Dynamic deduction method and system for distribution network structure based on time series dynamic analysis

Abstract

The present application relates to the technical field of power systems, and provides a method and system for dynamic deduction of distribution network structure based on time series dynamic analysis, which adopts a data processing algorithm based on deep learning to embed coding and principal component time series aggregation of the time queues of power generation side data, grid side data and user side data respectively, so as to intelligently obtain the short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation between the principal component time series aggregation representation of power generation side data features, the principal component time series aggregation representation of grid side data features and the principal component time series aggregation representation of user side data features. In this way, by integrating multi-source data from the power generation side, the grid side and the user side, the complex time series dynamic relationship in the distribution network can be captured more accurately, making the prediction of carbon emissions more accurate, thereby significantly improving the real-time, comprehensiveness and intelligence of carbon emission prediction.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and more specifically, to a method and system for dynamically deducing distribution network structure based on time series dynamic analysis. Background Art

[0002] Distribution networks play a crucial role in the power system, transmitting high-voltage electricity generated by power plants to the low-voltage grid for ultimate distribution to households and businesses. In the power industry, managing carbon emissions is crucial for optimizing the allocation of power generation resources and reducing environmental impact. Accurately forecasting carbon emissions allows for more efficient planning of power generation resources, particularly increasing the proportion of renewable energy, thereby promoting the sustainable use of green energy.

[0003] Traditional predictions of carbon emissions from distribution networks often only consider data from a single source, such as power generation or electricity consumption. However, modern power systems involve data from multiple levels, including the generation side, the grid side, and the user side. There are complex interactions between these data. For example, renewable energy output on the generation side is affected by weather conditions, while electricity consumption on the user side fluctuates over time and seasons. Traditional prediction methods fail to fully utilize the comprehensive information from these multi-source data, resulting in incomplete and inaccurate prediction results. In addition, traditional methods are usually based on static analysis of historical data and cannot reflect current dynamic changes in real time. The supply and demand relationship and operating status of the distribution network are constantly changing, and traditional prediction methods, due to the lack of a real-time data update mechanism, are unable to capture these time series changes in a timely manner, resulting in delayed and inaccurate prediction results.

[0004] Therefore, a dynamic deduction scheme for distribution network structure based on time series dynamic analysis is expected. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, this application provides a distribution network structure dynamic deduction method and system based on time series dynamic analysis.

[0006] According to one aspect of the present application, a method for dynamic deduction of distribution network structure based on time series dynamic analysis is provided, which includes:

[0007] Obtain the time queue of power generation side data, the time queue of grid side data and the time queue of user side data;

[0008] Embedding the time queue of the power generation side data, the time queue of the grid side data, and the time queue of the user side data respectively to obtain a time queue of the power generation side data embedded coding vector, a time queue of the grid side data embedded coding vector, and a time queue of the user side data embedded coding vector;

[0009] Performing dynamic principal component time series aggregation on the time queue of the power generation side data embedded coding vector, the time queue of the power grid side data embedded coding vector and the time queue of the user side data embedded coding vector respectively to obtain the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, including: extracting the sequence of the current power generation side data embedded coding vector and the historical power generation side data embedded coding vector from the time queue of the power generation side data embedded coding vector and then performing principal component extraction to obtain the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector; performing graph principal component propagation aggregation on the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector to obtain the power generation side data feature principal component time series aggregation representation vector;

[0010] Based on the spatiotemporal significant joint representation of the multi-source factors of the power grid between the principal component time series aggregation representation vector of the power generation side data characteristics, the principal component time series aggregation representation vector of the power grid side data characteristics and the principal component time series aggregation representation vector of the user side data characteristics, a short-term prediction value of carbon emissions is obtained.

[0011] In the above-mentioned method for dynamic deduction of distribution network structure based on dynamic analysis of time series, a time queue of power generation side data, a time queue of grid side data and a time queue of user side data are obtained, including: obtaining a time queue of the power generation side data, the power generation side data including the power generation of renewable energy and the power generation of traditional energy; obtaining a time queue of the grid side data, the grid side data including line load, substation operating status and voltage level; obtaining a time queue of the user side data, wherein the user side data includes electricity demand, electricity load and electricity consumption behavior.

[0012] In the above-mentioned method for dynamic deduction of distribution network structure based on dynamic analysis of time series, the time queue of the generation side data, the time queue of the grid side data and the time queue of the user side data are respectively embedded and encoded to obtain the time queue of the generation side data embedded coding vector, the time queue of the grid side data embedded coding vector and the time queue of the user side data embedded coding vector, including: using the generation side data embedding encoder to embed code each generation side data in the time queue of the generation side data to obtain the time queue of the generation side data embedded coding vector; using the grid side data embedding encoder to embed code each grid side data in the time queue of the grid side data to obtain the time queue of the grid side data embedded coding vector; using the user side data embedding encoder to embed code each user side data in the time queue of the user side data to obtain the time queue of the user side data embedded coding vector.

[0013] In the above-mentioned dynamic deduction method of distribution network structure based on time series dynamic analysis, the sequence of the current generation side data embedded coding vector and the historical generation side data embedded coding vector is extracted from the time queue of the generation side data embedded coding vector, and then the principal component extraction is performed to obtain the sequence of the current generation side data principal component representation vector and the historical generation side data principal component representation vector, including: extracting the current generation side data embedded coding vector from the time queue of the generation side data embedded coding vector, and performing principal component extraction based on the eigenvalue on the current generation side data embedded coding vector to obtain the current generation side data principal component representation vector; defining other generation side data embedded coding vectors in the time queue of the generation side data embedded coding vector as historical generation side data embedded coding vectors to obtain the sequence of the historical generation side data embedded coding vectors, and performing principal component extraction based on the eigenvalue on each historical generation side data embedded coding vector in the sequence of the historical generation side data embedded coding vector to obtain the sequence of the historical generation side data principal component representation vectors.

[0014] In the above-mentioned dynamic deduction method of distribution network structure based on time series dynamic analysis, the principal component propagation aggregation of the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector is performed to obtain the principal component time series aggregation representation vector of the power generation side data feature, including: calculating the absolute factor of the time series propagation attenuation entropy of each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vector relative to the current power generation side data principal component representation vector to obtain a sequence of absolute factors of the time series propagation attenuation entropy of the power generation side data; calculating the absolute factor of the time series propagation attenuation entropy of the power generation side data based on the time span between each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vector and the current power generation side data principal component representation vector. Perform time dimension modulation on each power generation side data time series propagation attenuation entropy absolute factor in the factor sequence to obtain a sequence of power generation side data time series span modulation propagation attenuation entropy factors; input the sequence of power generation side data time series span modulation propagation attenuation entropy factors into the information transmission screening module based on the gating function to obtain a sequence of power generation side data time series span modulation propagation attenuation weights; based on the sequence of power generation side data time series span modulation propagation attenuation weights, calculate the weighted sum of the sequence of historical power generation side data principal component representation vectors to obtain the historical power generation side data principal component significant transfer aggregation representation vector; calculate the positional sum of the historical power generation side data principal component significant transfer aggregation representation vector and the current power generation side data principal component representation vector to obtain the power generation side data feature principal component time series aggregation representation vector.

[0015] In the above-mentioned dynamic deduction method of distribution network structure based on time series dynamic analysis, the absolute factor of the time series propagation attenuation entropy of each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vector is calculated relative to the current power generation side data principal component representation vector to obtain a sequence of absolute factors of the time series propagation attenuation entropy of the power generation side data, including: dividing the eigenvalues of the corresponding positions of the historical power generation side data principal component representation vector and the current power generation side data principal component representation vector to obtain the power generation side data time series characteristic principal component attenuation vector; calculating the logarithmic function value with base two of the absolute value of each eigenvalue of the power generation side data time series characteristic principal component attenuation vector to obtain the power generation side data time series characteristic principal component attenuation logarithmic vector; calculating the positional point product of the historical power generation side data principal component representation vector and the power generation side data time series characteristic principal component attenuation logarithmic vector, and performing positional multiplication on the obtained point product vector to obtain the power generation side data time series propagation attenuation value; calculating the exponential function of the power generation side data time series propagation attenuation value with the natural constant e as the base to obtain the absolute factor of the power generation side data time series propagation attenuation entropy.

[0016] In the above-mentioned dynamic deduction method of distribution network structure based on time series dynamic analysis, each generation side data time series propagation attenuation entropy absolute factor in the sequence of the generation side data time series propagation attenuation entropy absolute factors is modulated in the time dimension based on the time span between each historical generation side data principal component representation vector in the sequence of the historical generation side data principal component representation vector and the current generation side data principal component representation vector to obtain a sequence of generation side data time series span modulation propagation attenuation entropy factors, including: subtracting the timestamp of the current generation side data principal component representation vector from the timestamp of the historical generation side data principal component representation vector and rounding down to obtain the generation side data time span value; calculating an exponential function with the natural constant e as the base and the generation side data time span value as the exponent to obtain the generation side data time span modulation value; dividing the generation side data time series propagation attenuation entropy absolute factor corresponding to the historical generation side data principal component representation vector by the generation side data time span modulation value to obtain the generation side data time series span modulation propagation attenuation entropy factor.

[0017] In the above-mentioned method for dynamic deduction of distribution network structure based on dynamic analysis of time series, the sequence of the generation side data timing span modulation propagation attenuation entropy factors is input into the information transmission screening module based on the gate function to obtain the sequence of the generation side data timing span modulation propagation attenuation weights, including: comparing each generation side data timing span modulation propagation attenuation entropy factor in the sequence of the generation side data timing span modulation propagation attenuation entropy factors with a predetermined threshold to obtain the sequence of the generation side data timing span modulation propagation attenuation weights; wherein, in response to the generation side data timing span modulation propagation attenuation entropy factor being greater than the predetermined threshold, the generation side data timing span modulation propagation attenuation entropy factor greater than the predetermined threshold is input into the sigmoid function; in response to the generation side data timing span modulation propagation attenuation entropy factor being less than or equal to the predetermined threshold, the generation side data timing span modulation propagation attenuation entropy factor less than or equal to the predetermined threshold is set to zero.

[0018] In the above-mentioned dynamic deduction method of distribution network structure based on dynamic time series analysis, a short-term prediction value of carbon emissions is obtained based on the spatiotemporal significant joint representation of the power grid multi-source factors between the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, including: inputting the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector into a spatial element joint deep analyzer based on a multi-layer perceptron model to obtain a spatiotemporal significant joint representation vector of the power grid multi-source factors as the spatiotemporal significant joint representation of the power grid multi-source factors; and obtaining the short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation of the power grid multi-source factors.

[0019] In the above-mentioned dynamic deduction method of distribution network structure based on time series dynamic analysis, the short-term prediction value of carbon emissions is obtained based on the spatiotemporal significant joint representation of the multi-source factors of the power grid, including: inputting the spatiotemporal significant joint representation vector of the multi-source factors of the power grid into a dynamic evolution module based on a decoder to obtain the short-term prediction value of the carbon emissions.

[0020] According to one aspect of the present application, a distribution network structure dynamic deduction system based on time series dynamic analysis is provided, comprising:

[0021] The distribution network data time queue acquisition module is used to obtain the time queue of the power generation side data, the time queue of the grid side data and the time queue of the user side data;

[0022] a distribution network data time queue embedding coding module, configured to embed code the time queue of the generation side data, the time queue of the grid side data, and the time queue of the user side data, respectively, to obtain a time queue of the generation side data embedded coding vector, a time queue of the grid side data embedded coding vector, and a time queue of the user side data embedded coding vector;

[0023] The distribution network data dynamic principal component time series aggregation module is used to perform dynamic principal component time series aggregation on the time queue of the power generation side data embedded coding vector, the time queue of the power grid side data embedded coding vector and the time queue of the user side data embedded coding vector to obtain the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, including: a principal component extraction unit, which is used to extract the sequence of the current power generation side data embedded coding vector and the historical power generation side data embedded coding vector from the time queue of the power generation side data embedded coding vector and then perform principal component extraction to obtain the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector; a graph principal component propagation aggregation unit, which is used to perform graph principal component propagation aggregation on the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector to obtain the power generation side data feature principal component time series aggregation representation vector;

[0024] The carbon emission short-term prediction module is used to obtain the short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation of the multi-source factors of the power grid between the principal component time series aggregation representation vector of the power generation side data characteristics, the principal component time series aggregation representation vector of the power grid side data characteristics, and the principal component time series aggregation representation vector of the user side data characteristics.

[0025] This application has significant technical effects:

[0026] The present application provides a method and system for dynamic deduction of distribution network structure based on time series dynamic analysis, which adopts a data processing algorithm based on deep learning to embed coding and principal component time series aggregation of the time queues of power generation side data, power grid side data and user side data respectively, so as to intelligently obtain the short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation between the principal component time series aggregation representation of power generation side data characteristics, the principal component time series aggregation representation of power grid side data characteristics and the principal component time series aggregation representation of user side data characteristics. In this way, by integrating multi-source data from the power generation side, power grid side and user side, and performing time series feature analysis and mutual correlation on them, the complex time series dynamic relationship in the distribution network can be captured more accurately, providing more complete information, making the prediction of carbon emissions more accurate, thereby significantly improving the real-time, comprehensiveness and intelligence of carbon emission prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0028] Figure 1 Flowchart of a method for dynamic deduction of distribution network structure based on time series dynamic analysis according to an embodiment of the present application.

[0029] Figure 2 Schematic diagram of data flow of a method for dynamic deduction of distribution network structure based on time series dynamic analysis according to an embodiment of the present application.

[0030] Figure 3 This is a flowchart of a time-series aggregation representation vector of principal components of multi-side data features according to an embodiment of the present application.

[0031] Figure 4 This is a flow chart for generating a short-term prediction value of carbon emissions according to an embodiment of the present application.

[0032] Figure 5 4 is a block diagram of a distribution network structure dynamic deduction system based on time series dynamic analysis according to an embodiment of the present application. DETAILED DESCRIPTION

[0033] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0034] The distribution network is essential in the power system, transmitting high-voltage electricity generated by power plants to the low-voltage grid and ultimately to end users, including homes and businesses. In the power industry, carbon emissions management is key to optimizing power resource allocation and reducing environmental impact. Therefore, accurate carbon emissions forecasting is crucial for planning power generation resources, particularly for increasing the utilization of renewable energy, which helps promote the sustainable development of green energy.

[0035] However, traditional methods for predicting carbon emissions from distribution networks often rely on a single data source, such as power generation or consumption data. This approach ignores the complex interactions between data at all levels of the power system. For example, renewable energy output on the power generation side is affected by weather, while power consumption on the user side varies over time and seasonally. These traditional methods fail to fully utilize the integrated information from multiple sources of data, resulting in limited comprehensiveness and accuracy of the prediction results. In addition, traditional methods are mostly based on static analysis of historical data and are unable to capture current dynamic changes in real time. Due to the continuous changes in the supply and demand conditions and operating status of the distribution network, and the lack of real-time data updates in traditional prediction methods, the prediction results are often delayed and inaccurate.

[0036] In response to the above technical problems, the technical concept of the present application is to obtain the time queue of power generation side data, the time queue of grid side data and the time queue of user side data, and adopt data mining and analysis algorithms based on deep learning to embed coding and principal component time series aggregation of the power generation side data, the grid side data and the user side data, so as to intelligently obtain the short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation between the principal component time series aggregation representation of power generation side data features, the principal component time series aggregation representation of grid side data features and the principal component time series aggregation representation of user side data features. In this way, by integrating multi-source data from the power generation side, the grid side and the user side, and performing time series feature analysis and mutual correlation on them, the complex time series dynamic relationship in the distribution network can be captured more accurately, providing more complete information, making the prediction of carbon emissions more accurate, thereby significantly improving the real-time, comprehensiveness and intelligence of carbon emission prediction.

[0037] Figure 1 Flowchart of a method for dynamic deduction of distribution network structure based on time series dynamic analysis according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the method for dynamic deduction of distribution network structure based on time series dynamic analysis according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the method for dynamic deduction of distribution network structure based on time series dynamic analysis according to an embodiment of the present application includes:

[0038] S110, obtaining a time queue of power generation side data, a time queue of grid side data, and a time queue of user side data;

[0039] S120, respectively embedding-encoding the time queue of the power generation side data, the time queue of the grid side data, and the time queue of the user side data to obtain a time queue of power generation side data embedded coding vectors, a time queue of grid side data embedded coding vectors, and a time queue of user side data embedded coding vectors;

[0040] S130, performing dynamic principal component time series aggregation on the time queue of the power generation side data embedded coding vector, the time queue of the power grid side data embedded coding vector, and the time queue of the user side data embedded coding vector, respectively, to obtain a principal component time series aggregation representation vector of power generation side data features, a principal component time series aggregation representation vector of power grid side data features, and a principal component time series aggregation representation vector of user side data features;

[0041] S140, based on the spatiotemporal significant joint representation of the multi-source factors of the power grid among the principal component time series aggregation representation vector of the power generation side data characteristics, the principal component time series aggregation representation vector of the power grid side data characteristics and the principal component time series aggregation representation vector of the user side data characteristics, obtain a short-term prediction value of carbon emissions.

[0042] In step S110, a time queue of power generation side data, a time queue of grid side data, and a time queue of user side data are obtained. Specifically, in an embodiment of the present application, obtaining a time queue of power generation side data, a time queue of grid side data, and a time queue of user side data includes: obtaining a time queue of the power generation side data, wherein the power generation side data includes the power generation of renewable energy and traditional energy; obtaining a time queue of the grid side data, wherein the grid side data includes line load, substation operating status, and voltage level; and obtaining a time queue of the user side data, wherein the user side data includes power demand, power load, and power consumption behavior.

[0043] It should be understood that the time series of data on the generation side specifically includes real-time and historical power generation data for renewable energy sources such as solar, wind, and hydropower, as well as real-time and historical power generation data for traditional energy sources such as coal, oil, and natural gas. The time series of data on the grid side specifically includes real-time and historical load conditions for each line in the grid, typically expressed in terms of current and power, as well as the operating status of substations (such as normal, faulty, or undergoing maintenance), and real-time and historical voltage values for each node in the grid. The time series on the user side specifically includes real-time and historical user electricity demand data, typically expressed in terms of electricity volume, as well as real-time and historical user load data (typically expressed as average power consumption over a specific time period), and electricity consumption behavior data, specifically peak and off-peak periods. Carbon emission prediction models based on data from these three time series can account for the combined impact of multiple factors, including the volatility of renewable energy generation, changes in grid load, and changes in user electricity consumption behavior. This comprehensive consideration helps improve prediction accuracy. Overall, by comprehensively analyzing data from these three time series, we can more accurately capture the dynamic factors affecting carbon emissions, providing more comprehensive basic data for short-term carbon emission forecasts. This multi-dimensional data integration can help the model understand complex relationships, thereby improving the accuracy and real-time nature of predictions.

[0044] In step S120, the time queue of the power generation side data, the time queue of the grid side data and the time queue of the user side data are respectively embedded and coded to obtain a time queue of the power generation side data embedded coding vector, a time queue of the grid side data embedded coding vector and a time queue of the user side data embedded coding vector. Specifically, in an embodiment of the present application, the time queue of the power generation side data, the time queue of the grid side data and the time queue of the user side data are respectively embedded and coded to obtain a time queue of the power generation side data embedded coding vector, a time queue of the grid side data embedded coding vector and a time queue of the user side data embedded coding vector, including: using a power generation side data embedding encoder to embed code each power generation side data in the time queue of the power generation side data to obtain a time queue of the power generation side data embedded coding vector; using a grid side data embedding encoder to embed code each grid side data in the time queue of the grid side data to obtain a time queue of the grid side data embedded coding vector; using a user side data embedding encoder to embed code each user side data in the time queue of the user side data to obtain a time queue of the user side data embedded coding vector.

[0045] It should be understood that each power generation side data in the time queue of the power generation side data usually contains a large number of features and variables, and there may be complex interactions between these features, which express important time series information of the power generation data, and this information is crucial for understanding the time series dynamic changes of electricity. Based on this, in order to map the high-dimensional features in the original data into a lower-dimensional space, so that the important characteristics of the data can be better captured in this space while removing redundant information, in the technical solution of the present application, a power generation side data embedding encoder is used to embed the encoding of each power generation side data in the time queue of the power generation side data to extract and mine the core features in the data, and obtain a time queue of power generation side data embedded encoding vectors.

[0046] Accordingly, considering that grid-side data usually contains multiple features (such as line load, substation status, and voltage level), these features may be high-dimensional. Directly using the original data for analysis may lead to high computational complexity and difficulty in model training. In order to convert the high-dimensional data and make the data easier to process and analyze, the technical solution of this application uses a grid-side data embedding encoder to embed-code each grid-side data in the time queue of the grid-side data to obtain a time queue of grid-side data embedding coding vectors. By embedding and coding the grid-side data, high-dimensional data can be compressed into low-dimensional embedding vectors, which can significantly reduce the computational complexity of the model, thereby making the training and inference process more efficient and more suitable for short-term prediction scenarios of carbon emissions.

[0047] Similarly, considering that user-side data usually includes multiple types of features, such as electricity demand, electricity consumption behavior, etc. These features may have different scales and distributions. Directly using the original data for analysis may make it difficult for the model to learn effective patterns. In order to convert these diverse data into a unified vector representation for subsequent processing, the technical solution of this application uses a user-side data embedding encoder to embed the individual user-side data in the time queue of the user-side data to obtain a time queue of user-side data embedded coding vectors. By embedding and coding the user-side data, it is possible to better understand the patterns and structures in the user-side data, while also reducing the computational complexity of the model, thereby improving the model's operational performance.

[0048] In step S130, dynamic principal component time series aggregation is performed on the time queue of the power generation side data embedded code vector, the time queue of the power grid side data embedded code vector, and the time queue of the user side data embedded code vector to obtain the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector, and the user side data feature principal component time series aggregation representation vector. Specifically, Figure 3 This is a flow chart of a method for dynamic deduction of distribution network structure based on dynamic time series analysis according to an embodiment of the present application, in which the time queue of the power generation side data embedded coding vector, the time queue of the power grid side data embedded coding vector and the time queue of the user side data embedded coding vector are respectively subjected to dynamic principal component time series aggregation to obtain a power generation side data feature principal component time series aggregation representation vector, a power grid side data feature principal component time series aggregation representation vector and a user side data feature principal component time series aggregation representation vector. Figure 3 As shown, dynamic principal component time series aggregation is performed on the time queue of the power generation side data embedded coding vector, the time queue of the power grid side data embedded coding vector and the time queue of the user side data embedded coding vector respectively to obtain the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, including: S131, extracting the sequence of the current power generation side data embedded coding vector and the historical power generation side data embedded coding vector from the time queue of the power generation side data embedded coding vector and performing principal component extraction to obtain the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector; S132, performing graph principal component propagation aggregation on the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector to obtain the power generation side data feature principal component time series aggregation representation vector.

[0049] It should be understood that the time queue of the power generation side data embedded coding vector, the time queue of the grid side data embedded coding vector and the time queue of the user side data embedded coding vector all have a mutual correlation and dependency relationship in the time dimension, and the features at different time points have different importance and influence in the entire time range. Therefore, in order to be able to integrate the data features at different times and form a continuous time series representation, so as to better understand the trends and patterns of the historical and current states of the power generation side, the grid side and the user side over time, in the technical solution of the present application, the time queue of the power generation side data embedded coding vector, the time queue of the grid side data embedded coding vector and the time queue of the user side data embedded coding vector are respectively subjected to dynamic principal component time series aggregation to obtain the power generation side data feature principal component time series aggregation representation vector, the grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector. In particular, the time queue of the power generation side data embedded coding vector is taken as an example for detailed explanation.

[0050] Specifically, first, a sequence of current and historical power-side data embedding code vectors is extracted from the time sequence of the power-side data embedding code vectors, and then principal component extraction is performed to obtain a sequence of current and historical power-side data principal component representation vectors. It is understood that principal component analysis (PCA) is a common dimensionality reduction technique that can reveal and identify key features. Specifically, PCA can identify the directions within a dataset that best reflect the overall trend of change, namely the principal components. These principal components capture key information in the data and help remove redundant information. Principal component extraction of current and historical features can provide historical information on how features change over time, providing a foundation for subsequent time series analysis and processing.

[0051] Next, the sequence of the principal component representation vector of the current power generation side data and the principal component representation vector of the historical power generation side data is subjected to graph principal component propagation aggregation to obtain the principal component time series aggregation representation vector of the power generation side data feature. That is, graph principal component propagation aggregation can capture the dynamic relationship of the power generation side data in time and space. By propagating the current and historical data on the graph, we can better understand the change pattern of the power generation side data over time and their mutual influence. Specifically, first, by calculating the absolute factor of the time series propagation attenuation entropy of each historical power generation side data principal component representation vector to quantify the degree of attenuation and information loss of the feature over time, this is very useful for understanding the importance of historical data at the current time point, thereby obtaining a sequence of the absolute factor of the time series propagation attenuation entropy of the power generation side data. To further capture and extract the ability to understand information transfer between features over long time intervals, the absolute propagation decay entropy factors of each generation-side data series are modulated in the time dimension based on the time span between the current generation-side data principal component representation vector and each historical generation-side data principal component representation vector. This ensures that even significant changes between nodes separated by long time spans are considered part of normal fluctuations. This allows the model to more appropriately handle feature correlations across different time spans, resulting in a sequence of generation-side data time-series span modulation propagation decay entropy factors. Subsequently, a gating function-based information transfer filtering module is used to feature-screen and adaptively weight the sequence of generation-side data time-series span modulation propagation decay entropy factors to identify the most influential propagation features and assign corresponding weights accordingly, thereby obtaining a sequence of generation-side data time-series span modulation propagation decay weights. Finally, the sequence of historical generation-side data principal component representation vectors is weighted and summed based on the sequence of propagation decay weights to fully integrate short-term and long-term feature trends and changes, resulting in an aggregated representation vector of significant principal component transfer for historical generation-side data. Finally, the historical aggregation representation vector and the principal component representation vector of the current power generation side data are added by position to fully integrate the long-term trend in the historical data and the immediate changes in the current data to generate the principal component time series aggregation representation vector of the power generation side data features.

[0052] Specifically, in an embodiment of the present application, a sequence of a current generation side data embedded coding vector and a historical generation side data embedded coding vector is extracted from a time queue of the generation side data embedded coding vector, and then principal component extraction is performed to obtain a sequence of a current generation side data principal component representation vector and a historical generation side data principal component representation vector, including: extracting the current generation side data embedded coding vector from the time queue of the generation side data embedded coding vector, and performing eigenvalue-based principal component extraction on the current generation side data embedded coding vector to obtain the current generation side data principal component representation vector; defining other generation side data embedded coding vectors in the time queue of the generation side data embedded coding vector as historical generation side data embedded coding vectors to obtain a sequence of historical generation side data embedded coding vectors, and performing eigenvalue-based principal component extraction on each historical generation side data embedded coding vector in the sequence of historical generation side data embedded coding vectors to obtain a sequence of historical generation side data principal component representation vectors.

[0053] Specifically, in an embodiment of the present application, the graph principal component propagation aggregation is performed on the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector to obtain the power generation side data feature principal component time series aggregation representation vector, including: calculating the time series propagation attenuation entropy absolute factor of each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vector relative to the current power generation side data principal component representation vector to obtain a sequence of the power generation side data time series propagation attenuation entropy absolute factors; calculating the time series propagation attenuation entropy absolute factors of the power generation side data based on the time span between each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vector and the current power generation side data principal component representation vector. The absolute factors of the time series propagation attenuation entropy of each power generation side data in the data are modulated in the time dimension to obtain a sequence of the time series span modulation propagation attenuation entropy factors of the power generation side data; the sequence of the time series span modulation propagation attenuation entropy factors of the power generation side data is input into the information transmission screening module based on the gating function to obtain a sequence of the time series span modulation propagation attenuation weights of the power generation side data; based on the sequence of the time series span modulation propagation attenuation weights of the power generation side data, the weighted sum of the sequence of the principal component representation vectors of the historical power generation side data is calculated to obtain the principal component significant transfer aggregation representation vector of the historical power generation side data; the positional sum of the principal component significant transfer aggregation representation vector of the historical power generation side data and the principal component representation vector of the current power generation side data is calculated to obtain the principal component time series aggregation representation vector of the power generation side data feature.

[0054] More specifically, in an embodiment of the present application, the absolute factors of the time series propagation attenuation entropy of each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vector relative to the current power generation side data principal component representation vector are calculated to obtain a sequence of absolute factors of the time series propagation attenuation entropy of the power generation side data, including: dividing the eigenvalues of the corresponding positions of the historical power generation side data principal component representation vector and the current power generation side data principal component representation vector to obtain the power generation side data time series characteristic principal component attenuation vector; calculating the logarithmic function value with base two of the absolute value of each eigenvalue of the power generation side data time series characteristic principal component attenuation vector to obtain the power generation side data time series characteristic principal component attenuation logarithmic vector; calculating the positional dot product of the historical power generation side data principal component representation vector and the power generation side data time series characteristic principal component attenuation logarithmic vector, and performing positional multiplication on the obtained dot product vector to obtain the power generation side data time series propagation attenuation value; calculating the exponential function of the power generation side data time series propagation attenuation value with the natural constant e as the base to obtain the absolute factors of the power generation side data time series propagation attenuation entropy.

[0055] More specifically, in an embodiment of the present application, based on the time span between each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vector and the current power generation side data principal component representation vector, each power generation side data time series propagation attenuation entropy absolute factor in the sequence of the power generation side data principal component representation vector is modulated in the time dimension to obtain a sequence of power generation side data time series span modulation propagation attenuation entropy factors, including: subtracting the timestamp of the current power generation side data principal component representation vector from the timestamp of the historical power generation side data principal component representation vector and rounding down to obtain the power generation side data time span value; calculating an exponential function with the natural constant e as the base and the power generation side data time span value as the exponent to obtain the power generation side data time span modulation value; dividing the power generation side data time series propagation attenuation entropy absolute factor corresponding to the historical power generation side data principal component representation vector by the power generation side data time span modulation value to obtain the power generation side data time series span modulation propagation attenuation entropy factor.

[0056] More specifically, in an embodiment of the present application, the sequence of the power generation side data timing span modulation propagation attenuation entropy factors is input into an information transmission screening module based on a gating function to obtain a sequence of the power generation side data timing span modulation propagation attenuation weights, including: comparing each power generation side data timing span modulation propagation attenuation entropy factor in the sequence of the power generation side data timing span modulation propagation attenuation entropy factors with a predetermined threshold to obtain a sequence of the power generation side data timing span modulation propagation attenuation weights; wherein, in response to the power generation side data timing span modulation propagation attenuation entropy factor being greater than the predetermined threshold, the power generation side data timing span modulation propagation attenuation entropy factor greater than the predetermined threshold is input into a sigmoid function; in response to the power generation side data timing span modulation propagation attenuation entropy factor being less than or equal to the predetermined threshold, the power generation side data timing span modulation propagation attenuation entropy factor less than or equal to the predetermined threshold is set to zero.

[0057] In the embodiment of the present application, specifically, dynamic principal component time series aggregation is performed on the time queue of the power generation side data embedded in the coding vector to obtain the power generation side data feature principal component time series aggregation representation vector, which can be expressed as:

[0058] ,

[0059] ,

[0060] ,

[0061] ,

[0062] ,

[0063] ,

[0064] ,

[0065] ,

[0066] in, A time queue for embedding the coding vector for the power generation side data, are the first, second, ..., and third time queues of the power generation side data embedded in the coding vector. ,..., The power generation side data is embedded in the coding vector, Yes Perform principal component extraction based on eigenvalues, for The corresponding principal component representation vector of the power generation side data is: for The transposed vector of is the length of the power generation side data embedding coding vector, yes The corresponding covariance matrix of the power generation side data, yes The corresponding principal component orthogonal matrix of the power generation side data is: for The transposed matrix of is the principal component vector of each power generation side data in the sequence of the principal component vector of the power generation side data, for The corresponding diagonal matrix of power generation side data, The diagonal elements of the matrix are The diagonal matrix of power generation side data, are the weight values of the principal component vectors of each power generation side data, Returns the maximum value corresponding to value, is the maximum approximate matching value, is the sequence of principal component representation vectors of the power generation side data, are the first, second, ..., and third principal component representation vectors of the power generation side data. ,..., The principal component representation vector of the power generation side data, for The eigenvalues at each position in , for The eigenvalues at each position in for The number of eigenvalues in , represents the logarithmic function value with base 2, represents the exponential function with the natural constant e as the base, for and The absolute factor of the attenuation entropy of the time series propagation of the power generation side data, and Respectively represent and The timestamp of the principal component representation vector of the power generation side data, To perform the floor rounding operation, yes and The data timing span between the power generation side modulates the propagation attenuation entropy factor, For masking, yes function, is a predetermined threshold, is the time series span modulation propagation attenuation weight of each power generation side data in the sequence of the time series span modulation propagation attenuation weight of the power generation side data, is the number of vectors in the sequence of principal component representation vectors of the power generation side data, is the principal component time series aggregation representation vector of the power generation side data feature. Similarly, the processing method of the time queue of the grid side data embedded in the coding vector and the time queue of the user side data embedded in the coding vector is also as shown in the above formula.

[0067] In step S140, a short-term prediction value of carbon emissions is obtained based on the spatiotemporal significant joint representation of the power grid multi-source factors among the principal component time series aggregation representation vector of the power generation side data features, the principal component time series aggregation representation vector of the power grid side data features, and the principal component time series aggregation representation vector of the user side data features. Specifically, Figure 4 This is a flow chart of obtaining a short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation of the power grid multi-source factors among the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector, and the user side data feature principal component time series aggregation representation vector in the distribution network structure dynamic deduction method based on time series dynamic analysis according to an embodiment of the present application. Figure 4 As shown, based on the spatiotemporal significant joint representation of the power grid multi-source factors among the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, a short-term prediction value of carbon emissions is obtained, including: S141, inputting the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector into a spatial element joint deep analyzer based on a multi-layer perceptron model to obtain the spatiotemporal significant joint representation vector of the power grid multi-source factors as the spatiotemporal significant joint representation of the power grid multi-source factors; S142, based on the spatiotemporal significant joint representation of the power grid multi-source factors, a short-term prediction value of carbon emissions is obtained.

[0068] In step S141, the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector, and the user side data feature principal component time series aggregation representation vector are input into a spatial element joint deep analyzer based on a multi-layer perceptron model to obtain a grid multi-source factor spatiotemporal significant joint representation vector as the grid multi-source factor spatiotemporal significant joint representation. It should be understood that in order to integrate and analyze data from different aspects, thereby obtaining a comprehensive and detailed grid state representation, so as to improve the accuracy of subsequent short-term carbon emission forecasting, in the technical solution of the present application, the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector, and the user side data feature principal component time series aggregation representation vector are input into a spatial element joint deep analyzer based on a multi-layer perceptron model to obtain a grid multi-source factor spatiotemporal significant joint representation vector. It is worth mentioning that the multi-layer perceptron model can capture the nonlinear relationship in the data and the interaction between different features in space. Therefore, through multi-layer perceptrons, the complex spatiotemporal interactions between data from the power generation side, grid side, and user side can be explored, so as to better understand how they jointly affect the overall performance of the power grid and improve the accuracy of the prediction of the power grid status.

[0069] In step S142, based on the spatiotemporal significant joint representation of the multi-source factors of the power grid, a short-term forecast value of the carbon emissions is obtained. Specifically, in an embodiment of the present application, based on the spatiotemporal significant joint representation of the multi-source factors of the power grid, a short-term forecast value of the carbon emissions is obtained, including: inputting the spatiotemporal significant joint representation vector of the multi-source factors of the power grid into a dynamic evolution module based on a decoder to obtain the short-term forecast value of the carbon emissions. That is, the spatiotemporal significant joint representation of the multi-source factors of the power grid obtained by spatiotemporally significant joint representation of the principal component of the power generation side data feature, the principal component of the power grid side data feature, and the principal component of the user side data feature is decoded to intelligently obtain the short-term forecast value of the carbon emissions. In this way, by integrating multi-source data from the power generation side, the power grid side, and the user side, and performing temporal feature analysis and mutual correlation on them, the complex temporal dynamic relationship in the distribution network can be more accurately captured, providing more complete information, making the prediction of carbon emissions more accurate, thereby significantly improving the real-time, comprehensiveness, and intelligence of the carbon emission prediction.

[0070] In particular, in the technical solution of the present application, the principal component time series aggregation representation vector of the power generation side data features, the principal component time series aggregation representation vector of the power grid side data features and the principal component time series aggregation representation vector of the user side data features respectively represent the time series propagation aggregation features of the power generation side data, the power grid measurement data and the user side data. In this way, when the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector are input into the spatial element joint deep analyzer based on the multi-layer perceptron model, considering that there are time series pattern distribution differences and source data modal differences between the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, this will cause the spatial element joint deep analysis based on the multi-layer perceptron model to have an imbalance in the element joint weights, thereby causing the characteristic manifold of the obtained power grid multi-source factor spatiotemporal significant joint representation vector to have expression sparsity and distribution imbalance in the high-dimensional feature space, thereby affecting the accuracy of the short-term prediction value of carbon emissions obtained through the dynamic evolution module based on the decoder.

[0071] Based on this, in a preferred embodiment, the spatiotemporal significant joint representation vector of the multi-source factors of the power grid is input into a dynamic evolution module based on a decoder to obtain a short-term prediction value of carbon emissions, including: performing L random sampling on the spatiotemporal significant joint representation vector of the multi-source factors of the power grid to obtain L power grid state random disturbance eigenvalues; arranging the L power grid state random disturbance eigenvalues in order from large to small to obtain a power grid state random disturbance ordered vector; calculating the self-inner product of the spatiotemporal significant joint representation vector of the multi-source factors of the power grid to obtain a power grid state energy modulation eigenvalue; and performing L random sampling on the power grid state random disturbance ordered vector based on the power grid state energy modulation eigenvalue. Perform semantic intensity scaling to obtain a semantically enhanced ordered vector of random disturbance of the power grid state; calculate the product between the semantically enhanced ordered vector of random disturbance of the power grid state and the transposed vector of the power grid multi-source factor spatiotemporal significant joint representation vector to obtain the power grid state best sample prompt orthogonal modulation matrix; calculate the matrix product between the power grid state best sample prompt orthogonal modulation matrix and the power grid multi-source factor spatiotemporal significant joint representation vector to obtain an optimized power grid multi-source factor spatiotemporal significant joint representation vector; input the optimized power grid multi-source factor spatiotemporal significant joint representation vector into the decoder-based dynamic evolution module to obtain the short-term prediction value of the carbon emissions.

[0072] The above optimization process is expressed as follows:

[0073] ,

[0074] ,

[0075] ,

[0076] ,

[0077] ,

[0078] ,

[0079] ,

[0080] ,

[0081] in, is the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, Indicates the mean calculation of the vector. is the mean of the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, is the spatiotemporal significant joint representation vector of the multi-source factors of the power grid. The eigenvalues at the positions, is the length of the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, is the variance of the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, is the Gaussian density probability function, express Random sampling, and Represents the spatiotemporal significant joint representation vector of the multi-source factors of the power grid The random sampling The random disturbance characteristic value of the power grid state, Express Arrange them in order from large to small. is the ordered vector of random disturbance of the power grid state, represents the square of the vector's two-norm, is the grid state energy modulation characteristic value, Indicates point multiplication by position, represents matrix multiplication, for The transposed vector of Prompt an orthogonal modulation matrix for the best example of the grid state, A spatiotemporal significant joint representation vector for optimized multi-source factors in power grids.

[0082] Accordingly, by performing spatial and temporal significant joint representation of the multi-source factors of the power grid, Random sampling is performed to obtain a set of new feature values based on the original features but containing slight variations. This introduces randomness to explore different aspects of the feature space and helps the model better generalize to unseen data by simulating the noise or variation in actual data. Next, these grid state random perturbation feature values are sorted from large to small to form an ordered grid state random perturbation vector to identify which perturbation directions may be most important for the current problem. This sorting not only helps the algorithm focus on the feature differences that best distinguish different categories, but also helps reveal hidden data structures or patterns. Subsequently, the inner product of the spatiotemporal significant joint representation vector of the multi-source factors of the power grid is calculated to obtain the power grid state energy modulation eigenvalue. The power grid state energy modulation eigenvalue reflects the overall strength or "energy" of the original feature vector and provides a benchmark for measuring the importance of the feature. Using this power grid state energy modulation eigenvalue as a reference point, the previously formed power grid state random perturbation ordered vector is semantically scaled to construct a semantically enhanced power grid state random perturbation ordered vector. In this way, those disturbances that appear to be more important in the original feature space are amplified, thereby enhancing the model's sensitivity to changes in key features, while also retaining relatively small but potentially useful disturbance information.

[0083] Furthermore, the product of the semantically enhanced ordered vector of random perturbations in the power grid state and the transposed vector representing the spatiotemporal saliency of the multi-source factors in the power grid is calculated to obtain the orthogonal modulation matrix of the best example prompt for the power grid state. This matrix aims to identify how features interact from two different perspectives. The resulting matrix can be considered a transformation tool for capturing and representing complex linear relationships between features. Finally, the final feature structure modulation is achieved by applying the constructed orthogonal modulation matrix of the best example prompt for the power grid state to the original spatiotemporal saliency of the multi-source factors in the power grid. This makes the features more consistent with the objective function, reduces redundant information, and maximizes the retention of useful information, thereby improving learning efficiency and prediction performance.

[0084] In summary, the feature manifold optimization process for the spatiotemporal significant joint representation vector of the multi-source factors of the power grid has been clarified. It extracts the best prompt examples based on fine-grained target retrieval of random perturbations, and then guides adversarial robust learning based on the best prompt examples to complete the structure of the high-dimensional manifold of the feature vector in the feature space, so as to improve the detail and structure of the feature expression, thereby improving the accuracy of the short-term prediction value of carbon emissions obtained by the decoder-based dynamic evolution module. In this way, by integrating multi-source data from the power generation side, the grid side, and the user side, and performing time series feature analysis and mutual correlation on them, the complex time series dynamic relationship in the distribution network can be more accurately captured, providing more complete information, making the prediction of carbon emissions more accurate, thereby significantly improving the real-time, comprehensiveness, and intelligence of carbon emission prediction.

[0085] In summary, a method for dynamic deduction of distribution network structure based on time series dynamic analysis based on an embodiment of the present application is explained, which adopts a data processing algorithm based on deep learning to embed coding and principal component time series aggregation of the time queues of power generation side data, grid side data and user side data respectively, so as to intelligently obtain the short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation between the principal component time series aggregation representation of power generation side data features, the principal component time series aggregation representation of grid side data features and the principal component time series aggregation representation of user side data features. In this way, by integrating multi-source data from the power generation side, grid side and user side, and performing time series feature analysis and mutual correlation on them, the complex time series dynamic relationship in the distribution network can be captured more accurately, providing more complete information, making the prediction of carbon emissions more accurate, thereby significantly improving the real-time, comprehensiveness and intelligence of carbon emission prediction.

[0086] Figure 5 FIG is a block diagram of a distribution network structure dynamic deduction system based on time series dynamic analysis according to an embodiment of the present application. Figure 5 As shown, the distribution network structure dynamic deduction system 100 based on time series dynamic analysis includes:

[0087] The distribution network data time queue acquisition module 110 is used to acquire the time queue of the power generation side data, the time queue of the grid side data and the time queue of the user side data;

[0088] a distribution network data time queue embedding coding module 120, configured to embed code the time queue of the generation side data, the time queue of the grid side data, and the time queue of the user side data, respectively, to obtain a time queue of the generation side data embedding coding vector, a time queue of the grid side data embedding coding vector, and a time queue of the user side data embedding coding vector;

[0089] The distribution network data dynamic principal component time series aggregation module 130 is used to perform dynamic principal component time series aggregation on the time queue of the power generation side data embedded coding vector, the time queue of the power grid side data embedded coding vector and the time queue of the user side data embedded coding vector to obtain the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, including: a principal component extraction unit, used to extract the sequence of the current power generation side data embedded coding vector and the historical power generation side data embedded coding vector from the time queue of the power generation side data embedded coding vector, and then perform principal component extraction to obtain the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector; a graph principal component propagation aggregation unit, used to perform graph principal component propagation aggregation on the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector to obtain the power generation side data feature principal component time series aggregation representation vector;

[0090] The carbon emission short-term prediction module 140 is used to obtain a short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation of the multi-source factors of the power grid between the principal component time series aggregation representation vector of the power generation side data characteristics, the principal component time series aggregation representation vector of the power grid side data characteristics, and the principal component time series aggregation representation vector of the user side data characteristics.

[0091] Here, those skilled in the art will understand that the specific operations of the various modules and units in the above-mentioned distribution network structure dynamic deduction system based on time series dynamic analysis have been referred to above. Figures 1 to 4 It has been introduced in detail in the description of the distribution network structure dynamic deduction method based on time series dynamic analysis, and therefore, its repeated description will be omitted.

Claims

1. A dynamic deduction method for distribution network structure based on time series dynamic analysis, characterized by: include: Obtain the time queue of power generation side data, the time queue of grid side data and the time queue of user side data; Embedding the time queue of the power generation side data, the time queue of the grid side data, and the time queue of the user side data respectively to obtain a time queue of the power generation side data embedded coding vector, a time queue of the grid side data embedded coding vector, and a time queue of the user side data embedded coding vector; Performing dynamic principal component time series aggregation on the time queue of the power generation side data embedded coding vector, the time queue of the power grid side data embedded coding vector and the time queue of the user side data embedded coding vector respectively to obtain the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, including: extracting the sequence of the current power generation side data embedded coding vector and the historical power generation side data embedded coding vector from the time queue of the power generation side data embedded coding vector and then performing principal component extraction to obtain the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector; performing graph principal component propagation aggregation on the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector to obtain the power generation side data feature principal component time series aggregation representation vector; Obtaining a short-term forecast value of carbon emissions based on a spatiotemporal significant joint representation of power grid multi-source factors among the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector, and the user side data feature principal component time series aggregation representation vector; The time-space significant joint representation of multi-source factors in the power grid is expressed as follows: V order =Order{S i } S i =∈ i ,i=1,2,...,L; u=average(V); Wherein, V is the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, average(·) represents the mean calculation of the vector, u is the mean of the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, v i is the eigenvalue of the i-th position in the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, d is the length of the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, σ 2 is the variance of the spatiotemporal significant joint representation vector of the multi-source factors of the power grid, is the Gaussian density probability function, L represents L random sampling, ∈ i and S i represents the random disturbance eigenvalue of the i-th grid state obtained by randomly sampling the spatiotemporal significant joint representation vector of the multi-source factors of the grid L times, Order{S i } indicates that S i Arrange them in order from large to small, V order is the ordered vector of random disturbance of the power grid state, represents the square of the second norm of the vector, e is the energy modulation characteristic value of the power grid state, ⊙ represents the point multiplication by position, Represents matrix multiplication, V T is the transposed vector of V, M t is the orthogonal modulation matrix of the best example of the power grid state, and V′ is the optimized spatiotemporal significant joint representation vector of the multi-source factors of the power grid.

2. The method for dynamic deduction of distribution network structure based on time series dynamic analysis according to claim 1 is characterized in that: Obtain the time queue of power generation side data, the time queue of grid side data, and the time queue of user side data, including: Acquiring a time queue of the power generation side data, wherein the power generation side data includes power generation of renewable energy and power generation of traditional energy; Obtaining a time queue of the grid-side data, the grid-side data including line load, substation operating status, and voltage level; A time queue of the user-side data is obtained, wherein the user-side data includes electricity demand, electricity load, and electricity usage behavior.

3. The method for dynamic deduction of distribution network structure based on time series dynamic analysis according to claim 2 is characterized in that: Embedding the time queue of the power generation side data, the time queue of the grid side data, and the time queue of the user side data to obtain a time queue of the power generation side data embedded coding vector, a time queue of the grid side data embedded coding vector, and a time queue of the user side data embedded coding vector, respectively, including: Using a power generation side data embedding encoder to embed code each power generation side data in the time queue of the power generation side data to obtain a time queue of the power generation side data embedding code vector; Using a grid-side data embedding encoder to embed-code each grid-side data in the time queue of the grid-side data to obtain a time queue of the grid-side data embedding coding vector; A user-side data embedding encoder is used to embed code each user-side data in the time queue of the user-side data to obtain the time queue of the user-side data embedding coding vector.

4. The method for dynamic deduction of distribution network structure based on time series dynamic analysis according to claim 3 is characterized in that: The method further comprises extracting a sequence of a current power generation side data embedding code vector and a sequence of a historical power generation side data embedding code vector from a time queue of the power generation side data embedding code vector and performing principal component extraction to obtain a sequence of a current power generation side data principal component representation vector and a sequence of a historical power generation side data principal component representation vector. Extracting the current power generation side data embedded coding vector from the time queue of the power generation side data embedded coding vector, and performing principal component extraction based on eigenvalues on the current power generation side data embedded coding vector to obtain the current power generation side data principal component representation vector; The other power generation side data embedded coding vectors in the time queue of the power generation side data embedded coding vector are defined as historical power generation side data embedded coding vectors to obtain a sequence of the historical power generation side data embedded coding vectors, and the principal component extraction based on the eigenvalue is performed on each historical power generation side data embedded coding vector in the sequence of the historical power generation side data embedded coding vector to obtain a sequence of the principal component representation vectors of the historical power generation side data.

5. The method for dynamic deduction of distribution network structure based on time series dynamic analysis according to claim 4 is characterized in that: Performing graph principal component propagation aggregation on the sequence of the principal component representation vector of the current power generation side data and the principal component representation vector of the historical power generation side data to obtain the principal component time series aggregation representation vector of the power generation side data feature, including: Calculating the absolute factor of the time series propagation attenuation entropy of each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vectors relative to the current power generation side data principal component representation vector to obtain a sequence of absolute factors of the time series propagation attenuation entropy of the power generation side data; Based on the time span between each historical power generation side data principal component representation vector in the sequence of historical power generation side data principal component representation vectors and the current power generation side data principal component representation vector, each power generation side data time series propagation attenuation entropy absolute factor in the sequence of power generation side data time series propagation attenuation entropy absolute factors is modulated in the time dimension to obtain a sequence of power generation side data time series span modulated propagation attenuation entropy factors; Inputting the sequence of the power generation side data time series span modulation propagation attenuation entropy factors into the information transmission screening module based on the gating function to obtain the sequence of the power generation side data time series span modulation propagation attenuation weights; Based on the sequence of the time series span modulation propagation attenuation weights of the power generation side data, calculating the weighted sum of the sequence of principal component representation vectors of the historical power generation side data to obtain the principal component significant transfer aggregation representation vector of the historical power generation side data; Calculating the positional sum of the historical power generation side data principal component significant transfer aggregate representation vector and the current power generation side data principal component representation vector to obtain the power generation side data characteristic principal component time series aggregate representation vector; The sequence of the power generation side data time series span modulation propagation attenuation entropy factors is input into the information transmission screening module based on the gating function to obtain the sequence of the power generation side data time series span modulation propagation attenuation weights, including: Each power generation side data timing span modulation propagation attenuation entropy factor in the sequence of the power generation side data timing span modulation propagation attenuation entropy factors is compared with a predetermined threshold to obtain a sequence of the power generation side data timing span modulation propagation attenuation weights; wherein, in response to the power generation side data timing span modulation propagation attenuation entropy factor being greater than the predetermined threshold, the power generation side data timing span modulation propagation attenuation entropy factor greater than the predetermined threshold is input into a sigmoid function; in response to the power generation side data timing span modulation propagation attenuation entropy factor being less than or equal to the predetermined threshold, the power generation side data timing span modulation propagation attenuation entropy factor less than or equal to the predetermined threshold is set to zero.

6. The method for dynamic deduction of distribution network structure based on time series dynamic analysis according to claim 5, characterized in that: Calculating the absolute factor of the time series propagation attenuation entropy of each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vectors relative to the current power generation side data principal component representation vector to obtain a sequence of absolute factors of the time series propagation attenuation entropy of the power generation side data, including: Dividing the eigenvalues of the corresponding positions of the principal component representation vector of the historical power generation side data and the principal component representation vector of the current power generation side data to obtain a principal component attenuation vector of the time series characteristics of the power generation side data; Calculating the base-two logarithmic function value of the absolute value of each eigenvalue of the principal component attenuation vector of the power generation side data time series feature to obtain the principal component attenuation logarithmic vector of the power generation side data time series feature; Calculate the position-by-position point product of the principal component representation vector of the historical power generation side data and the principal component attenuation logarithm vector of the time series feature of the power generation side data, and perform point-by-point multiplication on the obtained point product vector to obtain the power generation side data time series propagation attenuation value; An exponential function of the power generation side data time series propagation attenuation value with the natural constant e as the base is calculated to obtain the absolute factor of the power generation side data time series propagation attenuation entropy.

7. The method for dynamic deduction of distribution network structure based on time series dynamic analysis according to claim 6, characterized in that: Based on the time span between each historical power generation side data principal component representation vector in the sequence of the historical power generation side data principal component representation vectors and the current power generation side data principal component representation vector, each power generation side data time series propagation attenuation entropy absolute factor in the sequence of the power generation side data time series propagation attenuation entropy absolute factors is modulated in the time dimension to obtain a sequence of power generation side data time series span modulated propagation attenuation entropy factors, including: Subtracting the timestamp of the principal component representation vector of the current power generation side data from the timestamp of the principal component representation vector of the historical power generation side data and rounding down the result to obtain a time span value of the power generation side data; Calculating an exponential function with the natural constant e as the base and the time span value of the power generation side data as the exponent to obtain a time span modulation value of the power generation side data; The absolute factor of the generation side data time series propagation attenuation entropy corresponding to the principal component representation vector of the historical generation side data is divided by the time span modulation value of the generation side data to obtain the generation side data time series span modulation propagation attenuation entropy factor.

8. The method for dynamic deduction of distribution network structure based on time series dynamic analysis according to claim 7 is characterized in that: Based on the spatiotemporal significant joint representation of the power grid multi-source factors among the principal component time series aggregation representation vector of the power generation side data features, the principal component time series aggregation representation vector of the power grid side data features, and the principal component time series aggregation representation vector of the user side data features, a short-term prediction value of carbon emissions is obtained, including: Input the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector, and the user side data feature principal component time series aggregation representation vector into a spatial element joint depth analyzer based on a multi-layer perceptron model to obtain a power grid multi-source factor spatiotemporal significant joint representation vector as the power grid multi-source factor spatiotemporal significant joint representation; Based on the spatiotemporal significant joint representation of the multi-source factors of the power grid, a short-term forecast value of the carbon emissions is obtained.

9. The method for dynamic deduction of distribution network structure based on time series dynamic analysis according to claim 8, characterized in that: Based on the spatiotemporal significant joint representation of the multi-source factors of the power grid, a short-term prediction value of carbon emissions is obtained, including: inputting the spatiotemporal significant joint representation vector of the multi-source factors of the power grid into a dynamic evolution module based on a decoder to obtain the short-term prediction value of carbon emissions.

10. A distribution network structure dynamic deduction system based on time series dynamic analysis is characterized by: include: The distribution network data time queue acquisition module is used to obtain the time queue of the power generation side data, the time queue of the grid side data and the time queue of the user side data; a distribution network data time queue embedding coding module, configured to embed code the time queue of the generation side data, the time queue of the grid side data, and the time queue of the user side data, respectively, to obtain a time queue of the generation side data embedded coding vector, a time queue of the grid side data embedded coding vector, and a time queue of the user side data embedded coding vector; The distribution network data dynamic principal component time series aggregation module is used to perform dynamic principal component time series aggregation on the time queue of the power generation side data embedded coding vector, the time queue of the power grid side data embedded coding vector and the time queue of the user side data embedded coding vector to obtain the power generation side data feature principal component time series aggregation representation vector, the power grid side data feature principal component time series aggregation representation vector and the user side data feature principal component time series aggregation representation vector, including: a principal component extraction unit, which is used to extract the sequence of the current power generation side data embedded coding vector and the historical power generation side data embedded coding vector from the time queue of the power generation side data embedded coding vector and then perform principal component extraction to obtain the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector; a graph principal component propagation aggregation unit, which is used to perform graph principal component propagation aggregation on the sequence of the current power generation side data principal component representation vector and the historical power generation side data principal component representation vector to obtain the power generation side data feature principal component time series aggregation representation vector; The carbon emission short-term prediction module is used to obtain the short-term prediction value of carbon emissions based on the spatiotemporal significant joint representation of the multi-source factors of the power grid between the principal component time series aggregation representation vector of the power generation side data characteristics, the principal component time series aggregation representation vector of the power grid side data characteristics, and the principal component time series aggregation representation vector of the user side data characteristics.

Images