A smart grid online security operation and maintenance management system

Abstract

The present invention discloses a smart grid online safety operation and maintenance management system, which specifically relates to the technical field of grid operation and maintenance management, including an operation data acquisition module, an environmental data acquisition module, a data analysis module, a fault analysis module, a fault diagnosis module, and an automated operation and maintenance module; the operation data acquisition module acquires real-time operation data of grid equipment, the environmental data acquisition module acquires equipment environment data, the data analysis module analyzes the operation data and determines the fault demand; the fault analysis module generates an abnormal fluctuation index and a trend index, performs fuzzy reasoning to obtain the fault level and determines whether to trigger a multidimensional data support decision-making mechanism; the fault diagnosis module performs fault diagnosis, and the automated operation and maintenance module automatically manages operation and maintenance according to the diagnosis results and preset strategies; the present invention significantly reduces the demand for manual inspections and interventions, thereby reducing operation and maintenance costs, and can detect and handle faults in a timely manner, reducing economic losses and power outage risks caused by equipment failures.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid operation and maintenance management, and more specifically, to an online safe operation and maintenance management system for a smart power grid. Background Art

[0002] The power grid refers to an electric power system consisting of power generation equipment, substations, transmission lines and distribution lines, which is used to transmit electric energy from power plants to users. It achieves stable supply and consumption of electricity through a complex power transmission and distribution network, ensuring the power demand of social production and daily life. The power grid can cover the entire power transmission path from large-scale power plants to end users.

[0003] With the continuous expansion of the scale of power grids and the improvement of their intelligence, the operation and maintenance management of power equipment has become increasingly complex and important. Traditional power equipment operation and maintenance rely on manual regular inspections and passive repairs after failures, which usually have problems such as long response time, low maintenance efficiency, and poor fault prevention capabilities. In a complex power grid environment, the stability of equipment operation is directly related to the safety and reliability of the power system. Therefore, how to quickly and efficiently manage the operation and maintenance of power equipment has become a problem that needs to be solved urgently. Summary of the invention

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A smart grid online safety operation and maintenance management system, including an operation data acquisition module, an environmental data acquisition module, a data analysis module, a fault analysis module, a fault diagnosis module, and an automated operation and maintenance module;

[0006] The operation data acquisition module is used to collect various preset operation data of the target power grid equipment in real time, obtain the power grid equipment operation data set and transmit it to the data analysis module;

[0007] The environmental data acquisition module is used to collect various preset environmental data in the target power grid equipment operating environment in real time, obtain the power grid equipment environmental data set and transmit it to the fault diagnosis module when the multi-dimensional data support decision-making mechanism is triggered;

[0008] The data analysis module is used to perform data analysis operations based on the power grid equipment operation data set, and determine whether there is a need for fault analysis based on the data analysis results;

[0009] The fault analysis module performs fault analysis when there is a need for fault analysis, generates an abnormal fluctuation index and an abnormal trend index, and then performs fuzzy reasoning to obtain the fault level. When the fault level reaches the preset level range, the multi-dimensional data support decision-making mechanism is triggered;

[0010] The fault diagnosis module performs self-fault diagnosis or joint fault diagnosis on the target power grid equipment according to the fault level obtained by the fault analysis module and whether the multi-dimensional data support decision-making mechanism is triggered, and transmits the fault diagnosis results to the automated operation and maintenance module;

[0011] The automated operation and maintenance module performs operation and maintenance management based on fault diagnosis results and preset automated operation and maintenance strategies.

[0012] In a preferred embodiment, the data analysis module is used to perform data analysis operations based on the power grid equipment operation data set, which refers to:

[0013] The time series data of the target operation index are obtained from the power grid equipment operation data set, and then the mean value μ and standard deviation σ of the time series data are calculated respectively, and the hidden danger judgment interval is set as If the data points in the time series data do not fall into the hidden danger judgment interval When the data point in the time series data falls into the hidden danger judgment interval, the data point is marked as an abnormal point. , the data point is marked as a normal point, and finally the proportion of abnormal points to the total number of all data points is calculated.

[0014] In a preferred embodiment, judging whether there is a need for fault analysis based on the data analysis result means:

[0015] The proportion of abnormal points to the total number of all data points is compared with the preset normal distribution threshold. If the proportion of abnormal points to the total number of all data points is greater than or equal to the preset normal distribution threshold, it is determined that there is a need for fault analysis. If the proportion of abnormal points to the total number of all data points is less than the preset normal distribution threshold, it is determined that there is no need for fault analysis.

[0016] In a preferred embodiment, the confirmation logic of the target operation index is:

[0017] Use multiple candidate operating indicators X to train the machine learning model f(X). The machine learning model is a random forest model. The importance contribution of each feature, that is, the operating indicator, in predicting the probability of fault occurrence is calculated through the random forest model, and the candidate operating indicator with the largest contribution is selected as the target operating indicator.

[0018] In a preferred embodiment, the abnormal fluctuation index acquisition logic is:

[0019] Determine the interval by hidden danger Filter out all the data marked as outliers in the time series data, and set the outlier set as ; k is the number of outliers, and i is the time position corresponding to the abnormal point, and the standard deviation of the abnormal point time position is calculated. ;

[0020] Assume the distance between outliers is: ; Calculate the distance variance between outliers Evaluate the interval fluctuations between outliers;

[0021] Assume that the rate of change of the fluctuation amplitude of the abnormal point is the rate of change of the amplitude of the abnormal point and the previous normal point i:

[0022] i ; Indicates abnormal points The time position of the previous normal point corresponding to i, and then calculate the standard deviation of the rate of change of the fluctuation amplitude of the abnormal point ;

[0023] Finally, substitute it into the calculation formula of abnormal volatility index:

[0024] ; , , are all preset adjustment coefficients, T is the total length of the time series, Represents the abnormal volatility index.

[0025] In a preferred embodiment, the abnormal trend index acquisition logic is:

[0026] Determine the interval by hidden danger Filter out all the data marked as outliers in the time series data, and set the outlier set as ; k is the number of outliers, and i is the time position corresponding to the outlier point, and cubic polynomial regression is used to capture the changing trend of the outlier point:

[0027] ; is the time position corresponding to the abnormal point The time corresponding to i, , , , All are regression coefficients;

[0028] Select a sliding window w of fixed size in the time series, perform trend analysis on the outliers in each window, perform linear regression on the outliers in each window, and calculate the trend slope in each window : ; and are the mean of the time and the position of the outlier in the window, and then calculate the trend slope in all windows Standard Deviation ;

[0029] Finally, substitute it into the calculation formula of abnormal trend index:

[0030] ; , are preset proportional factors. Represents the abnormal trend index.

[0031] In a preferred embodiment, fuzzy reasoning refers to:

[0032] The abnormal fluctuation index and abnormal trend index of the target power grid equipment are taken as input variables, and the fault level of the target power grid equipment is taken as the output variable. The input variables are fuzzified and the values ​​of the input variables are converted into fuzzy sets. The output variables are fuzzified and the output variables are converted into fuzzy sets. Fuzzy rules are formulated to describe the fault levels under different data type combinations. The fuzzified input variables are inferred through fuzzy rules to obtain the fault level of the target power grid equipment.

[0033] In a preferred embodiment, self-fault diagnosis refers to: calculating the preset statistical characteristics of each item of operation data respectively, and then using threshold detection, if there is operation data exceeding the threshold, marking the operation data as a fault.

[0034] In a preferred embodiment, joint fault diagnosis refers to: extracting preset target features from the power grid equipment operation data set and the power grid equipment environment data set and combining them to form a feature vector, which is then transmitted to a pre-trained neural network model to output the fault type of the power grid equipment.

[0035] Technical effects and advantages of the present invention:

[0036] The present invention can realize automatic fault processing based on fault diagnosis results and preset automatic operation and maintenance strategies. When the system detects abnormalities in power grid equipment, it automatically analyzes the fault and takes corresponding operation and maintenance measures, which significantly improves the efficiency of fault processing, reduces response time, and avoids delays caused by manual intervention. By combining equipment operation data and environmental data, the present invention can perform more comprehensive fault analysis. Joint fault diagnosis can effectively distinguish between internal equipment problems and faults caused by environmental factors, thereby improving the accuracy of fault identification and reducing the occurrence of false alarms and missed alarms.

[0037] The preset automated operation and maintenance strategy ensures that the system takes the best maintenance measures in different fault scenarios. By realizing automated operation and maintenance, the present invention significantly reduces the need for manual inspection and intervention, thereby reducing operation and maintenance costs. At the same time, the system can detect and handle faults in a timely manner, reducing economic losses and power outage risks caused by equipment failures.

[0038] The present invention can monitor power grid equipment in real time and discover potential problems in equipment operation in advance by calculating abnormal fluctuation index and abnormal trend index. When signs of failure appear, the system can issue an early warning to remind operation and maintenance personnel to take measures in advance to avoid serious failures. The present invention combines environmental data (such as temperature, humidity, etc.) with equipment operation data to comprehensively analyze the equipment operation status. When environmental factors affect the equipment, the system can quickly identify and take corresponding measures. This makes operation and maintenance more comprehensive and intelligent, and effectively improves the stability and safety of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0040] Figure 1 This is a schematic diagram of a smart grid online security operation and maintenance management system in the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Example 1

[0043] A smart grid online safety operation and maintenance management system includes an operation data acquisition module, an environmental data acquisition module, a data analysis module, a fault analysis module, a fault diagnosis module, and an automated operation and maintenance module; the data acquisition module, the environmental data acquisition module, the data analysis module, the fault analysis module, the fault diagnosis module, and the automated operation and maintenance module are communicated with each other.

[0044] The operation data acquisition module is used to collect various preset operation data of the target power grid equipment in real time, obtain the power grid equipment operation data set and transmit it to the data analysis module; collection content: Current: record the real-time current value of the power grid equipment. Voltage: record the real-time voltage value of the power grid equipment. Example: Current acquisition: monitor the current flow of the equipment in real time through the current transformer (CT), the data point may be "Current: 250A". Voltage acquisition: monitor the voltage level of the equipment in real time through the voltage sensor, the data point may be "Voltage: 11kV". Temperature: record the working temperature of the power grid equipment. Example: Temperature acquisition: monitor the working temperature of the equipment through the temperature sensor, the data point may be "Temperature: 75℃".

[0045] Power: Record the actual power consumption of grid equipment. Power factor: Record the power factor of equipment to evaluate the efficiency of power use. Example: Power acquisition: Measure the power consumption of equipment through a power meter, and the data point may be "Power: 50kW". Power factor acquisition: Measure the power factor of equipment through a power factor meter, and the data point may be "Power factor: 0.95".

[0046] The environmental data acquisition module is used to collect various preset environmental data in the operating environment of the target power grid equipment in real time, obtain the power grid equipment environmental data set and transmit it to the fault diagnosis module when the multi-dimensional data support decision-making mechanism is triggered; the collection content: record the temperature of the environment around the equipment. Example: Environmental temperature collection: monitor the air temperature around the power grid equipment through the environmental temperature sensor. The data point may be "ambient temperature: 28℃". Record the humidity level of the equipment operating environment. Example: Environmental humidity collection: measure the ambient humidity through the humidity sensor. The data point may be "ambient humidity: 60%". Record the air pressure data around the equipment. Example: Air pressure collection: monitor the ambient air pressure through the air pressure sensor. The data point may be "air pressure: 1013hPa".

[0047] Record the light intensity around the device, especially the device with photosensors and lighting conditions. Example: Ambient light collection: Get the light intensity of the environment through the light sensor. The data point may be "Ambient light: 200lux". Record the wind speed data in the environment where the device is operating. Example: Wind speed collection: Measure the wind speed through the wind speed sensor. The data point may be "Wind speed: 15km / h". Monitor the gas concentration in the environment around the device, such as carbon dioxide (CO2), oxygen (O2) or harmful gases (such as ammonia). Example: CO2 concentration collection: Measure the concentration of carbon dioxide in the environment through the CO2 sensor. The data point may be "CO2 concentration: 400ppm".

[0048] The data analysis module is used to perform data analysis based on the power grid equipment operation data set, and judge whether there is a need for fault analysis according to the data analysis results; the data analysis module is used to perform data analysis based on the power grid equipment operation data set, which means: obtaining the time series data of the target operation index from the power grid equipment operation data set, and then calculating the mean value μ and standard deviation σ of the time series data respectively, setting the hidden danger judgment interval as If the data points in the time series data do not fall into the hidden danger judgment interval When the data point in the time series data falls into the hidden danger judgment interval, the data point is marked as an abnormal point. , the data point is marked as a normal point, and finally the proportion of abnormal points to the total number of all data points is calculated.

[0049] By calculating the mean and standard deviation of the time series data of the target operating indicators, and setting the hidden danger judgment interval , which can effectively identify data points that are outside the normal fluctuation range. This method helps to discover potential abnormal behaviors and provide early warning signals. Abnormal points are usually early signs of potential problems in equipment or systems. By monitoring these abnormal points, measures can be taken as early as possible to carry out maintenance and prevent problems from worsening or causing downtime. By calculating the proportion of abnormal points, the operating status and health of the equipment can be quantitatively evaluated, so as to make more scientific decisions.

[0050] Ideally, many natural phenomena and measurement data (including the operating data of power equipment) tend to be normally distributed, especially when the amount of data is large. According to the central limit theorem, the distribution of data tends to be normally distributed. For a sufficiently large data set, the data of power equipment may be closer to a normal distribution. The larger the amount of data, the more likely the sample distribution is to be close to a normal distribution. When the equipment is operating normally and the environment is stable, some operating data of the equipment (such as temperature, current, pressure, etc.) conform to the normal distribution. However, when the equipment is faulty or operating abnormally, these data may deviate from the normal distribution.

[0051] According to statistical principles, for most normally distributed data, approximately 99.7% of the data points will fall within This method is based on this principle and is used to identify abnormal points that are beyond the normal range. The standard deviation reflects the degree of discreteness of the data and is therefore used to define the boundaries of the normal operating range. Data points outside this range are considered abnormal, and the data analysis results can be used to determine whether there is a need for fault analysis: the proportion of the total number of data points occupied by the abnormal points is compared with the preset normal distribution threshold. If the proportion of the total number of data points occupied by the abnormal points is greater than or equal to the preset normal distribution threshold, it is determined that there is a need for fault analysis; if the proportion of the total number of data points occupied by the abnormal points is less than the preset normal distribution threshold, it is determined that there is no need for fault analysis.

[0052] The confirmation logic of the target operating indicator is: use multiple candidate operating indicators X to train the machine learning model f(X), the machine learning model is a random forest model, and the importance contribution of each feature, that is, the operating indicator, in predicting the probability of failure is calculated through the random forest model, and the candidate operating indicator with the largest contribution is selected as the target operating indicator.

[0053] Steps: Candidate operating indicator set: First, there are multiple candidate operating indicators X, each of which may have a certain impact on the occurrence of power grid equipment failures. Machine learning model training: Use these candidate indicators X to train a machine learning model f(X), and the commonly used model is random forest. The random forest model constructs multiple decision trees to determine the importance of each indicator in predicting the probability of failure.

[0054] Calculation of feature contribution: In the random forest model, each decision tree uses different features (i.e., candidate operating indicators) to make classification decisions. After the model training is completed, the random forest can evaluate the contribution of each feature in making decisions, which is called feature importance. Select the target operating indicator: By comparing the importance contribution of all candidate operating indicators, the operating indicator with the highest contribution is selected as the target operating indicator.

[0055] Feature Importance Calculation in Random Forest: Gini Impurity: Random forests usually measure the importance of a feature by calculating the reduction in Gini impurity of the decision tree node. The split of the node will lead to a reduction in impurity. The more impurity is reduced, the higher its importance. Output feature importance: Ultimately, the model assigns an importance weight to each feature, indicating the contribution of the feature in predicting the occurrence of a fault.

[0056] For example, data of various operating indicators, such as current, voltage, temperature and power factor, are collected from power grid equipment, and whether the equipment fails is recorded. These data are input into the random forest model for training. The goal of the model is to predict the probability of equipment failure based on these operating indicators. The model will learn the correlation between each indicator and the fault. After the training is completed, the model outputs the importance weight of each indicator. For example, the weight of current (I) is 0.45; the weight of voltage (V) is 0.20; the weight of temperature (T) is 0.25; the weight of power factor (PF) is 0.10; in this example, current (I) has the highest weight (0.45), so current can be selected as the target operating indicator. By evaluating the candidate operating indicators through the machine learning model, the most effective indicator for fault prediction can be quantitatively selected. This is more accurate and scientific than the selection method that relies solely on experience or rules. Selecting an indicator with the highest contribution as the target operating indicator can simplify the system monitoring process, concentrate resources on real-time monitoring of the most critical indicators, and improve the accuracy and efficiency of fault prediction. By selecting the most representative target operating indicators, potential problems in equipment operation can be discovered earlier, thus preventing major failures. The feature importance analysis provided by the random forest model makes the system's fault prediction results more interpretable, and operation and maintenance personnel can clearly know which operating indicators have the greatest impact on failures, so as to carry out targeted maintenance.

[0057] The fault analysis module performs fault analysis when there is a need for fault analysis, generates an abnormal fluctuation index and an abnormal trend index, and then performs fuzzy reasoning to obtain the fault level. When the fault level reaches the preset level range, the multi-dimensional data support decision-making mechanism is triggered;

[0058] The logic for obtaining the abnormal volatility index is:

[0059] Determine the interval by hidden danger Filter out all the data marked as outliers in the time series data, and set the outlier set as ; k is the number of outliers, and i is the time position corresponding to the abnormal point, and the standard deviation of the abnormal point time position is calculated. ; Standard deviation Indicates the degree of fluctuation of the distribution of abnormal points on the time axis. If the time positions of the abnormal points differ greatly, it means that the distribution of the abnormal points is uneven. The standard deviation It will be larger, which reflects that the anomalies may be randomly distributed or concentrated in certain specific time periods. It is used to measure whether the anomalies are evenly distributed on the time axis.

[0060] Assume the distance between outliers is: ; Calculate the distance variance between outliers Evaluate the interval fluctuation between anomalies; calculate the time distance between adjacent anomalies. If the distance between adjacent anomalies varies greatly, the variance It will increase, which indicates that the frequency of abnormal points is highly irregular and is used to evaluate the frequency fluctuation of abnormal points in the time series.

[0061] Assume that the rate of change of the fluctuation amplitude of the abnormal point is the rate of change of the amplitude of the abnormal point and the previous normal point i:

[0062] i ; Indicates abnormal points The time position of the previous normal point corresponding to i, and then calculate the standard deviation of the rate of change of the fluctuation amplitude of the abnormal point ; Fluctuation rate of change i measures the rate of change of the abnormal point relative to the previous normal point. If the standard deviation of this rate of change Larger means that the amplitude of the abnormal point changes dramatically, and there may be different degrees of abnormality, reflecting the amplitude fluctuation degree of the abnormal point, which will increase significantly when the amplitude changes dramatically.

[0063] Finally, substitute it into the calculation formula of abnormal volatility index:

[0064] ; , , All are preset adjustment coefficients, according to different analysis requirements. , , It can be adjusted to reflect the influence of time distribution, distance distribution or amplitude change rate on the abnormal fluctuation index. T is the total length of the time series. The total length of the time series T is a normalized parameter used to balance the differences in the lengths of different time series. In the formula, indicators such as standard deviation will be normalized according to T to ensure that different time series are comparable. Indicates the abnormal fluctuation index. The larger the abnormal fluctuation index, the more the abnormal points in the time series are unevenly distributed, or the time interval and amplitude fluctuation between the abnormal points are more volatile. The larger the abnormal fluctuation index, the more the abnormal points appear unevenly distributed on the time axis, and the amplitude fluctuates violently, which may indicate that the device is experiencing frequent and serious abnormal conditions.

[0065] The logic for obtaining the abnormal trend index is:

[0066] Determine the interval by hidden danger Filter out all the data marked as outliers in the time series data, and set the outlier set as ; k is the number of outliers, and i is the time position corresponding to the outlier point, and cubic polynomial regression is used to capture the changing trend of the outlier point:

[0067] ; is the time position corresponding to the abnormal point The time corresponding to i, , , , All are regression coefficients; is the cubic coefficient, capturing the nonlinear change trend of the outliers over time. It is the quadratic term coefficient, reflecting the acceleration or deceleration trend of the abnormal point. is the coefficient of the first-order term, which indicates the linear change of the outlier points over time. is a constant term, which indicates the position of the abnormal point in the initial state of the time series.

[0068] Select a fixed-size sliding window w in the time series to represent a local time range or subset in the time series. The sliding window is used for local trend analysis. Analysis is performed within each window to capture the local trend changes of the abnormal points within the window. Trend analysis is performed on the abnormal points within each window. Linear regression is performed on the abnormal points within each window to calculate the trend slope within each window. : ; and are the means of the time and the position of the outlier in the window, respectively, which are used to calculate the benchmark for regression in the window, so that the trend in each window can be adjusted with the local mean as a reference, and then the trend slope in all windows is calculated. Standard Deviation ; Standard deviation Measures the volatility of the trend change of abnormal points. If the standard deviation It is larger, indicating that the trend changes between different windows are more drastic and the trend fluctuations are larger.

[0069] Finally, substitute it into the calculation formula of abnormal trend index:

[0070] ; , These are preset proportional factors used to adjust the relative importance of trend changes and window trend volatility. Represents the abnormal trend index. The abnormal trend index is a comprehensive indicator to measure the trend change of abnormal points in time series. It combines the cubic polynomial regression trend of the abnormal points over time and the volatility of the trend slope in the sliding window to comprehensively evaluate the trend of the abnormal points. Through multiple trend terms (such as cubic and quadratic terms) and the standard deviation of the trend slope, the abnormal trend index can reflect whether the abnormal points show obvious upward, downward or nonlinear change trends, and whether these trends fluctuate violently. If the abnormal trend index is large, it means that the trend change of the abnormal points is significant. For example, the absolute values ​​of the cubic and quadratic regression coefficients are large, indicating that the abnormal points show obvious nonlinear trend changes in the time series, such as accelerated or decelerated growth.

[0071] Fuzzy reasoning refers to:

[0072] The abnormal fluctuation index and abnormal trend index of the target power grid equipment are taken as input variables, and the fault level of the target power grid equipment is taken as the output variable. The input variables are fuzzified and the values ​​of the input variables are converted into fuzzy sets. The output variables are fuzzified and the output variables are converted into fuzzy sets. Fuzzy rules are formulated to describe the fault levels under different data type combinations. The fuzzified input variables are inferred through fuzzy rules to obtain the fault level of the target power grid equipment.

[0073] In the fault diagnosis of power grid equipment, the abnormal fluctuation index and the abnormal trend index are used as input variables at the same time, and the fault level of the equipment is used as the output variable. First, the input variables are fuzzified, and the values ​​of the abnormal fluctuation index and the abnormal trend index are converted into fuzzy sets. Fuzzy sets usually consist of multiple labels, such as low, medium, and high, which are used to describe different degrees of volatility and trend. Next, the fault level is fuzzified, and the output variable (fault level) is also converted into a fuzzy set. The fault level can be divided into different levels such as slight, medium, and severe according to the operating conditions of the power grid equipment. Then fuzzy rules are formulated to describe how different combinations of abnormal fluctuation indexes and abnormal trend indexes affect the fault level. For example, if both the fluctuation index and the trend index are high, the fault level may be severe; if the fluctuation index is low but the trend index is high, the fault level may be medium. After the fuzzy rules are determined, the fuzzified input variables are inferred according to these rules, and the fault level of the power grid equipment is calculated. Finally, through the defuzzification process, the inference results are converted into specific fault levels so that corresponding maintenance or early warning measures can be implemented. Through this process, fuzzy reasoning can effectively handle the uncertainty and ambiguity in the operation data of power grid equipment and provide more accurate diagnosis and classification of equipment failures.

[0074] The fault diagnosis module performs self-fault diagnosis or joint fault diagnosis on the target power grid equipment according to the fault level obtained by the fault analysis module and whether the multi-dimensional data support decision-making mechanism is triggered, and transmits the fault diagnosis results to the automated operation and maintenance module; self-fault diagnosis means: calculating the preset statistical characteristics of each operation data separately, and then using the threshold detection. If there is an operation data that exceeds the threshold, the operation data is marked as a fault. Self-fault diagnosis is based on a single operation data set of the power grid equipment to determine whether there is a fault. Its core lies in monitoring the real-time operation data of the equipment, such as key indicators such as current, voltage, and temperature. For each operation data, preset statistical characteristics (such as mean value, standard deviation, threshold) to define its normal operating range. If a data exceeds the threshold, it is marked as a fault.

[0075] Self-fault diagnosis is a relatively simple diagnostic method. It only requires threshold detection based on a single data source. Therefore, it only marks whether a certain operating data exceeds the preset range, and can quickly and directly identify potential faults without complex analysis. This diagnostic method only focuses on the operation of the equipment itself and does not involve the influence of the external environment. Therefore, its output usually only marks that a specific operating data is out of the normal range, indicating that a certain operating parameter of the equipment is abnormal, but it cannot deeply identify the specific type or cause of the fault.

[0076] When the fault severity, i.e., the fault level, does not reach the preset level range, it is generally considered that the operation of the equipment is within the controllable range. In this case, the fault may be a temporary fluctuation or a minor equipment abnormality. Therefore, giving priority to using operating data to judge the fault type can improve the diagnostic efficiency and avoid unnecessary complex analysis. In low-severity fault states, environmental factors have relatively little impact on equipment operation. For example, a slight voltage fluctuation or a short-term equipment overload may have little to do with environmental conditions, but is more caused by the operating conditions of the equipment itself. At this time, a reasonable judgment can be made by relying only on equipment data. Environmental data is usually complex and contains a lot of redundant information. In the case of low fault severity, introducing environmental data may add unnecessary complexity and affect the efficiency of diagnosis. Judging only by operating data can simplify the process and save computing resources and time.

[0077] When the severity of the fault reaches the preset level range, the equipment may be in a high-risk state. At this time, a more comprehensive and accurate analysis is needed. Multi-dimensional data including environmental factors can help better diagnose and determine the type of fault. When the severity of the fault, that is, the fault level, does not reach the preset level range, the fault may not be serious and the equipment is in a low-risk state. At this time, simplified analysis through operating data can not only meet the needs, but also improve diagnostic efficiency and reduce the occupation of system resources.

[0078] Joint fault diagnosis refers to extracting preset target features from the grid equipment operation data set and the grid equipment environment data set and combining them to form a feature vector, which is then sent to the pre-trained neural network model to output the fault type of the grid equipment. Joint fault diagnosis not only uses the equipment's operation data, but also introduces environmental data, and uses these data combinations to perform more complex analysis. The operating status of the equipment is not only affected by internal conditions, but may also be affected by external environmental factors (such as temperature, humidity, air pressure, etc.). Therefore, joint fault diagnosis can accurately determine the root cause and type of the fault through more comprehensive data analysis.

[0079] Joint fault diagnosis not only analyzes the equipment's operating data, but also considers environmental factors. This multidimensional data can provide a panoramic view of the equipment's operation, making it possible to more accurately diagnose the specific type of equipment failure. This diagnostic method can distinguish between failures caused by internal equipment problems (such as component aging) and failures caused by environmental factors (such as excessive temperature and humidity). Joint fault diagnosis uses a neural network model for analysis. Neural networks can learn complex patterns and relationships from a large amount of training data, and output detailed fault types by inputting equipment operating data and environmental data. Because neural networks have nonlinear modeling capabilities, they can distinguish failures caused by various different reasons and can also take the impact of environmental factors into account. In joint fault diagnosis, the judgment of fault type is not only based on internal problems of the equipment, but also includes failures caused by the environment. Because environmental data is taken into account, the final diagnosis result can identify whether the equipment abnormality is caused by environmental factors, so the fault type will include environmental-related problems. For example, power grid equipment may fail due to excessive ambient temperature or excessive humidity, and neural networks can be trained to identify such environmental failures.

[0080] The grid equipment operation dataset includes real-time operation indicators of the equipment, such as current, voltage, power, temperature, etc. The grid equipment environment dataset includes environmental factors, such as ambient temperature, ambient humidity, wind speed, etc. The data is scaled to the same range to improve the stability and efficiency of model training. The dataset is then divided into training set, validation set, and test set.

[0081] Feature extraction: Extract useful features from raw data. For example, calculate statistical features (mean, standard deviation, etc.) through sliding windows. Feature fusion: Combine operational data features and environmental data features to form a comprehensive feature set.

[0082] Model selection: Neural network architecture: Choose an appropriate neural network model, such as a fully connected neural network, a convolutional neural network, or a recurrent neural network, depending on the characteristics of the data and the requirements of the task.

[0083] Model structure: Input layer: receives all feature data, including operation data and environmental data. Hidden layer: performs feature extraction and nonlinear mapping through multiple hidden layers. Activation functions (such as ReLU, Sigmoid) can be used to increase the nonlinear ability of the model. Output layer: usually a classification layer, used to output the probability distribution of fault types. For example, use the Softmax activation function to output multi-classification problems. Use the trained neural network model to predict new data. Output the probability distribution of fault types and select the category with the highest probability as the final diagnosis result. Deploy the trained model to the real-time monitoring system to analyze the operation and environmental data of power grid equipment in real time for fault detection and type judgment. According to the prediction results of the model, corresponding maintenance and repair operations can be performed later.

[0084] The automated operation and maintenance module performs operation and maintenance management based on the fault diagnosis results and the preset automated operation and maintenance strategies. An example is given below: Scenario: Automated operation and maintenance of transformer faults. Background: The transformer equipment in the power grid detects an abnormal situation. The system identifies the fault type of the transformer through the joint fault diagnosis module and passes the information to the automated operation and maintenance module. Fault diagnosis result transmission: The joint fault diagnosis module determines that the transformer has an overheating fault through operating data and environmental data. The diagnosis result is: the transformer has an overheating fault and the transformer temperature exceeds the safety threshold.

[0085] Automated operation and maintenance strategy triggering: The automated operation and maintenance strategy preset in the system stipulates that when the temperature of the transformer exceeds a certain safety threshold, the cooling equipment (such as the cooling fan) is automatically turned on and the load is reduced. The strategy performs different operations according to different fault types. For overheating faults, the system automatically performs cooling. The automated operation and maintenance module receives the fault diagnosis results and identifies that the transformer needs to be cooled. The system automatically performs the following operations: Turn on the cooling equipment: The automated control system sends a start command to the cooling equipment, and the cooling fan or cooling system automatically turns on to reduce the temperature of the transformer. Load adjustment: According to the preset automated operation and maintenance strategy, the load of the transformer is automatically reduced to prevent further equipment damage caused by overheating. Operation and maintenance status monitoring: The automated operation and maintenance module continuously monitors the temperature of the transformer. According to the automated strategy, when the temperature drops to a safe range, the system automatically turns off the cooling equipment and restores the load to normal levels.

[0086] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0087] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0088] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0089] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0090] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A smart grid online security operation and maintenance management system, characterized in that: It includes operation data collection module, environment data collection module, data analysis module, fault analysis module, fault diagnosis module and automated operation and maintenance module; The operation data acquisition module is used to collect various preset operation data of the target power grid equipment in real time, obtain the power grid equipment operation data set and transmit it to the data analysis module; The environmental data acquisition module is used to collect various preset environmental data in the target power grid equipment operating environment in real time, obtain the power grid equipment environmental data set and transmit it to the fault diagnosis module when the multi-dimensional data support decision-making mechanism is triggered; The data analysis module is used to perform data analysis operations based on the power grid equipment operation data set, and determine whether there is a need for fault analysis based on the data analysis results; The fault analysis module performs fault analysis when there is a need for fault analysis, generates an abnormal fluctuation index and an abnormal trend index, and then performs fuzzy reasoning to obtain the fault level. When the fault level reaches the preset level range, the multi-dimensional data support decision-making mechanism is triggered; The fault diagnosis module performs self-fault diagnosis or joint fault diagnosis on the target power grid equipment according to the fault level obtained by the fault analysis module and whether the multi-dimensional data support decision-making mechanism is triggered, and transmits the fault diagnosis results to the automated operation and maintenance module; The automated operation and maintenance module performs operation and maintenance management based on fault diagnosis results and preset automated operation and maintenance strategies; The logic for obtaining the abnormal volatility index is: Determine the interval by hidden danger Filter out all the data marked as outliers in the time series data, and set the outlier set as ; k is the number of outliers, and i is the time position corresponding to the abnormal point, and the standard deviation of the abnormal point time position is calculated. ; Assume the distance between outliers is: ; Calculate the distance variance between outliers Evaluate the interval fluctuations between outliers; Assume that the rate of change of the fluctuation amplitude of the abnormal point is the rate of change of the amplitude of the abnormal point and the previous normal point i: i ; Indicates abnormal points The time position of the previous normal point corresponding to i, and then calculate the standard deviation of the rate of change of the fluctuation amplitude of the abnormal point ; Finally, substitute it into the calculation formula of abnormal volatility index: ; , , are all preset adjustment coefficients, T is the total length of the time series, represents the abnormal volatility index; The logic for obtaining the abnormal trend index is: Determine the interval by hidden danger Filter out all the data marked as outliers in the time series data, and set the outlier set as ; k is the number of outliers, and i is the time position corresponding to the outlier point, and cubic polynomial regression is used to capture the changing trend of the outlier point: ; is the time position corresponding to the abnormal point The time corresponding to i, , , , All are regression coefficients; Select a sliding window w of fixed size in the time series, perform trend analysis on the outliers in each window, perform linear regression on the outliers in each window, and calculate the trend slope in each window : ; and are the mean of the time and the position of the outlier in the window, and then calculate the trend slope in all windows Standard Deviation ; Finally, substitute it into the calculation formula of abnormal trend index: ; , are preset proportional factors. Represents the abnormal trend index.

2. A smart grid online security operation and maintenance management system according to claim 1, characterized in that: The data analysis module is used to perform data analysis operations based on the power grid equipment operation data set, which refers to: The time series data of the target operation index is obtained from the power grid equipment operation data set, and then the mean value μ and standard deviation σ of the time series data are calculated respectively, and the hidden danger judgment interval is set as If the data points in the time series data do not fall into the hidden danger judgment interval When the data point in the time series data falls into the hidden danger judgment interval, the data point is marked as an abnormal point. , the data point is marked as a normal point, and finally the proportion of abnormal points to the total number of all data points is calculated.

3. A smart grid online security operation and maintenance management system according to claim 2, characterized in that: Judging whether there is a need for fault analysis based on data analysis results means: The proportion of abnormal points to the total number of all data points is compared with the preset normal distribution threshold. If the proportion of abnormal points to the total number of all data points is greater than or equal to the preset normal distribution threshold, it is determined that there is a need for fault analysis. If the proportion of abnormal points to the total number of all data points is less than the preset normal distribution threshold, it is determined that there is no need for fault analysis.

4. A smart grid online security operation and maintenance management system according to claim 3, characterized in that: The confirmation logic of the target operation indicator is: Use multiple candidate operating indicators X to train the machine learning model f(X). The machine learning model is a random forest model. The importance contribution of each feature, that is, the operating indicator, in predicting the probability of fault occurrence is calculated through the random forest model, and the candidate operating indicator with the largest contribution is selected as the target operating indicator.

5. A smart grid online security operation and maintenance management system according to claim 4, characterized in that: Fuzzy reasoning refers to: The abnormal fluctuation index and abnormal trend index of the target power grid equipment are taken as input variables, and the fault level of the target power grid equipment is taken as the output variable. The input variables are fuzzified and the values ​​of the input variables are converted into fuzzy sets. The output variables are fuzzified and the output variables are converted into fuzzy sets. Fuzzy rules are formulated to describe the fault levels under different data type combinations. The fuzzified input variables are inferred through fuzzy rules to obtain the fault level of the target power grid equipment.

6. A smart grid online security operation and maintenance management system according to claim 5, characterized in that: Self-fault diagnosis means: calculating the preset statistical characteristics of each operating data item separately, and then using threshold detection. If there is operating data that exceeds the threshold, the operating data is marked as a fault.

7. A smart grid online security operation and maintenance management system according to claim 6, characterized in that: Joint fault diagnosis refers to extracting preset target features from the power grid equipment operation data set and the power grid equipment environment data set and combining them to form a feature vector, which is then transmitted to a pre-trained neural network model to output the fault type of the power grid equipment.

Images