Power grid fault early warning system based on energy twinborn body

Abstract

The invention discloses a power grid fault early warning system based on energy twinborn bodies. Comprising a physical system, a data transmission layer and a digital twin system, the digital twinborn system pushes a twinborn body model to be synchronously updated in real time according to mobile data synchronized into a digital twinborn database by power equipment transaction data, and simulates and deduces various scenes of the twinborn system; according to the method, through construction of distributed energy twinborn bodies and source network load storage collaborative optimization operation deduction, the problems of heavy overload, voltage line crossing and the like of an actual power grid are found in time, and early warning is carried out on possible faults; a physical system, a data transmission layer and a digital twinning technology are combined, real-time monitoring, fault early warning and collaborative optimization of a power grid are realized through real-time acquisition, accurate modeling and intelligent analysis of data, and the system has relatively high intellectualization and self-adaption capabilities, has a wide application prospect in power grid management and scheduling, and is suitable for popularization and application. And particularly, the safety and the stability of power supply are ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid fault early warning, and more specifically, particularly relates to a power grid fault early warning system based on energy twins. Background Art

[0002] As a highly complex system, the power grid covers multiple links from power generation, transmission to distribution. With the changes in the energy structure, especially the large-scale access of renewable energy (such as wind power, photovoltaics, etc.), the operating environment of the power grid has become more complex. The power grid needs to cope with dynamically changing loads, external meteorological factors, and changes in the status of equipment, which makes the occurrence of power grid failures more unpredictable. Power grid failures usually manifest themselves in the following types:

[0003] Short circuit faults: such as single-phase grounding and three-phase short circuit, often occur in substations, transmission lines and other equipment.

[0004] Overload failure: The grid load exceeds the design range, resulting in equipment damage or power supply interruption.

[0005] Equipment failure: Due to aging or failure of equipment, some functions of the power grid fail.

[0006] Power quality issues: such as voltage fluctuations, unstable frequency, etc.

[0007] With the continuous development of the power industry and the advancement of digital transformation, digital twin technology provides a new solution to solve the various challenges faced by the power industry. Based on the digital twin model, the physical architecture of the power system is mapped to realize the digital modeling of power equipment, which is convenient for the management and monitoring of power equipment; at the same time, based on the power system, data is obtained from each power system operation module, and a virtual model is established to simulate the physical process in the real world, so as to realize real-time monitoring of power grid operation, collect, analyze and optimize data, realize the flow of data in the digital twin model, form a physical entity and a three-dimensional model to ensure consistent behavior through data, and show the synchronization state of virtual and real interaction. The issued instructions control the physical equipment through the digital twin model, so as to achieve the effect of virtual controlling the real, and provide a new technical framework and solution for the collaborative interaction between source, grid, load and storage.

[0008] Digital twin refers to the process of building a virtual model of a physical entity (such as power grid equipment and systems) and interacting with the physical entity in real time to simulate and predict the behavior of the physical system. In the field of power grids, digital twin technology can be applied to equipment status monitoring, fault prediction and diagnosis, and intelligent scheduling.

[0009] However, there are some problems with the existing technology: with the application of digital twin technology in the power system, a large amount of data is collected and processed, but data islands are caused, and the data cannot be transmitted synchronously, and the data cannot be processed quickly, and it is impossible to warn of power grid failures based on real-time data. Therefore, we propose a power grid fault warning system based on energy twins. Summary of the invention

[0010] In view of the problems existing in the prior art, the purpose of the present invention is to provide a power grid fault early warning system based on energy twins. Through the deduction method based on digital twin technology, panoramic monitoring of regional source, grid, load and storage resources is carried out to realize dynamic update of power grid topology and dynamic monitoring of equipment status; with the help of digital twin technology, the physical objects of the power system are mirrored in the virtual space, and the digital twin power grid and the physical power grid are corrected and updated through the virtual-reality interaction capability to regulate and control the power grid, and the power grid twins at various voltage levels are established; a source, grid, load and storage collaborative interaction model is constructed, and source, grid, load and storage collaborative interaction deduction is carried out in different scenarios to provide decision support for different business scenarios such as microgrids, virtual power plants, and demand response.

[0011] To achieve the above-mentioned object, the present invention provides the following technical solutions: a power grid fault early warning system based on energy twins, comprising a physical system, a data transmission layer and a digital twin system;

[0012] The physical system is used to collect data. The physical system obtains the new addition and replacement information of the power equipment according to the input data. The physical system also obtains the daily type data and environmental data through sensors. The daily type data includes the maintenance data and fault data of the power equipment. The physical system retrieves the photovoltaic data, wind power data and load history data of the distributed new energy stored in the database. The physical system transmits the collected data to the digital twin system through the data transmission layer.

[0013] The data transmission layer synchronously transmits the collected data to the digital twin system, and the digital twin system stores the collected data in a digital twin database. The digital twin system preprocesses and analyzes the collected data, and then calculates and processes the photovoltaic data, the wind power data, and the load history data to generate a photovoltaic model and a wind power model.

[0014] The calculation formula of the photovoltaic model is as follows:

[0015]

[0016] in, Expressed as the maximum power output of photovoltaic, The power output of each photovoltaic panel, ψ represents the conversion efficiency of the photovoltaic panel, α pv It is expressed as temperature coefficient, T represents the battery temperature, T ref It represents the set battery temperature, and S represents the area of ​​the photovoltaic panel;

[0017] The calculation formula of the wind power model is as follows:

[0018]

[0019] in, Expressed as the maximum power output of wind power, V w is the actual wind speed, Represents the wind speed of wind power input, Expressed as wind speed of wind power output, It is expressed as the reference wind speed for wind power. The power output of each wind power, ζ represents the conversion efficiency of wind power;

[0020] The digital twin system promotes the real-time synchronization update of the twin model based on the mobile data synchronized to the digital twin database according to the abnormal data of the power equipment, and carries out simulation and deduction of various scenarios of the twin system. The digital twin system performs collaborative optimization deduction of the power grid by simulating the power demand and supply in different scenarios.

[0021] Specifically, the data transmission layer includes a switch, a server and a firewall. The switch is used to realize device exchange between the physical system and the digital twin system, and is mainly used to forward data frames from the physical system to the digital twin system. The server is used to establish a communication connection between the physical system and the digital twin system. The firewall is used to realize security protection of the transmitted data and ensure stable data transmission.

[0022] Specifically, the synchronous transmission of the data transmission layer is that the physical system and the digital twin system share a common clock signal, and the clock information is transmitted through a physical connection or through a control signal. The physical system provides a clock signal while sending data, and the digital twin system receives data at an appropriate time according to the clock signal, or the physical system and the digital twin system are synchronized through the synchronization words in the data frame, and the digital twin system determines the data transmission timing by analyzing the synchronization words or other control information in the frame header.

[0023] Specifically, the preprocessing is used to remove noise, detect outliers, detect missing values, fill missing values ​​and normalize the data;

[0024] The noise removal is used to filter the noise in the data to improve the accuracy of the data, the outlier detection is used to detect abnormal data in the data and eliminate the abnormal data, the missing value detection is used to detect missing values ​​in the data, the missing value filling is used to fill the positions of the eliminated outliers and the missing values ​​in the data with the mean, and the normalization processing is used to scale the data to between [0,1].

[0025] Specifically, the noise removal is used to smooth and remove noise from the data, and the noise removal uses the following calculation formula to process the data:

[0026]

[0027] Among them, X t Represented as the original data corresponding to time point t in the data, S t-1 It is represented as the smoothing value corresponding to the time point t-1 in the data, α is the smoothing factor, which is used to assign weights to new data points and historical data points, and W f (a,b) represents the transformation structure of the data at scale a and position b, ψ * is the conjugate of the wavelet transform, a is the scale parameter, and b is the translation parameter.

[0028] Specifically, the detected abnormal value is detected by the following algorithm:

[0029]

[0030] Where Z represents the data cleaning calculation value, X represents the data, u represents the mean of the data group in the data, σ represents the standard deviation of the data group in the data, and n represents the number of data in the data group, and i represents the position of the data in the data group, x i is the i-th data in the data group, and the judgment for the anomaly is |Z|>2, then the data point is an outlier, and the outlier is eliminated.

[0031] Specifically, the normalization process maps the data between [0, 1], and the normalization formula is as follows:

[0032] Among them, X min Represents the minimum value in the data, X max Represented as the maximum value in the data;

[0033] The missing values ​​are filled by replacing the average value:

[0034] Among them, m is the amount of non-abnormal data.

[0035] Specifically, the analysis process is used to extract features from the data, and the extracted features are as follows:

[0036] The mean is calculated as follows:

[0037]

[0038] Among them, μ old is the mean of the previous data, μ new is the mean of the new data, N is the number of data;

[0039] And the calculation formula of the mean is as follows:

[0040]

[0041] Where μ is the calculated mean, w i is the weight value of the data, x i is the data value;

[0042] The standard deviation is calculated as follows:

[0043]

[0044] Where μ is the mean, x i is the data value, N is the number of data, and σ is the calculated standard deviation;

[0045] The variance is calculated as follows:

[0046]

[0047] Where μ is the mean, x i is the data value, N is the number of data, σ 2 is the calculated variance;

[0048] The maximum and minimum values ​​are calculated as follows:

[0049] Max new =max(Max old ,x new )

[0050] Min new =min(Min old ,x new ),

[0051] Among them, Max new is the new maximum value after calculation, Max old is the previous maximum value, Min new is the new minimum value after calculation, Min old is the previous minimum value, x new For newly added data.

[0052] Specifically, the digital twin system uses feature crossover technology to convert the mean value X 1 , standard deviation X 2 , Variance X 3 , maximum value X 4 and minimum value X 5 to integrate;

[0053]

[0054] Obtain new features of the feature fusion of the digital twin system;

[0055] Fault prediction is performed through a multi-layer perceptron, and the calculation formula of the multi-layer perceptron is:

[0056]

[0057] Among them, W 1 , W 2 …W k is the weight matrix of the fully connected layer, b 1 , b 2 …b k is the bias term, ReLU is the activation function, h 1 、h 2 …h k is the intermediate representation of each layer, the cross feature X cross Further learning is done through the fully connected layer, and the prediction result y is finally output.

[0058] Specifically, the digital twin model performs simulation based on real-time data streams and adjusts the behavior of the physical system through model predictive control, and the calculation formula of the model predictive control is as follows:

[0059]

[0060] Among them, u is the optimal control input, y k is the system output, y ref,k is the reference trajectory, Q and R are weight matrices, and N is the prediction time domain length.

[0061] Technical effects and advantages of the present invention:

[0062] The present invention constructs a distributed energy twin and a source-grid-load-storage collaborative optimization operation simulation to timely discover problems such as heavy overload and voltage crossing in the actual power grid, and warn of possible faults, thereby improving the dispatching decision-making capabilities in multiple scenarios and reducing potential operation risks.

[0063] By making full use of digital twin technology, the comprehensive collection and comprehensive analysis of power grid operation status data is realized. Through the construction of distributed energy twins, the arrangement of operation simulation scenarios and multi-state simulation, the weaknesses of the actual power grid are discovered in a timely manner, and possible faults are warned and preset for faults that have not occurred. The feasibility and availability of the dispatch optimization strategy are verified, which effectively promotes the coordinated operation of multiple distributed devices and power systems, significantly improves the dispatch decision-making capabilities in multiple scenarios, and reduces potential operation risks.

[0064] It combines the physical system, data transmission layer and digital twin technology, and realizes real-time monitoring, fault warning and collaborative optimization of the power grid through real-time data collection, precise modeling and intelligent analysis. It has strong intelligence and adaptive capabilities and has broad application prospects in power grid management and dispatching, especially in ensuring the safety and stability of power supply.

[0065] Real-time fault warning: The digital twin system can simulate and predict the fault data of power grid equipment and provide early warning before a fault occurs, thereby reducing sudden power outages and equipment damage.

[0066] Grid optimization: By simulating different load and power supply scenarios, it is possible to achieve coordinated optimization of grid operation, ensure energy supply and demand balance, and improve grid operation efficiency;

[0067] Data-driven decision support: With the help of big data analysis, machine learning and artificial intelligence technologies, the system can continuously learn and adjust the prediction model based on historical data to improve prediction accuracy.

[0068] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a schematic diagram of the system structure provided by the present invention;

[0070] Figure 2 It is a schematic diagram of the process provided by the present invention. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0072] like Figure 1 to Figure 2As shown, an embodiment of the present invention provides a power grid fault early warning system based on energy twins, including a physical system, a data transmission layer and a digital twin system;

[0073] The physical system is used to collect data. The physical system obtains the new addition and replacement information of the power equipment according to the input data. The physical system also obtains the daily type data and environmental data through sensors. The daily type data includes the maintenance data and fault data of the power equipment. The physical system retrieves the photovoltaic data, wind power data and load history data of the distributed new energy stored in the database. The physical system transmits the collected data to the digital twin system through the data transmission layer.

[0074] The data transmission layer synchronously transmits the collected data to the digital twin system, and the digital twin system stores the collected data in a digital twin database. The digital twin system preprocesses and analyzes the collected data, and then calculates and processes the photovoltaic data, the wind power data, and the load history data to generate a photovoltaic model and a wind power model.

[0075] The calculation formula of the photovoltaic model is as follows:

[0076]

[0077] in, Expressed as the maximum power output of photovoltaic, The power output of each photovoltaic panel, ψ represents the conversion efficiency of the photovoltaic panel, α pv It is expressed as temperature coefficient, T represents battery temperature, T ref It represents the set battery temperature, and S represents the area of ​​the photovoltaic panel;

[0078] The calculation formula of the wind power model is as follows:

[0079]

[0080] in, Expressed as the maximum power output of wind power, V w is the actual wind speed, Represents the wind speed of wind power input, Expressed as wind speed of wind power output, It is expressed as the reference wind speed for wind power. The power output of each wind power, ζ represents the conversion efficiency of wind power;

[0081] The digital twin system promotes the real-time synchronization update of the twin model based on the mobile data synchronized to the digital twin database according to the abnormal data of the power equipment, and carries out simulation and deduction of various scenarios of the twin system. The digital twin system performs collaborative optimization deduction of the power grid by simulating the power demand and supply in different scenarios.

[0082] In this embodiment, preferably, the data transmission layer includes a switch, a server and a firewall. The switch is used to realize device exchange between the physical system and the digital twin system, and is mainly used to forward data frames from the physical system to the digital twin system. The server is used to establish a communication connection between the physical system and the digital twin system. The firewall is used to realize security protection of the transmitted data and ensure the stability of data transmission.

[0083] It should be noted that the design of the data transmission layer ensures that the data exchange between the physical system and the digital twin system is not only fast and efficient, but also ensures the security and stability of the transmission process through the organic combination of switches, servers and firewalls. It can support the efficient operation of the power grid fault warning system, especially when facing massive data and potential security risks, and provide strong protection.

[0084] In this embodiment, preferably, the synchronous transmission of the data transmission layer is that the physical system and the digital twin system share a common clock signal, and the clock information is transmitted through a physical connection or through a control signal. The physical system provides a clock signal while sending data, and the digital twin system receives data at an appropriate time according to the clock signal, or the physical system and the digital twin system are synchronized through the synchronization word in the data frame, and the digital twin system determines the transmission timing of the data by analyzing the synchronization word or other control information in the frame header;

[0085] It should be noted that efficient and accurate synchronization between the physical system and the digital twin system can be achieved by sharing a common clock signal or using synchronization words. The shared clock signal is suitable for scenarios with extremely strict timing requirements, while the synchronization word method provides a more flexible and scalable solution. Both have their own advantages, and the appropriate synchronization method can be selected according to actual application requirements to ensure the efficiency of data transmission and the accuracy of timing.

[0086] In this embodiment, preferably, the preprocessing is used to remove noise, detect outliers, detect missing values, fill missing values ​​and normalize the data;

[0087] The noise removal is used to filter the noise in the data to improve the accuracy of the data, the abnormal value detection is used to detect abnormal data in the data and remove the abnormal data, the missing value detection is used to detect the missing values ​​in the data, the missing value filling is used to fill the position of the removed abnormal value and the position of the missing value in the data with the mean value, and the normalization processing is used to scale the data to between [0,1];

[0088] It should be noted that the quality of the data has been significantly improved through data preprocessing. Removing noise, detecting and eliminating outliers, detecting and filling missing values, and normalizing the data can ensure the accuracy, completeness and consistency of the data, laying a solid foundation for subsequent data analysis, modeling and prediction, and improving the performance of the model and the reliability of the results.

[0089] In this embodiment, preferably, the noise removal is used to smooth and remove noise from the data, and the noise removal uses the following calculation formula to process the data:

[0090]

[0091] Among them, X t Represented as the original data corresponding to time point t in the data, S t-1 It is represented as the smoothing value corresponding to the time point t-1 in the data, α is the smoothing factor, which is used to assign weights to new data points and historical data points, and W f (a,b) represents the transformation structure of the data at scale a and position b, ψ * is the conjugate of the wavelet transform, a is the scale parameter, and b is the translation parameter;

[0092] It should be noted that the denoising process using wavelet transform combined with smoothing factor can not only effectively remove the noise in the data, but also retain the key features of the data. Wavelet transform is a method that can analyze signals at multiple scales and resolutions, and can effectively capture the local features of the signal. At the same time, smoothing can remove high-frequency noise, thereby improving the quality of the data.

[0093] In this embodiment, preferably, the detected abnormal value is detected by the following algorithm:

[0094]

[0095] Where Z represents the data cleaning calculation value, X represents the data, u represents the mean of the data group in the data, σ represents the standard deviation of the data group in the data, and n represents the number of data in the data group, and i represents the position of the data in the data group, x iis the i-th data in the data set, and the judgment for abnormality is |Z|>2, then the data point is an outlier, and the outlier is removed;

[0096] It should be noted that by calculating the mean, standard deviation and other relevant statistics of the data group to determine the outliers, the extreme outliers in the data can be effectively removed.

[0097] In this embodiment, preferably, the normalization process maps the data between [0, 1], and the normalization formula is as follows:

[0098] Among them, X min Represents the minimum value in the data, X max Represented as the maximum value in the data;

[0099] The missing values ​​are filled by replacing the average value:

[0100] Among them, m is the amount of non-abnormal data;

[0101] It should be noted that scaling the data through normalization so that the data is mapped between [0,1] can reduce the dimension of the data, facilitate subsequent data calculations, improve calculation efficiency and reduce calculation complexity, and fill in missing values ​​through the mean to improve the completeness of the data.

[0102] In this embodiment, preferably, the analysis process is used to extract features from the data, and the extracted features are as follows:

[0103] The mean is calculated as follows:

[0104]

[0105] Among them, μ old is the mean of the previous data, μ new is the mean of the new data, N is the number of data;

[0106] And the calculation formula of the mean is as follows:

[0107]

[0108] Where μ is the calculated mean, w i is the weight value of the data, x i is the data value;

[0109] The standard deviation is calculated as follows:

[0110]

[0111] Where μ is the mean, x i is the data value, N is the number of data, and σ is the calculated standard deviation;

[0112] The variance is calculated as follows:

[0113]

[0114] Where μ is the mean, x i is the data value, N is the number of data, σ 2 is the calculated variance;

[0115] The maximum and minimum values ​​are calculated as follows:

[0116] Max new =max(Max old ,x new )

[0117] Min new =min(Min old ,x new ),

[0118] Among them, Max new is the new maximum value after calculation, Max old is the previous maximum value, Min new is the new minimum value after calculation, Min old is the previous minimum value, x new For newly added data;

[0119] It should be noted that.

[0120] In this embodiment, preferably, the digital twin system uses feature crossover technology to convert the mean value X 1 , standard deviation X 2 , Variance X 3 , maximum value X 4 and minimum value X 5 to integrate;

[0121]

[0122] Obtain new features of the feature fusion of the digital twin system;

[0123] Fault prediction is performed through a multi-layer perceptron, and the calculation formula of the multi-layer perceptron is:

[0124]

[0125] Among them, W 1 , W 2 …W k is the weight matrix of the fully connected layer, b 1、b 2 …b k is the bias term, ReLU is the activation function, h 1 、h 2 …h k is the intermediate representation of each layer, the cross feature X cross Further learning through the fully connected layer finally outputs the prediction result y;

[0126] It should be noted that feature crossover is to fuse the mean, standard deviation, variance, maximum value and minimum value in the data as new features to capture the interactions between them, which helps linear models or other traditional models to better model complex nonlinear relationships, and then predict fault information through multi-layer perceptron to obtain information about power grid faults.

[0127] In this embodiment, preferably, the digital twin model performs simulation based on the real-time data stream, and adjusts the behavior of the physical system through model predictive control, and the calculation formula of the model predictive control is as follows:

[0128]

[0129] Among them, u is the optimal control input, y k is the system output, y ref,k is the reference trajectory, Q and R are weight matrices, and N is the prediction time domain length;

[0130] It should be noted that model predictive control is used to optimize and adjust the behavior of physical systems. MPC is based on the dynamic model of the system. Through continuous optimization calculations, the control input is adjusted in real time so that the system output tracks the reference trajectory as much as possible, ensuring that the system behavior is consistent with the target reference trajectory, thereby achieving efficient system control and regulation.

[0131] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A power grid fault early warning system based on energy twins, characterized in that: Includes physical systems, data transmission layer and digital twin systems; The physical system is used to collect data. The physical system obtains the new addition and replacement information of the power equipment according to the input data. The physical system also obtains the daily type data and environmental data through sensors. The daily type data includes the maintenance data and fault data of the power equipment. The physical system retrieves the photovoltaic data, wind power data and load history data of the distributed new energy stored in the database. The physical system transmits the collected data to the digital twin system through the data transmission layer. The data transmission layer synchronously transmits the collected data to the digital twin system, and the digital twin system stores the collected data in a digital twin database. The digital twin system preprocesses and analyzes the collected data, and then calculates and processes the photovoltaic data, the wind power data, and the load history data to generate a photovoltaic model and a wind power model. The calculation formula of the photovoltaic model is as follows: in, Expressed as the maximum power output of photovoltaic, The power output of each photovoltaic panel, ψ represents the conversion efficiency of the photovoltaic panel, α pv It is expressed as temperature coefficient, T represents battery temperature, T ref It represents the set battery temperature, and S represents the area of ​​the photovoltaic panel; The calculation formula of the wind power model is as follows: in, Expressed as the maximum power output of wind power, V w is the actual wind speed, Represents the wind speed of wind power input, Expressed as wind speed of wind power output, It is expressed as the reference wind speed for wind power. The power output of each wind power, ζ represents the conversion efficiency of wind power; The digital twin system promotes the real-time synchronization update of the twin model based on the mobile data synchronized to the digital twin database according to the abnormal data of the power equipment, and carries out simulation and deduction of various scenarios of the twin system. The digital twin system performs collaborative optimization deduction of the power grid by simulating the power demand and supply in different scenarios.

2. The power grid fault early warning system based on energy twin according to claim 1 is characterized by: The data transmission layer includes switches, servers and firewalls. The switch is used to realize device exchange between the physical system and the digital twin system, and is mainly used to forward data frames from the physical system to the digital twin system. The server is used to establish a communication connection between the physical system and the digital twin system. The firewall is used to realize security protection of the transmitted data and ensure the stability of data transmission.

3. The power grid fault early warning system based on energy twin according to claim 1 is characterized by: The synchronous transmission of the data transmission layer is that the physical system and the digital twin system share a common clock signal, and the clock information is transmitted through a physical connection or through a control signal. The physical system provides a clock signal while sending data, and the digital twin system receives data at an appropriate time according to the clock signal, or the physical system and the digital twin system are synchronized through the synchronization words in the data frame, and the digital twin system determines the data transmission timing by analyzing the synchronization words or other control information in the frame header.

4. The power grid fault early warning system based on energy twin according to claim 1 is characterized in that: The preprocessing is used to remove noise, detect outliers, detect missing values, fill missing values ​​and normalize the data; The noise removal is used to filter the noise in the data to improve the accuracy of the data, the outlier detection is used to detect abnormal data in the data and eliminate the abnormal data, the missing value detection is used to detect missing values ​​in the data, the missing value filling is used to fill the positions of the eliminated outliers and the missing values ​​in the data with the mean, and the normalization processing is used to scale the data to between [0,1].

5. The power grid fault early warning system based on energy twin according to claim 4 is characterized in that: The noise removal is used to smooth and remove noise from the data. The noise removal uses the following calculation formula to process the data: Among them, X t Represented as the original data corresponding to time point t in the data, S t-1 It is represented as the smoothing value corresponding to the time point t-1 in the data, α is the smoothing factor, which is used to assign weights to new data points and historical data points, and W f (a,b) represents the transformation structure of the data at scale a and position b, ψ * is the conjugate of the wavelet transform, a is the scale parameter, and b is the translation parameter.

6. The power grid fault early warning system based on energy twin according to claim 4 is characterized in that: The detected outliers are detected by the following algorithm: Where Z represents the data cleaning calculation value, X represents the data, u represents the mean of the data group in the data, σ represents the standard deviation of the data group in the data, and n represents the number of data in the data group, and i represents the position of the data in the data group, x i is the i-th data in the data group, and the judgment for the anomaly is |Z|>2, then the data point is an outlier, and the outlier is eliminated.

7. The power grid fault early warning system based on energy twin according to claim 4 is characterized in that: The normalization process maps the data between [0, 1], and the normalization formula is as follows: Among them, X min Represents the minimum value in the data, X max Represented as the maximum value in the data; The missing values ​​are filled by replacing the average value: Among them, m is the amount of non-abnormal data.

8. The power grid fault early warning system based on energy twin according to claim 1 is characterized by: The analysis process is used to extract features from the data, and the extracted features are as follows: The mean is calculated as follows: Among them, μ old is the mean of the previous data, μ new is the mean of the new data, N is the number of data; And the calculation formula of the mean is as follows: Where μ is the calculated mean, w i is the weight value of the data, x i is the data value; The standard deviation is calculated as follows: Where μ is the mean, x i is the data value, N is the number of data, and σ is the calculated standard deviation; The variance is calculated as follows: Where μ is the mean, x i is the data value, N is the number of data, σ 2 is the calculated variance; The maximum and minimum values ​​are calculated as follows: Max new =max(Max old ,x new ) My new =min(Min old ,x new ), Among them, Max new is the new maximum value after calculation, Max old is the previous maximum value, Min new is the new minimum value after calculation, Min old is the previous minimum value, x new For newly added data.

9. The power grid fault early warning system based on energy twin according to claim 1 is characterized by: The digital twin system fuses the mean X1, standard deviation X2, variance X3, maximum value X4 and minimum value X5 through feature crossover technology; Obtain new features of the feature fusion of the digital twin system; Fault prediction is performed through a multi-layer perceptron, and the calculation formula of the multi-layer perceptron is: h1=ReLU(W1X cross +b1) h2=ReLU(W2h1+b2) y=W k h k-1 +b k , Among them, W1, W2…W k is the weight matrix of the fully connected layer, b1, b2…b k is the bias term, ReLU is the activation function, h1, h2…h k is the intermediate representation of each layer, the cross feature X cross Further learning is done through the fully connected layer, and the prediction result y is finally output.

10. The power grid fault early warning system based on energy twin according to claim 1, characterized in that: The digital twin model performs simulation based on real-time data streams and adjusts the behavior of the physical system through model predictive control, and the calculation formula of the model predictive control is as follows: Among them, u is the optimal control input, y k is the system output, y ref,k is the reference trajectory, Q and R are weight matrices, and N is the prediction time domain length.

Images