Power failure detection method and system for low-voltage transformer area in distribution network

Abstract

The invention discloses a power failure detection method and system for a low-voltage transformer area in a distribution network, and relates to the technical field of power system monitoring, and the method comprises the steps: collecting original data in real time, and carrying out the preprocessing; constructing a power failure risk state calculation model to calculate and classify power failure risk scores, and constructing a wind speed prediction model according to a classification result to calculate predicted wind speed data at a future time point; and generating a predicted power failure risk score of a future time point through the power failure risk state calculation model, and carrying out classified early warning. Through a wind speed prediction model and dynamic threshold adjustment, sudden risks caused by high wind speed can be accurately responded, the equipment outage probability caused by wind speed over-limit is effectively reduced, corresponding emergency response strategies are provided for different risk levels based on a grading early warning mechanism of future risk scoring, and the risk assessment efficiency is improved. Accuracy and response speed of power failure detection are optimized, and stable operation of a low-voltage transformer area in a distribution network in a complex environment is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of power system monitoring, and in particular to a method and system for detecting power outages in low-voltage distribution network areas. Background Art

[0002] With the rapid development of the power industry and the continuous advancement of smart grid technology, the management of medium and low voltage substations in the power distribution network system has gradually realized informatization and automation. However, power outage detection in medium and low voltage substations of the distribution network has always been a more difficult problem in the operation of the power grid, especially in the case of load fluctuations and environmental changes. The existing detection means and methods are often difficult to accurately judge the risk of power outages in real time. Traditional power outage detection mainly relies on manual inspections and single data monitoring. The detection response is relatively lagging, and it is difficult to accurately judge the complex multi-factor power outage causes. In recent years, the automation and intelligence level of distribution network substations has been improved. Some power outage detection schemes have introduced monitoring methods for multiple factors such as load and voltage, and even local wind speed monitoring to improve the accuracy of power outage judgment. However, most of these methods lack effective data integration and risk prediction methods, and it is difficult to provide early warning and classified response to future power outage risks. In addition, due to the differences in geographical conditions of the substations and complex environmental factors, such as dynamic changes in wind speed, temperature and humidity, the difficulty and response time of power outage detection are further increased.

[0003] In the prior art, power outage detection schemes are usually based on the setting of static thresholds. Once the grid operating parameters exceed the threshold, a power outage warning is issued, or manual troubleshooting is performed after an abnormality occurs in the equipment. However, these traditional methods have the following shortcomings: the accuracy of power outage detection is not high, and there is a lack of comprehensive analysis methods for the linkage of multiple factors, resulting in a high false alarm rate in the detection results. Most detection methods lack the prediction of power outage risks at future time points and can only passively respond to the current status, making it difficult to provide forward-looking decision support for grid managers. When responding to emergencies in extreme weather (such as strong winds), there is a lack of a dynamic response mechanism to environmental conditions, which often leads to detection delays. Current technology urgently needs a power outage detection method with real-time monitoring, dynamic evaluation and forward-looking prediction to improve the accuracy and response efficiency of power outage detection in low-voltage substations in distribution networks. Summary of the invention

[0004] In view of the problems existing in the above-mentioned existing power outage detection method and system for medium and low voltage distribution areas, the present invention provides a power outage detection method and system for medium and low voltage distribution areas. The problems of the existing power outage detection technology for medium and low voltage distribution areas, such as incomplete real-time monitoring, delayed risk prediction, and difficulty in coping with emergencies in extreme weather, are solved.

[0006] The technical solution provided by the present invention includes: a method for detecting power outages in low-voltage distribution network areas, which is performed according to the following steps:

[0007] S1 collects raw data in real time and performs preprocessing;

[0008] S2 builds a power outage risk status calculation model to calculate the power outage risk score and classify it, and builds a wind speed prediction model based on the classification results to calculate the predicted wind speed data at future time points;

[0009] S3 generates a predicted power outage risk score at a future time point through a power outage risk status calculation model, and performs classified warnings;

[0010] S4 stores all data and backs it up regularly.

[0011] As a preferred solution of the power outage detection method for the low-voltage distribution network area of ​​the present invention, the real-time acquisition of raw data includes:

[0012] Install sensors in the equipment in the medium and low voltage sections of the distribution network, set the collection frequency of each sensor, and collect raw data in real time;

[0013] The raw data includes wind speed, grid load, equipment health status and equipment power outage status at different time points.

[0014] As a preferred solution of the power outage detection method for the low-voltage distribution network area of ​​the present invention, the pre-processing includes:

[0015] A statistical anomaly detection algorithm is used to identify and remove noise and erroneous raw data, and a linear interpolation method is used to complete the missing raw data, and the completed raw data is standardized.

[0016] As a preferred solution of the power outage detection method for the low-voltage distribution network area of ​​the present invention, wherein: the construction of the power outage risk state calculation model to calculate the power outage risk score and classify it includes:

[0017] Construct an initial calculation model for power outage risk status, combining wind speed, grid load and equipment health status.

[0018] Calculate the initial score S(t) of power outage risk;

[0019] The Sigmoid function is introduced to map the wind speed into a nonlinear function, and the nonlinear mapping value σ(V(t)) of the wind speed at time point t is calculated. The formula is:

[0020]

[0021] Wherein, V(t) represents the wind speed at time point t, θ represents the curve adjustment coefficient, and ν represents the critical wind speed threshold;

[0022] The power correction and cosine function are introduced to simulate the fluctuation under high load, and the nonlinear correction value y(P(t)) of the load at time point t is calculated. The formula is:

[0023] y(P(t))=P(t) δ ·cos(ψ·P(t))

[0024] Wherein, P(t) represents the grid load at time point t, δ represents the power correction parameter of the load, and ψ represents the frequency regulation coefficient;

[0025] The exponential decay function is introduced to simulate the exponential decay effect of the equipment health status, and the exponential decay value g(T(t)) of the equipment health status at time point t is calculated. The formula is:

[0026] g(T(t))=e -λ(T(t)-μ)2

[0027] Wherein, T(t) represents the health status of the device at time point t, λ represents the adjustment coefficient of the health status of the device, and μ represents the health reference threshold;

[0028] Construct the final power outage risk state calculation model and calculate the final power outage risk score S`(t). The formula is:

[0029] S`(t)=σ(V(t))+y(P(t))+g(T(t))

[0030] Among them, S`(t) represents the final score of power outage risk at time point t;

[0031] The historical raw data is divided into a training set and a validation set, the CatBoost algorithm is used to train the final power outage risk state calculation model, and the parameters of CatBoost are set;

[0032] The training set includes input features and target labels, with wind speed, grid load and equipment health status as input features and equipment power outage status as target labels;

[0033] Use the validation set to test the prediction effect of the final power outage risk status calculation model, and adjust the parameters of CatBoost according to the validation results;

[0034] Input the real-time raw data into the final power outage risk status calculation model to obtain the real-time final power outage risk score;

[0035] According to the distribution of power outage data in the historical raw data, a real-time scoring threshold is set, and the final power outage risk score and the real-time scoring threshold are compared for classification processing;

[0036] If the final score of the power outage risk is greater than or equal to the real-time score threshold, the wind speed at the corresponding time point is taken as the potential risk data and extracted;

[0037] If the final score of the power outage risk is less than the real-time score threshold, the result is stored and data collection and calculation continue without any operation.

[0038] As a preferred solution of the power outage detection method for the low-voltage distribution network area of ​​the present invention, wherein: the wind speed prediction model is constructed according to the classification result to calculate the predicted wind speed data at the future time point, including:

[0039] Extract historical wind speed data of medium and low pressure areas from the central database;

[0040] Use automated scripts to detect and clean up abnormal data points, and use a moving average method to smooth historical wind speed data;

[0041] The generalized additive model is selected for wind speed prediction. The cubic spline function is used to construct smooth functions in the three time dimensions of hours, days and months. By combining the smooth functions, a wind speed prediction model is constructed to calculate the initial wind speed prediction value V at time point t. j (t);

[0042] Use historical wind speed data to train the wind speed prediction model and obtain a complete smooth function;

[0043] Adjust the smoothing function of the wind speed prediction model through cross-validation;

[0044] The extracted potential risk data are integrated into a potential risk data set, and the potential risk data set is used for local fitting to calculate the predicted wind speed data at future time points;

[0045] According to the design requirements of the power distribution equipment, set the wind speed threshold to determine the wind speed conditions at future time points;

[0046] If the wind speed at the future time point is greater than or equal to the wind speed threshold, the wind speed critical threshold in the wind speed nonlinear function is dynamically adjusted, otherwise no operation is performed;

[0047] The adjustment formula for the wind speed critical threshold in the dynamic adjustment wind speed nonlinear function is:

[0048] ν`=ν-k·(V j (t+Δt)-V r )

[0049] Among them, ν` represents the critical wind speed threshold after dynamic adjustment, k represents the wind speed adjustment coefficient, V j (t+Δt) represents the predicted wind speed at the future time point t+Δt, V rIndicates the wind speed threshold.

[0050] As a preferred solution of the power outage detection method for medium and low voltage areas in the distribution network of the present invention, wherein: the generating a predicted power outage risk score at a future time point through the power outage risk state calculation model and performing classification early warning includes,

[0051] Substituting the predicted wind speed data at the future time point into the final power outage risk state calculation model to generate the final predicted power outage risk score S`(t + Δt) at the future time point;

[0052] Setting a low early warning threshold A1 and a high early warning threshold A2, and performing classification early warning for the final predicted power outage risk score at the future time point;

[0053] If S`(t + Δt) < A1, then generate a first-level early warning to remind the monitoring personnel to strengthen monitoring and regularly check the wind speed change;

[0054] If A1 ≤ S`(t + Δt) < A2, then generate a second-level early warning to notify the operation and maintenance team to prepare emergency measures;

[0055] If A2 ≤ S`(t + Δt), then generate a third-level early warning to trigger a comprehensive emergency response process, dispatch operation and maintenance personnel to high-risk areas, and take preventive measures.

[0056] As a preferred solution of the power outage detection method for medium and low voltage areas in the distribution network of the present invention, wherein: the storing all data and regularly backing up includes,

[0057] Using a structured storage form to store all data generated during the generation process, establishing indexes for the content and related data generated in each round, establishing a data backup mechanism, regularly backing up the stored data, setting user permissions, and allowing managers to call the stored data.

[0058] The present invention also provides a power outage detection system for medium and low voltage areas in the distribution network based on the above method, which includes,

[0059] A data acquisition module for real-time collecting raw data and performing preprocessing;

[0060] A model construction module for constructing a power outage risk state calculation model to calculate the power outage risk score and perform classification;

[0061] A wind speed prediction module for constructing a wind speed prediction model according to the classification result to calculate the predicted wind speed data at a future time point;

[0062] A classification early warning module for generating a predicted power outage risk score at a future time point through the power outage risk state calculation model and performing classification early warning;

[0063] The data storage module is used to store all data and back up them regularly.

[0064] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of a power outage detection method for a medium and low voltage area in a distribution network when executing the computer program.

[0065] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for detecting power outages in low-voltage distribution network areas.

[0066] The beneficial effects of the present invention are as follows: through the wind speed prediction model and dynamic threshold adjustment, it can accurately respond to the sudden risks brought by high wind speed, effectively reduce the probability of equipment shutdown due to excessive wind speed, and provide corresponding emergency response strategies for different risk levels based on the graded early warning mechanism of future risk scores, optimize the accuracy and response speed of power outage detection, and ensure the stable operation of medium and low voltage substations in the distribution network in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0068] Figure 1 The figure is a flow chart of a method for detecting power outages in medium and low voltage distribution network areas.

[0069] Figure 2 The figure is a flow chart of the hierarchical early warning mechanism based on future risk scores.

[0070] Figure 3 This is a structural diagram of the power outage detection system in the medium and low voltage areas of the distribution network.

[0071] Figure 4 This is a structural diagram of the specific components of the modules of the power outage detection system in the medium and low voltage areas of the distribution network. DETAILED DESCRIPTION

[0072] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0073] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0074] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0075] Example 1

[0076] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and which provides a method for detecting power outages in a medium and low voltage area of ​​a distribution network. The method for detecting power outages in a medium and low voltage area of ​​a distribution network includes:

[0077] S1, real-time collection of raw data and preprocessing;

[0078] Specifically, real-time collection of raw data includes:

[0079] Install sensors in the equipment in the medium and low voltage sections of the distribution network, set the collection frequency of each sensor, and collect raw data in real time;

[0080] The raw data includes wind speed, grid load, equipment health status and equipment power outage status at different time points;

[0081] The device health status refers to the operating status of the device, and the value range is [0,1], where 0 indicates a fault state and 1 indicates normal operation;

[0082] The power-off status of the device includes 0 and 1, and the value range is [0,1], 0 indicates a normal state, and 1 indicates a power-off state.

[0083] By installing sensors on equipment in the low-voltage substations of the distribution network, key data such as wind speed, grid load, equipment health status and power outage status are collected in real time, achieving accurate monitoring and rapid response to the equipment operating status. The numerical representation of equipment health status and power outage status makes the identification of faults or power outages more intuitive and accurate, ensuring the timeliness of detection and the integrity of data, and providing a reliable data basis for subsequent power outage risk assessment and early warning.

[0084] Furthermore, preprocessing includes,

[0085] A statistical anomaly detection algorithm is used to identify and remove noise and erroneous raw data, and a linear interpolation method is used to complete the missing raw data, and the completed raw data is standardized.

[0086] This step effectively identifies and eliminates noise and erroneous data through a statistical anomaly detection algorithm to ensure data reliability, fills in missing data through linear interpolation to ensure data integrity, standardizes data, eliminates dimensional differences, and improves the accuracy and consistency of subsequent calculation models. Overall, this preprocessing process improves data quality and provides a more stable and accurate data basis for power outage risk assessment.

[0087] S2. Construct a power outage risk state calculation model to calculate the power outage risk score and classify it, and construct a wind speed prediction model based on the classification results to calculate the predicted wind speed data at future time points;

[0088] Specifically, constructing a power outage risk state calculation model to calculate the power outage risk score and classify it includes:

[0089] The initial calculation model of power outage risk state (SoR) is constructed, and the initial power outage risk score S(t) is calculated by combining wind speed, grid load and equipment health status. The formula is:

[0090] S(t)=α·V(t)+ζ·P(t)+γ·T(t)

[0091] Among them, α, ζ, and γ represent the weight coefficients of wind speed, grid load, and equipment health status, respectively. The weight coefficients are set according to actual conditions. V(t) represents the wind speed at time point t, P(t) represents the grid load at time point t, and T(t) represents the equipment health status at time point t.

[0092] The SoR refers to the SoR model in the power system that comprehensively evaluates the outage risk of equipment or systems by integrating multiple risk factors. The higher the SoR score, the higher the risk of the equipment, that is, the greater the possibility of outage or failure.

[0093] In extreme weather conditions, the effect of wind speed on the risk of power outage is nonlinear;

[0094] The Sigmoid function is introduced to map the wind speed into a nonlinear function, and the nonlinear mapping value σ(V(t)) of the wind speed at time point t is calculated to rapidly increase its risk impact when the wind speed exceeds a certain critical value. The formula is:

[0095]

[0096] Where V(t) represents the wind speed at time point t, θ represents the curve adjustment coefficient, which is used to adjust the steepness coefficient of the Sigmoid curve. The higher the value, the steeper the curve and the more sensitive the wind speed is to changes. ν represents the critical wind speed threshold, which is set so that the risk impact is significantly increased when the wind speed is higher than this value.

[0097] Compared with the linear weighting in the existing technology, the Sigmoid function better simulates the nonlinear characteristics of the impact of wind speed on risk. This can rapidly increase the SoR score when the wind speed approaches or exceeds the critical value, thereby more sensitively reflecting the risk of extreme weather.

[0098] The impact of load on risk lies not only in the load value itself, but also in the volatility of load when it approaches equipment saturation;

[0099] The power correction and cosine function are introduced to simulate the fluctuation under high load, and the nonlinear correction value y(P(t)) of the load at time point t is calculated. The formula is:

[0100] y(P(t))=P(t) δ ·cos(ψ·P(t))

[0101] Among them, P(t) represents the grid load at time point t, δ represents the power correction parameter of the load, which controls the nonlinear enhancement of the load on SoR in the high value area, and ψ represents the frequency adjustment coefficient, which controls the fluctuation frequency of the cosine wave to capture the periodic fluctuation effect of the load on the outage risk;

[0102] Through power correction, the impact of load on risk can be enhanced as the load increases, and the load fluctuation can be captured through the cosine term, which is suitable for the fluctuation effect on outage risk when the load is close to the limit;

[0103] The health status of the equipment has a large exponential decay effect on the downtime risk. When the health status drops to a certain value, the contribution to the risk increases significantly;

[0104] The exponential decay function is introduced to simulate the exponential decay effect of the equipment health status, and the exponential decay value g(T(t)) of the equipment health status at time point t is calculated. The formula is:

[0105] g(T(t))=e -λ(T(t)-μ)2

[0106] Where T(t) represents the health status of the device at time point t, λ represents the adjustment coefficient of the health status of the device, which is used to adjust the sensitivity of the health status of the device to the SoR score, so that the closer the health status value of the device is to the threshold, the faster the score increases, and μ represents the health reference threshold. When the health status of the device is lower than this threshold, the power outage risk of the device will increase significantly. The health reference threshold is set according to the actual situation.

[0107] Compared with the linear model, the exponential decay function can more effectively express the nonlinear impact of equipment health deterioration on risk, thereby significantly improving the SoR score when the equipment health is close to failure;

[0108] Construct the final power outage risk state calculation model and calculate the final power outage risk score S`(t). The formula is:

[0109] S`(t)=σ(V(t))+y(P(t))+g(T(t))

[0110] Among them, S`(t) represents the final score of power outage risk at time point t;

[0111] According to the historical original data, all coefficients in the formula are set to initial values;

[0112] The historical raw data is divided into a training set and a validation set, the CatBoost algorithm is used to train the final power outage risk state calculation model, and the parameters of CatBoost are set;

[0113] The parameters of CatBoost include learning rate, tree depth and number of iterations;

[0114] The training set includes input features and target labels, with wind speed, grid load and equipment health status as input features and equipment power outage status as target labels;

[0115] The target label refers to the specific feature patterns and conditions that lead to power outages that the CatBoost algorithm learns through the correspondence between input features and target labels during the training process. The target label serves as the "learning goal" of the model, allowing the model to identify conditions under which the risk of equipment power outage is higher and conditions under which the risk is lower;

[0116] Use the validation set to test the prediction effect of the final power outage risk status calculation model, and adjust the parameters of CatBoost according to the validation results;

[0117] Input the real-time raw data into the final power outage risk status calculation model to obtain the real-time final power outage risk score;

[0118] According to the distribution of power outage data in the historical raw data, a real-time scoring threshold is set, and the final power outage risk score and the real-time scoring threshold are compared for classification processing;

[0119] If the final score of the power outage risk is greater than or equal to the real-time score threshold, the wind speed at the corresponding time point is taken as the potential risk data and extracted;

[0120] If the final score of the power outage risk is less than the real-time score threshold, the result is stored and data collection and calculation continue without any operation.

[0121] The present invention constructs a power outage risk status calculation model, combines multi-factor data of wind speed, power grid load and equipment health status, and introduces nonlinear functions such as Sigmoid, power correction, cosine and exponential decay to achieve sensitive identification of extreme weather and equipment aging, thereby improving the accuracy of power outage risk scoring. The model is trained by the CatBoost algorithm so that the model can effectively identify power outage risk characteristics and update the score in real time, and classify risk conditions according to real-time thresholds, thereby accurately triggering the early warning mechanism and optimizing the emergency response. This method significantly improves the risk perception ability and response speed of low-voltage substations in the distribution network under complex conditions.

[0122] Furthermore, the wind speed prediction model is constructed based on the classification results to calculate the predicted wind speed data at future time points, including:

[0123] Extract historical wind speed data of medium and low pressure areas from the central database. The data must cover at least 1 year and include different meteorological conditions and seasonal characteristics.

[0124] Use automated scripts to detect and clean up abnormal data points (such as abnormally high or low values ​​during equipment maintenance), and use the moving average method to smooth historical wind speed data to eliminate the impact of sudden anomalies and ensure data point continuity;

[0125] The generalized additive model is selected for wind speed prediction. The cubic spline function is used to construct smooth functions in the three time dimensions of hours, days and months to capture the nonlinear changes of wind speed. By combining the smooth functions, a wind speed prediction model is constructed to calculate the initial wind speed prediction value V at time point t. j (t), the formula is:

[0126] V j (t)=β0+f1(hour(t))+f2(day(t))+f3(month(t))+ε

[0127] Among them, β0 represents the basic intercept of wind speed, f1(hour(t)), f2(day(t)), f3(month(t)) represent the smoothing functions of hour, day and month respectively, and ε represents the random error term;

[0128] Use historical wind speed data to train the wind speed prediction model and obtain a complete smooth function;

[0129] Adjust the smoothing function of the wind speed prediction model through cross-validation to ensure that the smoothing function has high accuracy in capturing the long-term trend and seasonal changes of historical wind speed data;

[0130] The generated smoothing function is based on historical wind speed data, which can reflect the periodic characteristics of wind speed in the distribution area in different time periods and is the basis for subsequent local adjustments;

[0131] The extracted potential risk data are integrated into a potential risk data set, and the potential risk data set is used for local fitting to calculate the predicted wind speed data at future time points;

[0132] The local fitting refers to re-optimizing the basic smoothing function of the wind speed prediction model so that the model can better fit the data of the high-risk period, and further enhance the prediction accuracy of the model in a specific key time period by retaining the model's ability to fit the overall trend. The local fitting uses the potential risk data set to fine-tune the smoothing function of the time dimensions such as hours, days, and months, while keeping the structure of the basic smoothing function unchanged;

[0133] According to the design requirements of the power distribution equipment, set the wind speed threshold to determine the wind speed conditions at future time points;

[0134] If the wind speed at a future time point is greater than or equal to the wind speed threshold, the wind speed critical threshold in the wind speed nonlinear function is dynamically adjusted to issue a risk warning in advance, otherwise no operation is performed;

[0135] The adjustment formula for the wind speed critical threshold in the dynamic adjustment wind speed nonlinear function is:

[0136] ν`=ν-k·(V j (t+Δt)-V r )

[0137] Among them, ν` represents the critical wind speed threshold after dynamic adjustment, k represents the wind speed adjustment coefficient, which indicates the adjustment strength of the risk threshold when the predicted wind speed exceeds the threshold, V j (t+Δt) represents the predicted wind speed at the future time point t+Δt, V r Indicates wind speed threshold;

[0138] When the predicted wind speed exceeds the wind speed threshold at any time in the future, the trigger threshold of SoR is automatically lowered to issue an early warning so that operation and maintenance personnel can arrange emergency measures in advance;

[0139] By dynamically adjusting the critical wind speed threshold, it is more sensitive in high wind weather and can effectively avoid the risk of equipment shutdown or damage caused by excessive wind speed.

[0140] Through the wind speed prediction model constructed based on historical data, combined with smoothing processing and local fitting in multiple time dimensions, the present invention can accurately predict the wind speed changes at future time points, and respond promptly to high wind speed risks. When the wind speed approaches or exceeds the equipment design threshold, the system automatically adjusts the wind speed critical threshold in the power outage risk status model, dynamically reduces the SoR trigger threshold, and thus issues an early warning in advance. This mechanism enhances the sensitivity and accuracy of the system under high wind speed conditions, effectively reduces the risk of equipment outage or damage caused by wind speed fluctuations, and provides a guarantee for early emergency preparation in operation and maintenance.

[0141] S3. Generate the predicted power outage risk score at a future time point through the power outage risk status calculation model, and conduct classified early warning;

[0142] Specifically, generating the predicted power outage risk score at a future time point through the power outage risk status calculation model and conducting classified early warning includes:

[0143] Substitute the predicted wind speed data at the future time point into the final power outage risk status calculation model to generate the final predicted power outage risk score S`(t + Δt) at the future time point;

[0144] Set a low early warning threshold A1 and a high early warning threshold A2, and conduct classified early warning for the final predicted power outage risk score at the future time point;

[0145] If S`(t + Δt) < A1, generate a first-level early warning to remind the monitoring personnel to strengthen monitoring and regularly check the wind speed changes;

[0146] If A1 ≤ S`(t + Δt) < A2, generate a second-level early warning to notify the operation and maintenance team to prepare emergency measures, such as checking and strengthening grid equipment and preparing a load transfer plan;

[0147] If A2 ≤ S`(t + Δt), generate a third-level early warning to trigger a comprehensive emergency response process, dispatch operation and maintenance personnel to high-risk areas, and take preventive measures, such as pre-disconnecting non-critical loads in high-risk areas.

[0148] By substituting the predicted wind speed data at the future time point into the power outage risk status model, generating a predicted score and setting three-level early warning thresholds, the present invention realizes the hierarchical management of potential risks. The low, medium, and high-level early warnings respectively guide monitoring, prevention, and comprehensive response measures to ensure appropriate actions are taken at different risk levels. This early warning mechanism improves the emergency response efficiency and accuracy of medium- and low-voltage distribution network areas, and provides an effective guarantee for timely preventing sudden power outage risks, optimizing resource allocation, and reducing the impact of faults.

[0149] S4. Store all data and perform regular backups;

[0150] Specifically, all data is stored and backed up regularly, including:

[0151] Use structured storage to store all data generated during the generation process, create indexes for each round of generated content and related data, establish a data backup mechanism, regularly back up stored data, set user permissions, and allow managers to call stored data.

[0152] This step systematically manages all data in the generation process through structured storage and indexing mechanisms, achieves efficient data storage and retrieval, ensures data security and integrity through regular backup and user permission settings, and guarantees that managers can flexibly call data within the scope of their authority. This solution greatly improves the security, reliability and operability of data management, and provides effective support for long-term storage and convenient call of data.

[0153] Example 2

[0154] Reference Figure 3 , which is a second embodiment of the present invention, provides a power outage detection system for a low-voltage distribution network area, comprising:

[0155] Data acquisition module, used to collect raw data in real time and perform preprocessing;

[0156] A model building module, used to build a power outage risk state calculation model to calculate the power outage risk score and classify it;

[0157] A wind speed prediction module is used to construct a wind speed prediction model based on the classification results to calculate the predicted wind speed data at future time points;

[0158] The classification warning module is used to generate a predicted power outage risk score at a future time point through a power outage risk status calculation model and to conduct a classification warning;

[0159] The data storage module is used to store all data and back up them regularly.

[0160] Reference Figure 4 , a specific module composition of a power outage detection system for a medium and low voltage distribution network is provided, including:

[0161] The data acquisition module (module of the prior art) includes sensors and hardware devices. In the present invention, the sensors and hardware devices used include wind speed sensors (such as RM Young 05103), current sensors (such as HTA series), voltage sensors (such as ABB's low-voltage voltage transformers), vibration sensors (such as IMI's CTC series sensors), industrial gateways (such as Siemens SIMATIC IOT2040), edge computing devices (such as NVIDIA JetsonNano), communication protocols (such as Modbus protocol), etc.;

[0162] The model building module introduces the Sigmoid function, power correction function and exponential decay function, which are respectively used to simulate the nonlinear effects of wind speed, grid load and equipment health status on risk scores. The wind speed is mapped to the nonlinear interval of the risk score through the Sigmoid function, which is used to quickly identify the sudden risks caused by excessive wind speed. The load factor adopts the power correction model to enhance the risk weight under high load, and combines the cosine function to simulate the impact of periodic fluctuations on equipment operation. The exponential decay model of equipment health status can accurately reflect the gradual increase in the risk of power outages caused by equipment aging;

[0163] The final model for risk score calculation is trained using machine learning algorithms (such as CatBoost). The historical operating data is divided into training sets and validation sets, and the model parameters are optimized to improve the accuracy of the score. The training features include wind speed, grid load, equipment health status, etc. The target label is the historical record of power outage status. The module also supports real-time data input to achieve dynamic update of the score and accurate assessment of power outage risk.

[0164] The wind speed prediction module ensures the accuracy of input data through data cleaning and feature extraction. The cleaning process includes outlier detection and moving average smoothing. The model adopts a generalized additive model (GAM), combines the time dimensions of hours, days and months, and captures the periodic characteristics of wind speed changes through cubic spline functions. The historical wind speed data is divided into a training set and a test set. The smoothing function of the model is adjusted through cross-validation to improve the prediction accuracy. During the period of high wind speed potential risk, the model introduces a local fitting method to optimize the data in the key time period to improve the prediction accuracy.

[0165] The wind speed forecast results are input into the power outage risk status model, which optimizes risk assessment through a mechanism that dynamically adjusts the wind speed critical threshold, making the system more sensitive and responsive to sudden weather conditions;

[0166] The classification warning module generates risk scores at future time points based on the output of the power outage risk status model, and triggers graded warnings according to the risk level, setting low warning thresholds, medium warning thresholds, and high warning thresholds, corresponding to the first, second, and third level warning mechanisms, respectively. The module can notify relevant personnel through the system interface, SMS, email, or automated push tools (such as DingTalk robots) to ensure the timeliness and effectiveness of risk response;

[0167] The data storage module stores all collected, generated and predicted data, and supports regular backup. All data is stored in a structured form, including timestamps and device identification indexes, to facilitate subsequent retrieval and analysis. Data storage supports time series databases (such as InfluxDB) for the storage of high-frequency collected data, and structured data can be stored in relational databases (such as PostgreSQL). The module is equipped with an automated backup function that can back up data to an off-site storage server or cloud storage service (such as Amazon S3 or Alibaba Cloud OSS) to ensure data security and redundancy. To improve data call efficiency, the module adopts a hierarchical user authority management mechanism that allows system administrators to call historical data and perform screening and analysis based on specific conditions.

[0168] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0169] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0170] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0171] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0172] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting power outages in low-voltage distribution network areas, characterized in that: include, S1 collects raw data in real time and performs preprocessing; S2 builds a power outage risk status calculation model to calculate the power outage risk score and classify it, and builds a wind speed prediction model based on the classification results to calculate the predicted wind speed data at future time points; S3 generates a predicted power outage risk score at a future time point through a power outage risk status calculation model, and performs classified warnings; S4 stores all data and backs it up regularly.

2. The method for detecting power outages in low-voltage distribution network areas according to claim 1, characterized in that: S1 The real-time collection of raw data includes: Install sensors in the equipment in the medium and low voltage sections of the distribution network, set the collection frequency of each sensor, and collect raw data in real time; The raw data includes wind speed, grid load, equipment health status and equipment power outage status at different time points.

3. The method for detecting power outages in low-voltage distribution network areas according to claim 1, characterized in that: The pre-processing described in S1 includes: A statistical anomaly detection algorithm is used to identify and remove noise and erroneous raw data, and a linear interpolation method is used to complete the missing raw data, and the completed raw data is standardized.

4. The method for detecting power outages in low-voltage distribution network areas according to claim 1, characterized in that: S2 constructs a power outage risk state calculation model to calculate the power outage risk score and classify it, including: Construct an initial calculation model for power outage risk status, combining wind speed, grid load and equipment health status. Calculate the initial score S(t) of power outage risk; The Sigmoid function is introduced to map the wind speed into a nonlinear function of the wind speed, and the nonlinear mapping value σ(V(t)) of the wind speed at time point t is calculated. The formula is: Wherein, V(t) represents the wind speed at time point t, θ represents the curve adjustment coefficient, and ν represents the critical wind speed threshold; The power correction and cosine function are introduced to simulate the fluctuation under high load, and the nonlinear correction value y(P(t)) of the load at time point t is calculated. The formula is: y(P(t))=P(t) δ ·cos(ψ·P(t)) Wherein, P(t) represents the grid load at time point t, δ represents the power correction parameter of the load, and ψ represents the frequency regulation coefficient; The exponential decay function is introduced to simulate the exponential decay effect of the equipment health status, and the exponential decay value g(T(t)) of the equipment health status at time point t is calculated. The formula is: g(T(t))=e -λ(T(t)-μ)2 Wherein, T(t) represents the health status of the device at time point t, λ represents the adjustment coefficient of the health status of the device, and μ represents the health reference threshold; Construct the final power outage risk state calculation model and calculate the final power outage risk score S`(t). The formula is: S`(t)=σ(V(t))+y(P(t))+g(T(t)) Among them, S`(t) represents the final score of power outage risk at time point t; The historical raw data is divided into a training set and a validation set, the CatBoost algorithm is used to train the final power outage risk state calculation model, and the parameters of CatBoost are set; The training set includes input features and target labels, with wind speed, grid load and equipment health status as input features and equipment power outage status as target labels; Use the validation set to test the prediction effect of the final power outage risk status calculation model, and adjust the parameters of CatBoost according to the validation results; Input the real-time raw data into the final power outage risk status calculation model to obtain the real-time final power outage risk score; Set a real-time scoring threshold according to the distribution of power outage data in the historical original data, and compare the final score of the power outage risk with the real-time scoring threshold for classification processing; If the final score of the power outage risk is greater than or equal to the real-time scoring threshold, the wind speed at the corresponding time point is taken as potential risk data and extracted; If the final score of the power outage risk is less than the real-time scoring threshold, the result is stored and data collection and calculation continue.

5. The method for detecting power outages in low-voltage distribution network areas according to claim 1, characterized in that: The calculation of the predicted wind speed data at future time points by constructing a wind speed prediction model according to the classification results described in S2 includes, Extract the historical wind speed data of medium and low voltage distribution areas from the central database; Use an automated script to detect and clean abnormal data points, and use the moving average method to smooth the historical wind speed data; The generalized additive model is selected for wind speed prediction. The cubic spline function is used to construct smooth functions in the three time dimensions of hours, days and months. By combining the smooth functions, a wind speed prediction model is constructed to calculate the initial wind speed prediction value V at time point t. j (t); Use the historical wind speed data to train the wind speed prediction model to obtain a complete smoothing function; Adjust the smoothing function of the wind speed prediction model through cross-validation; Integrate the extracted potential risk data into a potential risk data set, and use the potential risk data set for local fitting to calculate the predicted wind speed data at future time points; According to the design requirements of the power distribution equipment, set a wind speed threshold to judge the wind speed at future time points; If the wind speed at the future time point is greater than or equal to the wind speed threshold, dynamically adjust the wind speed critical threshold in the wind speed non-linear function; The adjustment formula for dynamically adjusting the wind speed critical threshold in the wind speed non-linear function is: n`=n-k·(V j (t+Δt)-V r ) Among them, ν` represents the critical wind speed threshold after dynamic adjustment, k represents the wind speed adjustment coefficient, V j (t+Δt) represents the predicted wind speed at the future time point t+Δt, V r Indicates the wind speed threshold.

6. The method for detecting power outages in low-voltage distribution network areas according to claim 1, characterized in that: The generation of the predicted power outage risk score at future time points by the power outage risk status calculation model described in S3 and the classification warning include, Substitute the predicted wind speed data at the future time point into the final power outage risk status calculation model to generate the predicted final power outage risk score S`(t+Δt) at the future time point; Set a low warning threshold A1 and a high warning threshold A2, and conduct classification warnings for the predicted final power outage risk score at future time points; If S`(t+Δt) < A1, generate a first-level warning to remind the monitoring personnel to strengthen monitoring and regularly check the wind speed changes; If A1 ≤ S`(t+Δt) < A2, generate a second-level warning to notify the operation and maintenance team to prepare emergency measures; If A2 ≤ S`(t+Δt), generate a third-level warning to trigger a comprehensive emergency response process, dispatch operation and maintenance personnel to high-risk areas, and take preventive measures.

7. The method for detecting power outages in low-voltage distribution network areas according to claim 1, characterized in that: The storage of all data and regular backup described in S4 include, Use a structured storage form to store all data generated during the generation process, establish indexes for the content and related data generated in each round, establish a data backup mechanism, regularly back up the stored data, set user permissions, and allow managers to call the stored data.

8. A power outage detection system for a distribution network medium and low voltage area based on the power outage detection method for a distribution network medium and low voltage area according to any one of claims 1 to 7, characterized in that: Include, A data collection module for real-time collecting original data and preprocessing it; A model construction module for constructing a power outage risk status calculation model to calculate the power outage risk score and classify it; A wind speed prediction module for constructing a wind speed prediction model according to the classification results to calculate the predicted wind speed data at future time points; A classification warning module for generating the predicted power outage risk score at future time points by the power outage risk status calculation model and conducting classification warnings; A data storage module for storing all data and regularly backing it up.

9. A computer device comprising: A memory and a processor; the memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of the power outage detection method for low-voltage substations in the distribution network described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting power outages in low-voltage substations in a distribution network described in any one of claims 1 to 7 are implemented.

Images