Multi-source data comprehensive prediction processing method for power grid system

Abstract

The invention belongs to the field of power grid data prediction and processing methods, and particularly relates to a multi-source data comprehensive prediction processing method for a power grid system. A sensor device is deployed at each node of a power grid system, power grid operation data are collected in real time, verification and standardization processing are carried out, and global and local standardized data sets are formed. On this basis, a global guidance model and a local learning model are constructed, and efficient prediction of data is realized through knowledge transmission. And meanwhile, a trust function and a trust threshold mechanism are introduced, so that the reliability of prediction data is ensured. And finally, according to the characteristics and trends of the global and local data, making a corresponding decision rule, and realizing real-time monitoring and optimal control of the power grid system. According to the method, the accuracy and the real-time performance of power grid data prediction are improved, the adaptability to special conditions of a power grid system is enhanced through the combination of the global model and the local model, and powerful support is provided for efficient, stable and safe operation of a smart power grid.

Description

Technical Field

[0001] The present invention belongs to the field of power grid data prediction and processing methods, and specifically is a multi-source data comprehensive prediction and processing method for a power grid system. Background Art

[0002] With the rapid development of smart grid systems, real-time monitoring, analysis and prediction of grid operation data have become particularly important. The grid system is a complex and large system involving multiple links such as power generation, transmission, transformation, distribution and consumption, and each link generates a huge amount of data. These data include not only basic electrical parameters such as voltage, current and power, but also information on equipment status, environmental factors, user behavior and other aspects. How to effectively collect, process and analyze these data is crucial to ensure the safe, stable and efficient operation of the grid.

[0003] Existing power grid data monitoring and analysis technologies mostly use traditional statistical methods or single prediction models. These methods are effective when processing small-scale, single-source data, but they often have the following shortcomings when facing large-scale, multi-source, high-dimensional power grid data:

[0004] Traditional technologies often lack effective data collection and integration mechanisms, making it difficult to achieve comprehensive monitoring of each node in the power grid system, resulting in incomplete and inconsistent data.

[0005] A single prediction model is difficult to adapt to the complexity and variability of the power grid system, and it is difficult to accurately predict the operating status and development trend of the power grid.

[0006] Existing technologies often lack the ability to adapt to special situations in the power grid system, such as sudden load changes, equipment failures, etc., making it difficult to achieve rapid response and adjustment to the operating status of the power grid.

[0007] Traditional technologies have limited capabilities in decision support and are unable to provide scientific and reasonable decision-making basis for the optimization and control of power grid operations.

[0008] Existing technologies are not good enough in real-time monitoring and dynamic adjustment, and cannot meet the real-time and dynamic requirements of smart grids.

[0009] With the development of artificial intelligence and big data technologies, existing technologies still have much room for improvement in the intelligent and automated processing of power grid data. Summary of the invention

[0010] In view of the shortcomings of the prior art, the present invention proposes a comprehensive prediction and processing method for multi-source data of power grid systems. The method deploys sensor equipment at each node of the power grid system, collects power grid operation data in real time, and performs verification and standardization processing to form global and local standardized data sets. On this basis, a global guidance model and a local learning model are constructed to achieve efficient data prediction through knowledge transfer. At the same time, trust functions and trust threshold mechanisms are introduced to ensure the reliability of predicted data. Finally, according to the characteristics and trends of global and local data, corresponding decision rules are formulated to achieve real-time monitoring and optimization control of the power grid system. This method not only improves the accuracy and real-time performance of power grid data prediction, but also enhances the adaptability to special situations of the power grid system through the combination of global and local models, providing strong support for the efficient, stable and safe operation of smart grids.

[0011] The technical solution of the present invention is as follows:

[0012] A multi-source data comprehensive prediction and processing method for a power grid system, comprising:

[0013] At each grid node in the grid system, sensor equipment is used to collect the operation data of the grid system, and the operation data is processed to obtain a local standardized data set at each node. The standardized data sets at each node are accumulated and integrated to form a global standardized data set.

[0014] A global guidance model is built in the central server, and the global guidance model is used to learn the global standardized data to obtain the global data prediction capability; a local learning model is built at each node, and the global guidance model transfers the data to the local learning model at each node through knowledge transfer to predict the data at the node;

[0015] An initialized trust function is added to the global guidance model and each local learning model, and the parameters of the trust function are generated according to the global standardized data set and the local standardized data set respectively; at the same time, the global and local trust thresholds are set respectively, and the prediction data with trust values ​​greater than the global and local trust thresholds are determined as the global standard prediction data and the local standard prediction data;

[0016] Analyze the characteristics and trends of global standard forecast data and local standard forecast data, set decision goals, determine decision indicator values, and formulate global and local decision rules respectively.

[0017] Furthermore, the processing of the operation data includes: preliminary verification of the collected data, sorting of the operation data at each node, and integration and accumulation of local standardized data sets into a global standardized data set;

[0018] Among them, the preliminary verification includes range check and mutation check.

[0019] Furthermore, the global guidance model is constructed as follows:

[0020] The global normalized data set obtained in S100 is used as the input vector, and the input vector is X t =[V t ,I t ,PF t ,P t ,Q t ], where V t is the voltage value at time t, I t is the current value at time t, PF t is the power factor at time t, P t is the active power at time t, Q t is the reactive power at time t;

[0021] Update Gate:

[0022] z t =σ(W z ·[h t-1 ,X t ]+b z );

[0023] Reset the gate:

[0024] r t =σ(W r ·[h t-1 ,X t ]+b r );

[0025] Candidate hidden states:

[0026]

[0027] Hidden state:

[0028]

[0029] Guidance Model:

[0030]

[0031] Where: z t represents the value of the update gate at time t; σ represents the sigmoid function, which is used to map the input value to between; W z represents the weight matrix of the update gate; h t-1 Indicates the hidden state of the previous moment; X t represents the input vector; b z represents the bias term of the update gate; r t represents the value of the reset gate at time t; Wr represents the weight matrix of the reset gate; b r represents the bias term of the reset gate; represents the candidate hidden state at time t; W h The weight matrix representing the candidate hidden state; b h represents the bias term of the candidate hidden state; ⊙ represents the element-wise product operation; h t represents the final hidden state at time t; represents the predicted value of the global future time t+Δt; f represents the activation function of the output layer; W 0 represents the weight matrix of the output layer; b 0 Represents the bias term of the output layer.

[0032] Furthermore, the local learning model is:

[0033]

[0034] in, represents the predicted value at the local future time t+Δt; X(local)t represents the local standardized data set at each node; α, β represent weight coefficients; γ represents the bias term of the local output layer.

[0035] Furthermore, the trust function is:

[0036]

[0037] in, and are the predicted values ​​of the global guidance model and the local learning model respectively; and θ and θ′ are global and local adjustment parameters, respectively, μ and μ′ are the means calculated based on the global standardized dataset or the local standardized dataset, respectively;

[0038] Among them, the global and local adjustment parameters are generated based on the global standardized data set and the local standardized data set respectively.

[0039] The beneficial effects of the present invention are as follows: 1) By deploying sensor equipment at each key node of the power grid system, the present method can collect and verify power grid operation data in real time, ensuring the accuracy and availability of the data. This comprehensive data collection and integration mechanism not only enhances the real-time monitoring capability of the power grid status, but also lays a solid foundation for further data analysis and prediction.

[0040] 2) The present invention significantly improves the accuracy and adaptability of data prediction by constructing a global guidance model and a local learning model. The global guidance model uses a centralized data set for learning to form a powerful prediction capability, while the local learning model can quickly adjust and optimize according to the special circumstances of each node. This collaborative working mechanism of the model not only improves the accuracy of the prediction, but also enhances the adaptability and response speed to the dynamic changes of the power grid system.

[0041] 3) The present invention further improves the reliability of the prediction data by introducing a trust function and setting a trust threshold. On this basis, the global and local decision rules formulated provide a scientific basis for real-time monitoring and optimization control of the power grid system. This method not only optimizes the operating efficiency and stability of the power grid system, but also realizes the intelligence and automation of power grid monitoring through real-time feedback and adjustment, meeting the needs of smart grids for efficient, stable and safe operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The overall flow chart of the present invention is shown.

[0043] Figure 2 The specific implementation of S100 data processing is demonstrated.

[0044] Figure 3 The specific process of constructing the S200 global guidance model and local learning model is demonstrated.

[0045] Figure 4 The process of determining the S300 global standard prediction data and local standard prediction data is demonstrated.

[0046] Figure 5 The S400 global and local decision rule determination process is demonstrated. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] See also Figure 1 , which shows a multi-source data comprehensive prediction and processing method for a power grid system provided by an embodiment of the present invention, comprising:

[0049] S100: At each grid node in the grid system, operating data of the grid system is collected through sensor equipment, the operating data is processed to obtain a local standardized data set at each node, and the standardized data set at each node is accumulated and integrated to form a global standardized data set;

[0050] S200: construct a global guidance model in the central server, use the global guidance model to learn global standardized data to obtain global data prediction capabilities; construct a local learning model at each node, and the global guidance model transfers data to the local learning model at each node through knowledge transfer to perform node data prediction;

[0051] S300: attaching an initialized trust function to the global guidance model and each local learning model, generating parameters of the trust function according to the global standardized data set and the local standardized data set respectively; and setting global and local trust thresholds respectively, and determining the prediction data that are respectively greater than the global and local trust thresholds as the global standard prediction data and the local standard prediction data;

[0052] S400: Analyze the characteristics and trends of global standard prediction data and local standard prediction data, set decision goals, determine decision indicator values, and formulate global and local decision rules respectively.

[0053] S100 involves the collection, verification and integration of power grid system operation data. Through data verification to eliminate outliers, verify to ensure the accuracy and rationality of the data, and finally integrate it into a standardized data set that can be used for subsequent analysis and modeling, ensuring the accuracy and availability of the data, and laying a solid foundation for further analysis and prediction of the power grid system.

[0054] In some possible embodiments, Figure 2 As shown, S100 specifically includes:

[0055] S110: Each node in the power grid system continuously collects real-time data through the sensor devices deployed therein (such as voltage sensors, current sensors, power sensors, etc.). These data are recorded in the form of time series, including but not limited to voltage value, current value, power factor, active power, reactive power, etc.

[0056] S120: Perform preliminary verification on the collected data to remove obviously abnormal or unreasonable data points, including range check and mutation check.

[0057] In the range check, check whether the collected data is within the safe range. For example, for voltage V, check whether V min ≤V i ≤V max , where V minand V max is the voltage safety range boundary.

[0058] In the mutation check, the difference rate between adjacent data points is calculated to check whether the difference rate is less than the preset threshold. For example, for the voltage V at the i-th node i , If the voltage value suddenly jumps from 230V to 500V at a certain moment and exceeds the set mutation threshold, the data point is considered abnormal and is removed.

[0059] S130: Arrange the operation data at each node to form a local standardized data set at each node.

[0060] S140: Integrate and accumulate a local standardized data set formed at each node into a unified global standardized data set for further analysis and processing. Ensure that the data from different data sources are synchronized in terms of timestamps, and then convert the data from different data sources into a unified format, including data type, unit, etc. And associate the data from different sources according to specific keys (such as timestamps, device IDs, etc.), and accumulate and integrate them into a unified global standardized data set.

[0061] S200 forms a global guidance model by design, and uses this model to learn the global standardized data set to obtain data prediction capabilities. A local learning model is formed at each node, and the output data of the global guidance model is transferred to the local learning model at each node through knowledge transfer, so that they can make predictions independently.

[0062] In some possible embodiments, Figure 3 As shown, S200 specifically includes:

[0063] S210: Construct a global guidance model in the central server, train the global guidance model using a global standardized data set, and update the model weights.

[0064] The global guidance model is constructed as follows:

[0065] The global normalized data set obtained in S100 is used as the input vector, and the input vector is X t =[V t ,I t ,PF t ,P t ,Q t ], where V t is the voltage value at time t, I t is the current value at time t, PF t is the power factor at time t, P t is the active power at time t, Qt is the reactive power at time t;

[0066] Update Gate:

[0067] z t =σ(W z ·[h t-1 ,X t ]+b z );

[0068] Reset the gate:

[0069] r t =σ(W r ·[h t-1 ,X t ]+b r );

[0070] Candidate hidden states:

[0071]

[0072] Hidden state:

[0073]

[0074] Guidance Model:

[0075]

[0076] Where: z t represents the value of the update gate at time t; σ represents the sigmoid function, which is used to map the input value to between; W z represents the weight matrix of the update gate; h t-1 Indicates the hidden state of the previous moment; X t represents the input vector; b z represents the bias term of the update gate; r t represents the value of the reset gate at time t; W r represents the weight matrix of the reset gate; b r represents the bias term of the reset gate; represents the candidate hidden state at time t; W h The weight matrix representing the candidate hidden state; b h represents the bias term of the candidate hidden state; ⊙ represents the element-wise product operation; h t represents the final hidden state at time t; represents the predicted value of the global future time t+Δt; f represents the activation function of the output layer; W 0 represents the weight matrix of the output layer; b 0 Represents the bias term of the output layer;

[0077] S220: Build a local learning model:

[0078]

[0079] in, represents the predicted value at the local future time t+Δt; X(local)t represents the local standardized data set at each node; α, β represent weight coefficients; γ represents the bias term of the local output layer.

[0080] S230: After receiving the data output by the global guidance model, the local learning model at each node is trained using a local standardized data set. During the training process, the local learning model continuously adjusts its own parameters to minimize the prediction error and improve performance. Through knowledge transfer, the local learning model at each node can benefit from the powerful learning ability and knowledge of the guidance model, thereby converging faster and improving the prediction accuracy. At the same time, since the global guidance model can be fine-tuned according to the local learning model output data, they can also better adapt to the special circumstances and changes of the local power grid system.

[0081] S300 defines a trust function, calculates the mean based on the global standardized data set and the local standardized data set, and generates the parameters of the trust function according to the mean and the preset adjustment parameters. The global future prediction value and the local future moment prediction value are substituted into the trust function to calculate the trust value, and the trust value threshold is set, and the prediction data with a trust value greater than the threshold is determined as the standard global and local prediction data.

[0082] In some possible embodiments, Figure 4 As shown, S300 specifically includes:

[0083] S310: Define a trust function, which is defined as:

[0084]

[0085] in, and are the predicted values ​​of the global guidance model and the local learning model respectively; and θ and θ′ are the global and local adjustment parameters, respectively, and μ and μ′ are the means calculated based on the global standardized dataset or the local standardized dataset, respectively.

[0086] S320: Generate global and local adjustment parameters based on the global standardized data set and the local standardized data set, respectively, to adjust the shape and characteristics of the trust function.

[0087] S330: respectively setting a global trust value threshold T and a local trust threshold T′, and determining the prediction data whose trust values ​​are respectively greater than the thresholds T and T′ as the global standard prediction data and the local standard prediction data.

[0088] S400 includes analyzing the characteristics and trends of global and local data, setting decision goals, determining decision indicator values, and formulating corresponding decision rules.

[0089] In some possible embodiments, Figure 5 As shown, the specific contents of S400 include:

[0090] S410: Global decision rule formulation. Analyze the characteristics and trends of the global standard prediction data and set the goals of the global decision rules, such as optimizing the stability of the power grid system and improving energy efficiency. Determine the specific global decision indicator value based on the goals and data characteristics. For example, the voltage fluctuation range, power factor range, etc. can be set as the global decision indicator value. Decision rules are formulated based on the global decision indicator value. For example, if the voltage fluctuation in the global standard prediction data exceeds the set decision indicator value, corresponding control measures are taken, such as adjusting the transformer gear, switching capacitors, etc.

[0091] S420: Local decision rule formulation. Analyze the characteristics and requirements of local standard prediction data. Consider the special conditions of the local power grid, such as load distribution, line impedance, etc. Determine the goals of local decision rules based on local characteristics and needs, such as ensuring local power supply reliability, optimizing local energy distribution, etc. Formulate specific local decision indicator values. For example, for a local node, you can set a current overload threshold, power limit, etc. Formulate local decision rules through local decision indicator values. For example, if the current in the local standard prediction data exceeds the overload threshold, the local protection device is activated to cut off part of the load.

[0092] S430: Apply the formulated global and local decision rules to the actual power grid system. Monitor the operation status of the power grid system in real time and make corresponding controls and adjustments according to the decision rules. Collect feedback data after the decision is executed and evaluate the effect of the decision rules. Adjust and optimize the decision rules according to the feedback data and actual operation conditions to improve the performance and reliability of the power grid system.

[0093] In practical applications, these decision rules are applied to the power grid system to monitor the operating status in real time and make control adjustments. At the same time, the decision rules are evaluated and optimized based on feedback data to improve the performance and reliability of the power grid system. For example, measures such as adjusting the transformer gear are taken according to the voltage fluctuation in the global data, and the protection device is activated to cut off part of the load according to the current overload of the local node.

[0094] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0095] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0096] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0097] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features of the present invention.

[0098] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for comprehensive prediction and processing of multi-source data for a power grid system, characterized in that: include: At each grid node in the grid system, sensor equipment is used to collect the operation data of the grid system, the operation data is processed to obtain a local standardized data set at each node, and the local standardized data set at each node is accumulated and integrated to form a global standardized data set; A global guidance model is built in the central server, and the global guidance model is used to learn the global standardized data to obtain the global data prediction capability; a local learning model is built at each node, and the global guidance model transfers the data to the local learning model at each node through knowledge transfer to predict the data at the node; An initialized trust function is added to the global guidance model and each local learning model, and the parameters of the trust function are generated according to the global standardized data set and the local standardized data set respectively. The global future prediction value and the local future moment prediction value are substituted into the trust function to calculate the trust value; at the same time, the global and local trust thresholds are set respectively, and the prediction data with trust values ​​greater than the global and local trust thresholds are determined as the global standard prediction data and the local standard prediction data; Analyze the characteristics and trends of global standard forecast data and local standard forecast data, set decision goals, determine decision indicator values, and formulate global and local decision rules respectively.

2. A method for comprehensive prediction and processing of multi-source data for a power grid system according to claim 1, characterized in that: The processing of the operation data includes: preliminary verification of the collected data, sorting the operation data at each node, and integrating and accumulating the local standardized data sets into a global standardized data set; Among them, the preliminary verification includes range check and mutation check.

3. A method for comprehensive prediction and processing of multi-source data for a power grid system according to claim 1, characterized in that: The global guidance model is constructed as follows: The global standardized data set is used as the input vector, and the input vector is X t =[V t ,I t ,PF t ,P t ,Q t ], where V t is the voltage value at time t, I t is the current value at time t, PF t is the power factor at time t, P t is the active power at time t, Q t is the reactive power at time t; Update Gate: z t =σ(W z ·[h t-1 ,X t ]+b z ); Reset the gate: r t =σ(W r ·[h t-1 ,X t ]+b r ); Candidate hidden states: Hidden state: Guidance Model: Where: z t represents the value of the update gate at time t; σ represents the sigmoid function, which is used to map the input value to between; W z represents the weight matrix of the update gate; h t-1 Indicates the hidden state of the previous moment; X t represents the input vector; b z represents the bias term of the update gate; r t represents the value of the reset gate at time t; W r represents the weight matrix of the reset gate; b r represents the bias term of the reset gate; represents the candidate hidden state at time t; W h The weight matrix representing the candidate hidden state; b h represents the bias term of the candidate hidden state; ⊙ represents the element-wise product operation; h t represents the final hidden state at time t; represents the predicted value of the global future time t+Δt; f represents the activation function of the output layer; W0 represents the weight matrix of the output layer; b0 represents the bias term of the output layer.

4. A method for comprehensive prediction and processing of multi-source data for a power grid system according to claim 3, characterized in that: The local learning model is: in, represents the predicted value at the local future time t+Δt; X(local)t represents the local standardized data set at each node; α, β represent weight coefficients; γ represents the bias term of the local output layer.

5. A method for comprehensive prediction and processing of multi-source data for a power grid system according to claim 1, characterized in that: The trust function is: in, and are the predicted values ​​of the global guidance model and the local learning model respectively; and θ and θ′ are global and local adjustment parameters, respectively, μ and μ′ are the means calculated based on the global and local standardized datasets, respectively; Among them, the global and local adjustment parameters are generated based on the global standardized data set and the local standardized data set respectively.

Images