Digital emergency plan generation method based on multi-source heterogeneous data fusion

Abstract

The invention belongs to the technical field of data processing, and particularly relates to a digital emergency plan generation method based on multi-source heterogeneous data fusion. The method comprises the following steps: firstly collecting Internet of Things sensor data, historical disaster text data and meteorological satellite remote sensing data, processing various types of data by adopting different fusion rules and generating log data, and performing secondary fusion through a fourth fusion rule to form final fusion data. Threshold logic division is carried out by calculating the correlation degree of the fused data and the initial data, and the data quality is guaranteed. And finally, constructing a digital plan model by using an improved hierarchical reinforcement learning framework, and inputting final fusion data to generate an emergency plan. Compared with a traditional method, multi-source data are fused, the data quality and reliability are improved, the generated emergency plan is more scientific and accurate, the emergency response capability of the power grid is effectively improved, fault loss is reduced, and stable operation of the power grid is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a method for generating a digital emergency plan based on multi-source heterogeneous data fusion. Background Art

[0002] In today's society that is highly dependent on electricity supply, the stable operation of the power grid is of vital importance. Once a power grid fails, such as natural disasters, equipment aging failures, external force damage, etc., it will have a serious impact on social production and life, cause huge economic losses, and even threaten the safety of public life. Therefore, it is of great significance to quickly, scientifically and effectively formulate and implement emergency plans to ensure the safety of the power grid and restore power supply. The traditional way of formulating emergency plans for power grids mainly relies on experience and a single type of data. For example, plans are formulated based solely on historical power grid failure data or the subjective judgment of operation and maintenance personnel. However, this method has many limitations. On the one hand, experience is often subjective and one-sided, and it is difficult to fully consider the complex and changeable power grid failure scenarios; on the other hand, a single data source cannot provide enough rich information, making the emergency plan lack of pertinence and effectiveness. Summary of the invention

[0003] In view of the technical problems existing in the above-mentioned background technology, the present invention proposes a digital emergency plan generation method based on multi-source heterogeneous data fusion.

[0004] In order to achieve the above object, the technical solution adopted by the present invention comprises the following steps:

[0005] S1. First, three types of multi-source heterogeneous data are collected, including IoT sensor data, historical disaster text data, and meteorological satellite remote sensing data;

[0006] S2, using three fusion rules to process each type of data respectively, forming the first, second and third types of fusion data, and synchronously generating log data containing data source, processing timestamp and confidence;

[0007] S3, fusing the three types of fused data and the log data for a second time through the fourth fusion rule to generate final fused data;

[0008] S4. After the final fusion, the final fused data needs to be compared with the initial data to see the correlation between the fused data and the initial data, and perform threshold logic division. If it does not exceed the threshold, a data credibility report is generated and data fusion is performed again. Otherwise, the next step is performed; the calculation method of the correlation is: , where F is the final fusion data, is the original data of the i-th category, is the weight factor, is the empirical coefficient;

[0009] S5. Build a digital emergency plan model, input the final fusion data, and predict the final data emergency plan.

[0010] Preferably, the three fusion rules in step S2 are respectively the first fusion rule, the second fusion rule and the third fusion rule; wherein the first fusion rule is a specific step for processing IoT sensor data, including: performing time series analysis on the original data collected by the IoT sensor, establishing a time series prediction model, and predicting the data at the next moment; comparing the predicted data with the actual collected data, and calculating the error; correcting the actual collected data according to the error, and fusing the corrected data to form a first type of fused data.

[0011] Preferably, the second fusion rule is a specific step for processing historical disaster text data, including: using the text classification algorithm in natural language processing technology to classify the historical disaster text data, annotate the key information in the text, extract information related to power grid failures, construct a semantic knowledge graph, and associate entities and relationships in the text data; and fuse the historical disaster text data after classification, annotation and knowledge graph construction.

[0012] Preferably, the third fusion rule is a specific step for processing meteorological satellite remote sensing data, including: extracting multispectral features of meteorological satellite remote sensing data; establishing a multispectral feature library, matching the extracted features with standard features in the feature library; fusing the remote sensing data according to the matching results to form a third type of fused data.

[0013] Preferably, the specific steps of performing secondary fusion according to the fourth fusion rule in step S3 include: aligning the three types of fused data according to time and space dimensions; fusing the aligned data in combination with the confidence data in the log data to form a final fusion set.

[0014] Preferably, the fourth fusion rule is calculated as follows:

[0015] , where t represents the current timestamp, s represents the three-dimensional space matrix, is the feature matrix after the three types of fusion data are aligned in time and space, is the spatial weight coefficient, is the time-effect attenuation compensation function; where the dynamic weight factor is generated based on the log data , are the historical data quality value and the real-time data quality value respectively, and k is the adjustment factor.

[0016] As a preference, when the data difference When Representing any two features, the dynamic weight factor is modified and calculated as follows: ,in is the mean of the three types of data, is the standard deviation of the feature.

[0017] As a preferred method, the feature matrix after the three types of fusion data are aligned in time and space , it also requires a delay Greater than the delay threshold When the compensation calculation is triggered, we get .

[0018] Preferably, the step S5 of constructing the digital emergency plan model is implemented by improving the hierarchical reinforcement learning framework, specifically: at the environmental perception layer, multi-dimensional feature extraction is performed on the final fused data, the features are encoded as state vectors, and input into the task planning layer; the task planning layer uses a deep reinforcement learning algorithm based on the received state vector to generate emergency task plans of different priorities; the execution decision layer generates a specific digital emergency plan through the policy network according to the output of the task planning layer.

[0019] Compared with the prior art, the advantages and positive effects of the present invention are that, in data collection, it integrates IoT sensors, historical disaster texts and meteorological satellite remote sensing data to break the limitations of single data and provide more comprehensive information. During data processing, multiple fusion rules are combined with log data for secondary fusion, and data quality and reliability are improved through time and space alignment and confidence weighting. Correlation calculation and threshold logic division are also introduced to ensure the accuracy of data input to the model. The model is constructed using an improved hierarchical reinforcement learning framework, integrating multi-dimensional features, and combining historical cases with real-time scenarios to generate plans. These innovations make emergency plans more scientific and accurate, greatly improve the emergency response capabilities of the power grid, reduce failure losses, and ensure the stable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Figure 1 It is a schematic diagram of the structural flow of the present invention. DETAILED DESCRIPTION

[0021] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.

[0023] Embodiment: In modern society, the stable operation of the power grid is the key to ensuring the normal operation of social production and life. Once a power grid fails, it will cause a series of serious problems, such as stagnation of industrial production, inconvenience in residents' lives, and even endangerment to public safety. The traditional way of formulating power grid emergency plans that relies on experience and single data is difficult to cope with complex and changeable fault scenarios, and the formulated plans lack pertinence and effectiveness. In order to achieve the effect of quickly, scientifically and effectively formulating power grid emergency plans, comprehensively improving the emergency response capabilities of the power grid, and reducing the losses caused by faults, the present invention adopts a digital emergency plan generation method based on multi-source heterogeneous data fusion. The specific scheme is as follows: Figure 1 As shown:

[0024] First, three types of multi-source heterogeneous data are collected. IoT sensors distributed at key nodes of the power grid are used to collect power parameter data such as voltage, current, and power in real time. At the same time, historical disaster text data are collected. These data come from documents and reports recorded after natural disasters in the past, covering detailed information such as the damage to the power grid, fault type, and emergency repair process when different types of disasters (such as earthquakes, floods, typhoons, etc.) occur. In addition, meteorological satellite remote sensing data is obtained, including meteorological information such as cloud thickness, precipitation intensity, wind speed and direction. These data can help predict meteorological disasters that may affect the power grid.

[0025] Three fusion rules are used to process each type of data respectively, forming the first, second and third types of fused data, and synchronously generating log data containing data source, processing timestamp and confidence. The specific steps of the first fusion rule to process IoT sensor data include: time series analysis of the original data collected by IoT sensors, establishment of a time series prediction model, and prediction of the next moment data; comparison of predicted data with actual collected data, calculation of errors; correction of actual collected data according to the errors, and fusion of the corrected data to form the first type of fused data. The specific steps of the second fusion rule to process historical disaster text data include: using the text classification algorithm in natural language processing technology to classify historical disaster text data, annotate key information in the text, and extract information related to power grid failures; construct a semantic knowledge graph to associate entities and relationships in text data; and fuse historical disaster text data after classification and annotation and knowledge graph construction. The specific steps of the third fusion rule for processing meteorological satellite remote sensing data include: extracting multispectral features of meteorological satellite remote sensing data; establishing a multispectral feature library and matching the extracted features with the standard features in the feature library; fusing the remote sensing data according to the matching results to form the third type of fused data.

[0026] Then, the fourth fusion rule is used to perform secondary fusion of the three types of fused data with the log data, and the three types of fused data are aligned according to the time and space dimensions to ensure the consistency of the data in time and space. Combined with the confidence data in the log data, the aligned data is fused to form the final fused data. Using the above secondary fusion method, from the perspective of data integrity, the IoT sensor data, historical disaster text data and meteorological satellite remote sensing data are integrated to present the operation status of the power grid in an all-round way. For example, in the event of a typhoon disaster, the sensor data reflects the real-time power parameters, the historical text provides the impact of past typhoons on the power grid, and the remote sensing data monitors the typhoon path and intensity, and the information after fusion is more comprehensive. In addition, combined with the confidence of the log data, reliable information can be screened. If the confidence of a sensor data is low, its weight will be reduced during fusion to avoid interference from erroneous data. Moreover, aligning the data according to the time and space dimensions can accurately associate different data.

[0027] The specific implementation is as follows: First, the fourth fusion rule is to align the three types of fusion data according to the time and space dimensions; combine the confidence data in the log data, and fuse the aligned data to form the final fusion set. The calculation method of the fourth fusion rule is:

[0028] , where t represents the current timestamp, s represents the three-dimensional space matrix, is the feature matrix after the three types of fusion data are aligned in time and space, is the spatial weight coefficient, is the time-effect attenuation compensation function; where the dynamic weight factor is generated based on the log data , are the historical data quality value and the real-time data quality value respectively, and k is the adjustment factor. When Representing any two features, the dynamic weight factor is modified and calculated as follows: ,in is the mean of the three types of data, is the standard deviation of the feature. For the feature matrix after the three types of fusion data are aligned in time and space , it also requires a delay Greater than the delay threshold When the compensation calculation is triggered, we get .

[0029] After the final fusion, the quality and reliability of the fused data are taken into consideration. In the process of data fusion, various processing steps may introduce errors or change the original characteristics of the data. If the difference between the fused data and the initial data is too large, it may mean that there is a problem in the fusion process, resulting in data distortion and failure to accurately reflect the actual situation. Therefore, it is necessary to compare and judge the final fused data with the initial data to see the correlation between the fused data and the initial data, and perform threshold logic division. If the threshold is not exceeded, a data credibility report is generated and data fusion is performed again. Otherwise, the next step is performed; the calculation method of the correlation is: , where F is the final fusion data, is the original data of the i-th category, is the weight factor, It is an empirical coefficient, which ensures the quality of data input into the subsequent digital emergency plan model.

[0030] Finally, a digital emergency plan model is constructed by improving the hierarchical reinforcement learning framework. In the environmental perception layer, multi-dimensional feature extraction is performed on the final fused data, and features such as power parameters, fault locations, meteorological conditions, and historical disaster experiences are encoded as state vectors and input into the task planning layer. Based on the received state vector, the task planning layer uses a deep reinforcement learning algorithm, combined with historical emergency case data and current scenario information, to generate emergency task plans of different priorities. For example, if it is detected that a power grid failure in a certain area is caused by severe weather and the impact range is large, the task planning layer will give priority to planning the deployment of emergency repair personnel and material transportation tasks, and determine a reasonable time window. Based on the output of the task planning layer, the execution decision layer generates a specific digital emergency plan through the policy network, including detailed emergency repair steps, personnel division of labor, material deployment plan, etc.

[0031] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A digital emergency plan generation method based on multi-source heterogeneous data fusion, characterized in that: The following steps are involved: S1. First, three types of multi-source heterogeneous data are collected, including IoT sensor data, historical disaster text data, and meteorological satellite remote sensing data; S2, using three fusion rules to process each type of data respectively, forming the first, second and third types of fusion data, and synchronously generating log data containing data source, processing timestamp and confidence; S3, fusing the three types of fused data and the log data for a second time through the fourth fusion rule to generate final fused data; S4. After the final fusion, the final fused data needs to be compared with the initial data to see the correlation between the fused data and the initial data, and perform threshold logic division. If it does not exceed the threshold, a data credibility report is generated and data fusion is performed again. Otherwise, the next step is performed; The calculation method of the correlation degree is: , where F is the final fusion data, is the original data of the i-th category, is the weight factor, is the empirical coefficient; S5. Build a digital emergency plan model, input the final fusion data, and predict the final data emergency plan.

2. According to claim 1, a method for generating a digital emergency plan based on multi-source heterogeneous data fusion is characterized in that: The three fusion rules in step S2 are the first fusion rule, the second fusion rule and the third fusion rule; The first fusion rule is a specific step of processing IoT sensor data including: Perform time series analysis on the raw data collected by IoT sensors, establish a time series prediction model, and predict the data at the next moment; Compare the predicted data with the actual collected data and calculate the error; The actual collected data is corrected according to the error, and the corrected data is fused to form the first type of fused data.

3. A method for generating a digital emergency plan based on multi-source heterogeneous data fusion according to claim 2, characterized in that: The specific steps of the second fusion rule for processing historical disaster text data include: The text classification algorithm in natural language processing technology is used to classify historical disaster text data, annotate key information in the text, and extract information related to power grid failures. Build a semantic knowledge graph to associate entities and relationships in text data; The historical disaster text data is integrated after classification and annotation and knowledge graph construction.

4. The method for generating a digital emergency plan based on multi-source heterogeneous data fusion according to claim 2 is characterized in that: The third fusion rule is a specific step of processing meteorological satellite remote sensing data, including: Extract multispectral features of meteorological satellite remote sensing data; Establish a multispectral feature library and match the extracted features with the standard features in the feature library; The remote sensing data are fused according to the matching results to form the third type of fused data.

5. The method for generating a digital emergency plan based on multi-source heterogeneous data fusion according to claim 1 is characterized in that: The specific steps of performing secondary fusion according to the fourth fusion rule in step S3 include: aligning the three types of fused data according to time and space dimensions; combining the confidence data in the log data, fusing the aligned data to form a final fusion set.

6. A method for generating a digital emergency plan based on multi-source heterogeneous data fusion according to claim 5, characterized in that: The calculation method of the fourth fusion rule is: , where t represents the current timestamp and s represents the three-dimensional space matrix. is the feature matrix after the three types of fusion data are aligned in time and space, is the spatial weight coefficient, is the time-effect attenuation compensation function; Among them, dynamic weight factors are generated based on log data , are the historical data quality value and the real-time data quality value respectively, and k is the adjustment factor.

7. A method for generating a digital emergency plan based on multi-source heterogeneous data fusion according to claim 6, characterized in that: When the data difference When Representing any two features, the dynamic weight factor is modified and calculated as follows: ,in is the mean of the three types of data, is the standard deviation of the feature.

8. The method for generating a digital emergency plan based on multi-source heterogeneous data fusion according to claim 6 is characterized in that: For the feature matrix of the three types of fusion data after spatiotemporal alignment , it also requires a delay Greater than the delay threshold When the compensation calculation is triggered, we get .

9. The method for generating a digital emergency plan based on multi-source heterogeneous data fusion according to claim 1 is characterized in that: The step S5 of constructing the digital emergency plan model is achieved by improving the hierarchical reinforcement learning framework, specifically: In the environment perception layer, multi-dimensional features are extracted from the final fused data, the features are encoded into state vectors, and input into the task planning layer; The task planning layer uses a deep reinforcement learning algorithm based on the received state vector to generate emergency task plans of different priorities; The executive decision layer generates a specific digital emergency plan through the strategy network based on the output of the task planning layer.

Images