Artificial Intelligence-Based Risk Classification and Early Warning Method for Power Engineering

Abstract

This invention relates to the field of data processing technology, specifically to an artificial intelligence-based method for risk classification and early warning in power engineering. The method includes: acquiring environmental factor data and equipment operation data for each dimension at each time point, and marking the time of failure; obtaining a risk coefficient for the environmental factor data based on the relative changes in environmental factor data for each dimension at the time of failure, combined with the deviation of the environmental factor data; obtaining a risk coefficient for the equipment operation data based on the equipment operation data for each dimension at each historical time point before the time of failure and the time interval between the historical time point and the time of failure, combined with correlation and risk coefficients; and constructing a risk classification model based on the risk coefficients and the changes in data for the same dimension at each historical time point before each time of failure, thereby providing risk classification and early warning for power engineering. This invention can effectively identify high-risk equipment with potential failures, and the early warning effect is good.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to an artificial intelligence-based method for risk classification and early warning in power engineering. Background Technology

[0002] Mine power systems involve numerous devices in complex environments, where even minor faults or environmental changes can trigger serious accidents. Risk early warning systems enable timely detection of potential risks and the implementation of preventative measures. Furthermore, mine power systems are often located in remote areas with sparse personnel, making timely fault detection difficult. Therefore, real-time monitoring and risk prediction effectively improve safety, reduce losses, and ensure stable production. Risk early warning systems not only protect personnel and equipment but also reduce economic losses and improve system reliability; thus, risk early warning systems are crucial for mine power engineering.

[0003] Generally, historical data for mine power systems includes equipment operation data and environmental factor data at each moment, and a binary classification method is used to determine whether a fault has occurred at each moment. By analyzing historical data, existing classification algorithms, such as the KNN algorithm, can predict the risk status at the current moment by calculating the similarity between real-time data and historical data.

[0004] Although the probability of mining equipment failure is low, a failure can have serious consequences. In such cases, risk warnings for mining equipment often face the problem of data imbalance, especially in binary classification models where there are fewer failure records and a larger number of normally operating devices. However, not all moments without failures represent a risk level of zero. The KNN algorithm is highly sensitive to data imbalance. If the model does not handle data imbalance (e.g., oversampling or undersampling), it may overlook equipment in a high-risk period before failure, potentially leading to most devices being predicted as low-risk while ignoring a few high-risk devices. Because binary classification exacerbates this data imbalance, the KNN algorithm tends to predict low-risk devices and ignore high-risk ones, resulting in poor warning effectiveness and an inability to effectively identify potentially high-risk devices at risk of failure. Summary of the Invention

[0005] To address the technical problem that existing classification models lack balancing mechanisms, resulting in poor early warning performance and an inability to effectively identify high-risk equipment with potential faults, the present invention aims to provide an artificial intelligence-based power engineering risk classification and early warning method. The specific technical solution adopted is as follows:

[0006] In the historical data of mining power engineering, environmental factor data and equipment operation data for each dimension are acquired at each time point, and the time of each equipment failure is marked.

[0007] Based on the relative changes of environmental factor data in each dimension at the time of failure, and combined with the deviations of environmental factor data in each dimension at historical times before the time of failure, the risk coefficient of environmental factor data in each dimension at each time of failure is obtained.

[0008] Based on the equipment operation data of each dimension at each historical moment before the failure time and the time interval between the historical moment and the failure time, combined with the correlation between the equipment operation data and environmental factor data at each historical moment before the failure time and the risk coefficient, the risk coefficient of the equipment operation data of each dimension at each failure time is obtained.

[0009] By using risk coefficients to synthesize the data changes in the same dimension at each historical time before each fault, the risk level at each historical time before each fault is obtained. Based on the risk level, a risk classification model is constructed for risk classification and early warning of power engineering.

[0010] Preferably, the step of obtaining the risk coefficient of environmental factor data for each dimension at each fault time based on the relative changes of environmental factor data for each dimension at the fault time and the deviations of environmental factor data for each dimension at historical times prior to the fault time specifically includes:

[0011] Based on the deviation of environmental factor data for each dimension from preset values ​​at all fault times, the correlation index between environmental factor data for each dimension and fault occurrence is obtained.

[0012] Based on the deviation between the environmental factor data of each dimension at the previous historical time of each failure and the preset environmental failure threshold, and combined with the correlation index, the risk coefficient of the environmental factor data of each dimension at each failure time is obtained.

[0013] Preferably, the step of obtaining the correlation index between the environmental factor data of each dimension and the occurrence of the fault based on the deviation of the environmental factor data of each dimension at all fault times from the preset value specifically includes:

[0014] At each fault time, the absolute value of the difference between the normalized value and the preset value of the environmental factor data for each dimension is used as the environmental change factor for each dimension at each fault time.

[0015] The mean of environmental change factors at all fault times under each dimension is used as the correlation index between environmental factor data and fault occurrence in each dimension.

[0016] Preferably, the step of obtaining the risk coefficient of environmental factor data for each dimension at each fault time based on the deviation between the environmental factor data of each dimension at the previous historical time and the preset environmental fault threshold, combined with the correlation index, specifically includes:

[0017] For any dimension of environmental factor data, the negative correlation coefficient between the environmental factor data of the previous historical time and the preset environmental fault threshold at each fault time is used as the environmental deviation factor at that dimension at each fault time. The normalized result of the product between the environmental deviation factor and the correlation index corresponding to the environmental factor data of the same dimension is used as the risk coefficient of the environmental factor data of that dimension at each fault time.

[0018] Preferably, the step of obtaining the risk coefficient of the equipment operation data for each dimension at each time point before the failure, based on the equipment operation data for each dimension at each historical time point before the failure and the time interval between the historical time point and the failure time, combined with the correlation between the equipment operation data and environmental factor data at each historical time point before the failure and the risk coefficient, specifically includes:

[0019] Use any dimension of device operation data as the selected operation data, and use any moment of device failure as the target failure moment;

[0020] Based on the correlation coefficient between the selected operating data at each historical time before the target failure time and the environmental factor data of each dimension, the correlation characteristic factors between the selected operating data and the environmental factor data of each dimension at the target failure time are determined.

[0021] Based on the correlation characteristic factors between the selected operating data at the target failure time and the environmental factor data of each dimension, as well as the risk coefficients corresponding to the environmental factor factors of the same dimension, the basic risk indicators of the selected operating data at the target failure time are obtained.

[0022] Based on the historical differences of the selected operating data at each historical time before the target failure time, and combined with the time interval between the target failure time and each previous time, the failure change index of the selected operating data at the target failure time is obtained.

[0023] The risk coefficient of the selected operating data at the target failure time is obtained by normalizing the product between the basic risk index and the failure change index of the selected operating data at the target failure time.

[0024] Preferably, the step of obtaining the basic risk index of the selected operating data at the target failure time based on the correlation characteristic factors between the selected operating data at the target failure time and the environmental factor data of each dimension, as well as the risk coefficients corresponding to the environmental factor factors of the same dimension, specifically includes:

[0025] The correlation feature factors between the selected operating data at the time of the target failure and the environmental factor data of each dimension are used as the weight coefficients of the environmental factor data of each dimension at the time of the target failure. Using the weight coefficients, the negative correlation coefficients obtained by weighting and averaging the risk coefficients corresponding to the environmental factor data of each dimension at the time of the target failure are used as the basic risk indicators of the selected operating data at the time of the target failure.

[0026] Preferably, the step of obtaining the fault change index of the selected operating data at the target fault time based on the historical differences of the selected operating data at each historical time before the target fault time, combined with the time interval between the target fault time and each previous time, specifically includes:

[0027] Each historical moment before the target failure moment is used as a reference moment. The time interval between each reference moment and the target failure moment is negatively correlated to obtain the time decay weight of each reference moment.

[0028] The ratio of the absolute value of the difference between the selected running data at each reference time and the adjacent previous reference time to the selected running data at each reference time is used as the trend factor of the selected running data at each reference time.

[0029] By using the time decay weight of each reference time, the change trend factor is weighted and summed to obtain the fault change index of the selected operating data at the target fault time.

[0030] Preferably, the step of using risk coefficients to synthesize the data changes in the same dimension at each historical time before each failure time to obtain the risk level at each historical time before each failure time specifically includes:

[0031] The historical moment preceding any equipment failure time is selected as the selected moment, and each historical moment before the selected moment is used as the reference moment.

[0032] The data difference index for each control time in each dimension is obtained based on the data differences between the engineering monitoring data at each control time and the selected time in each dimension; where the engineering monitoring data includes environmental factor data and equipment operation data;

[0033] By using the risk coefficient of the engineering monitoring data for each dimension at the selected time corresponding to the fault time, the data difference indicators of each control time at the same dimension are weighted and summed, and then normalized to obtain the risk level of each control time.

[0034] Preferably, obtaining the data difference index for each control time in each dimension based on the data differences between the engineering monitoring data at each control time and the selected time in each dimension specifically includes:

[0035] The absolute values ​​of the differences between the engineering monitoring data at each control time and the selected time in each dimension are negatively correlated to obtain the data difference index at each control time in each dimension.

[0036] Preferably, the step of constructing a risk grading model based on the risk level specifically includes:

[0037] Based on the risk level of each historical moment prior to each failure, the risk level of each historical moment prior to each failure is classified; the risk level at the time of failure is the highest risk level.

[0038] The environmental factor data, equipment operation data, and equipment failure status of each failure moment and each previous historical moment are used to construct the feature vector of each moment. The feature vector of each moment is used as input and the risk level of the corresponding moment is used as output to train the classification model. The trained classification model is the risk grading model.

[0039] The embodiments of the present invention have at least the following beneficial effects:

[0040] This invention first collects monitoring data from different dimensions at both the environmental and equipment levels in mine power engineering, providing a data foundation for subsequent comprehensive risk analysis. Then, the first aspect analyzes the performance of environmental factors, considering changes and deviations in environmental factor data at the time of failure, and assesses the risk exhibited by each dimension of environmental factor data at the time of failure. Further, the second aspect analyzes the performance of equipment factors, considering the correlation between equipment and environmental factors before the time of failure. Combined with the risk coefficient of environmental factors, environmental influences can be effectively eliminated to obtain the true basic risk of the equipment itself. Then, considering time decay, the risk exhibited by each dimension of equipment operation data at the time of failure is assessed. Finally, by combining the risk analysis results from both aspects with the changes in environmental factor data or equipment operation data at the corresponding dimensions, the risk level at each historical moment is comprehensively characterized. This allows for a more comprehensive assessment of the risk level in mine power engineering, resulting in a better evaluation effect and a more effective early warning system for the constructed risk classification model, effectively identifying high-risk equipment with potential failures. Attached Figure Description

[0041] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart of the steps of a power engineering risk classification and early warning method based on artificial intelligence provided by the present invention;

[0043] Figure 2 This is a flowchart illustrating the steps of the method for obtaining the risk coefficient of environmental factor data for each dimension at each fault moment, as provided by the present invention.

[0044] Figure 3 This is a flowchart of the steps for obtaining the risk coefficient of selected operating data at the target fault moment provided by the present invention;

[0045] Figure 4 This is a flowchart of the steps for obtaining fault change indicators provided by the present invention;

[0046] Figure 5 This is a flowchart of the sub-steps of step S400 provided by the present invention. Detailed Implementation

[0047] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an artificial intelligence-based power engineering risk classification and early warning method proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0049] The following description, in conjunction with the accompanying drawings, details a specific scheme for an artificial intelligence-based risk classification and early warning method for power engineering provided by this invention.

[0050] Please see Figure 1 The diagram illustrates a flowchart of a power engineering risk classification and early warning method based on artificial intelligence, according to an embodiment of the present invention. The method includes the following steps:

[0051] Step S100: In the historical data of the mine power engineering, acquire environmental factor data and equipment operation data for each dimension at each time point, and mark the time of each equipment failure.

[0052] In the database of mining power engineering, historical monitoring data is retrieved to identify monitoring data at various dimensions before and during equipment failure. By analyzing the characteristics of monitoring data at and before failure in historical data, a data foundation is provided for accurately assessing the risk level at each historical moment before failure.

[0053] Specifically, environmental factor data includes ambient temperature, humidity, air pressure, and wind speed data at various locations within the mine. One dimension of environmental factor data corresponds to one type of data for a specific location; for example, temperature data for a location is one dimension of environmental factor data. Equipment operation data refers to data generated during the operation of various electrical equipment, including substations, generators, distribution cabinets, and power transformers. Therefore, equipment operation data includes current data, power data, equipment temperature, vibration data, etc. One dimension of equipment operation data corresponds to one type of data for a specific piece of equipment; for example, current data for a substation is one dimension of equipment operation data.

[0054] It should be noted that, in order to avoid the different units of different types of data affecting the results of multi-dimensional data analysis, the data collected in this embodiment for each dimension are all standardized data. The z-score standardization method can be used for processing, which will not be described in detail here.

[0055] At this point, we can obtain environmental factor data and equipment operation data for each dimension at every moment. It should be noted that these monitoring data are all historical data from the mine power engineering project, and we can directly retrieve the fault status of a specific electrical device at a specific moment from the database. That is, the moment when each equipment failure occurs is recorded as the fault moment, and the equipment that exhibited the faulty phenomenon is identified as the faulty equipment.

[0056] It should be understood that in the process of real-time monitoring of mine power engineering, various sensors collect monitoring data from various dimensions, and then transmit the data to the control system or database for storage via wireless or wired networks.

[0057] Because different environmental factors and equipment operations have varying impacts on each failure, a detailed analysis of the environmental factors of the mine power system prior to each failure is necessary to determine the risk coefficient of each environmental factor. Furthermore, it is required to analyze the risk impact of various environmental factors on the operational data of each piece of equipment in the mine power system before each failure, and assign a corresponding risk coefficient to the operational data of each piece of equipment. In this way, the risk of the mine power system at each historical moment can be comprehensively assessed by observing the changes in the values ​​of environmental factors and equipment operational risk coefficients. Specifically, step S200 assesses the risk coefficient of environmental factors based on the characteristic performance of various environmental factor data; step S300 assesses the risk coefficient of equipment operation based on the characteristic performance of various equipment operational data; and step S400 integrates the risk analysis results from both aspects to comprehensively assess the risk performance of the mine power system at each historical moment.

[0058] Step S200: Based on the relative changes in environmental factor data for each dimension at the time of failure, and combined with the deviations in environmental factor data for each dimension at historical times prior to the time of failure, obtain the risk coefficient of environmental factor data for each dimension at each time of failure.

[0059] Specifically, this step mainly analyzes the correlation between environmental factors and the failure when the failure occurs, and quantifies the risk contribution of environmental factors. That is, by analyzing the relative changes of environmental factor data in each dimension when the failure occurs, and combining the deviation of environmental factor data at historical moments before the failure with the data at the time of the failure, the risk level of environmental factor data in each dimension at the time of the failure is assessed.

[0060] As a concrete example, such as Figure 2 As shown, the method for obtaining the risk coefficient of environmental factor data for each dimension at each fault time can be implemented by steps S201 to S203.

[0061] Step S201: Based on the deviation of environmental factor data for each dimension from preset values ​​at all fault times, obtain the correlation index between environmental factor data for each dimension and fault occurrence.

[0062] In mining power engineering, changes in environmental factors, such as temperature, humidity, air pressure, and wind speed, can directly or indirectly affect the normal operation of electrical equipment. If environmental factors exceed normal ranges, the equipment may be subjected to unsuitable pressure or load, leading to malfunctions. For example, excessively high temperatures may cause equipment overheating, resulting in short circuits or aging of insulation materials; excessive humidity may cause corrosion or short circuits in electrical equipment.

[0063] Therefore, if an environmental factor changes significantly compared to normal (e.g., abnormal temperature increase or decrease, significant increase or decrease in humidity, etc.), and the equipment malfunctions at this time, it usually indicates that the environmental factor may be highly correlated with the occurrence of the malfunction.

[0064] Specifically, the first step is to use the absolute value of the difference between the normalized value and the preset value of the environmental factor data for each dimension at each fault time as the environmental change factor for each dimension at each fault time.

[0065] In this embodiment, the environmental factor data can be normalized using the minimax normalization method, without any limitation. The value range of each environmental factor data after normalization is (0,1). Therefore, in this embodiment, a preset value is set as the median of the value range, representing the normal median value of the environmental factor data. That is, the preset value in this embodiment is 0.5.

[0066] Furthermore, the environmental change factor for each dimension at each fault time reflects the relative degree of change of environmental factor data for each dimension at each fault time. The larger the value, the more abnormal the change of environmental factor data for that dimension is when the fault occurs.

[0067] The second step is to use the mean of the environmental change factors at all fault times under each dimension as the correlation index between the environmental factor data of each dimension and the occurrence of the fault.

[0068] Specifically, for each dimension, the mean of the environmental change factors at all fault times reflects the balance of the degree of abnormal change in environmental factor data under that dimension. The larger the value, the greater the correlation between the change in environmental factor data at the time of fault occurrence and the fault occurrence event under that dimension, and the larger the value of the corresponding correlation index.

[0069] Thus, each dimension of environmental factor data corresponds to a correlation index, which represents the degree of overall correlation between environmental factor data and faults under that dimension.

[0070] Step S202: Based on the correlation index between environmental factor data and fault occurrence in each dimension and the total number of fault occurrences, filter the environmental factor data of each dimension for the previous historical moment of each fault to obtain the environmental fault threshold for each environmental dimension.

[0071] To determine the failure threshold for each environmental factor, it is necessary to select the data most strongly correlated with failure occurrence. This primarily involves filtering environmental factor data prior to failure, focusing on the most critical environmental factors for each dimension before failure. These data play a significant role in the pre-failure process, more accurately reflecting the precursory signs of equipment failure. The maximum possible data manifestation of the failure, i.e., the environmental failure threshold, is adaptively determined using the selected environmental factor data. This reduces the impact of extreme values ​​on the results, making the determined threshold more stable and reliable. In other embodiments, the implementer can set a custom environmental failure threshold based on the specific implementation scenario.

[0072] In this embodiment, the first step is to use environmental factor data of any one dimension as an example. The environmental factor data of any one dimension is denoted as the target dimension environmental factor data. The moment immediately preceding the fault time of each fault occurrence is recorded as the neighboring moment of each fault time. The neighboring moment represents the critical time point before the fault occurs.

[0073] The second step is to round up the product of the correlation index corresponding to the environmental factor data in the target dimension and the total number of equipment failures to obtain the number of features of the environmental factor data in the target dimension.

[0074] The number of features indicates that when the correlation between environmental factor data in the target dimension and the fault is greater, more historical fault condition data values ​​should be selected to adaptively determine the threshold.

[0075] The third step involves arranging all nearby time points in descending order of their environmental factor data, based on the target dimension. Then, a specific number of nearby time points are selected according to this arrangement, and the mean of the environmental factor data from these selected nearby time points is calculated as the environmental fault threshold. This adaptive determination of the environmental fault threshold stably reflects the data values ​​of the typical critical state before a fault occurs.

[0076] Step S203: Based on the deviation between the environmental factor data of each dimension at the previous historical time of each fault and the preset environmental fault threshold, and in conjunction with the correlation index, obtain the risk coefficient of the environmental factor data of each dimension at each fault time.

[0077] In mining power engineering, the correlation between environmental factors and fault occurrence reflects the degree to which changes in environmental factors contribute to fault occurrence. A high correlation indicates that changes in environmental factors have a significant impact on fault occurrence. When environmental factor data approaches the fault threshold, even minor fluctuations can trigger a fault, leading to a rapid increase in the risk factor.

[0078] Based on the correlation between environmental factors and failure occurrence, since the failure threshold for environmental factors is a critical value, when environmental factors approach this threshold, the equipment's capacity is nearing its limit. At this point, fluctuations or minor changes in environmental factors may trigger a failure. Therefore, when the value of environmental factors at the moment before a failure occurs is close to the threshold (i.e., close to the threshold), it indicates that the equipment is already on the verge of danger, the probability of failure caused by environmental factors increases, and the risk factor rises accordingly.

[0079] Specifically, for any dimension of environmental factor data, the negative correlation coefficient between the environmental factor data of the previous historical time before each failure time and the preset environmental failure threshold is used as the environmental deviation factor for that dimension at each failure time. The normalized result of the product between the environmental deviation factor and the correlation index corresponding to the environmental factor data of the same dimension is used as the risk coefficient of the environmental factor data of that dimension at each failure time. In this embodiment, each time before each failure time is a historical time.

[0080] As a concrete example, taking environmental factor data of any dimension and any failure time as an illustration, the risk coefficient of the environmental factor data of the a-th dimension at the h-th failure time can be expressed by the formula:

[0081]

[0082] in, This represents the risk coefficient of the environmental factor data in the a-th dimension at the h-th failure time. This represents the environmental factor data for the a-th dimension. This indicates the h-th fault time, which is also the time when the h-th fault occurs. This represents the environmental factor data from the previous historical moment before the h-th fault. This represents the preset environmental fault threshold, which in this embodiment is the environmental fault threshold of the environmental factor data of the a-th dimension. This represents the correlation index between the environmental factor data of the a-th dimension and the occurrence of the failure. This represents an exponential function with base e. This is the normalization function.

[0083] The environmental deviation factor is the number of environmental factors in the a-th dimension at the h-th fault time. It reflects the degree to which the environmental factor data in the a-th dimension deviates from the threshold before the h-th fault occurs. The larger the value, the greater the correlation between the environmental factor data in this dimension and the fault, and the higher the potential triggering risk of the environmental factors for the fault at the time of the fault.

[0084] Thus, the risk coefficient corresponding to the environmental factor data represents the magnitude of the potential triggering risk of the corresponding dimension of environmental factor data for a failure under a single failure event.

[0085] Step S300: Based on the equipment operation data of each dimension at each historical time before the failure time and the time interval between the historical time and the failure time, combined with the correlation between the equipment operation data and environmental factor data at each historical time before the failure time and the risk coefficient, the risk coefficient of the equipment operation data of each dimension at each failure time is obtained.

[0086] Specifically, this step mainly analyzes the risk of each dimension of equipment operation data under different environmental influences before the failure occurs, requiring the separation of environmental interference to focus on the abnormal risk manifestations of the equipment itself. As a concrete example, such as... Figure 3 As shown, the example is taken with equipment operation data of any dimension and the time of any equipment failure. That is, the equipment operation data of any dimension is used as the selected operation data, and the time of any equipment failure is used as the target failure time. The method of obtaining the risk coefficient of the selected operation data under the target failure time can be implemented by steps S301 to S304.

[0087] Step S301: Based on the correlation coefficient between the selected operating data at each historical time before the target failure time and the environmental factor data of each dimension, determine the correlation feature factors between the selected operating data at the target failure time and the environmental factor data of each dimension.

[0088] The Pearson correlation coefficient measures the linear relationship between two variables. It ranges from -1 to 1. The closer the value is to 1, the stronger the positive correlation between the two variables; the closer the value is to -1, the stronger the negative correlation; a value of 0 indicates that there is no linear relationship.

[0089] Before a failure occurs, the larger the absolute value of the Pearson correlation coefficient between environmental factor data and equipment operation data in each dimension, the stronger the correlation between them. For example, excessively high temperature or humidity may directly affect the normal operation of electrical equipment. In this case, the equipment operation data, such as power and temperature, will show a strong correlation with changes in environmental factors.

[0090] Based on this, the selected operating data at each historical moment before the target failure time are used to form a selected operating data sequence, and the environmental factor data at each moment before the target failure time in each dimension are used to form an environmental data sequence in each dimension. The absolute value of the Pearson correlation coefficient between the selected operating data sequence and the environmental data sequence in each dimension is used as the correlation characteristic factor between the selected operating data and the environmental factor data in each dimension at the target failure time.

[0091] The relevant characteristic factor characterizes the degree of correlation between the operating data of a device and various types of environmental factors before the failure occurs, under the event corresponding to the failure at the target failure time. The larger the value, the stronger the correlation between the two, and thus the greater the influence of the environment on the device operation.

[0092] It should be noted that, in this embodiment, all historical moments before each fault moment refer to the moments between two adjacent fault moments. That is, the historical moments corresponding to each fault moment include all moments between the current fault moment and the adjacent previous fault moment. The purpose is to analyze the data distribution before a single fault occurs.

[0093] By following the same method, we can obtain the degree of correlation between environmental factor data and equipment operation data in each dimension, which is the correlation characteristic factor.

[0094] Step S302: Based on the correlation characteristic factors between the selected operating data at the target failure time and the environmental factor data of each dimension, as well as the risk coefficients corresponding to the environmental factor factors of the same dimension, the basic risk index of the selected operating data at the target failure time is obtained.

[0095] When assessing the risk coefficient of equipment operating data, it is necessary to isolate the interference of other environmental factors and accurately identify the risk contribution of the equipment operation itself. There is a complex correlation between environmental factors and equipment operation; directly calculating the risk coefficient will introduce the synergistic effects of environmental factors, leading to an overestimation or underestimation of the true risk. For example, an increase in temperature may simultaneously affect the equipment's heat dissipation efficiency and humidity sensor readings. If the influence of temperature is not eliminated, fluctuations in equipment operating data may be incorrectly attributed to its own malfunction rather than environmental changes.

[0096] Based on this, by removing the correlation between environmental factors and equipment operating data, the fluctuations in equipment operating data (such as current, voltage, and temperature) before each failure are analyzed. Large fluctuations typically indicate system anomalies, such as component wear, overload, or insulation aging, directly increasing the risk of failure. These fluctuations may be driven by internal faults or external environmental factors, but regardless of the cause, abnormal changes signify decreased equipment stability, making it more prone to failure. Therefore, the greater the fluctuation amplitude, the higher the risk coefficient, serving as a key indicator for fault early warning and more accurately reflecting the true risk level.

[0097] Specifically, before each failure occurs, the risk coefficient of each dimension of equipment operation data is assessed. Firstly, considering the correlation between environmental factor data and equipment operation data, and combining the risk level shown by the corresponding environmental factor data, the comprehensive interference of environmental factors on equipment operation is analyzed to preliminarily assess the basic risk performance of the equipment's own operating status.

[0098] Based on this, the correlation feature factors between the selected operating data at the time of the target failure and the environmental factor data of each dimension are used as the weight coefficients of the environmental factor data of each dimension at the time of the target failure. Using the weight coefficients, the negative correlation coefficients obtained by weighting and averaging the risk coefficients corresponding to the environmental factor data of each dimension at the time of the target failure are used as the basic risk indicators of the selected operating data at the time of the target failure.

[0099] As a concrete example, if we take the equipment operation data of the x-th dimension at the h-th failure time as the selected operation data at the target failure time, then the basic risk index of the selected operation data at the target failure time can be expressed by the formula:

[0100]

[0101] in, This indicates the basic risk indicators of the selected operational data at the time of the target failure. This represents the device operation data in the xth dimension, which is also the selected operation data; This indicates the time of the h-th fault. This represents the risk coefficient of the environmental factor data in the a-th dimension at the h-th failure time. This represents the correlation feature factor between the environmental factor data in dimension a and the equipment operation data in dimension x at the h-th fault time. This represents the total number of dimensions for all environmental factor data. This is the normalization function.

[0102] Weighting coefficient The larger the value, the greater the correlation between environmental factor data and equipment operation data. Consequently, the higher the failure risk level of environmental factor data in this dimension, the greater the risk impact on equipment operation data in that dimension. This reflects the comprehensive impact of environmental factors across all dimensions. Furthermore, after negative correlation processing, the resulting basic risk index represents the degree of equipment failure risk after eliminating environmental interference. This provides a data foundation for subsequently refining the variability of equipment operation data and obtaining the final risk coefficient.

[0103] Step S303: Based on the historical differences of the selected operating data at each historical time before the target failure time, and combined with the time interval between the target failure time and each previous time, obtain the failure change index of the selected operating data at the target failure time.

[0104] First, assess the risk coefficient of each dimension of equipment operation data before each failure occurs. Second, assess the degree of data fluctuation of each dimension of equipment operation data before each failure occurs. The greater the data fluctuation, the higher the risk of abnormal equipment operation.

[0105] As a concrete example, such as Figure 4 As shown, the method for obtaining fault change indicators can be implemented by steps S3031 to S3033.

[0106] Step S3031: Take each historical time before the target failure time as a reference time, perform negative correlation processing on the time interval between each reference time and the target failure time, and obtain the time decay weight of each reference time.

[0107] Specifically, the time decay weight at the i-th reference time before the h-th fault time. It can be represented as: ,in This represents the i-th reference time before the h-th fault time. This represents the time interval between the h-th fault time and the corresponding i-th reference time. This represents an exponential function with the natural constant e as its base. This indicates the normalization method, such as the minimax normalization method.

[0108] The longer the time interval between the reference time and the target failure time, the lower the reference value of the equipment operation data at the corresponding reference time; therefore, the smaller the value of the time decay weight. Conversely, the shorter the time interval between the reference time and the target failure time, the greater the reference value of the equipment operation data at the corresponding reference time; therefore, the larger the value of the time decay weight.

[0109] Step S3032: The ratio of the absolute value of the difference between the selected running data at each reference time and the adjacent previous reference time to the selected running data at each reference time is used as the trend factor of the selected running data at each reference time.

[0110] For the i-th reference time before the target failure time, calculate the absolute value of the difference between the selected operating data and the (i-1)-th reference time, and then use the ratio of the absolute value of the difference to the selected operating data at the i-th reference time as the trend factor of the selected operating data at the i-th reference time.

[0111] The trend factor measures the relative fluctuation of the device operation data in that dimension at each historical moment by comparing the difference between the selected operation data (the xth dimension's equipment operation data) and the original data at adjacent historical moments. The greater the relative fluctuation, the larger the value of the corresponding trend factor.

[0112] Step S3033: Using the time decay weight of each reference time, the change trend factor is weighted and summed to obtain the fault change index of the selected operating data at the target fault time.

[0113] Specifically, taking the equipment operation data of the xth dimension as an example, that is, in this embodiment, the equipment operation data of the xth dimension is used as the selected operation data, and the hth fault time is used as the target fault time. Then, the calculation formula of the fault change index of the selected operation data at the target fault time can be expressed as:

[0114]

[0115] in, This indicates the fault change index of selected operational data at the target fault time. This represents the device operation data in the xth dimension, which is also the selected operation data; This indicates the h-th fault time; N represents the total number of historical times preceding the target fault time. This represents the time decay weight of the i-th reference time before the h-th fault time. The factor representing the trend of change in the x-th dimension of the device operation data at the i-th reference time.

[0116] The time decay weight characterizes the reference value of the data fluctuation of the selected operating data at the corresponding reference time. The larger the value, the greater the degree of data fluctuation of the selected operating data at the corresponding reference time, indicating that the risk of equipment failure is higher, and the corresponding failure change index value is larger.

[0117] The fault variation index indicates the degree of fluctuation in the overall equipment operating data before the fault. The larger the value, the more drastic the fluctuation in the equipment operating data under the corresponding dimension before the equipment fault, and the higher the risk of abnormality of the equipment itself in that dimension.

[0118] Step S304: Normalize the product between the basic risk index and the fault change index of the selected operating data at the target fault time to obtain the risk coefficient of the selected operating data at the target fault time.

[0119] The normalization method can be the minimax normalization method, which will not be discussed further here.

[0120] By following the same method, we can obtain the risk coefficient of each dimension of equipment operation data at each fault moment. This not only retains the risk information of the equipment's own operational fluctuations but also eliminates the interference of environmental factors, more accurately reflecting the contribution of the equipment's own state to the fault and providing reliable data for subsequent risk level assessment.

[0121] Step S400: Use the risk coefficient to synthesize the data changes in the same dimension at each historical time before each fault time to obtain the risk level at each historical time before each fault time. Construct a risk classification model based on the risk level for risk classification and early warning of power engineering.

[0122] In mining power engineering, assessing the risk at each historical moment can be done by analyzing changes in environmental and equipment operating factors. Given that a historical moment occurs before a specific historical failure, and the risk coefficients of the environmental and equipment operating factors for that failure are known, then if, prior to the failure, the risk coefficient of a factor at a given historical moment is already high, the smaller the difference between that factor value and the value at the moment preceding the failure, the more likely that factor was the cause of the equipment failure, leading to an increased risk of failure. Therefore, this indicates a higher risk at that historical moment.

[0123] Therefore, by analyzing the differences between historical factor changes and those prior to historical failures, the risk at that moment can be assessed, providing a basis for failure prediction and early warning. As a concrete example, such as... Figure 5 As shown, the sub-steps of step S400 specifically include steps S401 to S404.

[0124] Step S401: Select the historical moment preceding any equipment failure moment as the selected moment, and use each historical moment before the selected moment as the reference moment.

[0125] The selected time point represents a moment before a failure occurs, reflecting objects in the time proximity to the failure, and the data at the selected time point is in a critical state. Each control time point represents a historical moment when risk analysis is about to be conducted.

[0126] Step S402: Obtain the data difference index for each control time in each dimension based on the data differences between the engineering monitoring data at each control time and the selected time in each dimension; wherein, the engineering monitoring data includes environmental factor data and equipment operation data.

[0127] Specifically, in this step, when comprehensively assessing the risk situation at each moment, it is necessary to consider all factors in the data. This means integrating the deviations and risk levels of environmental factor data and equipment operation data for each dimension to fully reflect the risk level at each historical moment before the equipment failure. First, both environmental factor data and equipment operation data are recorded as engineering monitoring data. The total number of dimensions of engineering monitoring data is the sum of the total number of dimensions of environmental factor data and equipment operation data. Similarly, at each failure moment, each dimension of engineering monitoring data corresponds to a risk coefficient value, representing the degree of risk of the engineering monitoring data in that dimension.

[0128] Specifically, a negative correlation is applied to the absolute values ​​of the differences between the engineering monitoring data at each control time and the selected time in each dimension to obtain the data difference index for each control time in each dimension. The negative correlation can be performed using a negative exponential function, for example... ,in This represents an exponential function with base e. This indicates the object to be processed for negative correlation; here, it represents the calculation result of the absolute value of the corresponding difference.

[0129] Step S403: Using the risk coefficient of the engineering monitoring data for each dimension at the selected time corresponding to the fault time, the data difference index of each control time at the same dimension is weighted and summed, and then normalized to obtain the risk level of each control time.

[0130] As a concrete example, taking the time immediately preceding the h-th failure time as the selected time, and using the n-th control time as an example, the risk level of the n-th control time can be expressed as:

[0131]

[0132] in, This indicates the risk level at the nth control time corresponding to the hth failure time. This represents the risk coefficient of the engineering monitoring data in the y-th dimension at the h-th fault time. This represents the engineering monitoring data in the y-th dimension of the time interval preceding the h-th fault time interval. This represents the engineering monitoring data in the y-th dimension at the nth control time. Let y be the data difference index at the nth control time point in the y-th dimension. This represents the set of all dimensions of engineering monitoring data. This represents the normalization method, which can specifically be minimization normalization.

[0133] Differences between engineering monitoring data at the control time and the selected time in each dimension This reflects the data difference between the reference time and the critical state value of the fault. The greater the difference, the larger the corresponding data difference index value. At the same time, the larger the risk coefficient value of the data under the dimension, the higher the risk level of the data under that dimension, and thus the higher the risk of the corresponding historical time.

[0134] Step S404: Construct a risk classification model based on the risk level for risk classification and early warning of power engineering projects.

[0135] In mining power engineering, assessing the detailed risk levels at each historical moment is crucial for accurately describing the risk situation at that time and providing a basis for subsequent decision-making and early warning. Given the risk level at each moment preceding every historical fault in a mining power engineering project, to facilitate management and operation, the risk level corresponding to each moment is converted into discrete risk levels, providing a clearer distinction between risks.

[0136] The first step is to classify the risk level of each historical moment before each failure by utilizing the risk level of each historical moment before each failure; the risk level at the failure moment is the highest risk level.

[0137] As a specific example, this embodiment divides the risk level into 10 levels. The risk level of each historical moment before each failure moment is multiplied by 10 and rounded up to obtain the risk level of each historical moment before each failure moment.

[0138] It should be noted that the risk level mentioned above is not calculated for each fault time and the time immediately preceding the fault time. In this embodiment, the fault time and the time immediately preceding it can be assigned the highest risk level, indicating that the risk of a fault occurring at these two times is the highest.

[0139] Based on this, the risk level at each moment can be obtained. The higher the risk level, the higher the risk at each moment corresponding to the historical data in the mine power project. By multiplying by 10 and rounding up, over-refinement can be avoided, preventing unnecessary complex judgments due to excessively small risk differences. Summarizing it into 10 levels makes risk assessment more practical and operable, helps to simplify and quantify risk assessment, and thus provides a clear basis for the safety management of mine power projects.

[0140] The second step involves constructing a feature vector for each time step using environmental factor data, equipment operation data, and equipment failure status from each time step and previous historical time steps. The feature vector for each time step is used as input, and the risk level for the corresponding time step is used as output to train a classification model. The trained classification model is a risk grading model.

[0141] In simple terms, all historical device operation data and environmental data are compiled into a dataset, with each data point representing a sample. Features include device status and environmental conditions. These datasets are divided into training and testing sets. During the training phase, the KNN algorithm calculates the distances between different samples, finds the K most similar neighbors to the current sample, and predicts its risk level.

[0142] More specifically, a sample dataset is constructed by combining environmental factor data from various dimensions, equipment operation data from various dimensions, and equipment fault status at a given moment. Here, equipment fault status refers to the data value indicating whether a fault has occurred at that moment. In this embodiment, it is marked with 0 and 1 values, where 0 indicates no equipment fault at that moment, and 1 indicates that a equipment fault has occurred at that moment.

[0143] Furthermore, in this embodiment, the KNN algorithm is selected as the classification model for training the data. The training process is a well-known technique and will not be described in detail here.

[0144] In the early warning system, based on real-time data input, the KNN algorithm compares real-time samples with historical data. By judging the risk level distribution of its K nearest neighbors, it predicts the risk level at the current moment. When a high-risk situation is detected, such as risk level 7 or higher, no specific restrictions are imposed; implementers can define limitations based on the specific implementation scenario. The system can also trigger an early warning, alerting relevant personnel to perform equipment maintenance. Through this method, mine power engineering can effectively perform real-time monitoring and risk classification early warning, improving overall safety and response efficiency.

[0145] In summary, due to the varying impacts of different environmental factors and equipment operation on each failure, a detailed analysis of the environmental factors of the mine power system prior to each failure is necessary to determine the risk coefficient of each environmental factor. Furthermore, it is crucial to analyze the risk impact of various environmental factors on the operational data of each piece of equipment in the mine power system before each failure, and assign a corresponding risk coefficient to the operational data of each piece of equipment. In this way, by observing the changes in the values ​​of environmental factors and equipment operational risk coefficients, a comprehensive assessment of the risk of the mine power system at each historical moment can be conducted, providing data support for subsequent risk classification and early warning. Finally, by determining the environmental factor risk before each failure and the equipment failure risk value under the influence of environmental factors, the risk level of each historical moment in the mine power project is determined, ultimately enabling the application of the KNN algorithm for risk classification and early warning in subsequent mine power projects.

[0146] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An artificial intelligence-based power engineering risk grading early warning method, characterized in that, The method comprises the following steps: In the historical data of the mine power engineering, the environmental factor data of each dimension and the equipment operation data of each dimension are respectively acquired at each time, and the failure time of each occurrence of equipment failure is marked; According to the relative change of the environmental factor data of each dimension at the failure time, in combination with the deviation of the environmental factor data of each dimension at the historical time before the failure time, a risk coefficient of the environmental factor data of each dimension at each failure time is obtained; According to the equipment operation data of each dimension at each historical time before the failure time and the time interval between the historical time and the failure time, in combination with the correlation between the equipment operation data and the environmental factor data at each historical time before the failure time and the risk coefficient, a risk coefficient of the equipment operation data of each dimension at each failure time is obtained; The risk degree of each historical time before each failure time is obtained by comprehensively utilizing the risk coefficient and the data change of the same dimension at each historical time before each failure time, and a risk grading model is constructed according to the risk degree, which is used for power engineering risk grading early warning; The risk coefficient of the environmental factor data of each dimension at each failure time is obtained according to the relative change of the environmental factor data of each dimension at the failure time, in combination with the deviation of the environmental factor data of each dimension at the historical time before the failure time, and specifically comprises: According to the deviation of the environmental factor data of each dimension at all failure times from a preset value, an index of the correlation degree between the environmental factor data of each dimension and the occurrence of failure is obtained; According to the deviation between the environmental factor data of each dimension at the previous historical time of each failure time and a preset environmental failure threshold, in combination with the correlation degree index, a risk coefficient of the environmental factor data of each dimension at each failure time is obtained, which comprises: for the environmental factor data of any one dimension, a negative correlation coefficient of the difference between the environmental factor data at the previous historical time of each failure time and the preset environmental failure threshold is taken as an environmental deviation factor of the dimension at each failure time, and a normalized result of the product between the environmental deviation factor and the correlation degree index corresponding to the environmental factor data of the same dimension is taken as the risk coefficient of the environmental factor data of the dimension at each failure time; The risk coefficient of the equipment operation data of each dimension at each failure time is obtained according to the equipment operation data of each dimension at each historical time before the failure time and the time interval between the historical time and the failure time, in combination with the correlation between the equipment operation data and the environmental factor data at each historical time before the failure time and the risk coefficient, and specifically comprises: The equipment operation data of any one dimension is taken as selected operation data, and the failure time of any occurrence of equipment failure is taken as a target failure time; Based on the correlation coefficient between the selected operation data and the environmental factor data of each dimension at each historical time before the target failure time, a correlation feature factor between the selected operation data and the environmental factor data of each dimension at the target failure time is determined; According to the correlation characteristic factor between the selected operation data at the target fault moment and the environmental factor data of each dimension and the risk coefficient corresponding to the environmental factor of the same dimension, a basic risk index of the selected operation data at the target fault moment is obtained; According to the historical difference of the selected operation data at each historical moment before the target fault moment, and in combination with the time interval between the target fault moment and each moment before it, a fault change index of the selected operation data at the target fault moment is obtained; The product of the basic risk index and the fault change index of the selected operation data at the target fault moment is normalized to obtain a risk coefficient of the selected operation data at the target fault moment; The historical difference of the selected operation data at each historical moment before the target fault moment is obtained, and in combination with the time interval between the target fault moment and each moment before it, a fault change index of the selected operation data at the target fault moment is obtained, specifically including: Each historical moment before the target fault moment is taken as a reference moment, and the time interval between each reference moment and the target fault moment is negatively correlated to obtain a time decay weight of each reference moment; The ratio of the absolute value of the difference of the selected operation data between each reference moment and the adjacent previous reference moment to the selected operation data of each reference moment is taken as a change trend factor of the selected operation data of each reference moment; The change trend factor is weighted and summed by using the time decay weight of each reference moment to obtain a fault change index of the selected operation data at the target fault moment; The risk degree of each historical moment before each fault moment is obtained by comprehensively using the risk coefficient and the data change of the same dimension at each historical moment before each fault moment, specifically including: A historical moment before any occurrence of equipment failure is taken as a selected moment, and each historical moment before the selected moment is taken as a control moment; The absolute value of the difference between the engineering monitoring data of each control moment and the selected moment in each dimension is negatively correlated to obtain a data difference index of each control moment in each dimension; wherein the engineering monitoring data includes environmental factor data and equipment operation data; The risk degree of each control moment is obtained by weighted summing and normalizing the data difference index of each control moment in the same dimension by using the risk coefficient of the engineering monitoring data of each dimension at the corresponding fault moment of the selected moment.

2. The power engineering risk grading and early warning method based on artificial intelligence according to claim 1, characterized in that, The correlation degree index of the environmental factor data of each dimension and the occurrence of failure is obtained according to the deviation of the environmental factor data of each dimension at all fault moments from a preset value, specifically including: At each fault moment, the absolute value of the difference between the normalized value of the environmental factor data of each dimension and the preset value is taken as the environmental change factor of each dimension at each fault moment; The mean value of the environmental change factors of all fault moments of each dimension is taken as the correlation degree index of the environmental factor data of each dimension and the occurrence of failure.

3. The power engineering risk grading and early warning method based on artificial intelligence according to claim 1, characterized in that, The basic risk index of the selected operation data at the target fault time is obtained according to the correlation characteristic factor between the selected operation data at the target fault time and the environmental factor data of each dimension and the risk coefficient corresponding to the environmental factor of the same dimension, and specifically comprises: The correlation characteristic factor between the selected operation data at the target fault time and the environmental factor data of each dimension is taken as a weight coefficient of the environmental factor data of each dimension at the target fault time, and the negative correlation coefficient of the weighted average of the risk coefficients corresponding to the environmental factor data of each dimension at the target fault time is taken as the basic risk index of the selected operation data at the target fault time by using the weight coefficient.

4. The power engineering risk grading and early warning method based on artificial intelligence according to claim 1, characterized in that, The risk grading model is constructed according to the risk degree, and specifically comprises: The risk levels of each historical time before each fault time are divided by using the risk degree of each historical time before each fault time; and the risk level of the fault time is the highest risk level; The environmental factor data, the equipment operation data and the equipment fault state of each historical time before each fault time are used to form a feature vector of each time; the feature vector of each time is taken as an input, and the risk level of the corresponding time is taken as an output, so that a classification model is trained, and the trained classification model is a risk grading model.

Images