Source network charged carbon panoramic monitoring method based on multi-time scale situation awareness

Abstract

The invention discloses a source network charged carbon panoramic monitoring method based on multi-time scale situation awareness, and belongs to the technical field of power system monitoring and situation awareness. Comprising the following steps: step S101: data acquisition and convergence; step S102, a situation awareness step; step S103, a situation understanding step; step S104, a situation prediction step; and S105, a decision support step. According to the method, full-link real-time monitoring and optimization of a power supply, a power grid, a load, energy storage and carbon emission are realized through the steps of data acquisition and convergence, situation detection, situation understanding, situation prediction and decision support. The problems of insufficient data integration, single time scale and poor multi-target coordination in the prior art are solved, the method has the advantages of global coverage, multi-time scale prediction, multi-target coordination optimization and the like, and the safe, low-carbon and economic operation level of a power system is improved.

Description

Technical Field

[0001] This invention relates to the field of power system monitoring and situational awareness technology, specifically a method for panoramic monitoring of source-grid-load carbon based on multi-timescale situational awareness. Background Technology

[0002] Source-grid-load carbon panoramic monitoring refers to the comprehensive, real-time, and multi-dimensional monitoring and perception of the operating status and data information of the entire process of power generation, grid (transmission, transformation and distribution, etc.), load, power conversion and storage equipment, and carbon emissions in the power system. This aims to achieve a panoramic understanding of the physical characteristics, energy flow, and carbon flow of the power system, and provide basic support for the safe, low-carbon, and economical operation of the power system.

[0003] Research on power system situational awareness and low-carbon monitoring began earlier abroad. Developed countries in Europe and America emphasize the application of multi-source data fusion and situational awareness technologies in smart grid construction. For example, smart grid projects supported by the U.S. Department of Energy employ advanced sensing and data analytics technologies to achieve real-time monitoring and risk prediction of grid operation. In terms of low-carbon monitoring, Europe has established a pan-European energy monitoring network to conduct real-time statistical analysis of electricity carbon emissions from member states. Mainstream methods include multi-source data integration based on big data, situational prediction using machine learning algorithms (such as support vector machines and random forests), and achieving a balance between low-carbon and economic decisions through the construction of multi-objective optimization models.

[0004] Despite extensive research and practice both domestically and internationally, the following shortcomings still exist in the field of source-grid-charged carbon panoramic monitoring: First, there is a lack of data integration and panoramic perception. Data from multiple sources, including power generation, grid, load, and storage, are scattered across different stages and devices. Existing technologies lack efficient means to integrate this heterogeneous data, making it difficult to achieve panoramic, real-time, and accurate perception of the physical state and carbon flow information of the power system. This results in blind spots in the understanding of the system's operating status.

[0005] Secondly, current research focuses on a single time scale and lacks the ability to coordinate perception of short and medium time scales. Furthermore, when predicting the situation, the model algorithms do not have sufficient accuracy in fitting the evolution trend of the system state across multiple time scales, making it difficult to effectively identify potential risks.

[0006] Third, existing technologies are unable to achieve efficient synergistic optimization among multiple objectives such as safety, low carbon, and economy. They often focus on a single objective and cannot meet the needs of source-grid-load-storage synergistic optimization decision-making under the goal of deep low carbon, resulting in insufficient scientificity and comprehensiveness in decision-making.

[0007] In summary, conducting research on source-grid-load carbon panoramic monitoring methods based on multi-timescale situational awareness is of great necessity for addressing issues such as insufficient data integration, a single timescale for situational awareness, and poor coordination in multi-objective decision-making, and for promoting the deep low-carbon, safe, and efficient operation of new power systems. Summary of the Invention

[0008] This invention aims to at least solve one of the technical problems existing in the prior art. Therefore, one objective of this invention is to propose a source-grid-load carbon panoramic monitoring method based on multi-timescale situational awareness. This method can effectively achieve panoramic monitoring of the entire process of power system power sources, grids, loads, energy conversion and storage equipment, and carbon emissions.

[0009] To address the aforementioned problems, this invention provides a source-grid charged carbon panoramic monitoring method based on multi-timescale situational awareness, comprising the following steps: Step S101: Data acquisition and aggregation step. Through sensors and monitoring devices deployed on the power supply side, grid side, load side and energy storage side, multi-source heterogeneous data are collected, including power output, grid power flow, load data, energy storage status and carbon emission data. The data is then transmitted to the data platform via communication links for cleaning, standardization and fusion processing to form a unified data resource pool covering the dimensions of source, grid, load, storage and carbon. Step S102: Situational awareness step. Based on a unified data resource pool, real-time perception of the physical state and carbon flow information of the power system is achieved through state estimation and data fusion technology. The state estimation adopts the weighted least squares method to solve the state variables of the entire network. The data fusion adopts the Kalman filter algorithm for electrical quantities with high real-time requirements and the DS evidence theory for equipment state and carbon flow data. Step S103: Situational understanding step. Based on the situational awareness results, construct a multi-dimensional assessment system for safety, low carbon and economy, and conduct correlation analysis, including assessing grid safety through N-1 fault simulation, assessing low carbon progress through carbon emission accounting model, and assessing economic operation benefits through cost accounting model. Step S104: Situation prediction step. Based on the situation understanding results, a multi-timescale prediction model is adopted, including short-timescale prediction and medium-to-long-timescale prediction. The short-timescale prediction uses an LSTM neural network model to predict the system state in the next 1-24 hours, and the medium-to-long-timescale prediction uses an ARIMA model to predict the trend in the next 1-7 days. Step S105: Decision support step. Based on the situation prediction results, a multi-objective optimization model is constructed under the premise of safety constraints. The objective function is to minimize carbon emissions and operating costs. The model is solved using a genetic algorithm, and the source-grid-load-storage collaborative optimization strategy is output and distributed to each execution unit to achieve collaborative optimization.

[0010] Preferably, the data acquisition and aggregation step in step S101 specifically includes: Phasor measurement units (PMUs) and new energy power prediction devices are deployed on the power supply side to collect power output and carbon emission intensity data from coal-fired power plants, natural gas power plants, wind farms, and photovoltaic power plants. Deploy intelligent substation monitoring devices on the power grid side to collect power flow, bus voltage and equipment status data of transmission lines, and extract network topology data from the information system; Deploy smart meters and load monitoring terminals on the load side to collect load data from industrial users, residential users, and electric vehicle charging piles; Collect charging and discharging status and energy storage capacity data on the energy storage side; Data is transmitted to the data platform via 5G slicing or fiber optic leased lines, and then undergoes cleaning, standardization and fusion processing in sequence: the cleaning stage uses the differentiated 3σ criterion to remove outliers, the standardization stage unifies data units and timestamps, and the fusion stage constructs a spatiotemporal correlation model based on graph neural network (GNN), using power grid nodes as graph nodes and establishing data mapping relationships through a spatiotemporal attention mechanism.

[0011] Preferably, the situational awareness step in step S102 includes a state estimation process, the objective function of which is: in, z i For measurement value, h i (x) For the measurement equation, w i To measure the weight, x A vector of state variables; ; For the PQ node, the measurement equation is: in, U The node voltage amplitude, δ The node voltage phase; In data fusion, the formula for the Kalman filter algorithm is: in, This is the state estimate. A Here is the state transition matrix. P k Let covariance matrix be the variance matrix. K k For Kalman gain, Q and R These are the covariances of process noise and measurement noise, respectively. The formula for the DS evidence theory is: in, K The conflict coefficient, Let i be the basic probability assignment for the i-th piece of evidence.

[0012] Preferably, in the situation prediction step S104, the short-term prediction uses an LSTM neural network model, whose input is a time series composed of historical power flow, power output, load, carbon emissions, and equipment status data. Through training, it learns the time evolution pattern and outputs future hourly prediction values. The medium- and long-term prediction uses an ARIMA model, whose formula is: in, For autoregressive operators, For moving average operators, For difference operators, The noise is white noise, and prediction is performed by determining the difference order d, the autoregression order p, and the moving average order q.

[0013] Preferably, the decision support step in step S105 is based on safety constraints, including that the voltage deviation after an N-1 fault is no greater than 5% and the line load rate is no greater than 80%. The multi-objective optimization model is solved using a genetic algorithm, and the output strategies include adjusting the output plans of wind farms and photovoltaic power plants, guiding electric vehicles to stagger their charging times, and optimizing the charging and discharging time of energy storage.

[0014] Preferably, the method further includes a data verification step, which is integrated into the situation awareness step. This step involves verifying the consistency of the measurement data, encrypting the security data using the national cryptographic SM4 algorithm, completing the integrity verification by linear interpolation to fill in the missing data, and correcting the deviation data based on the 3σ criterion.

[0015] Preferably, the data update frequency of the unified data resource pool is consistent with that of the source data, with high-frequency data at 50Hz and low-frequency data at 15 minutes to 1 hour to ensure real-time performance.

[0016] The source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness of the present invention has the following advantages compared with the prior art: (1) This method breaks down the barriers between multi-source heterogeneous data by collecting data across multiple links of source, grid, load and storage, and realizes full-domain perception of power physical state and carbon flow information, thus solving the problem of cognitive blind spots caused by insufficient traditional data integration.

[0017] (2) This method adopts short and medium time scale prediction models to match the hourly real-time scheduling and daily and weekly plan optimization requirements, respectively, improves the fitting degree of the system state evolution trend, more accurately identifies potential risks in different scenarios, and makes up for the limitations of single time scale research.

[0018] (3) The decision optimization of this method closely follows actual needs. With safety as the premise, a multi-objective collaborative optimization model of safety, low carbon and economy is constructed. Through scientific algorithms, a collaborative strategy for all links is output, avoiding the single-objective tendency of traditional technology and meeting the comprehensive decision-making needs of deep low carbon and efficient operation of the new power system. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a diagram illustrating the method architecture of the present invention; Figure 2 This is a flowchart of the power system situational awareness and decision-making execution process for multiple links of the power source, grid, and load in this invention. Detailed Implementation

[0021] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0022] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0023] The present invention will now be described in further detail with reference to the accompanying drawings.

[0024] like Figure 1As shown, this method is divided into three layers: the data source layer, the situation awareness engine, and the decision support layer. The data source layer collects multi-source data, including measurement information, equipment status, network topology, and equipment parameters, through various terminal devices such as TTUs, DTUs, FTUs, PMUs, instrument transformers, automated terminal equipment, and smart meters, and aggregates this data using a SCADA system. The situation awareness engine comprises three stages: situation perception, situation understanding, and situation prediction. Situation perception focuses on data acquisition, processing, and visualization, providing a comprehensive situational insight into the power grid from both breadth and depth. Situation understanding uncovers potential data information, and situation prediction, based on the former two stages, anticipates potential system security risks. The decision support layer, with low-carbon, safe, and economical goals, achieves panoramic perception and decision support for deep low-carbon power grid operation, providing a basis for decision-making in the coordinated optimization of power generation, grid, load, and storage.

[0025] like Figure 2 As shown, this method first comprehensively collects multi-source data from the power source side (coal-fired power, natural gas, wind power, photovoltaic), the grid side (energy system, information system), and the load side (industrial load, residential load, electric vehicle) from the environmental state dimension. Then, through the situational awareness stage, it sequentially proceeds through situational perception (achieving comprehensive data capture and initial insight), situational understanding (deep analysis of data correlations and inherent patterns), and situational prediction (predicting system evolution trends and potential risks), forming a comprehensive understanding of the power system's operating status. Based on this, it enters the decision-making and execution stage: on the one hand, through the experience training module, it conducts similar weather identification and similar trend identification, and after training sample selection and neural network training, it optimizes the situational awareness model in conjunction with the predicted goals; on the other hand, it outputs decisions and drives action execution, ultimately achieving coordinated and optimized operation of the entire power source, grid, load, and storage system, supporting the achievement of power system safety, low-carbon, and economic goals.

[0026] The present invention provides a source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness, comprising: (I) Data Collection and Aggregation First, sensors are deployed on the power supply side at coal-fired power plants, natural gas power plants, wind farms, and photovoltaic power stations to collect data such as power output and carbon emission intensity. On the grid side, monitoring devices are deployed at substations and transmission lines to collect data such as power flow, voltage, and equipment status, and network topology data is extracted from the information system. On the load side, smart terminals are deployed at industrial users, residential users, and electric vehicle charging piles to collect load data. On the energy storage side, data such as charging and discharging status and energy storage capacity are collected. This multi-source heterogeneous data is transmitted to a data platform via 5G slicing, fiber optic leased lines, and other communication links. The platform adopts a three-level processing flow of "cleaning, standardization, and fusion." 1. Cleaning phase: Differentiated 3σ criteria are adopted for data with different sampling frequencies. For high-frequency data (such as PMU), the mean and standard deviation are calculated with a statistical window of 1 minute to remove outliers that exceed the range of [μ-3σ, μ+3σ]. For low-frequency data (such as smart meters), the statistical window is 1 hour. At the same time, secondary verification is carried out in combination with power system operation rules to avoid the accidental deletion of valid data. 2. Standardization Phase: Based on GB / T 19862-2019 Power System Dispatch Automation System, unified data units are implemented, with active power standardized as "MW", voltage as "kV", carbon emissions as "tCO2", and timestamps standardized as UTC+8. 3. Integration Phase: A spatiotemporal correlation model based on Graph Neural Network (GNN) is constructed, using grid nodes as graph nodes and lines as edges. Power output, grid power flow, load data, and carbon flow data are used as node features. A spatiotemporal attention mechanism (time window set at 5 minutes, spatial attention weights allocated based on node electrical distance) is used to establish mapping relationships between data, ultimately forming a unified data resource pool covering the dimensions of "source, grid, load, storage, and carbon." The data update frequency is consistent with the source data (50Hz for high-frequency data, 15 minutes to 1 hour for low-frequency data).

[0027] (ii) Situational awareness Power grid data situational awareness technology is the core support for achieving comprehensive, accurate, and real-time perception of the power system. Its technical system relies on massive intelligent terminals (PMUs, smart meters, etc.), advanced measurement systems (AMIs), and high-speed communication technologies to build a "comprehensive, deeply accurate, and real-time reliable" power carbon data perception network, providing a high-quality data foundation for collaborative perception of power sources, grids, loads, and energy storage.

[0028] On the power supply side, phasor measurement units (PMUs) and new energy power prediction devices are deployed to collect data such as output and carbon emission intensity of power sources such as coal-fired power, wind power, and photovoltaic power in real time. On the grid side, relying on the digital measurement system of smart substations, electrical and status quantities such as power flow of transmission lines, bus voltage, and equipment temperature are collected. On the load side, data on electricity load and electricity consumption patterns of users such as residents, industries, and electric vehicles are obtained through smart meters (AMIs) and load monitoring terminals. At the same time, carbon emission monitoring sensors are configured at key nodes to realize the direct measurement of carbon flow data.

[0029] For areas with weak measurement capabilities, state estimation and data fusion techniques are used to compensate for missing data. State estimation involves establishing a power grid mathematical model, combining it with existing measurement data, and using the weighted least squares method to solve for the state variables (node ​​voltage amplitude and phase) of the entire network. The formula is as follows: in, For measurement value, For the measurement equation, For measurement weights. for State variables.

[0030] For PQ nodes: Where U is the node voltage amplitude and δ is the node voltage phase.

[0031] Measurement data is aggregated by local edge computing nodes (deployed in 220kV and above substations) and then transmitted to the Measurement Data Management System (MDMS) via high-speed communication channels (5G, fiber optic private network). MDMS performs multi-dimensional verification of the data. ① Consistency verification: Verify the data deviation between different measuring devices at the same node. ② Security verification: Data encryption verification based on the national cryptographic algorithm SM4 is used to prevent data tampering; ③ Integrity verification: Based on the data packet loss rate statistics, lost data is filled by linear interpolation (time interval < 5 minutes). ④ Accuracy verification: Identify bad data based on the 3σ criterion, correct biased data by combining state estimation results, and finally form standardized data resources.

[0032] Ultimately, this will form standardized data resources that serve the power grid and users, providing multi-dimensional and high-quality data input for subsequent situational awareness.

[0033] In terms of technical implementation, a multi-source data fusion algorithm is adopted to perform hierarchical fusion of various data types, including electrical quantities, equipment status quantities, and carbon flow. For electrical quantities with high real-time requirements, a Kalman filter algorithm is used for dynamic fusion to correct the measured values ​​in real time. The formula is as follows: in, This is the state estimate. Here is the state transition matrix. Let covariance matrix be the variance matrix. For Kalman gain, , These are the process noise and measurement noise covariances, respectively, to ensure the real-time performance and accuracy of electrical quantities.

[0034] For static or slowly varying data such as equipment status and carbon flow, DS evidence theory is used for fusion. By establishing a basic probability allocation function for multi-source data and using evidence synthesis rules to handle uncertainties between data, the formula is as follows: in, The conflict coefficient, Assign a basic probability to the i-th piece of evidence to achieve a comprehensive judgment of equipment status and carbon flow information.

[0035] (III) Situational Understanding Situational understanding reveals the inherent laws and potential contradictions in power grid operation by constructing a multi-dimensional assessment system and correlation analysis models. Its technical architecture revolves around three core objectives: "safety, low carbon, and economy," establishing multi-level assessment indicators and correlation analysis models.

[0036] In terms of safety, a method combining N-1 fault simulation and equipment load rate analysis is employed to calculate the load rate of each device in the power grid in real time. By combining historical fault data and equipment health status, the system's resilience to interference during single-device failures is assessed. In terms of low-carbon performance, a carbon emission accounting model is established, comprehensively considering factors such as power source type, generation capacity, and operating time to accurately calculate the system's total carbon emissions. This is then compared with regional low-carbon targets to assess progress towards these targets. Furthermore, the impact of different power source output combinations on carbon emissions is analyzed to identify potential areas for low-carbon operation. In terms of economics, a cost accounting model is constructed, covering power generation costs (coal-fired power, natural gas power, and maintenance costs of renewable energy generation), grid loss costs (power loss costs based on power flow calculations), and ancillary service costs. Combined with electricity market transaction prices, the system's economic efficiency is evaluated. Finally, correlation analysis algorithms are used to uncover the inherent relationships between safety, low-carbon, and economic indicators.

[0037] (iv) Situation forecast (1) Short timescale prediction (LSTM neural network model) For short-term timescale predictions of the next 1-24 hours, a Long Short-Term Memory (LSTM) neural network model is used, which is specifically designed to handle the long-short-term dependency problem of time series data.

[0038] First, a multi-dimensional feature vector is constructed, including historical data on power flow, power output, load, carbon emissions, and equipment status. This data is arranged chronologically as a sequence and used as model input. The core of the LSTM model is the memory unit, which includes an input gate, a forget gate, and an output gate. The input gate determines how new information is added to the memory unit, the forget gate controls the retention of old information in the memory unit, and the output gate determines how the information in the memory unit is output to the hidden layer. During training, the historical time-series data of source, grid, load, and carbon emissions are divided into training and validation sets. The model's weight parameters are continuously adjusted using the backpropagation algorithm, enabling the model to learn the temporal evolution patterns in the data. After training, by inputting the current and historical multi-dimensional feature vectors, the model can output predicted values ​​for power flow, power output, load, and carbon emissions for the next few hours. Furthermore, through a risk identification algorithm, it identifies potential risks such as equipment overload risk and carbon emission exceedance risk, providing accurate auxiliary decision-making basis for real-time grid dispatching.

[0039] (2) Medium and long-term scale prediction (ARIMA model) To address the planned optimization needs for the next 1-7 days, an Autoregressive Moving Average (ARIMA) model is used to analyze the medium- to long-term trend of the power grid status. The ARIMA model formula is: in, For autoregressive operators, For moving average operators, For difference operators, It is white noise. By determining the difference order d, the autoregression order p, and the moving average order q, ARIMA can predict future daily or weekly trends in grid output, load, and carbon emissions, providing a basis for grid maintenance plans and renewable energy consumption plans.

[0040] (v) Decision support With safety, low carbon emissions, and economic efficiency as core objectives, a multi-objective optimization decision-making model is constructed based on situational prediction results. The model is built under the premise of safety constraints (e.g., voltage deviation ≤5% after N-1 fault, line load rate ≤80%), with the objective functions being minimizing carbon emissions and minimizing operating costs. A genetic algorithm is used to solve the model, outputting source-grid-load-storage collaborative optimization strategies, such as adjusting the output plans of wind farms and photovoltaic power plants, guiding off-peak charging of electric vehicles, and optimizing the charging and discharging time of energy storage. These decision instructions are then distributed to each execution unit to achieve collaborative optimization execution.

[0041] Finally, for all aspects of this invention, mature products and technologies from the prior art are used.

[0042] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness, characterized in that: Includes the following steps: Step S101: Data acquisition and aggregation step. Through sensors and monitoring devices deployed on the power supply side, grid side, load side and energy storage side, multi-source heterogeneous data are collected, including power output, grid power flow, load data, energy storage status and carbon emission data. The data is then transmitted to the data platform via communication links for cleaning, standardization and fusion processing to form a unified data resource pool covering the dimensions of source, grid, load, storage and carbon. Step S102: Situational awareness step. Based on the unified data resource pool, real-time perception of the physical state and carbon flow information of the power system is achieved through state estimation and data fusion technology. The state estimation adopts the weighted least squares method to solve the state variables of the entire network. The data fusion adopts the Kalman filter algorithm for electrical quantities with high real-time requirements and the DS evidence theory is adopted for equipment state and carbon flow data. Step S103: Situational understanding step. Based on the situational awareness results, construct a multi-dimensional assessment system for safety, low carbon and economy, and conduct correlation analysis, including assessing grid safety through N-1 fault simulation, assessing low carbon progress through carbon emission accounting model, and assessing economic operation benefits through cost accounting model. Step S104: Situation prediction step. Based on the situation understanding results, a multi-timescale prediction model is adopted, including short-timescale prediction and medium-to-long-timescale prediction. The short-timescale prediction uses an LSTM neural network model to predict the system state in the next 1-24 hours, and the medium-to-long-timescale prediction uses an ARIMA model to predict the trend in the next 1-7 days. Step S105: Decision support step. Based on the situation prediction results, a multi-objective optimization model is constructed under the premise of safety constraints. The objective function is to minimize carbon emissions and operating costs. The model is solved using a genetic algorithm, and the source-grid-load-storage collaborative optimization strategy is output and distributed to each execution unit to achieve collaborative optimization.

2. The source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness according to claim 1, characterized in that: The data acquisition and aggregation steps in step S101 specifically include: Phasor measurement units (PMUs) and new energy power prediction devices are deployed on the power supply side to collect power output and carbon emission intensity data from coal-fired power plants, natural gas power plants, wind farms, and photovoltaic power plants. Deploy intelligent substation monitoring devices on the power grid side to collect power flow, bus voltage and equipment status data of transmission lines, and extract network topology data from the information system; Deploy smart meters and load monitoring terminals on the load side to collect load data from industrial users, residential users, and electric vehicle charging piles; Collect charging and discharging status and energy storage capacity data on the energy storage side; The data is transmitted to the data platform via 5G slicing or fiber optic leased lines, and then undergoes cleaning, standardization and fusion processing in sequence: the cleaning stage uses the differentiated 3σ criterion to remove outliers, the standardization stage unifies the data units and timestamps, and the fusion stage constructs a spatiotemporal correlation model based on graph neural network (GNN), with power grid nodes as graph nodes, and establishes data mapping relationships through a spatiotemporal attention mechanism.

3. The source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness according to claim 1, characterized in that: The situation awareness step in step S102 includes a state estimation process, the objective function of which is: ; in, z i For measurement value, h i (x) For the measurement equation, w i To measure the weight, x A vector of state variables; ； For the PQ node, the measurement equation is: ; in, U The node voltage amplitude, δ The node voltage phase; In the data fusion process, the formula for the Kalman filter algorithm is: ; in, This is the state estimate. A Here is the state transition matrix. P k Let covariance matrix be the variance matrix. K k For Kalman gain, Q and R These are the covariances of process noise and measurement noise, respectively. The formula for the DS evidence theory is: ; in, K The conflict coefficient, Let i be the basic probability assignment for the i-th piece of evidence.

4. The source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness according to claim 1, characterized in that: In the situation prediction step S104, the short-term prediction uses an LSTM neural network model, whose input is a time series composed of historical power flow, power output, load, carbon emissions, and equipment status data. Through training, it learns the time evolution pattern and outputs future hourly prediction values. The medium- and long-term prediction uses an ARIMA model, whose formula is: ; in, For autoregressive operators, For moving average operators, For difference operators, The noise is white noise, and prediction is performed by determining the difference order d, the autoregression order p, and the moving average order q.

5. The source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness according to claim 1, characterized in that: The decision support step in step S105 is based on safety constraints, including that the voltage deviation after an N-1 fault is no greater than 5% and the line load rate is no greater than 80%. The multi-objective optimization model is solved using a genetic algorithm, and the output strategies include adjusting the output plans of wind farms and photovoltaic power plants, guiding electric vehicles to stagger their charging times, and optimizing the charging and discharging time of energy storage.

6. The source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness according to claim 1, characterized in that: The method also includes a data verification step, which is integrated into the situation awareness step. The measurement data is verified for consistency, security using the national cryptographic SM4 algorithm, integrity by linear interpolation to complete lost data, and accuracy by correcting deviation data based on the 3σ criterion.

7. The source-grid-charged carbon panoramic monitoring method based on multi-timescale situational awareness according to claim 1, characterized in that: The data update frequency of the unified data resource pool is consistent with that of the source data, with high-frequency data at 50Hz and low-frequency data at 15 minutes to 1 hour to ensure real-time performance.

Images