An event-triggered microgrid energy scheduling method and system
By constructing a dynamic graph structure in the microgrid, adjusting the edge weights using the Pearson correlation coefficient, and combining it with a spatiotemporal graph neural network for event monitoring, the problem of inaccurate monitoring caused by the fixed graph structure is solved, and accurate monitoring and scheduling of key events in the microgrid are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TIANJIN PORT (GROUP) COMPANY
- Filing Date
- 2026-05-13
- Publication Date
- 2026-06-09
AI Technical Summary
In existing technologies, when graph neural networks use a fixed graph structure for event monitoring in microgrids, they cannot adapt to the state changes of dynamic systems, resulting in the neglect of the coupling relationship of key variables and affecting the accuracy of event monitoring.
By dynamically adjusting edge weights through the calculation of Pearson correlation coefficients between variables, a dynamic graph structure is constructed. This is combined with a spatiotemporal graph neural network for event monitoring, generating a dynamic adjacency matrix and strengthening the coupling relationship between key variables.
It enables accurate monitoring and scheduling of critical events in the dynamic system of microgrids, improving the accuracy and reliability of event monitoring.
Smart Images
Figure CN122178569A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of microgrid dispatching technology, and in particular to a microgrid energy dispatching method and system based on event triggering. Background Technology
[0002] In the operation of microgrids, timely and accurate monitoring of key events that affect the safe, stable, or economical operation of the system, and triggering corresponding dispatch strategies, is the core of ensuring the efficient and reliable operation of microgrids.
[0003] In existing technologies, graph neural network (GNN) structures are used to monitor and determine microgrid events. However, the core of a GNN is its graph structure, which is a topological structure composed of nodes and edges. In traditional GNNs, the graph structure is usually fixed, with the number of nodes and edge connections predefined and unchanged. In dynamic systems like microgrids, the operating state of the microgrid changes the strength of dependencies between variables. For example, under normal operating conditions, the correlation between photovoltaic (PV) output and load is relatively strong. However, under power imbalance conditions, when PV output suddenly drops, the correlation between load and energy storage output increases sharply. If a fixed graph structure is used, the GNN will continuously focus on the coupling between PV processing and load, but will ignore the more critical correlation between energy storage output and load during power imbalance, leading to feature extraction failure and affecting the accuracy of microgrid event monitoring. Summary of the Invention
[0004] To address the aforementioned technical problems, this invention provides an event-triggered microgrid energy dispatching method and system to resolve the issues existing in the prior art.
[0005] This invention provides an event-triggered microgrid energy dispatching method, the method comprising the following steps: S1: Collect multi-source operation data of the microgrid through the data acquisition unit; S2: Perform data preprocessing on the multi-source operating data; S3: Monitor and judge events of the microgrid based on the preprocessed multi-source operation data, and perform microgrid energy dispatch; S3 specifically comprises: S3.1: constructing time series data from the preprocessed multi-source operational data; S3.2: inputting the time series data into a fixed threshold rule channel for event monitoring to obtain a first event monitoring result; S3.3: inputting the time series data into a spatiotemporal graph neural network model for event monitoring to obtain a second event monitoring result; S3.3 specifically comprises: converting the time series data into a dynamic adjacency matrix for input to the spatiotemporal graph neural network model; converting the time series data into a dynamic adjacency matrix specifically comprises: defining each variable in the time series data as a node in a graph structure, defining the connection relationship between nodes as an edge, calculating the Pearson correlation coefficient between every two variables within a fixed window; determining the edge weight based on the Pearson correlation coefficient between every two variables within a fixed window; generating a dynamic adjacency matrix based on the edge weight; The dynamic adjacency matrix is input into the spatiotemporal graph neural network model, and the spatiotemporal graph neural network model outputs the event judgment result as the second event monitoring result; S3.4: Integrate the first event monitoring results and the second event monitoring results to obtain the final event monitoring results of the microgrid, and perform microgrid energy dispatch.
[0006] Preferably, variable X i X j The Pearson correlation coefficient ρ at time t ij The formula for calculating (t) is: ; In the formula, For variable X i X j The covariance at time t For variable X i The variance at time t For variable X j The variance at time t.
[0007] Preferably, the formula for calculating edge weights is: ; In the formula, For variable X i X j The edge weight of the formed edge.
[0008] Preferably, in step S1, data acquisition units are deployed at each key node of the microgrid. These data acquisition units include: a smart combiner monitoring unit installed at the photovoltaic array outlet to capture real-time output information of the photovoltaic power source; a controller integrated within the wind turbine generator set to acquire real-time output information of the wind power source; a battery management system embedded in the energy storage battery pack to collect information on the energy storage system's state of charge, charging / discharging power limits, and health status; smart meters and load monitoring terminals distributed at each load center or feeder circuit to measure and sense the total load power demand; and a synchronous phasor measurement device or remote terminal unit deployed at network hub points to collect system frequency, key node voltage amplitude and phase angle, and network power flow distribution.
[0009] Preferably, the data acquisition unit is also connected to an external information system via a secure communication link to periodically acquire ultra-short-term and short-term forecast data of renewable energy output, load forecast data, and real-time electricity market price information.
[0010] Preferably, in step S2, the preprocessing operation includes data frame integrity verification, data cleaning, and data smoothing.
[0011] Preferably, the data frame integrity verification specifically involves: performing integrity verification on each received data packet, and confirming that no bit errors or packet loss occurred during the transmission of the data packet by verifying its built-in check code; The data cleaning process specifically involves: employing a multi-level judgment logic to identify outlier data points. The first level is a physical limit check based on prior knowledge, setting a theoretically reasonable range of values for each state variable. Any measurement value exceeding this range will be initially identified as an outlier. The second level is anomaly detection based on statistical regularities. By establishing a sliding observation window on the time axis, the statistical characteristics of the data within the observation window are calculated. Any data point exceeding three standard deviations from the mean will be confirmed as an outlier. For confirmed outliers, a linear interpolation method based on time series is used for repair. If multiple consecutive data points are missing, the data in the time period containing the missing data points will be marked as invalid. The data smoothing process specifically involves using a digital filtering algorithm to smooth the data.
[0012] Preferably, in step S3.1, the window length of the time series data is 60 seconds and the sampling period is 1 second.
[0013] Preferably, step S3.2 specifically involves: executing a cyclic task at a preset frequency; receiving the time series data within each judgment cycle; then traversing the event triggering condition library; locating the corresponding variable value from the time series data based on the triggering conditions in the event triggering condition library; performing logical operations based on the variable value and a fixed threshold in the triggering conditions; if the logical operation result is true, the triggering condition is triggered; otherwise, the triggering condition is not triggered, thereby outputting the event judgment result as the first event monitoring result.
[0014] According to another aspect of the present invention, an event-triggered microgrid energy dispatching system is provided, the system employing the above-described event-triggered microgrid energy dispatching method, the system comprising: The data acquisition unit collects multi-source operation data of the microgrid; The data preprocessing unit is used to perform data preprocessing operations on the multi-source running data; The event monitoring and dispatching unit is used to monitor and judge events in the microgrid based on preprocessed multi-source operating data, and to perform energy dispatching in the microgrid.
[0015] The present invention has the following technical effects: In this invention, when using a graph neural network model for microgrid event monitoring, the correlation coefficients between variables are calculated during the graph structure construction process. As the operating conditions of the microgrid change, the correlation coefficients also change, leading to changes in the correlation coefficients between variables and updates in the edge weights. Through this scheme, a dynamic graph structure can be generated, and the dynamic graph structure always strengthens the coupling relationship of the most critical variables. Based on the dynamic graph structure, accurate monitoring of microgrid events can be achieved. Attached Figure Description
[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0017] Figure 1 This is a flowchart of an event-triggered microgrid energy dispatching method provided in an embodiment of the present invention; Figure 2 This is a flowchart provided by an embodiment of the present invention for monitoring and judging events in a microgrid and scheduling energy in a microgrid based on preprocessed multi-source operating data. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0019] Example 1, Figure 1 A flowchart of an event-triggered microgrid energy dispatching method is shown, such as... Figure 1 As shown, an event-triggered microgrid energy dispatching method includes the following steps: S1: Collect multi-source operation data of the microgrid through the data acquisition unit; Data acquisition units are deployed at various key nodes of the microgrid. These units include, but are not limited to: intelligent combiner monitoring units installed at the photovoltaic array outlets to capture real-time photovoltaic power output information; controllers integrated within wind turbine generators to acquire real-time wind power output information; battery management systems embedded in energy storage battery packs to collect information on the energy storage system's state of charge, charging and discharging power limits, and health status; smart meters and load monitoring terminals distributed at various load centers or feeder circuits to measure and sense the total load power demand; and synchronous phasor measurement devices or remote terminal units deployed at network hubs to acquire system frequency, key node voltage amplitude and phase angle, and network power flow distribution with high precision. Furthermore, these data acquisition units are connected to external information systems via secure communication links to periodically or on-demand acquire ultra-short-term and short-term renewable energy output forecasts, load forecasts, and real-time electricity market price information.
[0020] S2: Perform data preprocessing on the multi-source operating data; Raw multi-source operational data inevitably contains noise, outliers, and missing values introduced by factors such as measurement errors, environmental interference, and communication jitter, making it unsuitable for direct and accurate analysis and decision-making. Therefore, it is necessary to preprocess the raw multi-source operational data to improve its quality and usability.
[0021] In this step, the preprocessing operations include data frame integrity verification, data cleaning, and data smoothing. The data frame integrity verification specifically involves performing integrity verification on each received data packet, and confirming that no bit errors or packet loss occurred during the transmission of the data packet by verifying its built-in check code.
[0022] The data cleaning process specifically involves: employing a multi-level judgment logic to identify outlier data points. The first level is a physical limit check based on prior knowledge, setting a theoretically reasonable value range for each state variable. Any measurement value exceeding this range, such as a negative state of charge or exceeding 100%, or a system frequency significantly lower or higher than the rated value, will be initially identified as an outlier. The second level is anomaly detection based on statistical regularities. By establishing a sliding observation window on the time axis, the statistical characteristics of the data within the observation window, such as the mean and standard deviation, are calculated. Any data point exceeding three standard deviations from the mean will be confirmed as an outlier. For confirmed outliers, linear interpolation based on time series is preferentially used for repair. This involves using adjacent normal data points to generate replacement values, thereby maintaining the continuity of the data sequence. If multiple consecutive data points are missing and cannot be repaired by interpolation, the data in the time period containing these missing data points will be marked as invalid.
[0023] The data smoothing process specifically involves using a digital filtering algorithm to smooth the data. This digital filtering algorithm can effectively reduce the influence of high-frequency noise, making the data curve more smoothly reflect the true trend while retaining its main dynamic characteristics.
[0024] S3: Monitor and judge events of the microgrid based on the preprocessed multi-source operation data, and perform microgrid energy dispatch; In this step, a multi-layered event verification system is constructed for the accurate identification of microgrid events during event monitoring and judgment. Specifically, such as... Figure 2 As shown, S3 specifically includes: S3.1: Construct time series data from the preprocessed multi-source runtime data; To capture the temporal evolution and cross-variable coupling relationship of the multi-source operational data, multi-source operational data over a past period is saved and formed into time series data of a fixed length; in this step, the window length of the time series data is 60 seconds and the sampling period is 1 second.
[0025] S3.2: Input the time series data into a fixed threshold rule channel for event monitoring to obtain the first event monitoring result; In this step, the fixed threshold rule channel includes an event trigger condition library, which stores the trigger conditions for all events. Each trigger condition is a structured data object, and each data object contains at least the following elements: an event identifier, used to locate the event in a subsequent flag register; an event description, used to provide a textual description of the event, such as describing the event as an emergency event due to excessively low system frequency; an associated variable, used to specify the specific state variable on which the trigger condition depends; comparison operators, used to define the logical relationship between the variable value and the fixed threshold, such as less than (<), greater than (>), less than or equal to (<=), etc.; a fixed threshold, which is a specific numerical value used as a benchmark for determining whether an event is triggered; a priority, used to define the processing priority of the event, typically with emergency events having a higher priority than economic events; and a validity flag indicating whether the rule is effective under the current operating conditions, for example, in grid-connected mode, the rule for an islanded mode switching event can be set to invalid.
[0026] S3.2 specifically involves: executing a cyclic task at a preset frequency; receiving the time series data within each judgment cycle; then traversing the event triggering condition library; locating the corresponding variable value from the time series data based on the triggering conditions in the event triggering condition library; and performing logical operations based on the variable value and a fixed threshold in the triggering conditions. If the logical operation result is true, the triggering condition is triggered; otherwise, the triggering condition is not triggered, thereby outputting the event judgment result as the first event monitoring result.
[0027] S3.3: Input the time series data into the spatiotemporal graph neural network model for event monitoring to obtain the second event monitoring result; Specifically, S3.3 is as follows: The time series data is converted into a dynamic adjacency matrix, which is then used as input to the spatiotemporal graph neural network model. The core of a Graph Neural Network (GNN) is the graph structure, a topological structure composed of nodes and edges. In traditional GNNs, the graph structure is usually fixed, with the number of nodes and edge connections predefined and unchanged. However, in dynamic systems like microgrids, the operating state of the microgrid alters the dependence strength between variables. For example, under normal operating conditions, the correlation between photovoltaic (PV) output and load is relatively strong. However, under power imbalance conditions, when PV output suddenly drops, the correlation between load and energy storage output increases dramatically. If a fixed graph structure is used, the GNN will continuously focus on the coupling between PV processing and load, but will ignore the more critical correlation between energy storage output and load during power imbalance, leading to feature extraction failure. Therefore, this embodiment proposes a dynamic graph structure construction method for microgrids, allowing the graph structure to evolve dynamically with the real-time operating conditions of the microgrid and dynamically adjusting the edge connection weights between nodes, ensuring that the graph structure always focuses on the coupling relationship of the most critical variable at the moment. Specifically, in this step, converting the time series data into a dynamic adjacency matrix involves: Each variable in the time series data is defined as a node in a graph structure, and the connection between nodes is defined as an edge. The Pearson correlation coefficient between any two variables within a fixed window is calculated. Wherein, variable X i X j The Pearson correlation coefficient ρ at time t ij The formula for calculating (t) is: ; In the formula, For variable X i X j The covariance at time t For variable X i The variance at time t For variable X j The variance at time t.
[0028] The edge weights are determined based on the Pearson correlation coefficient between each pair of variables within a fixed window. In traditional graph neural network structures, edge weights are fixed. However, in the dynamic graph construction process of this step, the edge weights are determined by the absolute value of the correlation coefficient matrix; the specific formula is as follows: ; In the formula, For variable X i X j The boundary weights formed; A dynamic adjacency matrix is generated based on the edge weights.
[0029] In this step, by calculating the correlation coefficient between variables, the correlation coefficient will change when the operating conditions of the microgrid change, causing the correlation coefficient between variables to change accordingly, and the edge weights will be updated. Through this scheme, a dynamic graph structure can be generated, and the dynamic graph structure always strengthens the coupling relationship of the most critical variables.
[0030] The dynamic adjacency matrix is input into the spatiotemporal graph neural network model, and the spatiotemporal graph neural network model outputs the event judgment result as the second event monitoring result; The structure and training process of the spatiotemporal graph neural network model are existing technologies and will not be discussed in detail in this embodiment.
[0031] S3.4: Combine the first event monitoring results and the second event monitoring results to obtain the final event monitoring results of the microgrid, and perform microgrid energy dispatch; The fusion rule for the first event monitoring result and the second event is as follows: if the first event monitoring result and the second event monitoring result are the same, then the event monitoring result is taken as the final event monitoring result of the microgrid and marked as high confidence; if only the second event monitoring result is triggered, then the final event monitoring result of the microgrid is set as an event trend warning. Since the graph neural network model has strong predictive power, microgrid energy dispatch is not performed first. If only the first event monitoring result is triggered, then the first event monitoring result is set as the final event monitoring result of the microgrid, marked as low confidence, and reported to the operation and maintenance personnel for further prediction.
[0032] Example 2: The present invention also provides an event-triggered microgrid energy dispatching system, which adopts an event-triggered microgrid energy dispatching method according to Example 1. The system includes: The data acquisition unit collects multi-source operation data of the microgrid; The data preprocessing unit is used to perform data preprocessing operations on the multi-source running data; The event monitoring and dispatching unit is used to monitor and judge events in the microgrid based on preprocessed multi-source operating data, and to perform energy dispatching in the microgrid.
[0033] Example 3: The present invention also provides an electronic device, including one or more processors and a memory.
[0034] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.
[0035] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may execute the program instructions to implement an event-triggered microgrid energy dispatching method based on any embodiment of this application described above, and / or other desired functions. Various contents such as initial extrinsic parameters and thresholds may also be stored in the computer-readable storage medium.
[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.
Claims
1. A microgrid energy dispatching method based on event triggering, characterized in that, The method includes the following steps: S1: Collect multi-source operation data of the microgrid through the data acquisition unit; S2: Perform data preprocessing operations on the multi-source operating data; S3: Monitor and judge events of the microgrid based on the preprocessed multi-source operation data, and perform microgrid energy dispatch; S3 specifically comprises: S3.1: constructing time series data from the preprocessed multi-source operational data; S3.2: inputting the time series data into a fixed threshold rule channel for event monitoring to obtain a first event monitoring result; S3.3: inputting the time series data into a spatiotemporal graph neural network model for event monitoring to obtain a second event monitoring result; S3.3 specifically comprises: converting the time series data into a dynamic adjacency matrix for input to the spatiotemporal graph neural network model; converting the time series data into a dynamic adjacency matrix specifically comprises: defining each variable in the time series data as a node in a graph structure, defining the connection relationship between nodes as an edge, calculating the Pearson correlation coefficient between every two variables within a fixed window; determining the edge weight based on the Pearson correlation coefficient between every two variables within a fixed window; generating a dynamic adjacency matrix based on the edge weight; The dynamic adjacency matrix is input into the spatiotemporal graph neural network model, and the spatiotemporal graph neural network model outputs the event judgment result as the second event monitoring result; S3.4: Integrate the first event monitoring results and the second event monitoring results to obtain the final event monitoring results of the microgrid, and perform microgrid energy dispatch.
2. The event-triggered microgrid energy dispatching method according to claim 1, characterized in that, VARIABLE X i X j The Pearson correlation coefficient ρ at time t ij The formula for calculating (t) is: ； In the formula, For variable X i X j The covariance at time t, For variable X i The variance at time t For variable X j The variance at time t.
3. The microgrid energy dispatching method based on event triggering according to claim 2, characterized in that, The formula for calculating edge weight is: ； In the formula, For variable X i X j The edge weight of the formed edge.
4. The event-triggered microgrid energy dispatching method according to claim 1, characterized in that, In step S1, data acquisition units are deployed at various key nodes of the microgrid. These data acquisition units include: a smart combiner monitoring unit installed at the photovoltaic array outlet to capture real-time output information of the photovoltaic power source; a controller integrated inside the wind turbine generator to acquire real-time output information of the wind power source; a battery management system embedded in the energy storage battery pack to collect information on the state of charge, charging and discharging power limits, and health status of the energy storage system; smart meters and load monitoring terminals distributed at various load centers or feeder circuits to measure and sense the power demand of the total load; and a synchronous phasor measurement device or remote terminal unit deployed at network hub points to collect system frequency, voltage amplitude and phase angle of key nodes, and network power flow distribution.
5. The event-triggered microgrid energy dispatching method according to claim 4, characterized in that, The data acquisition unit is also connected to an external information system via a secure communication link to periodically acquire ultra-short-term and short-term forecast data of renewable energy output, load forecast data, and real-time electricity market price information.
6. The microgrid energy dispatching method based on event triggering according to claim 1, characterized in that, In step S2, the preprocessing operation includes data frame integrity verification, data cleaning, and data smoothing.
7. The microgrid energy dispatching method based on event triggering according to claim 6, characterized in that, The data frame integrity verification specifically involves: performing integrity verification on each received data packet, and confirming that no bit errors or packet loss occurred during the transmission of the data packet by verifying its built-in check code; The data cleaning process specifically involves: using multi-level judgment logic to identify abnormal data points. The first level is a physical limit check based on prior knowledge, setting a theoretically reasonable range of values for each state variable. Any measurement value that exceeds this range will be initially identified as an abnormal point. The second level is anomaly detection based on statistical regularity. By setting up a sliding observation window on the time axis, the statistical characteristics of the data within the observation window are calculated. Any data point that exceeds three standard deviations from the mean will be identified as an outlier. For the identified outliers, a linear interpolation method based on time series is used for repair. If multiple consecutive data points are missing, the data in the time period where the multiple missing data points are located will be marked as invalid. The data smoothing process specifically involves using a digital filtering algorithm to smooth the data.
8. The event-triggered microgrid energy dispatching method according to claim 1, characterized in that, In step S3.1, the window length of the time series data is 60 seconds, and the sampling period is 1 second.
9. A microgrid energy dispatching method based on event triggering according to claim 8, characterized in that, S3.2 specifically involves: executing a cyclic task at a preset frequency; receiving the time series data within each judgment cycle; then traversing the event triggering condition library; locating the corresponding variable value from the time series data based on the triggering conditions in the event triggering condition library; and performing logical operations based on the variable value and a fixed threshold in the triggering conditions. If the logical operation result is true, the triggering condition is triggered; otherwise, the triggering condition is not triggered, thereby outputting the event judgment result as the first event monitoring result.
10. An event-triggered microgrid energy dispatching system, characterized in that, The system employs an event-triggered microgrid energy dispatching method as described in any one of claims 1-9, and the system comprises: The data acquisition unit collects multi-source operation data of the microgrid; The data preprocessing unit is used to perform data preprocessing operations on the multi-source running data; The event monitoring and dispatching unit is used to monitor and judge events in the microgrid based on preprocessed multi-source operating data, and to perform energy dispatching in the microgrid.