A cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints

Abstract

This invention discloses a cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints, belonging to the field of energy system control technology. This invention connects an edge gateway to several energy devices to perform data cleaning, anomaly detection, and missing value completion, generating a standardized time-series dataset. It establishes sub-model knowledge graphs for different energy devices and associates these graphs to construct a physical model knowledge graph for the entire energy system. An LSTM load forecasting model is then established to perform hourly load forecasting of the energy system, obtaining the corresponding load forecast results. Using the physical constraints of all energy devices and the load forecast results as modeling elements, a multi-objective optimization model is constructed to determine the optimal device combination and operating parameters. The decision results are visualized and output, and then sent to the edge gateway. This enables collaborative control of several energy devices based on cloud modules, achieving efficient and stable operation of the energy system.

Description

Technical Field

[0001] This invention relates to the field of energy system control technology, and more specifically to a cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints. Background Technology

[0002] Traditional energy system control suffers from problems such as heterogeneous equipment and non-standard data. Optimizing collaborative control among multiple devices is difficult. When AI models execute decisions, they only provide the final decision result and lack the reasoning of the decision-making process, resulting in poor interpretability. Maintenance personnel have insufficient trust in key decision-making scenarios. It is difficult to balance the real-time performance and reliability of cloud and edge collaborative control, which can easily lead to network outages and loss of control. Furthermore, the lack of deep integration of physical constraints into the management of the energy system leads to deviations between load forecasting and optimization decisions and actual operating patterns, resulting in insufficient overall energy efficiency and operational stability. Therefore, there is an urgent need for a cloud-edge collaborative intelligent control system for energy systems that integrates physical constraints. Summary of the Invention

[0003] The purpose of this invention is to provide a cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints, in order to solve the problems in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a cloud-edge collaborative intelligent control system for energy systems embedded with physical constraints, the system comprising: The edge operation module is used to connect the edge gateway to several types of energy devices, perform data cleaning, anomaly detection and missing value completion for the corresponding energy devices, and generate corresponding standardized time series datasets. The knowledge graph module is used to establish knowledge graphs for the sub-models corresponding to different energy devices, to associate the knowledge graphs of different sub-models, and to construct the physical model knowledge graph corresponding to the energy system composed of all energy devices. The load forecasting module is used to establish an LSTM load forecasting model, perform hourly load forecasting on the energy system based on the LSTM load forecasting model, and obtain the corresponding load forecasting results. The decision output module is used to construct a multi-objective optimization model by taking the physical constraints of all energy equipment in the energy system and the load forecast results as modeling elements, and to make decisions on the optimal equipment combination and operating parameters based on the multi-objective optimization model, and to visualize the decision results. The collaborative control module is used to send decision results to the edge gateway and perform collaborative control on several energy devices based on the cloud module.

[0005] Furthermore, the edge operation module includes: The access unit is used to determine and enable a number of edge gateways, determine a number of categories of energy devices that need to be monitored, connect all energy devices to the enabled edge gateways, and store the corresponding call addresses of each device at the edge gateways. The standardized unit is used to perform data cleaning, anomaly detection, and missing value completion on the energy device data obtained by the edge gateway, and to generate a standardized time-series dataset for each energy device based on the time series.

[0006] Furthermore, the knowledge graph module includes: Sub-graph units are used to obtain the physical information and physical laws of each energy device. Based on the preset graph generation rules, the physical information of the devices is mapped to the blank graph to construct the corresponding sub-model knowledge graph. The graph association unit is used to set registration points, match the registration points with the sub-model knowledge graph, set different registration points to be in free registration lines, and set mounting threads at both ends of the registration lines. The physical laws of the equipment are used as registration conditions. Based on the registration conditions, the mounting threads are operated to select the free registration lines to establish the connection relationship between the corresponding registration points, and construct the physical model knowledge graph of the energy system composed of all energy equipment.

[0007] Furthermore, the map association unit includes: The thread operation unit is used to handle the mounting thread. The mounting thread includes a mounting end, a message end, and an association end. The mounting end is used to determine the mounting target of the registration line, which is the registration point. The message end is used to store the target information of the mounting target. The association end is used to associate data between registration points of the same type of registration line. The registration unit is used to attach the two ends of the free registration line to the corresponding attachment target, read the target information of the current attachment target from the message end, and send the target information to all registration points with data association based on the association end; The storage unit is used to determine the registration point for each received target information, synchronize the mounting data of the corresponding registration line of the sender to all the corresponding registration points of the receiver, and store it in the registration point. When other registration lines of the same type need to be mounted, the mounting data stored in the registration point is called.

[0008] Furthermore, the load forecasting module includes: The modeling unit is used to obtain historical load data, weather data, and time dimension data to build and train an LSTM load prediction model. The load forecasting unit is used to set the model call address of the LSTM load forecasting model, store the LSTM load forecasting model on the cloud server, and each energy device in the energy system operates the LSTM load forecasting model based on the model call address to predict its own hourly load, and use the hourly load as the load forecasting result.

[0009] Furthermore, the decision output module includes: The model building unit is used to combine the graph content recorded in the physical model knowledge graph with the equipment information of pre-deployed redundant equipment, determine the physical constraints of energy equipment, set the physical constraints of each energy equipment and the load prediction results as modeling elements, and build a multi-objective optimization model based on the modeling elements. The decision unit is used to determine the equipment clusters composed of a number of energy devices in the energy system that satisfy the multi-objective optimization model, to determine each equipment cluster as the optimal equipment combination, to obtain the operating parameters of the energy devices under each optimal equipment combination, and to take the optimal equipment combination and the operating parameters of the corresponding included energy devices as the decision result. The visualization output unit is used to establish the reasoning chain for each decision result, set up several visual display points on the reasoning chain, and visualize part of the decision result based on the visual display points.

[0010] Furthermore, the visualization output unit includes: The link generation unit is used to process the decision result into several data segments based on time sequence, establish a sub-inference link between two sequentially adjacent data segments, store the previous data segment and the current data segment at the beginning and end of the path respectively, and store the inference content of the previous data segment to obtain the current data segment on the sub-inference path. Integrate all sub-inference links to obtain the inference link. The visual display unit is used to set the connection position of adjacent sub-inference links as a visual display point, take the inference content of the sub-inference link termination position corresponding to each visual display point as the decision result of the corresponding part, and display it based on the visual display point.

[0011] Furthermore, the collaborative control module includes: The data distribution unit is used to deploy the cloud module and create storage points for all decision results in the cloud module. The storage points are used to store the decision results and the communication addresses of all energy devices included in the corresponding optimal device combination. Based on the communication addresses, the decision results are distributed to the edge gateway at the corresponding optimal device combination. The analysis unit is used to analyze the local optimal operation and maintenance control objectives of different optimal equipment combinations under the energy system, and to determine the global operation and maintenance control objectives corresponding to the entire energy system. The collaborative control unit is used to establish a collaborative relationship function between the local optimal operation and maintenance control objective and the global operation and maintenance control objective. The collaborative relationship function is used to characterize the number of energy devices that meet the local optimal operation and maintenance control objective in their respective optimal equipment combinations when the energy system meets the global operation and maintenance control objective. The corresponding energy devices are controlled based on the collaborative relationship function.

[0012] Furthermore, the analysis unit includes: The local analysis unit is used to analyze the optimal energy efficiency range of each optimal equipment combination in the energy system. When the operating data is in the optimal energy efficiency range, the optimal equipment combination achieves the local optimal operation and maintenance control target in the energy system. The global analysis unit is used to analyze the energy efficiency requirements of the entire energy system during operation and to determine whether the energy system can achieve the global operation and maintenance control objectives if all the optimal equipment combinations under the energy system maintain their current operating conditions. If so, no action is taken; If not, determine the optimal equipment combination that needs operational adjustments, and adjust the operational data of the optimal equipment combination within its respective optimal energy efficiency range until the energy system reaches the overall operation and maintenance control target.

[0013] Furthermore, the collaborative control unit includes: The relational function unit is used to construct the relational function between energy devices in each optimal equipment combination that satisfies the local optimal operation and maintenance control objective, using the optimal energy efficiency range of each optimal equipment combination as the function parameter. The relational function of each optimal equipment combination that satisfies the global operation and maintenance control objective is used as the collaborative relational parameter to construct the collaborative relational function. The control unit is used to set the activation number of each energy device through the cooperative relationship function, create a control thread for each optimal device combination, register the control thread to the cloud module, and the cloud module compiles the control files for the activation number of each energy device in the optimal device combination. The control thread uses the control files to control the corresponding energy devices. Autonomous units are used to distribute control threads to edge gateways and establish corresponding local area networks at the location of each edge gateway. When the cloud module fails to manage any edge gateway, the control thread of the corresponding edge gateway is switched from the cloud module to the local area network, and the local area network performs local control over the corresponding edge gateway.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention eliminates heterogeneous differences by standardizing the data collected by each edge gateway, embedding physical constraints to construct a physical model knowledge graph, making the operation of energy equipment conform to the actual operating rules of the equipment, improving the rationality of decision-making, establishing an LSTM load prediction model for accurate load prediction and combining it with a multi-objective optimization model, outputting the optimal equipment combination and its corresponding operating parameters, and visualizing the decision-making reasoning process of the model to enhance the credibility of key application scenarios. The setting of cloud-edge collaboration and network outage autonomy takes into account both control real-time performance and reliability, ensuring uninterrupted critical loads and improving the overall energy efficiency and operational stability of the system. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a system block diagram of the present invention.

[0017] Figure 2 This is a block diagram of the module structure of the collaborative control module of the present invention.

[0018] Figure 3 This is a schematic diagram of the operation of the autonomous unit of the collaborative control unit in the collaborative control module of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 As shown, a cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints is disclosed. The system includes: The edge operation module is used to connect the edge gateway to several types of energy devices, perform data cleaning, anomaly detection and missing value completion for the corresponding energy devices, and generate corresponding standardized time series datasets. The knowledge graph module is used to establish knowledge graphs for the sub-models corresponding to different energy devices, to associate the knowledge graphs of different sub-models, and to construct the physical model knowledge graph corresponding to the energy system composed of all energy devices. The load forecasting module is used to establish an LSTM load forecasting model, perform hourly load forecasting on the energy system based on the LSTM load forecasting model, and obtain the corresponding load forecasting results. The decision output module is used to construct a multi-objective optimization model by taking the physical constraints of all energy equipment in the energy system and the load forecast results as modeling elements, and to make decisions on the optimal equipment combination and operating parameters based on the multi-objective optimization model, and to visualize the decision results. The collaborative control module is used to send decision results to the edge gateway and perform collaborative control on several energy devices based on the cloud module.

[0021] It should be noted that by connecting each energy device to an edge gateway, collecting its own operational data, and processing it to generate a standardized dataset, which serves as the data basis for the independent operation of each energy device, a physical model knowledge graph corresponding to the energy system composed of all energy devices is established. The load prediction results obtained from the LSTM load prediction model and the physical constraints corresponding to the energy devices are used together as decision parameters for the multi-objective optimization model, resulting in a visualized output of the decision results. On the one hand, the visualized output of the decision results can clearly show the reasoning chain of the decision, allowing relevant operators to understand and trust it, avoiding the drawback of insufficient interpretability of traditional AI decision-making. The combined use of edge gateways and cloud modules solves the problems of difficulty in multi-device collaborative optimization and difficulty in balancing real-time performance and reliability during collaborative control, improving the overall energy efficiency of the energy system and ensuring that critical loads are never interrupted.

[0022] It should be further explained that, in the specific implementation process, the edge operation module includes: The access unit is used to determine and enable a number of edge gateways, determine a number of categories of energy devices that need to be monitored, connect all energy devices to the enabled edge gateways, and store the corresponding call addresses of each device at the edge gateways. The standardized unit is used to perform data cleaning, anomaly detection, and missing value completion on the energy device data obtained by the edge gateway, and to generate a standardized time-series dataset for each energy device based on the time series.

[0023] It should be noted that data cleaning includes: performing field parsing and format normalization on raw data from energy devices of different types (chillers, water pumps, cooling towers, fans, meters, sensors, etc.) and different protocols (Modbus, BACnet, DL / T645, IEC104, and over 30 other industrial protocols) collected by the edge gateway; unifying timestamp precision, data units, encoding methods, and data types; eliminating heterogeneous data differences; deduplicating based on the device's unique identifier and collection timestamp; deleting redundant sampling points uploaded repeatedly by the same device at the same time; and removing illegal data that does not conform to the preset collection logic. Anomaly detection includes: setting upper and lower limit thresholds according to the device's rated parameters and operating specifications; judging whether indicators such as voltage, current, power, temperature, and energy consumption exceed the limits; marking data points exceeding the limits; identifying anomalies such as sudden changes, drifts, spikes, and drops using time-series characteristics (rate of change, fluctuation amplitude, periodicity); and using time-series prediction models (such as ARIMA and LSTM). The process involves predicting residuals, modeling the normal operating range, identifying outliers where residuals exceed confidence intervals, and marking outliers with their type, occurrence time, and severity to differentiate between minor fluctuations, general anomalies, and severe faults, providing a basis for subsequent processing. Missing value completion includes: first, determining the missing pattern, and then selecting an appropriate completion method based on the pattern. For example, for a small number of short-term missing values, nearest neighbor, linear interpolation, or mean / median filling are used; for continuous long-term missing values, time-series prediction methods are used for completion, performing periodic interpolation based on historical data (daily / weekly cycles), predicting the missing segment values ​​using a time-series prediction model, and combining this with data from similar equipment for correlation completion; finally, the completed data undergoes range verification and reasonableness assessment to ensure that the completed values ​​remain within the normal operating range of the equipment and do not introduce new anomalies.

[0024] Furthermore, after the above three steps of data cleaning, anomaly detection, and missing value completion, the data is sorted according to device ID + timestamp to generate a standardized time series dataset. Each data entry in the standardized time series dataset includes: a unified time axis, standardized values, data source identifier, and cleaning / anomaly / completion markers.

[0025] It should be further explained that, in the specific implementation process, the knowledge graph module includes: Sub-graph units are used to obtain the physical information and physical laws of each energy device. Based on the preset graph generation rules, the physical information of the devices is mapped to the blank graph to construct the corresponding sub-model knowledge graph. The graph association unit is used to set registration points, match the registration points with the sub-model knowledge graph, set free registration lines for different registration points, and set mounting threads at both ends of the registration lines. The physical laws of the equipment are used as registration conditions. Based on the registration conditions, the mounting threads are operated to select and establish the connection relationship between the corresponding registration points of the free registration lines, and construct the physical model knowledge graph of the energy system composed of all energy equipment.

[0026] It should be noted that registration points, registration lines, and mounting threads are all virtual operation addresses registered in the virtual machine. Registration points and mounting threads correspond to a single number of virtual operation addresses in the virtual machine, while registration lines consist of a series of consecutive virtual operation addresses. The processing of registration points, registration lines, and mounting threads is all based on the virtual machine.

[0027] It should be further explained that, in the specific implementation process, the map association unit includes: The thread operation unit is used to handle the mounting thread. The mounting thread includes a mounting end, a message end, and an association end. The mounting end is used to determine the mounting target of the registration line, which is the registration point. The message end is used to store the target information of the mounting target. The association end is used to associate data between registration points of the same type of registration line. The registration unit is used to attach the two ends of the free registration line to the corresponding attachment target, read the target information of the current attachment target from the message end, and send the target information to all registration points with data association based on the association end; The storage unit is used to determine the registration point for each received target information, synchronize the mounting data of the corresponding registration line of the sender to all the corresponding registration points of the receiver, and store it in the registration point. When other registration lines of the same type need to be mounted, the mounting data stored in the registration point is called.

[0028] It should be noted that the role of the graph association unit is to perform associations between several sub-model knowledge graphs to construct the final physical model knowledge graph used to connect all energy devices. Each sub-model knowledge graph is assigned a registration point, and free registration lines are attached to the registration points corresponding to the defined attachment targets, realizing graph association between any two sub-model knowledge graphs. Registration lines correspond to different categories. When a registration line of a certain category completes its attachment, the attachment data of the attached registration line is synchronized to other registration points with data associations. These other registration points pre-store their data, and when an attachment requirement occurs, they attach the registration point to the appropriate free registration line, improving the efficiency of graph association. The physical model knowledge graph establishes the physical dependencies between different energy devices, as well as the execution order and priority of different energy devices.

[0029] Furthermore, the physical laws governing different energy equipment include: the similarity law of water pumps: power ∝ speed³, the energy efficiency curve of chillers: characterizing the COP value under different load rates, with the highest energy efficiency usually occurring at a load rate of 60%–80%, the heat dissipation model of cooling towers: based on the theoretical heat dissipation limit of wet-bulb temperature, where the water temperature cannot be reduced below the wet-bulb temperature, and the thermodynamic inertia model: the time delay characteristics of temperature changes, used to predict the effect after adjustment.

[0030] It should be further explained that, in the specific implementation process, the load forecasting module includes: The modeling unit is used to obtain historical load data, weather data, and time dimension data to build and train an LSTM load prediction model. The load forecasting unit is used to set the model call address of the LSTM load forecasting model, store the LSTM load forecasting model on the cloud server, and each energy device in the energy system operates the LSTM load forecasting model based on the model call address to predict its own hourly load, and use the hourly load as the load forecasting result.

[0031] The LSTM load forecasting model predicts the hourly load of energy equipment over a period of 24 hours. Once the prediction accuracy of the LSTM load forecasting model is trained to ≥95%, training is stopped, resulting in the LSTM load forecasting model used to perform the forecasting task. Weather data includes temperature, humidity, and irradiance of the environment in which the energy system is located, while time dimension data includes different time periods such as holidays and weekdays.

[0032] It should be further explained that, in the specific implementation process, the decision output module includes: The model building unit is used to combine the graph content recorded in the physical model knowledge graph with the equipment information of pre-deployed redundant equipment, determine the physical constraints of energy equipment, set the physical constraints of each energy equipment and the load prediction results as modeling elements, and build a multi-objective optimization model based on the modeling elements. The decision unit is used to determine the equipment clusters composed of a number of energy devices in the energy system that satisfy the multi-objective optimization model, to determine each equipment cluster as the optimal equipment combination, to obtain the operating parameters of the energy devices under each optimal equipment combination, and to take the optimal equipment combination and the operating parameters of the corresponding included energy devices as the decision result. The visualization output unit is used to establish the reasoning chain for each decision result, set up several visual display points on the reasoning chain, and visualize part of the decision result based on the visual display points.

[0033] It should be further explained that, in the specific implementation process, the visualization output unit includes: The link generation unit is used to process the decision result into several data segments based on time sequence, establish a sub-inference link between two sequentially adjacent data segments, store the previous data segment and the current data segment at the beginning and end of the path respectively, and store the inference content of the previous data segment to obtain the current data segment on the sub-inference path. Integrate all sub-inference links to obtain the inference link. The visual display unit is used to set the connection position of adjacent sub-inference links as a visual display point, take the inference content of the sub-inference link termination position corresponding to each visual display point as the decision result of the corresponding part, and display it based on the visual display point.

[0034] The reasoning content displayed at each visual display point includes, but is not limited to, the load forecast curve and confidence interval of the energy system (equipment), the energy consumption comparison bar chart of each candidate energy dispatch scheme, the physical feasibility verification of the optimal scheme, and the textual description of the decision basis.

[0035] It should be noted that while traditional AI models built on artificial intelligence technology have high prediction accuracy, their decision-making process is uninterpretable, only providing the final execution parameters. This makes them unsuitable for critical scenarios such as power grid dispatching and ICU temperature control due to a lack of trust. Therefore, they are often only used as auxiliary references in practical applications. In contrast, this invention establishes a corresponding inference link based on the decision results of each optimal device group, segmenting the decision-making process. Each sub-inference link stores the inference content of a time-series segment and displays it at the corresponding visual point. This provides relevant operation and maintenance personnel of the energy system with a step-by-step decision-making process, allowing them to conduct auxiliary analysis based on their own operation and maintenance experience, making the decision results more trustworthy.

[0036] It should be further explained that, in the specific implementation process, the collaborative control module includes: For the module structure of the collaborative control module, please refer to [link / reference]. Figure 2 As shown; The data distribution unit is used to deploy the cloud module and create storage points for all decision results in the cloud module. The storage points are used to store the decision results and the communication addresses of all energy devices included in the corresponding optimal device combination. Based on the communication addresses, the decision results are distributed to the edge gateway at the corresponding optimal device combination. The analysis unit is used to analyze the local optimal operation and maintenance control objectives of different optimal equipment combinations under the energy system, and to determine the global operation and maintenance control objectives corresponding to the entire energy system. The collaborative control unit is used to establish a collaborative relationship function between the local optimal operation and maintenance control objective and the global operation and maintenance control objective. The collaborative relationship function is used to characterize the number of energy devices that meet the local optimal operation and maintenance control objective in their respective optimal equipment combinations when the energy system meets the global operation and maintenance control objective. The corresponding energy devices are controlled based on the collaborative relationship function.

[0037] It should be further explained that, in the specific implementation process, the analysis unit includes: The local analysis unit is used to analyze the optimal energy efficiency range of each optimal equipment combination in the energy system. When the operating data is in the optimal energy efficiency range, the optimal equipment combination achieves the local optimal operation and maintenance control target in the energy system. The global analysis unit is used to analyze the energy efficiency requirements of the entire energy system during operation and to determine whether the energy system can achieve the global operation and maintenance control objectives if all the optimal equipment combinations under the energy system maintain their current operating conditions. If so, no action is taken; If not, determine the optimal equipment combination that needs operational adjustments, and adjust the operational data of the optimal equipment combination within its respective optimal energy efficiency range until the energy system reaches the overall operation and maintenance control target.

[0038] It should be noted that each optimal equipment combination includes a number of energy devices. Each energy device has a corresponding optimal energy efficiency range set based on historical operating status analysis. The optimal energy efficiency range is a range value. When the corresponding operating data of each optimal equipment combination is within the range value, they can maintain their good operation. The entire energy system is composed of a number of optimal equipment combinations. Under the premise of ensuring that the energy system can meet the energy efficiency requirements, the operation of the energy devices under each optimal equipment combination is adjusted accordingly. The adjustment range is still kept within their respective optimal energy efficiency ranges. This ensures that the entire energy system meets the global optimal optimization goal, while each subset (optimal equipment combination) under the energy system can also achieve its own local optimal optimization goal.

[0039] It should be further explained that, in the specific implementation process, the collaborative control unit includes: The relational function unit is used to construct the relational function between energy devices in each optimal equipment combination that satisfies the local optimal operation and maintenance control objective, using the optimal energy efficiency range of each optimal equipment combination as the function parameter. The relational function of each optimal equipment combination that satisfies the global operation and maintenance control objective is used as the collaborative relational parameter to construct the collaborative relational function. The control unit is used to set the activation number of each energy device through the cooperative relationship function, create a control thread for each optimal device combination, register the control thread to the cloud module, and the cloud module compiles the control files for the activation number of each energy device in the optimal device combination. The control thread uses the control files to control the corresponding energy devices. Autonomous units are used to distribute control threads to edge gateways and establish corresponding local area networks at the location of each edge gateway. When the cloud module fails to manage any edge gateway, the control thread of the corresponding edge gateway is switched from the cloud module to the local area network, and the local area network performs local control over the corresponding edge gateway.

[0040] For details regarding the relevant working processes of autonomous units, please refer to [link / reference]. Figure 3 As shown.

[0041] It should be noted that the control thread, which is uniformly registered in the cloud module, is distributed to each edge gateway, enabling independent control of the energy devices in each optimal device combination. When the cloud module experiences network failure, data anomalies, or unauthorized external access to the corresponding edge gateway, the control thread is switched to the local area network where the corresponding edge gateway is located. This achieves autonomous operation without network interruption, ensuring uninterrupted control of the energy devices and maintaining the long-term stable operation of the energy system. Autonomous operation without network interruption can support up to 72 hours. During this period, data communication between the edge gateway and the cloud module is restored.

[0042] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints, characterized in that, The system includes: The edge operation module is used to connect the edge gateway to several types of energy devices, perform data cleaning, anomaly detection and missing value completion for the corresponding energy devices, and generate corresponding standardized time series datasets. The knowledge graph module is used to establish knowledge graphs for the sub-models corresponding to different energy devices, to associate the knowledge graphs of different sub-models, and to construct the physical model knowledge graph corresponding to the energy system composed of all energy devices. The load forecasting module is used to establish an LSTM load forecasting model, perform hourly load forecasting on the energy system based on the LSTM load forecasting model, and obtain the corresponding load forecasting results. The decision output module is used to construct a multi-objective optimization model by taking the physical constraints of all energy equipment in the energy system and the load forecast results as modeling elements, and to make decisions on the optimal equipment combination and operating parameters based on the multi-objective optimization model, and to visualize the decision results. The collaborative control module is used to send decision results to the edge gateway and perform collaborative control on several energy devices based on the cloud module.

2. The cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 1, characterized in that, The edge operation module includes: The access unit is used to determine and enable a number of edge gateways, determine a number of categories of energy devices that need to be monitored, connect all energy devices to the enabled edge gateways, and store the corresponding call addresses of each device at the edge gateways. The standardized unit is used to perform data cleaning, anomaly detection, and missing value completion on the energy device data obtained by the edge gateway, and to generate a standardized time-series dataset for each energy device based on the time series.

3. The cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 2, characterized in that, The knowledge graph module includes: Sub-graph units are used to obtain the physical information and physical laws of each energy device. Based on the preset graph generation rules, the physical information of the devices is mapped to the blank graph to construct the corresponding sub-model knowledge graph. The graph association unit is used to set registration points, match the registration points with the sub-model knowledge graph, set different registration points to be in free registration lines, and set mounting threads at both ends of the registration lines. The physical laws of the equipment are used as registration conditions. Based on the registration conditions, the mounting threads are operated to select the free registration lines to establish the connection relationship between the corresponding registration points, and construct the physical model knowledge graph of the energy system composed of all energy equipment.

4. The cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 3, characterized in that, The map association unit includes: The thread operation unit is used to handle the mounting thread. The mounting thread includes a mounting end, a message end, and an association end. The mounting end is used to determine the mounting target of the registration line, which is the registration point. The message end is used to store the target information of the mounting target. The association end is used to associate data between registration points of the same type of registration line. The registration unit is used to attach the two ends of the free registration line to the corresponding attachment target, read the target information of the current attachment target from the message end, and send the target information to all registration points with data association based on the association end; The storage unit is used to determine the registration point for each received target information, synchronize the mounting data of the corresponding registration line of the sender to all the corresponding registration points of the receiver, and store it in the registration point. When other registration lines of the same type need to be mounted, the mounting data stored in the registration point is called.

5. The cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 4, characterized in that, The load forecasting module includes: The modeling unit is used to obtain historical load data, weather data, and time dimension data to build and train an LSTM load prediction model. The load forecasting unit is used to set the model call address of the LSTM load forecasting model, store the LSTM load forecasting model on the cloud server, and each energy device in the energy system operates the LSTM load forecasting model based on the model call address to predict its own hourly load, and use the hourly load as the load forecasting result.

6. The cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 5, characterized in that, The decision output module includes: The model building unit is used to combine the graph content recorded in the physical model knowledge graph with the equipment information of pre-deployed redundant equipment, determine the physical constraints of energy equipment, set the physical constraints of each energy equipment and the load prediction results as modeling elements, and build a multi-objective optimization model based on the modeling elements. The decision unit is used to determine the equipment clusters composed of a number of energy devices in the energy system that satisfy the multi-objective optimization model, to determine each equipment cluster as the optimal equipment combination, to obtain the operating parameters of the energy devices under each optimal equipment combination, and to take the optimal equipment combination and the operating parameters of the corresponding included energy devices as the decision result. The visualization output unit is used to establish the reasoning chain for each decision result, set up several visual display points on the reasoning chain, and visualize part of the decision result based on the visual display points.

7. The cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 6, characterized in that, The visualization output unit includes: The link generation unit is used to process the decision result into several data segments based on time sequence, establish a sub-inference link between two sequentially adjacent data segments, store the previous data segment and the current data segment at the beginning and end of the path respectively, and store the inference content of the previous data segment to obtain the current data segment on the sub-inference path. Integrate all sub-inference links to obtain the inference link. The visual display unit is used to set the connection position of adjacent sub-inference links as a visual display point, take the inference content of the sub-inference link termination position corresponding to each visual display point as the decision result of the corresponding part, and display it based on the visual display point.

8. The cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 7, characterized in that, The collaborative control module includes: The data distribution unit is used to deploy the cloud module and create storage points for all decision results in the cloud module. The storage points are used to store the decision results and the communication addresses of all energy devices included in the corresponding optimal device combination. Based on the communication addresses, the decision results are distributed to the edge gateway at the corresponding optimal device combination. The analysis unit is used to analyze the local optimal operation and maintenance control objectives of different optimal equipment combinations under the energy system, and to determine the global operation and maintenance control objectives corresponding to the entire energy system. The collaborative control unit is used to establish a collaborative relationship function between the local optimal operation and maintenance control objective and the global operation and maintenance control objective. The collaborative relationship function is used to characterize the number of energy devices that meet the local optimal operation and maintenance control objective in their respective optimal equipment combinations when the energy system meets the global operation and maintenance control objective. The corresponding energy devices are controlled based on the collaborative relationship function.

9. The cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 8, characterized in that, The analysis unit includes: The local analysis unit is used to analyze the optimal energy efficiency range of each optimal equipment combination in the energy system. When the operating data is in the optimal energy efficiency range, the optimal equipment combination achieves the local optimal operation and maintenance control target in the energy system. The global analysis unit is used to analyze the energy efficiency requirements of the entire energy system during operation and to determine whether the energy system can achieve the global operation and maintenance control objectives if all the optimal equipment combinations under the energy system maintain their current operating conditions. If so, no action is taken; If not, determine the optimal equipment combination that needs operational adjustments, and adjust the operational data of the optimal equipment combination within its respective optimal energy efficiency range until the energy system reaches the overall operation and maintenance control target.

10. A cloud-edge collaborative intelligent control system for energy systems with embedded physical constraints according to claim 8, characterized in that, The collaborative control unit includes: The relational function unit is used to construct the relational function between energy devices in each optimal equipment combination that satisfies the local optimal operation and maintenance control objective, using the optimal energy efficiency range of each optimal equipment combination as the function parameter. The relational function of each optimal equipment combination that satisfies the global operation and maintenance control objective is used as the collaborative relational parameter to construct the collaborative relational function. The control unit is used to set the activation number of each energy device through the cooperative relationship function, create a control thread for each optimal device combination, register the control thread to the cloud module, and the cloud module compiles the control files for the activation number of each energy device in the optimal device combination. The control thread uses the control files to control the corresponding energy devices. Autonomous units are used to distribute control threads to edge gateways and establish corresponding local area networks at the location of each edge gateway. When the cloud module fails to manage any edge gateway, the control thread of the corresponding edge gateway is switched from the cloud module to the local area network, and the local area network performs local control over the corresponding edge gateway.

Images