Data fusion method and system for multi-source microgrid

Abstract

The present invention discloses a data fusion method and system for a multi-source microgrid, which belongs to the field of microgrid technology, and specifically includes: through layered data collection and preprocessing, using a deep learning enhanced spatiotemporal data fusion model and a dynamic adaptive strategy, intelligently aggregating multi-source operation and maintenance data; formulating an intelligent data distribution strategy, dynamically optimizing the data transmission path and bandwidth resource allocation; integrating an AI-based real-time monitoring system to monitor the microgrid status, and once an abnormality or potential risk is found, automatically warning and feeding back to the dispatching system and operators through a visual interface, thereby improving the microgrid operation and maintenance efficiency and safety.

Description

Technical Field

[0001] The present invention belongs to the technical field of microgrids, and in particular to a data fusion method and system for multi-source microgrids. Background Art

[0002] A microgrid is a new type of small-scale power generation and distribution network, which consists of distributed power sources, energy storage devices, energy conversion devices, related loads, and monitoring and protection devices to form a single controllable power generation and distribution system. The multiple distributed power sources and related loads of the microgrid form a network according to a certain topological structure and are connected to the conventional power grid through static switches. Compared with traditional large power grids, microgrids can achieve self-control, protection and management of the system, and can operate independently or in parallel with external power grids.

[0003] With the rapid development of distributed energy and Internet of Things technologies, microgrids, as an important part of smart grids, contain a variety of power generation units, energy storage devices and load nodes, generating a large amount of heterogeneous data with high real-time requirements. How to effectively integrate this data has become a key issue in the optimization and intelligent control of microgrids. In the existing technology, data fusion methods often have problems such as low processing efficiency, insufficient accuracy, and poor real-time performance, which makes it difficult to meet the complex and changeable operation requirements of microgrids.

[0004] For example, a Chinese patent application with publication number CN115688444A discloses a microgrid state estimation method and system for multi-source measurement terminal data fusion, including: by acquiring SCADA measurement sequence data and PMU measurement sequence data of the SCADA measurement terminal and the PMU measurement terminal at the corresponding preset measurement points respectively, and also acquiring AMI measurement sequence data of the preset measurement point of the AMI measurement terminal at the end of the microgrid, based on the network topology of the microgrid, searching for multiple preset measurement points of the SCADA measurement terminal associated with the AMI measurement terminal to form a SCADA measurement point set, and determining the SCADA measurement point with the closest electrical distance to the AMI measurement terminal as the SCADA measurement point most associated with the AMI measurement terminal, estimating the AMI current measurement value according to the SCADA measurement sequence data corresponding to the SCADA measurement point most associated with the AMI measurement terminal and the AMI measurement sequence data, and performing microgrid state estimation according to the SCADA measurement sequence data, the PMU measurement sequence data and the AMI current measurement value, thereby improving the calculation accuracy of the microgrid state estimation.

[0005] The above existing technologies all have the following problems: 1) strong dependence and insufficient flexibility; 2) the data processing and fusion methods are relatively traditional, resulting in certain limitations in processing complex, nonlinear, and high-dimensional microgrid data; 3) although the fusion of multi-source data is considered, there are deficiencies in real-time and dynamic adaptability. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention proposes a data fusion method and system for multi-source microgrids. Through layered data acquisition and preprocessing, a deep learning-enhanced spatiotemporal data fusion model and a dynamic adaptive strategy are used to intelligently aggregate multi-source operation and maintenance data. An intelligent data distribution strategy is formulated to dynamically optimize data transmission paths and bandwidth resource allocation. A real-time monitoring system with integrated AI monitors the status of the microgrid. Once an abnormality or potential risk is found, an automatic warning is issued and feedback is given to the dispatching system and operators through a visual interface, thereby improving the operation and maintenance efficiency and safety of the microgrid.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The data fusion method for multi-source microgrid includes:

[0009] Step S1: collect hierarchical data of the microgrid through the intelligent sensor network and perform preprocessing;

[0010] Step S2: input the preprocessed hierarchical data into the pre-trained deep learning enhanced spatiotemporal data fusion model, and perform intelligent aggregation through a dynamic adaptive strategy to obtain multi-source operation and maintenance data;

[0011] Step S3: Formulate an intelligent data distribution strategy, and dynamically allocate multi-source operation and maintenance data transmission paths and bandwidth resources according to the intelligent data distribution strategy;

[0012] Step S4: Utilize multi-source operation and maintenance data to establish a real-time monitoring system with integrated AI to monitor the operating status of the microgrid in real time. When an abnormality or potential risk is detected, the early warning mechanism is automatically triggered, and the early warning results are fed back to the dispatching system and operators in a visual manner. At the same time, the operators take response measures based on the feedback information.

[0013] Specifically, the specific steps of step S2 include:

[0014] S2.1: Dynamically load the pre-trained deep learning enhanced spatiotemporal data fusion model, input the pre-processed layered data into the spatiotemporal data fusion model in real time, and use a dynamic adaptive strategy to adjust the parameters of the spatiotemporal data fusion model according to the real-time changes of the layered data. The specific formula of the dynamic adaptive strategy is:

[0015] ;

[0016] in, represents the parameter value of parameter group g at time step t+1, represents the parameter value of parameter group g at time step t, where g represents the parameter group of the spatiotemporal data fusion model. represents the global learning rate, represents the deviation correction value of the first-order moment estimate of parameter group g at time step t, represents the deviation correction value of the second-order moment estimate of parameter group g at time step t, Represents a positive number, Indicates that Crop to The gradient clipping function of the interval, represents the gradient of the parameter group g at time step t, Indicates the maximum absolute value of all gradient elements in the current batch. , Respectively represent the gradient Normalized minimum and maximum values;

[0017] S2.2: In the spatiotemporal data fusion model, the multi-dimensional attention fusion mechanism is used to perform feature fusion and obtain feature fusion data;

[0018] S2.3: Intelligently aggregate feature fusion data to generate multi-source operation and maintenance data, and evaluate the generated operation and maintenance data.

[0019] Specifically, the specific steps of S2.3 include:

[0020] S2.31: Obtain the feature fusion data in S2.2, load the pre-built convolutional neural network model, and input the feature fusion data into the convolutional neural network model;

[0021] S2.32: Use the improved gradient descent optimization algorithm to adjust the convolutional neural network model parameters. The formula is:

[0022] ;

[0023] in, and Respectively represent the convolutional neural network model parameter values ​​at time t+1 and time t, represents the momentum term at time t, and represents the hyperparameter that adjusts the effect of the momentum term, represents the learning rate adjustment factor, Represents the gradient vector at time t;

[0024] S2.33: The output of the adjusted convolutional neural network model is used to obtain the result of intelligent aggregation. Based on the result of intelligent aggregation and combined with business needs, multi-source operation and maintenance data is generated, and the quality of the generated operation and maintenance data is evaluated by calculating the accuracy.

[0025] Specifically, the specific steps of step S3 include:

[0026] S3.1: Develop intelligent data distribution strategies and dynamically adjust the distribution strategies based on historical data and real-time network status;

[0027] S3.2: Monitor the network status and bandwidth usage in real time, and select the optimal transmission path using the routing algorithm based on the adjusted intelligent data distribution strategy and real-time network status. The formula is:

[0028] ;

[0029] in, represents the comprehensive performance of the jth transmission path, represents the transmission delay of the jth transmission path, represents the bandwidth utilization of the jth transmission path, represents the packet loss rate of the jth transmission path, represents the energy consumption cost of the jth transmission path, represents the stability of the jth transmission path, , , , , u represents the weight factor;

[0030] S3.3: Based on the selected optimal transmission path, a dynamic bandwidth allocation algorithm is used to allocate bandwidth resources;

[0031] S3.4: Deploy the developed intelligent data distribution strategy and dynamic allocation mechanism into the actual environment and monitor the data transmission process and resource usage in real time;

[0032] S3.5: Collect feedback data, and optimize and adjust the distribution strategy and dynamic allocation mechanism based on the collected feedback data.

[0033] Specifically, the specific steps of S3.3 include:

[0034] S3.31: Set token bucket parameters and dynamically adjust token bucket parameters based on the bandwidth capacity of the optimal transmission path;

[0035] S3.32: When packet X arrives:

[0036] like , then packet X is allowed to be sent and removed from the token bucket Tokens;

[0037] like , then the data packet X is queued and waits, where Indicates the number of tokens in the token bucket;

[0038] S3.33: Monitor the effect of network bandwidth allocation in real time, and dynamically adjust the parameters of the token bucket and the dynamic bandwidth allocation algorithm according to the monitoring results and the real-time network status.

[0039] Specifically, the intelligent data distribution strategy in S3.1 includes the priority, schedule and distribution method of data distribution.

[0040] The data fusion system for multi-source microgrid includes: data acquisition module, intelligent aggregation module, data distribution module and decision support module;

[0041] The data acquisition module is used to collect the power quality, equipment status and environmental parameter data of the microgrid in real time through smart sensors deployed at different levels of the microgrid, and perform preprocessing;

[0042] The intelligent aggregation module is used to fuse the pre-processed data using a pre-trained spatiotemporal data fusion model, and dynamically adjust the weights and parameters of the data fusion according to the real-time changes of the layered data and the prediction results of the model;

[0043] The data distribution module is used to formulate an intelligent data distribution strategy according to network conditions and data types, and to perform dynamic resource allocation according to the intelligent data distribution strategy;

[0044] The decision support module evaluates the state of the microgrid based on the intelligently aggregated data and provides optimization operation suggestions to operators.

[0045] Specifically, the intelligent aggregation module includes: a dynamic adaptive unit and an intelligent aggregation unit;

[0046] The dynamic adaptive unit is used to dynamically adjust the weights and parameters of data fusion according to the real-time changes of the data and the prediction results of the model by using an improved gradient descent optimization algorithm to improve the accuracy and robustness of data fusion;

[0047] The intelligent aggregation unit is used to perform intelligent aggregation processing on the fused layered data to form multi-source operation and maintenance data.

[0048] Specifically, the data distribution module includes: an intelligent distribution strategy unit and a resource allocation unit;

[0049] The intelligent distribution strategy unit is used to formulate an intelligent data distribution strategy based on data priority, urgency and network status factors;

[0050] The resource allocation unit is used to dynamically allocate data transmission paths and bandwidth resources using a dynamic bandwidth allocation algorithm according to an intelligent data distribution strategy.

[0051] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a data fusion method for a multi-source microgrid when executing the computer program.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. The present invention proposes a data fusion system for a multi-source microgrid, and optimizes and improves the architecture, operation steps and processes. The system has the advantages of simple process, low investment and operation costs, and low production work costs.

[0054] 2. The present invention proposes a data fusion method for multi-source microgrids. Through hierarchical data collection and preprocessing of intelligent sensor networks, combined with deep learning enhanced spatiotemporal data fusion models and dynamic adaptive strategies, intelligent aggregation of multi-source operation and maintenance data is achieved, thereby improving the real-time and accuracy of the microgrid energy dispatching system; through intelligent data distribution strategies, data transmission paths and bandwidth resources are dynamically optimized to ensure the timely transmission and quality of operation and maintenance data, and reduce the transmission of redundant information; at the same time, the integrated AI real-time monitoring system established can monitor the status of the microgrid in real time, promptly discover and warn of abnormalities or potential risks, and help the dispatching system and operators respond quickly through a visual feedback mechanism; the intelligent dispatching decision support system and real-time feedback mechanism realize efficient and accurate energy dispatching, effectively improving the safe operation level and emergency response capability of the microgrid. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic diagram of a data fusion method for a multi-source microgrid according to the present invention;

[0056] Figure 2 This is a principle flow chart of the data fusion method for multi-source microgrids of the present invention;

[0057] Figure 3 It is a multi-source operation and maintenance data allocation and transmission flow chart of the data fusion method for multi-source microgrid of the present invention;

[0058] Figure 4 This is a diagram of the data fusion system architecture for a multi-source microgrid according to the present invention. DETAILED DESCRIPTION

[0059] Example 1

[0060] See also Figure 1-Figure 3 , an embodiment of the present invention provides: a data fusion method for a multi-source microgrid, comprising the following steps:

[0061] Step S1: collect hierarchical data of the microgrid through the intelligent sensor network and perform preprocessing;

[0062] Among them, the hierarchical data includes power quality data, equipment status data, and environmental parameters.

[0063] Furthermore, the specific steps of step S1 include:

[0064] S1.1: Determine the key levels and areas that need to be monitored in the microgrid, such as power generation, transmission, distribution, and user side, and select temperature sensors, current sensors, voltage sensors, and vibration sensors according to monitoring requirements. At the same time, plan the topology of the sensor network to ensure that data can be transmitted efficiently and reliably;

[0065] S1.2: Deploy smart sensor nodes at key locations of the microgrid, configure the parameters of the sensor nodes, such as sampling frequency, data transmission mode, and communication protocol, ensure that the sensor nodes can work stably, and establish communication connections with edge computing nodes or central processing systems. The data transmission mode refers to wired / wireless.

[0066] S1.3: The sensor nodes collect the operation data of each level of the microgrid in real time according to the preset sampling frequency and monitoring range. During the data collection process, the sensor nodes perform preliminary data verification and filtering to ensure the accuracy and reliability of the data, and transmit the collected data to the edge computing node or the central processing system through the communication network;

[0067] S1.4: In the edge computing node or the central processing system, the received original hierarchical data is preprocessed, wherein the preprocessing steps include data cleaning, data compression, and data conversion. Furthermore, data cleaning includes removing noise and outliers, data compression includes reducing storage and transmission costs, and data conversion includes converting analog signals into digital signals.

[0068] Step S2: input the preprocessed hierarchical data into the pre-trained deep learning enhanced spatiotemporal data fusion model, and perform intelligent aggregation through a dynamic adaptive strategy to obtain multi-source operation and maintenance data;

[0069] Step S3: Formulate an intelligent data distribution strategy, and dynamically allocate multi-source operation and maintenance data transmission paths and bandwidth resources according to the intelligent data distribution strategy;

[0070] Step S4: Utilize multi-source operation and maintenance data to establish a real-time monitoring system integrating AI to monitor the operating status of the microgrid in real time. When an abnormality or potential risk is detected, the early warning mechanism is automatically triggered, and the early warning results are fed back to the dispatching system and operators in a visual manner. At the same time, the operators take response measures based on the feedback information. The specific process of establishing a real-time monitoring system integrating AI based on known data is the prior art content in this field, and is not the creative solution of this application, so it will not be repeated here.

[0071] Furthermore, response measures include: emergency fault handling, load adjustment and optimization, energy scheduling and balancing, safety isolation and recovery, and information communication and coordination.

[0072] The specific steps of step S2 include:

[0073] S2.1: Dynamically load the pre-trained deep learning enhanced spatiotemporal data fusion model, input the pre-processed layered data into the spatiotemporal data fusion model in real time, and use a dynamic adaptive strategy to adjust the parameters of the spatiotemporal data fusion model according to the real-time changes of the layered data. The specific formula of the dynamic adaptive strategy is:

[0074] ;

[0075] in, represents the parameter value of parameter group g at time step t+1, represents the parameter value of parameter group g at time step t, where g represents the parameter group of the spatiotemporal data fusion model. represents the global learning rate, represents the deviation correction value of the first-order moment estimate of parameter group g at time step t, represents the deviation correction value of the second-order moment estimate of parameter group g at time step t, Represents a positive number, Indicates that Crop to The gradient clipping function of the interval, represents the gradient of the parameter group g at time step t, Indicates the maximum absolute value of all gradient elements in the current batch. , Respectively represent the gradient Normalized minimum and maximum values;

[0076] It should be noted that the present invention uses the gradient clipping function based on the existing calculation formula Clipping the gradient helps prevent the gradient explosion problem, that is, when the gradient is large, it will cause the parameter update to be too large, thus destroying the stability of the model; introducing a global learning rate that changes over time , allowing the learning rate to be adjusted as needed during the training process. For example, the learning rate is gradually reduced through the annealing strategy to achieve more detailed parameter adjustment in the later stage of training. The dynamic learning rate helps the model to converge quickly in the early stage of training and avoid overfitting in the later stage. Different parameter groups are represented by the index g, which implies that different optimization strategies can be applied to different parameter groups. Therefore, the modified formula improves the stability, flexibility and effectiveness of the optimizer by introducing improvements in gradient clipping, dynamic learning rate and parameter grouping, which helps to achieve better training results on complex problems.

[0077] Furthermore, the construction process of the deep learning enhanced spatiotemporal data fusion model includes:

[0078] S2.11: Collect spatiotemporal data from different sources, such as sensor data, satellite images, and social media data, and perform preprocessing steps such as cleaning, denoising, missing value processing, and standardization on the collected spatiotemporal data to ensure data quality. At the same time, perform spatial registration and temporal synchronization as needed to eliminate differences between different data sources;

[0079] S2.12: Extract features from the preprocessed spatiotemporal data, where the features include temporal features and spatial features. The temporal features include timestamps and time intervals, and the spatial features include geographic locations and spatial distributions.

[0080] S2.13: Design the RNN deep learning model architecture, configure the number of network layers and parameters, introduce the attention mechanism and design the feature fusion strategy, wherein the introduction of the attention mechanism is to enable the model to focus on the more important parts of the data, and the feature fusion strategy adopts a weighted fusion method to effectively integrate features from different sources. However, the weighted fusion calculation formula is the prior art content in this field and is not the inventive solution of this application, and will not be elaborated here;

[0081] S2.14: Use the labeled spatiotemporal data to train the RNN model, optimize the model parameters through the back propagation algorithm, and use the cross-validation method to prevent overfitting. At the same time, adjust the hyperparameters of the RNN model as needed, such as learning rate, batch size, and number of iterations, to obtain a deep learning enhanced spatiotemporal data fusion model, wherein the back propagation algorithm and the cross-validation method are both prior art contents in this field, and are not the inventive solutions of the present application, and are not described in detail here;

[0082] S2.15: Use the test set to evaluate the accuracy of the deep learning enhanced spatiotemporal data fusion model, and tune the deep learning enhanced spatiotemporal data fusion model based on the evaluation results, such as adjusting the network structure and optimizing the feature fusion strategy;

[0083] S2.16: Deploy the trained deep learning enhanced spatiotemporal data fusion model to actual application scenarios. At the same time, monitor the performance of the deep learning enhanced spatiotemporal data fusion model and maintain and update it according to actual conditions.

[0084] S2.2: In the spatiotemporal data fusion model, the multi-dimensional attention fusion mechanism is used to perform feature fusion and obtain feature fusion data;

[0085] Furthermore, the specific steps of S2.2 include:

[0086] S2.21: Perform preprocessing operations such as cleaning, normalization, and standardization on the spatiotemporal data to ensure the consistency and comparability of the data, and use a feature extraction method based on a convolutional neural network to extract features of the spatiotemporal data, such as timestamp, spatial position, and attribute value. The calculation formula of the feature extraction method based on the convolutional neural network is the prior art content in this field, which is not the inventive solution of the present application and is not described in detail here;

[0087] S2.22: Design a hybrid attention mechanism to capture the interactions and dependencies between different dimensions, including temporal, spatial, and attribute dimensions, and calculate the importance weight of each feature;

[0088] S2.23: Weighted fusion of features in time, space, and attribute dimensions based on attention weights.

[0089] S2.3: Intelligently aggregate feature fusion data to generate multi-source operation and maintenance data, and evaluate the generated operation and maintenance data.

[0090] The specific steps of S2.3 include:

[0091] S2.31: Obtain the feature fusion data in S2.2, load the pre-built convolutional neural network model, and input the feature fusion data into the convolutional neural network model;

[0092] S2.32: Use the improved gradient descent optimization algorithm to adjust the convolutional neural network model parameters. The formula is:

[0093] ;

[0094] in, and Respectively represent the convolutional neural network model parameter values ​​at time t+1 and time t, represents the momentum term at time t, and represents the hyperparameter that adjusts the effect of the momentum term, represents the learning rate adjustment factor, Represents the gradient vector at time t;

[0095] It should be noted that in the present invention, by In the form of, the learning rate will be adaptively adjusted according to the size of the momentum term. When the momentum term is large, the learning rate will be relatively small, which helps to slow down the update speed and improve stability when approaching the optimal solution; moreover, the momentum term It helps to accelerate convergence in the relevant direction and reduce oscillation. The momentum term accumulates past gradient information, making the update direction more stable. , the gradient term can be scaled to suit different learning needs.

[0096] S2.33: The output of the adjusted convolutional neural network model is used to obtain the result of intelligent aggregation. Based on the result of intelligent aggregation and combined with business needs, multi-source operation and maintenance data is generated, and the quality of the generated operation and maintenance data is evaluated by calculating the accuracy.

[0097] The specific steps of step S3 include:

[0098] S3.1: Develop intelligent data distribution strategies and dynamically adjust the distribution strategies based on historical data and real-time network status;

[0099] Furthermore, the specific steps of S3.1 include:

[0100] S3.11: Based on the results of demand analysis, design a preliminary data distribution strategy, including distribution time, distribution volume, and distribution method;

[0101] S3.12: Based on the characteristics of historical data and business needs, a time series prediction model is constructed, and the time series prediction model is trained using the historical data to obtain a trained time series prediction model, wherein the construction and training process of the time series prediction model are both prior art contents in this field, and are not the inventive solutions of the present application, and are not described in detail here;

[0102] S3.13: Collect network status data and data source status data in real time, and input the real-time data into the trained time series prediction model to predict the network status and data source status in the future;

[0103] S3.14: Dynamically adjust the distribution strategy based on the prediction results, such as adjusting the distribution priority, changing the distribution path, increasing or decreasing the distribution volume;

[0104] S3.15: Deploy the adjusted distribution strategy to the actual environment and start implementing data distribution.

[0105] S3.2: Monitor the network status and bandwidth usage in real time, and select the optimal transmission path using the routing algorithm based on the adjusted intelligent data distribution strategy and real-time network status. The formula is:

[0106] ;

[0107] in, represents the comprehensive performance of the jth transmission path, represents the transmission delay of the jth transmission path, Indicates the bandwidth utilization of the jth transmission path, indicating the bandwidth usage on the current path, represents the ratio of the remaining available bandwidth of the jth transmission path, represents the packet loss rate of the jth transmission path, represents the energy consumption cost of the jth transmission path, represents the stability of the jth transmission path, indicating the ability of the path to maintain connectivity and stable performance over a period of time. , , , , u represents the weight factor;

[0108] S3.3: Based on the selected optimal transmission path, a dynamic bandwidth allocation algorithm is used to allocate bandwidth resources;

[0109] S3.4: Deploy the developed intelligent data distribution strategy and dynamic allocation mechanism into the actual environment and monitor the data transmission process and resource usage in real time;

[0110] S3.5: Collect feedback data, and optimize and adjust the distribution strategy and dynamic allocation mechanism based on the collected feedback data, including adjusting the priority, schedule, distribution method, transmission path, and bandwidth allocation.

[0111] The specific steps of S3.3 include:

[0112] S3.31: Set token bucket parameters and dynamically adjust token bucket parameters based on the bandwidth capacity of the optimal transmission path;

[0113] S3.32: When packet X arrives:

[0114] like , then packet X is allowed to be sent and removed from the token bucket Tokens;

[0115] like , then the data packet X is queued and waits, where Indicates the number of tokens in the token bucket;

[0116] S3.33: Monitor the effect of network bandwidth allocation in real time, and dynamically adjust the parameters of the token bucket and the dynamic bandwidth allocation algorithm according to the monitoring results and the real-time network status.

[0117] The intelligent data distribution strategy in S3.1 includes the priority, schedule and distribution method of data distribution.

[0118] Example 2

[0119] See also Figure 4 Another embodiment provided by the present invention is a data fusion system for a multi-source microgrid, comprising:

[0120] Data collection module, intelligent aggregation module, data distribution module, decision support module;

[0121] The data acquisition module is used to collect the power quality, equipment status and environmental parameter data of the microgrid in real time and perform preprocessing through smart sensors deployed at different levels of the microgrid;

[0122] The intelligent aggregation module is used to fuse the pre-processed data using the pre-trained spatiotemporal data fusion model, and dynamically adjust the weights and parameters of data fusion according to the real-time changes of the layered data and the prediction results of the model;

[0123] Data distribution module, used to formulate intelligent data distribution strategy according to network conditions and data types, and to dynamically allocate resources according to the intelligent data distribution strategy;

[0124] The decision support module evaluates the microgrid status based on the intelligently aggregated data and provides operators with suggestions for optimal operation.

[0125] The intelligent aggregation module includes: a dynamic adaptive unit and an intelligent aggregation unit;

[0126] Dynamic adaptive unit, which is used to dynamically adjust the weights and parameters of data fusion according to the real-time changes of data and the prediction results of the model, using the improved gradient descent optimization algorithm to improve the accuracy and robustness of data fusion;

[0127] The intelligent aggregation unit is used to perform intelligent aggregation processing on the fused layered data to form multi-source operation and maintenance data.

[0128] The data distribution module includes: intelligent distribution strategy unit and resource allocation unit;

[0129] Intelligent distribution strategy unit, used to formulate intelligent data distribution strategy based on data priority, urgency and network status factors;

[0130] The resource allocation unit is used to dynamically allocate data transmission paths and bandwidth resources based on the intelligent data distribution strategy using a dynamic bandwidth allocation algorithm to ensure that key data can be transmitted first.

[0131] Example 3

[0132] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a data fusion method for a multi-source microgrid are implemented. For details, please refer to the above method embodiment, which will not be repeated here.

[0133] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention, and all of these are within the protection of the present invention.

Claims

1. A data fusion method for a multi-source microgrid, characterized in that: include: Step S1: collect hierarchical data of the microgrid through the intelligent sensor network and perform preprocessing; Step S2: input the preprocessed hierarchical data into the pre-trained deep learning enhanced spatiotemporal data fusion model, and perform intelligent aggregation through a dynamic adaptive strategy to obtain multi-source operation and maintenance data; Step S3: Formulate an intelligent data distribution strategy, and dynamically allocate multi-source operation and maintenance data transmission paths and bandwidth resources according to the intelligent data distribution strategy; Step S4: Using multi-source operation and maintenance data, establish a real-time monitoring system integrated with AI to monitor the operation status of the microgrid in real time. When an abnormality or potential risk is detected, the early warning mechanism is automatically triggered, and the early warning results are fed back to the dispatching system and operators in a visual way. At the same time, the operators take response measures based on the feedback information; The specific steps of step S2 include: S2.1: Dynamically load the pre-trained deep learning enhanced spatiotemporal data fusion model, input the pre-processed layered data into the spatiotemporal data fusion model in real time, and use a dynamic adaptive strategy to adjust the parameters of the spatiotemporal data fusion model according to the real-time changes of the layered data. The specific formula of the dynamic adaptive strategy is: ； in, represents the parameter value of parameter group g at time step t+1, represents the parameter value of parameter group g at time step t, where g represents the parameter group of the spatiotemporal data fusion model. represents the global learning rate, represents the deviation correction value of the first-order moment estimate of parameter group g at time step t, represents the deviation correction value of the second-order moment estimate of parameter group g at time step t, Represents a positive number, Indicates that Crop to The gradient clipping function of the interval, represents the gradient of the parameter group g at time step t, Indicates the maximum absolute value of all gradient elements in the current batch. , Respectively represent the gradient Normalized minimum and maximum values; S2.2: In the spatiotemporal data fusion model, the multi-dimensional attention fusion mechanism is used to perform feature fusion and obtain feature fusion data; S2.3: Intelligently aggregate feature fusion data to generate multi-source operation and maintenance data, and evaluate the generated operation and maintenance data; The specific steps of step S3 include: S3.1: Develop intelligent data distribution strategies and dynamically adjust the distribution strategies based on historical data and real-time network status; S3.2: Monitor the network status and bandwidth usage in real time, and select the optimal transmission path using the routing algorithm based on the adjusted intelligent data distribution strategy and real-time network status. The formula is: ； in, represents the comprehensive performance of the jth transmission path, represents the transmission delay of the jth transmission path, represents the bandwidth utilization of the jth transmission path, represents the packet loss rate of the jth transmission path, represents the energy consumption cost of the jth transmission path, represents the stability of the jth transmission path, , , , , u represents the weight factor; S3.3: Based on the selected optimal transmission path, a dynamic bandwidth allocation algorithm is used to allocate bandwidth resources; S3.4: Deploy the developed intelligent data distribution strategy and dynamic allocation mechanism into the actual environment and monitor the data transmission process and resource usage in real time; S3.5: Collect feedback data, and optimize and adjust the distribution strategy and dynamic allocation mechanism based on the collected feedback data.

2. The data fusion method for a multi-source microgrid according to claim 1, characterized in that: The specific steps of S2.3 include: S2.31: Obtain the feature fusion data in S2.2, load the pre-built convolutional neural network model, and input the feature fusion data into the convolutional neural network model; S2.32: Use the improved gradient descent optimization algorithm to adjust the convolutional neural network model parameters. The formula is: ； in, and Respectively represent the convolutional neural network model parameter values ​​at time t+1 and time t, represents the momentum term at time t, and represents the hyperparameter that adjusts the effect of the momentum term, represents the learning rate adjustment factor, Represents the gradient vector at time t; S2.33: The output of the adjusted convolutional neural network model is used to obtain the result of intelligent aggregation. Based on the result of intelligent aggregation and combined with business needs, multi-source operation and maintenance data is generated, and the quality of the generated operation and maintenance data is evaluated by calculating the accuracy.

3. The data fusion method for a multi-source microgrid according to claim 2, characterized in that: The specific steps of S3.3 include: S3.31: Set token bucket parameters and dynamically adjust token bucket parameters based on the bandwidth capacity of the optimal transmission path; S3.32: When packet X arrives: like , then packet X is allowed to be sent and removed from the token bucket Tokens; like , then the data packet X is queued and waits, where Indicates the number of tokens in the token bucket; S3.33: Monitor the effect of network bandwidth allocation in real time, and dynamically adjust the parameters of the token bucket and the dynamic bandwidth allocation algorithm according to the monitoring results and the real-time network status.

4. The data fusion method for a multi-source microgrid according to claim 3, characterized in that: The intelligent data distribution strategy in S3.1 includes the priority, schedule and distribution method of data distribution.

5. A data fusion system for a multi-source microgrid, which is used to implement the data fusion method for a multi-source microgrid according to any one of claims 1 to 4, characterized in that: include: Data collection module, intelligent aggregation module, data distribution module, decision support module; The data acquisition module is used to collect the power quality, equipment status and environmental parameter data of the microgrid in real time through smart sensors deployed at different levels of the microgrid, and perform preprocessing; The intelligent aggregation module is used to fuse the pre-processed data using a pre-trained spatiotemporal data fusion model, and dynamically adjust the weights and parameters of the data fusion according to the real-time changes of the layered data and the prediction results of the model; The data distribution module is used to formulate an intelligent data distribution strategy according to network conditions and data types, and to perform dynamic resource allocation according to the intelligent data distribution strategy; The decision support module evaluates the state of the microgrid based on the intelligently aggregated data and provides optimization operation suggestions to operators.

6. The data fusion system for multi-source microgrid according to claim 5, characterized in that: The intelligent aggregation module includes: a dynamic adaptive unit and an intelligent aggregation unit; The dynamic adaptive unit is used to dynamically adjust the weights and parameters of data fusion according to the real-time changes of the data and the prediction results of the model by using an improved gradient descent optimization algorithm to improve the accuracy and robustness of data fusion; The intelligent aggregation unit is used to perform intelligent aggregation processing on the fused layered data to form multi-source operation and maintenance data.

7. The data fusion system for multi-source microgrid according to claim 6, characterized in that: The data distribution module includes: an intelligent distribution strategy unit and a resource allocation unit; The intelligent distribution strategy unit is used to formulate an intelligent data distribution strategy based on data priority, urgency and network status factors; The resource allocation unit is used to dynamically allocate data transmission paths and bandwidth resources using a dynamic bandwidth allocation algorithm according to an intelligent data distribution strategy.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the data fusion method for a multi-source microgrid according to any one of claims 1 to 4 are implemented.

Images