure data of oilfield network power equipment is imported to probability type data training algorithm, generates probability type failure prediction model;Finally, the real-time data of oilfield network power equipment is imported into probability type failure prediction model, compares, makes prediction and early warning to failure.The application proposes a new idea of building historical failure model on the basis of original failure threshold judgment, makes a more effective early warning mechanism for the possibility of equipment failure, is superior to the existing method, with the advantages of strong universality, convenient to build etc..

Description

Technical Field

[0001] This invention relates to the field of computer fault early warning methods, and more particularly to a fault early warning method for oilfield power grid equipment. Background Technology

[0002] Equipment fault early warning technology plays a beneficial role in extending the service life of oilfield power grid equipment by providing early warnings of poor equipment conditions before failures occur. Oil drilling sites are typically located in remote, outdoor areas. Currently, oil drilling often uses grid-driven motors as the primary power source for oil extraction. On-site grid power conditions frequently face challenges such as long-distance overhead lines, unstable grid voltage, heavy loads, and high harmonics. Prolonged operation under these conditions leads to heavy loads on grid equipment, making it prone to failure and wear. Therefore, designing an efficient fault early warning method for oilfield power grid equipment is both necessary and urgent.

[0003] Previous predictive methods targeting data fault thresholds typically only addressed two types of data: first, protection and fault flags within instruments, often collected and judged in the form of switch signals; and second, thresholds for parameters such as current, voltage, and harmonics. Both methods have their limitations. Methods targeting protection and fault flags can usually only detect the occurrence of faults, but cannot predict them in advance. While early warning methods targeting parameter thresholds are often based on calculations of the equipment's rated parameters, the operating conditions of oilfield power grid equipment are extremely complex, with significant equipment wear and tear. Many devices, after operating for a period of time, can no longer function normally within their designed rated range. Therefore, early warning methods based on parameter thresholds have certain limitations in oilfield power grid equipment. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a fault early warning method based on oilfield power grid equipment.

[0005] The objective of this invention is achieved through the following technical solution: a fault early warning method based on oilfield power grid equipment, comprising the following steps:

[0006] (1) Fault data training algorithm construction: Based on the requirements of oilfield power grid equipment fault early warning, a probabilistic data training algorithm suitable for oilfield power grid is constructed.

[0007] (2) Fault early warning model training: Import historical fault data of oilfield network equipment into the probabilistic data training algorithm built in step (1) to generate a probabilistic fault prediction model;

[0008] (3) Fault warning: Import the real-time data of the oilfield network equipment into the probabilistic fault prediction model generated in step (2) for comparison, and make predictions and warnings for faults.

[0009] Further, step (1) includes the following sub-steps:

[0010] (1.1) Define parameters: Based on the fault early warning requirements of oilfield power grid equipment, extract the core data set G for fault prediction. e The core data set G at the time of the failure e The values ​​of each element in G are defined as follows: en The real-time data of each element is defined as G. rn The safe interval of each element in the core data is defined as G. sn Let N be the number of times a fault occurs, and let G be the value of each element in the core data during N faults. en Exceeding the safe range G s The degree of n is defined as G bn The early warning accuracy is defined as um, and the interval for triggering fault early warning is defined as G. umn The probability of each element in the model triggering a fault is defined as G. tn n is a sequence of 1, 2, 3, 4...;

[0011] (1.2) Using the parameters defined in step (1.1), the following algorithm is used to obtain historical fault data: G bn The initial value is 0, when G en Exceeding G sn At that time, G bn +1; G umn The interval is G sn The upper and lower limits multiplied by um, G umn =G sn *um; G tn =G bn / N.

[0012] Furthermore, step (2) specifically involves: counting the number of historical faults into N, and the value G of each element in the core data from the N faults. en Exceeding the safe range G sn number of times G bn Reset to zero, when the historical data of the core data G en Exceeding the safe range G sn At that time, the number G bn +1; Fault warning trigger interval G umn The value is the safe interval G sn The upper and lower limits are multiplied by the early warning accuracy um to obtain the probability of each element's fault occurrence G. tn The value is determined by G bn Dividing by the number of historical failures yields the fault warning model generated based on the actual historical failures.

[0013] Furthermore, step (3) specifically involves: when G tn When G > um, rn Exceeding G umn This triggers a fault warning; when G tn <um or G tn When =um, let the accumulation parameter n = 0, G rn Exceeding G umn Accumulate over time, then divide the accumulated value n by G. tn The number of times the data ≤ um is obtained as m. If (m+1) / 2 > 0.5, a fault warning is triggered.

[0014] The beneficial effects of this invention are that, by constructing a fault early warning algorithm and model for oilfield power grid equipment, it can effectively protect oilfield power grid equipment, predict dangerous equipment behavior in advance, and extend the service life of the equipment. The effect is superior to existing methods based on fault thresholds for prediction, and it has the characteristics of strong versatility and convenient implementation. Attached Figure Description

[0015] Figure 1 This is an algorithm block diagram of the fault early warning method for oilfield power grid equipment according to the present invention. Detailed Implementation

[0016] This invention analyzes historical fault data of oilfield power grid equipment, builds a reasonable fault early warning model through a fault early warning algorithm, compares the real-time data of oilfield power grid equipment with the model, and issues fault warnings for data that does not conform to the model.

[0017] This invention proposes a fault early warning method based on oilfield power grid equipment, comprising the following steps:

[0018] 1. Fault Data Training Algorithm Construction: Based on the requirements for fault early warning of oilfield power grid equipment, a probabilistic data training algorithm suitable for oilfield power grids is constructed; specifically, it includes the following sub-steps:

[0019] 1.1. Based on the fault early warning requirements of oilfield power grid equipment, extract the core data set G for fault prediction. e The core data set G at the time of the failure e The values ​​of each element in G are defined as follows: en (a sequence of n, where n is 1, 2, 3, 4...), the real-time data of each element is defined as G. rn (a sequence of n = 1, 2, 3, 4... and G) en (corresponding to the sequence), the safe interval of each element in the core data is defined as G. sn (a sequence of n = 1, 2, 3, 4... and G) en(corresponding to the sequence), the number of fault occurrences is defined as N, and the value G of each element in the core data of N faults is... en Exceeding the safe range G sn The number of times is defined as G bn (a sequence of n = 1, 2, 3, 4... and G) en (corresponding to the sequence), the early warning accuracy is defined as um, and the interval for triggering the fault early warning is defined as G. umn (a sequence of n = 1, 2, 3, 4... and G) en (corresponding to the sequence), the probability of each element in the model triggering a fault is defined as G. tn (a sequence of n = 1, 2, 3, 4... and G) en (The sequence corresponds to).

[0020] 1.2. Using the parameters defined in step 1.1), the following algorithm is used to obtain historical fault data: G bn The initial value is 0, when G en Exceeding G sn Time G bn +1; G umn The interval is G sn The upper and lower limits multiplied by um, G umn =G sn *um; G tn =G bn / N.

[0021] 2. Fault early warning model training: Import historical fault data of oilfield power grid equipment into the fault data training algorithm in step 1 to generate a probabilistic fault prediction model.

[0022] The number of historical failures is counted as N, and the value G of each element in the core data from N failures is... en Exceeding the safe range G sn number of times G bn Reset to zero, when the historical data of the core data G is zeroed out. en Exceeding the safe range G sn At that time, the number G bn +1; Fault warning trigger interval G umn The value is the safe interval G sn The upper and lower limits are multiplied by the early warning accuracy um to obtain the probability of each element's fault occurrence G. tn The value is determined by G bn Dividing by the number of historical failures yields the fault warning model generated based on the actual historical failures.

[0023] 3. Fault Early Warning: Fault early warning is verified through implementation data. Real-time data from the oilfield network equipment is imported into step 2 from historical fault data G. enThe generated fault model is compared to predict and warn of faults: when real-time data G rn Exceeding the safe range G sn When G triggers a fault warning; tn When G > um, if rn Exceeding G umn When G triggers a fault warning; tn When <um, accumulate the sum, divide the accumulated value by the number of accumulations to get m, and if (m+1) / 2>0.5, a fault warning is triggered.

[0024] Compared to previous prediction methods that only target data fault thresholds, the fault early warning algorithm proposed in this invention, based on oilfield power grid equipment, has higher prediction accuracy and can more effectively protect oilfield power grid equipment. However, this prediction algorithm has the problem of not being able to accurately measure elements in the model whose probability of triggering a fault is less than the prediction accuracy in actual prediction. Therefore, a fuzzy algorithm must be used to minimize the probability of false triggering and improve the prediction accuracy as much as possible.

[0025] Implementation Examples

[0026] An embodiment of the invention is implemented on a machine equipped with an Intel Core i7-3770 central processing unit, an NVIDIA GTX760 graphics processor, and 32GB of memory. According to Figure 1 The algorithm shown is used to develop a computer program that derives a fault prediction model from historical fault data. Real-time equipment data is then compared with the fault model, and a fault is triggered when the model reaches the required accuracy for a fault warning. Compared to existing methods, this method can more accurately predict equipment faults and issue warnings as soon as a fault occurs, significantly reducing the equipment failure rate and extending its service life.

Claims

1. A fault early warning method based on oilfield power grid equipment, characterized in that, Includes the following steps: (1) Fault data training algorithm construction: Based on the requirements of oilfield power grid equipment fault early warning, a probabilistic data training algorithm suitable for oilfield power grid is constructed; step (1) includes the following sub-steps: (1.1) Define parameters: Based on the fault early warning requirements of oilfield power grid equipment, extract the core data set for fault prediction. The core data set at the time of the failure The values ​​of each element in the array are defined as follows: Define the real-time data of each element as Define the safe range of each element in the core data as Define the number of failures as N, and then define the values ​​of each element in the core data from the N failures. Beyond the safe zone The number of times is defined as The early warning accuracy is defined as um, and the interval for triggering fault early warning is defined as... The probability of each element in the model triggering a fault is defined as follows: n is a sequence of 1, 2, 3, 4...; (1.2) Using the parameters defined in step (1.1), the following algorithm is used to obtain historical fault data: The initial value is 0, when Exceeding hour, +1; The interval is Multiply the upper and lower limits by um. = *um; = / N; (2) Fault early warning model training: Import historical fault data of oilfield power equipment into the probabilistic data training algorithm built in step (1) to generate a probabilistic fault prediction model; the specific steps (2) are: count the number of historical faults into N, and the values ​​of each element in the core data of N faults Beyond the safe zone Number of times Reset to zero, when the core data is historical data. Beyond the safe zone At time, number of times +1; Fault warning trigger interval The value is within the safe range The upper and lower limits are multiplied by the warning accuracy um to obtain the probability of each element's failure. The value is determined by Divide by the number of historical failures to obtain the fault warning model based on the actual historical failures; (3) Fault warning: Import the real-time data of the oilfield network equipment into the probabilistic fault prediction model generated in step (2) for comparison, and make predictions and warnings for faults.

2. The method according to claim 1, characterized in that, The specific content of step (3) is as follows: When > um, exceeds it will trigger a fault warning; when < um or = um, set the accumulation parameter n = 0, exceeds accumulate when it exceeds, divide the accumulated value n by the number of times of < um data to get m. If (m + 1) / 2 > 0.5, trigger a fault warning.

Images