Digital twin service platform construction method based on cloud edge end collaboration

Abstract

The invention relates to the technical field of digital twinning, in particular to a digital twinning service platform construction method based on cloud edge end collaboration. The method comprises the following steps: acquiring energy system basic data through an edge end device; carrying out preliminary state modeling on the edge end device by utilizing the energy system basic data, constructing a virtual energy system model, and transmitting the virtual energy system model to the cloud server; controlling the cloud server to receive the virtual energy system model uploaded by the edge end device, and performing digital twinborn model construction on the virtual energy system model to obtain an energy system digital twinborn model; and performing energy scheduling optimization according to the digital twin model of the energy system to obtain energy scheduling data of the energy system, and sending the energy scheduling data to an edge end device to perform energy system scheduling optimization operation. According to the invention, through multi-scene simulation and strategy correction, a power supply gap can be predicted and adjusted in advance, and stable operation of energy supply under various meteorological conditions and load fluctuation conditions is realized.

Description

Technical Field

[0001] The present invention relates to the field of digital twin technology, and in particular to a method for constructing a digital twin service platform based on cloud-edge collaboration. Background Art

[0002] In the modern rural energy system, with the promotion of new energy technologies such as photovoltaic power generation, wind power generation and energy storage equipment in rural areas, the rural energy structure is gradually developing in a diversified and sustainable direction. However, the energy demand in rural areas has significant volatility, and the supply of new energy is greatly affected by weather and seasons. Therefore, how to achieve efficient management and optimal utilization of rural new energy has become an urgent problem to be solved.

[0003] Rural new energy systems are usually composed of multiple decentralized small photovoltaic power stations, wind power stations and energy storage equipment. These energy devices can make full use of local natural resources, reduce rural areas' dependence on traditional fossil energy, and reduce greenhouse gas emissions. However, due to the scattered layout of rural new energy equipment and limited communication networks, traditional energy management systems are difficult to uniformly monitor and dispatch these devices in real time, resulting in greater difficulty in matching energy supply and demand, low equipment utilization, and energy waste. Summary of the invention

[0004] In order to solve the above-mentioned technical problems, the present invention proposes a method for constructing a digital twin service platform based on cloud-edge collaboration to solve at least one of the above-mentioned technical problems.

[0005] The present application provides a method for constructing a digital twin service platform based on cloud-edge collaboration, the method comprising:

[0006] S1. Obtain basic data of energy system through edge devices;

[0007] S2, using the basic data of the energy system to perform preliminary state modeling at the edge device and construct a virtual energy system model for transmission to the cloud server;

[0008] S3, controlling the cloud server to receive the virtual energy system model uploaded by the edge device, and constructing a digital twin model of the virtual energy system model to obtain a digital twin model of the energy system;

[0009] S4. Optimize energy scheduling based on the digital twin model of the energy system to obtain energy scheduling data of the energy system and send it to the edge device for energy system scheduling optimization.

[0010] In the present invention, the data collection process is completed at the edge, which reduces the frequency and delay of data transmission to the cloud and improves the system response speed. Preliminary modeling at the edge shares the computing burden of the cloud server and improves the resource utilization of the entire system. The preliminary state modeling at the edge enables the system to respond to local events, such as load fluctuations or equipment state changes, more quickly, providing support for real-time control and local optimization. The construction of a cloud-based digital twin model can break geographical restrictions and achieve collaborative management and optimization of cross-regional energy systems. The cloud performs dynamic scheduling optimization based on the digital twin model, which can adapt to real-time changes in supply and demand, energy price fluctuations, etc., to achieve efficient and low-cost operation of the energy system. After the scheduling data is transmitted to the edge, it can form a closed-loop feedback with the actual operating status, support the system's self-optimization and continuous learning, and improve the stability and adaptability of long-term operation.

[0011] Optionally, S1 includes:

[0012] S11. Collect basic data of photovoltaic energy system in real time through sensors deployed on photovoltaic inverter equipment;

[0013] S12. Collect basic data of wind power energy system in real time through sensors deployed on wind turbine equipment;

[0014] S13. Collect basic data of energy storage system in real time through sensors deployed on energy storage battery equipment;

[0015] S14, performing a benchmark check on the photovoltaic energy system basic data, the wind power energy system basic data, and the energy storage energy system basic data to obtain energy system benchmark check data;

[0016] S15. Perform time synchronization according to the energy system benchmark verification data to obtain energy system basic data.

[0017] The present invention directly obtains data from the photovoltaic inverter to ensure the real-time data collection, so that the system can quickly respond to fluctuations in photovoltaic power generation. Real-time collection of wind power data enables the system to quickly adapt to changes in wind speed and direction, and effectively optimize the utilization rate of wind power. By monitoring the charging and discharging of energy storage batteries, it is ensured that the batteries operate in the best condition and the energy efficiency of the energy storage system is improved. Energy data from different sources are verified to remove deviations caused by equipment differences and ensure consistency between data from different systems. Benchmark verification helps to improve the overall quality of system data and avoid model errors caused by erroneous data. Synchronous timestamps eliminate data delays caused by equipment time differences, ensuring the real-time and accuracy of analysis and scheduling.

[0018] Optionally, S2 includes:

[0019] S21. Perform daily energy trend analysis using basic data of the energy system at the edge device to obtain daily energy trend data;

[0020] S22. Perform power generation and power consumption matching analysis based on daily energy trend data to obtain power generation supply and demand deviation data;

[0021] S23, classifying energy status according to the power generation supply and demand deviation data to obtain energy status classification data;

[0022] S24, analyzing the state of the energy storage system according to the basic data of the energy system to obtain energy storage capacity estimation data;

[0023] S25. Perform edge state simulation based on energy system basic data, energy state classification data, and energy storage capacity estimation data to obtain a virtual energy system model for transmission to a cloud server.

[0024] In the present invention, daily energy trend analysis is used to predict daily power generation and power consumption trends in advance, helping the system to better perform energy scheduling. Power generation resources and energy storage equipment are reasonably allocated according to the supply and demand matching results to avoid unnecessary energy loss and improve overall energy efficiency. Through energy state classification, the system state can be divided into categories such as high supply and demand balance, high load or low load, which is convenient for targeted management and scheduling. Real-time estimation of energy storage status can help the system reasonably arrange charging and discharging plans, avoid excessive use, and extend the service life of energy storage equipment. State simulation and construction of virtual models at the edge reduce the computing load on the cloud and improve the overall efficiency of the system.

[0025] Optionally, the power generation and power consumption matching analysis includes:

[0026] Conduct demand response elasticity analysis based on daily energy trend data to obtain response elasticity score data;

[0027] The flexibility score of power generation equipment is scored based on daily energy trend data and basic energy system data to obtain the flexibility score data of power generation equipment;

[0028] Perform multi-scenario supply and demand matching simulation based on daily energy trend data, response elasticity score data, and power generation equipment flexibility score data to obtain multi-scenario supply and demand simulation data;

[0029] Generate a short-term balance plan based on multi-scenario supply and demand simulation data to obtain short-term balance plan data;

[0030] Make real-time deviation adjustments based on short-term balance plan data to obtain real-time deviation adjustment numbers;

[0031] Supply and demand stability assessment is performed based on multi-scenario supply and demand simulation data and real-time deviation adjustment numbers to obtain supply and demand stability assessment data;

[0032] The power generation supply and demand deviation is generated according to the response elasticity score data, the power generation equipment flexibility score data and the supply and demand stability assessment data to obtain the power generation supply and demand deviation data.

[0033] The elasticity score in the present invention provides an important basis for adaptive scheduling, and the system can adapt to dynamic load changes more flexibly and optimize the scheduling strategy. The flexibility score helps to identify the startup and response characteristics of different power generation equipment, so that the system can select appropriate power generation resources when supply and demand change, and enhance the reliability of power supply. Through multi-scenario supply and demand simulation, the system can respond to different load conditions, weather conditions, power generation fluctuations, etc. in advance, and improve the adaptability of the system in complex environments. The short-term balance plan can formulate response strategies according to different scenarios to ensure timely response when demand changes suddenly and maintain supply and demand balance. Real-time deviation adjustment can make instant adjustments according to the actual supply and demand deviation to ensure power supply quality and system stability. Through supply and demand stability evaluation, the system can select different optimization strategies under different stability states to enhance the intelligent management capabilities of the system. Through accurate supply and demand deviation data, the system can perform refined scheduling according to different states to improve the stability and power supply quality of the system.

[0034] Optionally, S3 includes:

[0035] S31, controlling the cloud server to receive the virtual energy system model uploaded by the edge device;

[0036] S32, initializing the digital twin model according to the virtual energy system model to obtain an initialized digital twin model;

[0037] S33, optimizing the initialized digital twin model to obtain a digital twin optimization model;

[0038] S34, performing multi-source data fusion according to the digital twin optimization model to obtain a digital twin fusion model;

[0039] S35. Perform scene simulation according to the digital twin fusion model to obtain scene simulation data;

[0040] S36. Modify the digital twin fusion model according to the scenario simulation data to obtain the digital twin model of the energy system.

[0041] In the present invention, the model data uploaded after edge preprocessing is smaller than the original data, which reduces the pressure of redundant data transmission and saves bandwidth. The cloud receives model data of multiple edge devices, which helps to synchronize and integrate data in different regions and provide data support for global analysis. The cloud digital twin model is quickly initialized based on the virtual model of the edge, which greatly reduces the model reconstruction time and improves the efficiency of system construction. The optimization of the digital twin model can improve the prediction accuracy and response speed of the model, so that the model can more accurately simulate the behavior of the actual energy system. The fused model can adapt to changes in different time, place and environmental factors, and its adaptability and stability are significantly improved. Through scenario simulation, it is possible to predict situations such as high load, high demand or equipment failure, issue early warnings in time, and reduce potential risks. Correction based on scenario simulation data can ensure that the error between the digital twin model and the actual system is minimized, thereby improving the accuracy of the model.

[0042] Optionally, the digital twin model optimization includes:

[0043] Performing a first digital twin model optimization on the initialized digital twin model to obtain a first digital twin optimization model;

[0044] Performing a second digital twin model optimization on the initialized digital twin model to obtain a second digital twin optimized model;

[0045] Cross-validation and error correction are performed based on the first digital twin optimization model and the second digital twin optimization model to obtain a digital twin optimization model, wherein the first digital twin model optimization and the second digital twin model optimization are different digital twin model optimization methods.

[0046] The first optimization method in the present invention focuses on improving the response speed of the model so that the model can reflect the dynamic changes of the system more promptly. The second optimization method focuses on the adaptability of the model. The optimized model can cope with more complex system changes and better adapt to different environmental conditions. By using different optimization methods, the uncertainty factors in the model can be effectively identified and reduced, and the stability of the model under complex conditions can be improved. Through cross-validation, the model results of two different optimization methods are compared and verified to identify and correct the errors in the model to make the model more robust. After cross-validation, the system combines the advantages of the two optimization models to form a digital twin optimization model with better performance, achieving a balance in accuracy and response speed. The error correction step ensures that the model error will not accumulate over time, and the model can continue to maintain high accuracy, improving the ability to truly reflect the actual system.

[0047] Optionally, the first digital twin model optimization is specifically as follows:

[0048] Select the parameter space to be optimized for the initialized digital twin model to obtain the parameter space data to be optimized;

[0049] Selecting an optimization target based on the parameter space data to be optimized to obtain optimization target selection data;

[0050] Select data according to the optimization target to generate evaluation criteria, and obtain optimized evaluation standard data;

[0051] Perform Gaussian process regression calculation based on the initialized digital twin model to obtain the initial proxy model;

[0052] Perform Bayesian iterative optimization based on the initial proxy model to obtain an updated proxy model;

[0053] Convergence judgment is performed based on the update agent model to obtain the optimal parameter data of the digital twin;

[0054] Optimization is performed based on the digital twin optimal parameter data and the initialized digital twin model to obtain the first digital twin optimization model.

[0055] In the present invention, by selecting the parameter space to be optimized, the optimization process is focused on the parameters that have the greatest impact on the model, unnecessary calculations are reduced, and the effective use of optimization resources is ensured. According to the parameter space data, a clear optimization target is selected to make the optimization process more directional and purposeful. By setting evaluation criteria (such as mean square error, accuracy, etc.), it is ensured that the improvement of the optimized model in the target index is significant. The initial proxy model is constructed by Gaussian process regression, and the model performance can be approximately evaluated without directly calculating the objective function, reducing the computational burden. By gradually approaching the optimal solution through Bayesian iterative optimization, it is ensured that the parameter optimization of the model gradually converges to the global optimum, rather than the local optimum. Through convergence judgment, it is ensured that the model can be stable and unchanged after reaching the optimal parameter configuration, avoiding excessive iteration of the optimization process. Under the guidance of the optimal parameter data, the initialization model is optimized, which significantly improves the prediction accuracy, response speed and robustness of the model.

[0056] Optionally, the second digital twin model optimization is specifically as follows:

[0057] Perform baseline performance evaluation on the initialized digital twin model to obtain model baseline performance data;

[0058] Select the maximum deviation optimization target according to the model baseline performance data to obtain the maximum deviation optimization target data;

[0059] Adjust the control parameters of the initialized digital twin model according to the maximum deviation optimization target data to obtain control parameter adjustment data, wherein the control parameter adjustment includes model learning rate adjustment, model smoothing coefficient adjustment, and model loss coefficient adjustment;

[0060] According to the control parameters, the data is adjusted and the digital twin model is initialized to perform optimization iteration and model adjustment to obtain the second digital twin optimization model.

[0061] In the present invention, the performance shortcomings and main error sources of the model are identified through baseline evaluation data, providing a basis for target setting and optimization direction. Baseline performance data helps the optimization process to select the optimization direction more targeted, ensuring that the optimization process can solve the key problems currently existing in the model. By setting the maximum deviation optimization target, the optimization process can focus on reducing model deviation and improving the accuracy and stability of the model. By adjusting the learning rate, the model can adaptively adjust the optimization step size according to the speed of system changes, ensuring the rapid response capability of the system. Smoothing coefficient adjustment can reduce the volatility of model output, reduce the interference of noise and abnormal data, and make the output of the model more stable and reliable. Through the adjustment of the loss coefficient, the model can meet the optimization target while minimizing energy consumption and improve the energy-saving efficiency of the model. With the support of control parameters, the model is continuously adjusted through multiple rounds of iterative optimization, so that the accuracy and response speed of the model are gradually improved. In the optimization iteration process, the parameters of the model are continuously adjusted to reduce the accumulation of deviations to ensure that the prediction accuracy of the model remains stable during long-term operation.

[0062] Optionally, S4 includes:

[0063] S41. Perform energy demand forecasting based on the energy system digital twin model to obtain energy demand forecast data;

[0064] S42. Evaluate the energy production capacity of the energy system digital twin model to obtain energy production capacity evaluation data;

[0065] S43, performing energy supply and demand balance calculation based on energy demand forecast data and energy production capacity assessment data to obtain energy supply and demand balance table data;

[0066] S44, generating an energy dispatch optimization strategy based on the energy supply and demand balance table data to obtain energy dispatch optimization strategy data;

[0067] S45, performing simulation verification according to the energy dispatch optimization strategy data to obtain dispatch verification data;

[0068] S46. Correct the energy dispatch optimization strategy data according to the dispatch verification data to obtain energy system energy dispatch data;

[0069] S47. Send the energy system energy scheduling data to the edge device to perform energy system scheduling optimization operations.

[0070] In the present invention, demand forecasting is performed through the historical data and real-time information of the digital twin model to ensure accurate prediction of future energy demand and to respond to fluctuations in advance. Through production capacity assessment, the actual production upper limit of the current energy system is understood to avoid scheduling plans that exceed the equipment load. Supply and demand balance calculations ensure that the system can adjust supply in a timely manner to meet demand changes in different time periods and avoid supply and demand imbalances. Scheduling optimization strategies based on supply and demand balance data can accurately match production and demand and improve the accuracy of scheduling. Simulation verification can detect the feasibility of scheduling plans in advance to ensure that the plans can achieve the expected results in actual implementation. Through the correction process, deviations caused by errors or uncertainties are eliminated, making scheduling data more accurate. The optimized scheduling data is fed back to the edge to complete scheduling execution, forming a closed-loop control from cloud decision-making to edge execution, thereby improving the overall efficiency of the system.

[0071] The purpose of the present invention is to deploy sensors at the edge end, collect data from photovoltaic, wind power and energy storage equipment in real time, perform preliminary state modeling on the original data at the edge end, and form a virtual energy system model, which can not only reduce the amount of data transmission and transmission delay, but also reduce the computing pressure on the cloud, thereby realizing distributed data processing on the edge and cloud. After completing the basic data processing at the edge end, only the virtual model is uploaded to the cloud, which reduces the transmission demand of redundant data and greatly saves bandwidth resources. At the same time, streamlined data transmission improves the efficiency of cloud reception and processing, enabling the system to complete multiple tasks in a shorter time and ensuring the real-time nature of the data. After receiving the virtual energy system model at the edge end, the cloud constructs a more accurate digital twin model. The digital twin model is optimized for multiple rounds in the cloud to ensure that the model can more accurately reflect the state of the real energy system and achieve a balance in terms of predictive ability, flexibility and adaptability. Multi-scenario simulation of the fused digital twin model in the cloud can predict the performance of the system in different supply and demand scenarios in advance, thereby discovering and correcting errors in advance. Based on the scheduling optimization strategy generated by the cloud digital twin model, the system can achieve accurate energy supply and demand balance. Using data such as demand forecasts and production capacity assessments, the cloud generates the optimal scheduling strategy and transmits it to the edge for execution after simulation verification, thereby achieving efficient resource allocation and real-time scheduling in a dynamic environment. During the scheduling optimization process, the system reduces unnecessary power generation and energy storage consumption, achieves accurate allocation of energy resources, and maximizes system efficiency. The optimization strategy generated after multiple rounds of verification and adjustment ensures that the system operates at the lowest cost while meeting demand, improves energy utilization efficiency, reduces operating costs, and achieves energy-saving goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings:

[0073] Figure 1 A flowchart of a method for constructing a digital twin service platform based on cloud-edge collaboration according to an embodiment is shown;

[0074] Figure 2 A flowchart showing the steps of a method for acquiring basic data of an energy system according to an embodiment is shown;

[0075] Figure 3 A flowchart showing a method for constructing a virtual energy system model according to an embodiment of the present invention is provided;

[0076] Figure 4 A flowchart showing a method for constructing a digital twin model of an energy system according to an embodiment of the present invention is shown;

[0077] Figure 5 A flowchart showing a method for energy scheduling of an energy system according to an embodiment of the present invention is provided;

[0078] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0079] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0080] In addition, the drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0081] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0082] A new energy system has been built in a rural area, including photovoltaic power generation, wind power generation and energy storage battery equipment, to meet the daily electricity needs of villagers in the area and agricultural irrigation and other load demands. Based on this new energy system, the cloud-edge collaborative digital twin service platform is used to achieve real-time monitoring, optimized scheduling and efficient utilization of the rural energy system.

[0083] The daily electricity demand fluctuates greatly. The peak electricity demand during the day occurs at 8:00-12:00 in the morning and 18:00-22:00 in the afternoon, with a total demand of 1000kWh. The peak power is 600kWh (under sunny conditions), and the actual output is greatly affected by the sunlight. The power drops to 400kWh under cloudy conditions. The peak power is 400kWh, and the actual output varies according to the wind speed. The average daily output is about 300kWh. The capacity is 500kWh, which is used to supplement the power supply when the power generation is insufficient.

[0084] Edge sensors collect data from new energy equipment in real time, generate virtual models and upload them to the cloud. Photovoltaic equipment sensors collect voltage, current and light intensity every 10 seconds. On sunny days, the power of photovoltaic equipment is 600kWh, and on cloudy days it is 400kWh. The power data are 580kWh, 595kWh and 605kWh respectively, and the edge end performs denoising, cleaning and time synchronization on these data. The wind speed sensor collects data every 10 seconds and calculates the power output. In the morning and evening when the wind speed is high, the peak power of wind power generation can reach 400kWh. The wind speed samples are 7.8m / s and 8.1m / s, which are converted into power outputs of 300kWh and 310kWh. The energy storage device collects battery voltage, charge and discharge current and remaining power. The current power is 250kWh. The edge end preprocesses the photovoltaic, wind power and energy storage data to form a preliminary virtual energy model and transmits it to the cloud server.

[0085] Based on historical data and meteorological data, the photovoltaic and wind power generation trends of the day are predicted. The photovoltaic power generation is expected to be 550kWh throughout the day, and the wind power generation is expected to be 280kWh. Combined with the predicted power generation and demand data, supply and demand matching analysis is performed. The total rural electricity demand is 1000kWh, and the photovoltaic and wind power generation is expected to be 830kWh in total, resulting in a power supply gap of 170kWh. The current energy storage capacity is 250kWh, which can provide additional power during the peak demand in the evening to make up for the lack of power generation. The virtual energy system model generated by the edge is uploaded to the cloud to provide basic data support for cloud optimization and scheduling.

[0086] The cloud builds an initial digital twin model based on the model uploaded by the edge. The daytime demand is 1000kWh, photovoltaic and wind power generation totals 830kWh, the power supply gap is 170kWh, and the energy storage device can provide 150kWh of supplementary power. Bayesian optimization is performed on the output parameters of photovoltaic and wind power generation to reduce errors. The initial proxy model is constructed through Gaussian process regression, and the power supply error is optimized to ±5%. The learning rate and smoothing coefficient in the digital twin model are adjusted to further reduce output fluctuations, and the power supply error is optimized to ±2%. Real-time weather forecast data (such as sunny to cloudy) is integrated with the current digital twin model, and daytime power generation is expected to drop to 600kWh. The simulation results show that on cloudy days, the system will have a power supply gap of 200kWh, and it is recommended to start the energy storage system to supplement it.

[0087] According to weather forecasts and historical data, the electricity demand for the day is predicted to be 1000kWh and 800kWh at night. The maximum power supply capacity of photovoltaic and wind power is about 830kWh. The power supply gap during the day is 170kWh, and the energy storage system will provide 150kWh as a supplement. The energy storage will release 150kWh of electricity at night to ensure stable power supply. Photovoltaic and wind power output at full power, and the energy storage system is charged during the day. The energy storage system releases 150kWh of electricity at night to supplement demand. The scheduling strategy has been verified by simulation to meet the needs of the whole day, and the actual power supply gap during the day is reduced to 50kWh. The energy storage release time is adjusted to the peak period to further reduce the gap. The scheduling strategy is sent to the edge end, which monitors and adjusts the equipment in real time to ensure the balance of supply and demand. The edge end uploads the scheduling execution data to the cloud, and the cloud adjusts the strategy based on the feedback data to form a closed-loop control.

[0088] See also Figures 1 to 5 The present application provides a method for constructing a digital twin service platform based on cloud-edge collaboration, the method comprising:

[0089] S1. Obtain basic data of energy system through edge devices;

[0090] Specifically, the edge device collects basic data through various sensors installed in the energy system, including temperature sensors, current sensors, voltage sensors, etc., to collect relevant data on the operating status of the energy system in real time. In the electric energy system, the edge device collects basic data from sensors such as voltage, current, and temperature in real time. For example: voltage data: 230V, 235V, 240V, current data: 10A, 15A, 20A, temperature data: 45℃, 50℃, 47℃.

[0091] S2, using the basic data of the energy system to perform preliminary state modeling at the edge device and construct a virtual energy system model for transmission to the cloud server;

[0092] Specifically, using the processed basic data, a preliminary state model is generated at the edge according to specific physical rules and system operation characteristics. This state model performs a simple simulation of the basic operation of the system. The state model is expanded into a virtual energy system model, which includes predictions of key parameters of system operation and simulates the system's operating performance under different operating conditions. Due to the limited transmission bandwidth at the edge, the generated virtual model is compressed and data is packaged to reduce the amount of transmitted data. The compressed virtual energy system model is uploaded to the cloud server through the network.

[0093] Specifically, the edge uses the processed basic data to preliminarily establish a simulation model. For example, the preliminary model based on voltage, current, and temperature predicts the energy consumption of the equipment under different load conditions: Load 1 (70%): Energy consumption = 100kWh, Load 2 (80%): Energy consumption = 120kWh, Load 3 (90%): Energy consumption = 150kWh. The edge expands the state model into a virtual energy system model to include predictions under different operating conditions. For example, the predicted load is covered from 60% to 100%, and the corresponding energy consumption data is generated. Due to the limited bandwidth of the edge, the model data is packaged in a compressed format to reduce the amount of transmission. Compress the original model data to 20% of the size. Upload the compressed virtual model to the cloud server.

[0094] S3, controlling the cloud server to receive the virtual energy system model uploaded by the edge device, and constructing a digital twin model of the virtual energy system model to obtain a digital twin model of the energy system;

[0095] Specifically, the cloud server decompresses and parses the received virtual model data, converts it into a standardized format, and prepares for deep processing. On the cloud server, the model is feature analyzed to extract the key operating parameters and characteristic behaviors of the energy system. Using the extracted characteristic data, a complete digital twin model is constructed through further high-precision calculation and simulation. The model accurately simulates the actual operating conditions of the energy system and has higher prediction accuracy and adaptability. The constructed digital twin model is parameterized to ensure the accuracy and stability of the model and the subsequent energy scheduling optimization effect.

[0096] Specifically, the cloud server decompresses and parses the received model data. For example, the compressed model data is restored to its original form for analysis. Key features in the virtual model (such as energy consumption, load relationship) are extracted for further analysis. For example, the relationship between voltage, temperature and energy consumption is identified. Based on the virtual model and extracted features, a digital twin model is constructed. For example, the cloud server uses the parsed voltage, current and temperature data to calculate a more accurate energy consumption model to predict the load demand in the next 24 hours. The digital twin model is optimized to ensure its ability to simulate the real system. For example, at 90% load, the optimized model predicts an energy consumption of 148kWh, instead of 150kWh in the preliminary state modeling, with a reduced error rate.

[0097] S4. Optimize energy scheduling based on the digital twin model of the energy system to obtain energy scheduling data of the energy system and send it to the edge device for energy system scheduling optimization.

[0098] Specifically, the digital twin model is used to predict the future operation of the energy system and identify the fluctuations and bottlenecks in energy demand. Based on the prediction results, the cloud server optimizes the energy scheduling strategy and determines the optimal energy distribution and operation plan through the optimization algorithm. Generate a scheduling optimization plan for the energy system, including specific implementation plans such as energy distribution and load adjustment, to ensure the efficient operation of the energy system. The generated energy scheduling data is transmitted back to the edge device to ensure that the edge can perform scheduling operations according to the optimized plan to improve the operating efficiency and stability of the system. Determine future demand based on the load forecast of the digital twin model. For example, it is predicted that the maximum load demand in the next 24 hours will be 90%, and the average will be 75%. According to the load forecast, the scheduling of the energy system is optimized. For example, it is set to reduce the load during non-peak hours to reduce energy consumption. Assume that during the time when the load reaches more than 80%, the energy consumption can be reduced from 120kWh to 115kWh through scheduling optimization. Generate an optimized scheduling data plan, for example: off-peak period (load ≤ 70%): execute low power mode, energy consumption = 95kWh, peak period (load > 80%): schedule load, energy consumption = 115kWh, and transmit the optimized scheduling plan back to the edge device so that the optimized operation can be performed on the edge device to improve the operating efficiency of the energy system.

[0099] Optionally, S1 includes:

[0100] S11. Collect basic data of photovoltaic energy system in real time through sensors deployed on photovoltaic inverter equipment;

[0101] Specifically, sensors such as voltage, current, power, and temperature are installed on photovoltaic inverter equipment. The sensors detect key parameters such as voltage and current of photovoltaic panels in real time, and the equipment uploads the collected data regularly. For example, the output voltage of each photovoltaic panel is measured (such as 320V, 315V, 310V). The output current of each photovoltaic panel is measured (such as 8A, 7.5A, 8.2A).

[0102] S12. Collect basic data of wind power energy system in real time through sensors deployed on wind turbine equipment;

[0103] Specifically, sensors are deployed at key locations such as blades and towers of wind turbines to collect data such as wind speed, rotation speed, voltage, and power generation. The sensors record wind speed, generator rotation speed, voltage, and other indicators in real time. For example, the wind speed sensor reads the current wind speed (such as 5.2m / s, 5.5m / s, 6.0m / s). Measure the rotation speed of the generator (such as 1500rpm, 1550rpm, 1520rpm). Record the power generation per hour in real time (such as 2kWh, 2.5kWh).

[0104] S13. Collect basic data of energy storage system in real time through sensors deployed on energy storage battery equipment;

[0105] Specifically, voltage, current, temperature, and charge and discharge rate sensors are installed on the energy storage battery. The sensors collect the voltage, current, temperature and other parameters of the energy storage battery in real time. For example, the voltage of the energy storage battery is recorded (such as 480V, 475V, 490V). The current charge and discharge rate is recorded (such as 5% / h charging, 3% / h discharging). The battery temperature is monitored (such as 30°C, 32°C, 31°C).

[0106] S14, performing a benchmark check on the photovoltaic energy system basic data, the wind power energy system basic data, and the energy storage energy system basic data to obtain energy system benchmark check data;

[0107] Specifically, the preliminary collected data of photovoltaic, wind power, and energy storage systems are aligned, and inconsistent readings are aligned to a standardized interval. The voltage, current, power and other parameters of each system are compared with the preset benchmark parameters of the equipment to ensure data accuracy. For example: compared with the standard output voltage (such as 315V), if the difference is within a reasonable range (±5V), the collected data is considered valid. Compare the collected power generation with the expected power generation (based on wind speed and rotation speed) to generate verification data.

[0108] S15. Perform time synchronization according to the energy system benchmark verification data to obtain energy system basic data.

[0109] Specifically, the timestamps of data from different devices are aligned based on the energy system benchmark data. For example, a synchronization mechanism is used to ensure that the collection time of photovoltaic, wind power, and energy storage data is within a unified time base (such as an error of less than ±1 second) to ensure data synchronization.

[0110] Optionally, S2 includes:

[0111] S21. Perform daily energy trend analysis using basic data of the energy system at the edge device to obtain daily energy trend data;

[0112] Specifically, in the edge device, the basic data of photovoltaic, wind power, energy storage and other systems collected within a day are summarized to calculate the power generation and power consumption trends for the whole day. Generate a power generation trend curve based on the power generation per hour of the day (such as photovoltaic and wind power system data). For example, the photovoltaic power generation per hour: 3kWh, 4kWh, 5kWh, 4kWh, 3kWh, etc., are summarized into a trend curve. Similarly, the power consumption data is processed to generate a power consumption trend curve for the whole day. Compare the changes in the power generation and power consumption curves, observe the peak and valley periods, and provide a basis for subsequent power generation and power consumption matching analysis.

[0113] Specifically, energy data is decomposed into long-term trends, seasonal fluctuations, and random residuals to extract stable daily trend information. Moving average calculations are performed on power generation and power consumption data to smooth random fluctuations in the time series. If the data has large non-stationarity, the data is differentiated to remove the trend component. Time series decomposition methods (such as STL decomposition) are used to separate long-term trends, seasonal components, and residuals. Based on power generation equipment and load data, hourly power generation and load trends are calculated. Power generation and load data are grouped by time period (such as hour), and hourly means are calculated to obtain trend data. The hourly data is then averaged over a sliding window to further smooth the trend. In response to data with large fluctuations (such as wind power), exponential smoothing (such as EWMA) is used to reduce the impact of sudden changes. Future demand is predicted based on the results of daily trend analysis, and possible abnormal trends are identified. The power generation and load trends for the next day are predicted based on past trend data (such as moving average or weighted average). The prediction errors of the past few days are calculated to adjust the prediction model and reduce the prediction errors of future trends. The results of the above analysis processes are summarized to generate daily energy trend data.

[0114] S22. Perform power generation and power consumption matching analysis based on daily energy trend data to obtain power generation supply and demand deviation data;

[0115] Specifically, based on the daily energy trend data, the power generation and power consumption at different time points are compared to identify the peak and valley of power supply, and the supply and demand deviation is calculated. The difference between power generation and power consumption is compared every hour. For example, if the power generation in a certain hour is 4kWh and the power consumption is 3.5kWh, the supply and demand deviation is +0.5kWh (excess). If the power generation is 2kWh and the power consumption is 3kWh, the supply and demand deviation is -1kWh (insufficient). The supply and demand deviation of each hour is recorded as a set of supply and demand deviation data for subsequent classification of energy status.

[0116] S23, classifying energy status according to the power generation supply and demand deviation data to obtain energy status classification data;

[0117] Specifically, according to the supply and demand deviation data, the energy status is classified into three states, such as "surplus", "deficient" and "balanced". Set specific classification standards, such as a deviation within ±0.5kWh is considered "balanced"; a deviation greater than +0.5kWh is "surplus"; a deviation less than -0.5kWh is "deficient". Traverse the supply and demand deviation every hour and determine the status according to the classification standards. For example: 10:00-deviation +0.7kWh, marked as "surplus". 14:00-deviation -1.2kWh, marked as "deficient". 18:00-deviation 0.3kWh, marked as "balanced". Generate an hourly energy status data set for analyzing the actual supply and demand status of the system.

[0118] S24, analyzing the state of the energy storage system according to the basic data of the energy system to obtain energy storage capacity estimation data;

[0119] Specifically, the basic data of the energy storage system is used to calculate the current energy storage capacity and the operating time under the current supply and demand status. According to the current charging rate and battery capacity of the energy storage system, the current remaining energy storage is estimated. If the energy storage battery capacity is 100kWh and the current charging rate is 50%, the estimated current energy storage capacity is 50kWh. According to the current electricity demand, the power supply time that can be supported under this energy storage state is calculated. For example, if the current supply and demand deviation is -2kWh / h, the current 50kWh energy storage can provide continuous power supply for about 25 hours. The analysis results of the energy storage system status are recorded as energy storage capacity estimation data.

[0120] S25. Perform edge state simulation based on energy system basic data, energy state classification data, and energy storage capacity estimation data to obtain a virtual energy system model for transmission to a cloud server.

[0121] Specifically, based on the basic data of the energy system, the energy state classification data and the energy storage capacity estimation data, a virtual simulation of the system is performed at the edge. The basic data, energy state classification data and energy storage capacity estimation data are used as input to establish the parameters of the virtual energy system model. The simulation model simulates the dynamic changes of the actual system under different supply and demand conditions. For example, the change of the charging rate of the energy storage system in the "surplus" state and the change of the discharge rate in the "insufficient" state are simulated. The simulation outputs the virtual energy system model data, including dynamic data such as hourly power generation, power consumption, and energy storage status, which are provided to the cloud server for further analysis and optimization.

[0122] Optionally, the power generation and power consumption matching analysis includes:

[0123] Conduct demand response elasticity analysis based on daily energy trend data to obtain response elasticity score data;

[0124] Specifically, based on daily energy trend data, analyze the response elasticity of electricity load to changes in energy prices or power generation, and quantify the response capacity of each electricity load under different conditions. Divide the electricity load into multiple categories (such as peak load, valley load, elastic load, etc.), and analyze the adjustable range of each load. For example, the load can be adjusted by 20% during peak periods and by 10% during valley periods. Different scores are assigned according to the degree of response elasticity (based on threshold judgment or weight formula), and the higher the score, the stronger the load response elasticity. For example, the elasticity score of peak load is 80, and that of valley load is 50. Generate response elasticity score data for each load category for subsequent multi-scenario analysis.

[0125] The flexibility score of power generation equipment is scored based on daily energy trend data and basic energy system data to obtain the flexibility score data of power generation equipment;

[0126] Specifically, combined with daily energy trend data and basic energy system data, the output flexibility of power generation equipment (such as photovoltaics, wind power, and energy storage systems) is analyzed to quantify the rapid adjustment capabilities of each power generation equipment. The startup and shutdown time of various types of power generation equipment, as well as the rate of increase and decrease in power generation, are analyzed (based on threshold judgment or weight formula). For example, it takes 5 minutes for wind power equipment to start, photovoltaics can be adjusted instantly, and the energy storage system completes output adjustment within 2 minutes. Scores are assigned according to the degree of flexibility of the power generation equipment. For example, the flexibility score of the energy storage system is 90, photovoltaics is 70, and wind power is 60. Generate flexibility score data for each power generation equipment to support supply and demand matching simulation.

[0127] Perform multi-scenario supply and demand matching simulation based on daily energy trend data, response elasticity score data, and power generation equipment flexibility score data to obtain multi-scenario supply and demand simulation data;

[0128] Specifically, based on daily energy trend data, response elasticity score data, and power generation equipment flexibility score data, a supply and demand matching simulation under different scenarios (such as peak load, valley load, and sudden load) is constructed. Set a variety of supply and demand matching scenarios, for example: Scenario 1: Peak load, power generation equipment is at full load. Scenario 2: Valley load, power generation equipment is adjusted on demand. Scenario 3: Sudden load fluctuations, power generation equipment responds quickly. In each scenario, the supply and demand balance state is calculated based on the response elasticity score and the power generation flexibility score. For example, in the peak load scenario, the power supply output of the energy storage system is increased to match the demand. Record the supply and demand matching data under each scenario and generate multi-scenario supply and demand simulation data.

[0129] Generate a short-term balance plan based on multi-scenario supply and demand simulation data to obtain short-term balance plan data;

[0130] Specifically, a short-term supply and demand balance plan is developed based on multi-scenario supply and demand simulation data to ensure system balance under different demand conditions. A power supply plan is generated in hours, for example, increasing the discharge of energy storage systems during peak load periods and reducing the output of power generation equipment during off-peak periods. According to the short-term forecast results in the supply and demand simulation data, the details of load adjustment and power generation scheduling are arranged to develop a short-term balance plan. Short-term balance plan data is generated for real-time adjustment.

[0131] Make real-time deviation adjustments based on short-term balance plan data to obtain real-time deviation adjustment numbers;

[0132] Specifically, based on the short-term balancing plan data, the actual supply and demand situation is monitored in real time, and deviation adjustments are made. Detect the deviation between the actual supply and demand and the short-term balancing plan, and record the current deviation value. For example, if the expected load is 8kWh, but the actual load is 9kWh, the deviation is +1kWh. Adjust the output of the power generation equipment and the energy storage system according to the deviation value to eliminate the deviation. For example, when the deviation is +1kWh, increase the discharge of the energy storage system to compensate. Record the deviation adjustment number for use in stability assessment.

[0133] Supply and demand stability assessment is performed based on multi-scenario supply and demand simulation data and real-time deviation adjustment numbers to obtain supply and demand stability assessment data;

[0134] Specifically, combine the multi-scenario supply and demand simulation data and the real-time deviation adjustment number to evaluate the stability of the system supply and demand to ensure stability in various scenarios. Analyze the multi-scenario supply and demand simulation results and the actual deviation adjustment, and calculate the amplitude of the system supply and demand fluctuations. For example, record the system deviation recovery time and deviation size in an emergency scenario. According to the supply and demand fluctuations, different stability scores are given, and the higher the score, the stronger the system stability. Generate supply and demand stability evaluation data to provide a basis for the generation of supply and demand deviations.

[0135] The power generation supply and demand deviation is generated according to the response elasticity score data, the power generation equipment flexibility score data and the supply and demand stability assessment data to obtain the power generation supply and demand deviation data.

[0136] Specifically, based on the response elasticity score, the power generation equipment flexibility score and the supply and demand stability assessment data, the power generation supply and demand deviation data is generated. Considering the response elasticity, power generation flexibility and system stability, the supply and demand deviation is calculated according to the actual situation. For example, under peak load conditions, the deviation is -1.5kWh, and under low load conditions, the deviation is +0.8kWh. The supply and demand deviation of each time period is recorded as the power generation supply and demand deviation data, and the results of the supply and demand matching analysis are output.

[0137] Optionally, S3 includes:

[0138] S31, controlling the cloud server to receive the virtual energy system model uploaded by the edge device;

[0139] Specifically, the cloud server receives the virtual energy system model uploaded by the edge device and performs data parsing and preparation. The server verifies the integrity and accuracy of the uploaded model data. The virtual energy system model is decompressed and parsed, and the basic information of each device (such as power, status, running time, etc.) is extracted and formatted to ensure that it can be applied to the subsequent construction of the digital twin model.

[0140] S32, initializing the digital twin model according to the virtual energy system model to obtain an initialized digital twin model;

[0141] Specifically, based on the virtual energy system model, an initial digital twin model is constructed, which is used to accurately simulate the current state of the energy system. Key data (such as equipment status, power generation, and load demand) are extracted from the virtual model and used as the initial parameters of the digital twin model. Using these initial parameters, an initial digital twin model is constructed in the cloud so that the model can accurately reflect the overall operating status of the current energy system. An initialized digital twin model is generated to record the initial information of the current state of the system.

[0142] S33, optimizing the initialized digital twin model to obtain a digital twin optimization model;

[0143] Specifically, the initialized digital twin model is further optimized, and the model parameters are dynamically adjusted to improve the accuracy and responsiveness of the model. In the digital twin model, the response speed, power adjustment range and operation mode of the device are dynamically adjusted to optimize its response effect. For example, the charging and discharging rate of the energy storage device and the power generation power of the photovoltaic device are adjusted. Through the real-time data feedback mechanism, the model can automatically adjust its own parameters according to the actual operation of the system to enhance the adaptability of the model. Generate an optimized digital twin model to provide a basis for multi-source data fusion.

[0144] S34, performing multi-source data fusion according to the digital twin optimization model to obtain a digital twin fusion model;

[0145] Specifically, data from different sources (such as weather, load, market data, etc.) are integrated to achieve multi-source data fusion in the digital twin optimization model. Data from different sources are aligned in time and space. For example, meteorological data and power generation load data are matched according to a unified timestamp. Multi-source data are integrated in the digital twin model so that the model can accurately simulate the impact of the external environment on the energy system. For example, the photovoltaic power generation is predicted in combination with meteorological data, and the operation strategy of the energy storage system is optimized in combination with market data. A digital twin fusion model containing multi-source data is generated with higher environmental perception capabilities.

[0146] S35. Perform scene simulation according to the digital twin fusion model to obtain scene simulation data;

[0147] Specifically, based on the digital twin fusion model, the energy system behavior under different operating scenarios is simulated to generate scenario simulation data. Set multiple operating scenarios (such as peak load, emergencies, and severe weather) to simulate system operation under different conditions. In each scenario, the digital twin fusion model is used to calculate the operating status and resource allocation of the system. For example, simulate how energy storage devices support system balance at peak load, or the impact of reduced photovoltaic power generation on the system in severe weather. Record the simulation data of each scenario, including system status, load changes, and response measures, to provide a basis for subsequent model corrections.

[0148] S36. Modify the digital twin fusion model according to the scenario simulation data to obtain the digital twin model of the energy system.

[0149] Specifically, the digital twin fusion model is corrected according to the scenario simulation data to ensure that its simulation of the real energy system is more accurate and comprehensive. Compare the scenario simulation data with the actual system operation data, calculate the errors and deviations of the model, and identify the inaccurate parts of the model. Adjust the parameters in the digital twin fusion model, such as response time, power range, and equipment operation efficiency, to reduce errors. Through multiple iterative corrections, a more accurate digital twin model of the energy system is generated, which can be used for real-time monitoring and optimization of the energy system. Compare the simulation data with the actual system operation data. For example, in scenario 1, the actual power generation is 8.5kWh, but the simulation power is only 8kWh, resulting in an error of 0.5kWh. Adjust the power generation range of the photovoltaic system to make the simulation consistent with the actual, and set the allowable power generation range to 7.5-8.5kWh.

[0150] Optionally, the digital twin model optimization includes:

[0151] Performing a first digital twin model optimization on the initialized digital twin model to obtain a first digital twin optimization model;

[0152] Specifically, the initialized digital twin model is optimized in the first way, mainly for the power generation and load characteristics of the energy system, to ensure that the model can accurately simulate the matching of power generation and load. Based on real-time power generation data, the power generation range of photovoltaic and wind power systems is optimized. For example, the power output of the photovoltaic system is set to 5-10kW, and the output of the wind power system is set to 3-5kW. According to the volatility of the power load, the load part of the model is optimized to ensure that the model can cope with changes in load peaks and valleys. The response time parameters of each device are adjusted in the model. For example, the charge and discharge response time of the energy storage system is set to 30 seconds to quickly respond to load changes. After the optimization is completed, the first digital twin optimization model is generated, which mainly improves the simulation effect of power generation and load characteristics.

[0153] Performing a second digital twin model optimization on the initialized digital twin model to obtain a second digital twin optimized model;

[0154] Specifically, the initialized digital twin model is optimized in the second way, mainly for the impact of the external environment on the energy system, so that the model can better adapt to environmental changes. Meteorological data (such as temperature, sunshine, and wind speed) are integrated into the model to make its response to environmental changes more accurate. For example, when the sunshine intensity increases, the photovoltaic power generation power automatically increases. Adjust the charging and discharging strategy of the energy storage system according to market prices or peak electricity demand. For example, during peak electricity price periods, the energy storage system discharges to support loads first; charging is performed during valley periods. Add equipment failure prediction capabilities to the model to simulate the impact of equipment failures on the system. Assuming that the power output of the wind power system decreases when the wind speed exceeds 15m / s, the model automatically adjusts to reflect the failure. After completing the environmental adaptability optimization, generate the second digital twin optimization model to make the model more accurate in responding to external factors such as meteorology and the market.

[0155] Cross-validation and error correction are performed based on the first digital twin optimization model and the second digital twin optimization model to obtain a digital twin optimization model, wherein the first digital twin model optimization and the second digital twin model optimization are different digital twin model optimization methods.

[0156] Specifically, through cross-validation, the results of the first and second digital twin optimization models are compared and analyzed, and the existing errors are corrected to obtain a more accurate digital twin optimization model. The results of the two optimization models are compared with the actual data to detect the errors of the two in various scenarios. For example, the load error of the first optimization model in the high-load period is +0.8kW, and the error of the second optimization model under the same conditions is -0.5kW. The error of the first optimization model in the photovoltaic power generation simulation on a sunny day is +0.3kW, and the error of the second optimization model is -0.2kW. Combining the advantages of both, the power generation characteristics of the first optimization model and the environmental adaptability of the second optimization model are combined, and the power generation power, load response and external conditions are adjusted within the error value range. For example: in the high-load period, the load response speed is adjusted in combination with the data of the first optimization model; in the sunny period, the photovoltaic power range is corrected using the data of the second optimization model. Through multiple cross-validations and corrections, the generated digital twin optimization model can more accurately reflect the actual situation of the system, including power generation characteristics, load response and environmental adaptability.

[0157] Optionally, the first digital twin model optimization is specifically as follows:

[0158] Select the parameter space to be optimized for the initialized digital twin model to obtain the parameter space data to be optimized;

[0159] Specifically, determine the key parameter range in the model, that is, the parameter space to be optimized, to cover the main variables that affect the performance of the digital twin model. Select key power generation, load and response speed parameters in the initialization model. For example: photovoltaic power generation power range: 5-10kW, wind power generation power range: 3-5kW, energy storage system charge and discharge rate: 1-3kW / h, generate parameter space data to be optimized, including the initial range and upper and lower limits of each parameter to limit the optimization range.

[0160] Selecting an optimization target based on the parameter space data to be optimized to obtain optimization target selection data;

[0161] Specifically, based on the parameter space data to be optimized, define the main optimization objectives so that the model can maximize or minimize performance in specific aspects. Set the goal of model optimization, for example: maximize the power output stability of power generation equipment (such as keeping the fluctuation of photovoltaic and wind power output within ±5%). Minimize load response time (such as response time <1 minute). Generate optimization target selection data and clarify the optimization direction of each parameter to be optimized.

[0162] Select data according to the optimization target to generate evaluation criteria, and obtain optimized evaluation standard data;

[0163] Specifically, data is selected based on the optimization goal, and evaluation criteria for measuring the optimization effect are defined. Specific measurement criteria and thresholds are defined, for example, the evaluation criteria for power output fluctuation is set to within ±5%. The evaluation criteria for load response time is set to no more than 60 seconds. Optimization evaluation standard data is generated so that the performance of the agent model can be continuously evaluated during the optimization process to see if it meets the requirements.

[0164] Perform Gaussian process regression calculation based on the initialized digital twin model to obtain the initial proxy model;

[0165] Specifically, the Gaussian process regression method is used to generate an initial proxy model based on the initialized digital twin model. According to the data in the initialization model, the parameter space data to be optimized is input into the Gaussian process regression process to fit a preliminary proxy model. The initial proxy model provides a mapping relationship between parameters and targets, which can be quickly estimated as the basis for optimization. The initial proxy model is generated for further Bayesian iterative optimization.

[0166] Perform Bayesian iterative optimization based on the initial proxy model to obtain an updated proxy model;

[0167] Specifically, based on the initial proxy model, the Bayesian optimization method is used to gradually update the proxy model to continuously approach the optimization target. Each iteration adjusts the key parameters of the model, such as increasing the upper limit of the photovoltaic power range and reducing the response time of the wind power system. According to the feedback of the proxy model, the parameter space is gradually reduced, and the proxy model of each round is optimized to improve the accuracy. The updated proxy model is generated and gets closer to the optimal performance of the actual system through multiple iterations.

[0168] Convergence judgment is performed based on the update agent model to obtain the optimal parameter data of the digital twin;

[0169] Specifically, the optimization results of the updated proxy model are judged for convergence to determine whether the proxy model has achieved the preset optimization goal. According to the optimization evaluation standard data, detect whether the model meets the optimization requirements. For example: if the power fluctuations of photovoltaic and wind power are both within the range of ±5%, and the load response time has stabilized within 60 seconds, the model is considered to have converged. If the model has achieved the optimization goal, the optimal parameters are output; otherwise, Bayesian iterative optimization continues. The optimal parameter data of the digital twin is obtained for constructing the first digital twin optimization model. For example, after the tenth round of iterations, the model parameter results are: photovoltaic power 7.8kW, wind power 4.1kW, energy storage charge and discharge rate 2kW / h. The proxy model results show that the response time is stable at 59 seconds and the power fluctuation is ±4%.

[0170] Optimization is performed based on the digital twin optimal parameter data and the initialized digital twin model to obtain the first digital twin optimization model.

[0171] Specifically, the optimal parameter data of the digital twin is applied to the initialized digital twin model to generate the first digital twin optimization model. The optimal parameter values ​​are applied to the power generation, load and response time parameters of the model. Simulation runs are performed in the optimization model to verify its stability and response speed in different scenarios to ensure that the optimal parameters have indeed improved the performance of the model. The first digital twin optimization model is generated, which has better power generation output stability and load response performance. Applying the optimal parameter data, the photovoltaic power is 7.8kW, the wind power is 4.1kW, and the energy storage charge and discharge rate is 2kW / h. The simulation test shows that the model response time is kept within 59 seconds, and the power fluctuation is ±4%, which meets the optimization criteria.

[0172] Optionally, the second digital twin model optimization is specifically as follows:

[0173] Perform baseline performance evaluation on the initialized digital twin model to obtain model baseline performance data;

[0174] Specifically, a baseline performance test is performed on the initialized digital twin model to obtain the initial performance data of the system for optimization reference. The model is tested for baseline performance of load response time, power output stability, and environmental adaptability. Average response time: 70 seconds, power fluctuation range: ±6%, power output deviation under weather changes: ±10%, and the response of the initialized model under different load and weather conditions, including output stability and adaptability to environmental changes, is recorded. Generate model baseline performance data as a performance benchmark before optimization.

[0175] Select the maximum deviation optimization target according to the model baseline performance data to obtain the maximum deviation optimization target data;

[0176] Specifically, based on the baseline performance data, set the main goals of model optimization to reduce the maximum deviation of the system and improve the adaptability and response speed of the model. Optimize the power fluctuation range to within ±4%. Reduce the average response time to 60 seconds. Optimize the power output deviation under weather conditions to within ±5%. Select the maximum deviation in the baseline performance and set the optimization direction and range. Generate the maximum deviation optimization target data for the control parameter adjustment step.

[0177] Adjust the control parameters of the initialized digital twin model according to the maximum deviation optimization target data to obtain control parameter adjustment data, wherein the control parameter adjustment includes model learning rate adjustment, model smoothing coefficient adjustment, and model loss coefficient adjustment;

[0178] Specifically, according to the maximum deviation optimization target, the control parameters of the initialized model are adjusted to ensure that the model can be gradually improved in the optimization direction. The adjusted control parameters include: the model learning rate is adjusted from 0.01 to 0.005 to reduce volatility. The smoothing coefficient is adjusted from 0.8 to 0.9 to reduce response noise. The loss coefficient is adjusted from 0.02 to 0.015 to reduce error accumulation. According to the direction set by the optimization target, the learning rate, smoothing coefficient and loss coefficient of the model are adjusted to enhance the adaptability and robustness of the model to data changes. Generate control parameter adjustment data and use the optimized parameter values ​​as input for subsequent optimization iterations.

[0179] According to the control parameters, the data is adjusted and the digital twin model is initialized to perform optimization iteration and model adjustment to obtain the second digital twin optimization model.

[0180] Specifically, the control parameters are used to adjust the data, and optimization iterations are performed on the initialization model. By gradually updating the model parameters, the deviation is continuously reduced to achieve the optimization goal. After the fifth round of iteration, the model performance data is: Average response time: 63 seconds,

[0181] Power fluctuation range: ±4.5%, power output deviation under weather conditions: ±6%. After the 10th round of iteration, the model performance was further optimized to: average response time: 59 seconds, power fluctuation range: ±4%, power output deviation under weather conditions: ±5%. Each round of iteration gradually adjusts the control parameters of the model based on the current performance feedback to converge to the optimization target. After each round of iteration, the learning rate, smoothing coefficient and loss coefficient in the model are adjusted to gradually approach the optimal parameters. After multiple rounds of optimization and adjustment, the second digital twin optimization model is generated to achieve a stable response and a lower deviation level, which meets the maximum deviation optimization target.

[0182] Optionally, S4 includes:

[0183] S41. Perform energy demand forecasting based on the energy system digital twin model to obtain energy demand forecast data;

[0184] Specifically, the energy demand in the future time period is predicted by regression model or time-shifted exponential smoothing using the historical data and current state of the digital twin model of the energy system. The historical load, seasonal changes and usage pattern data are extracted from the digital twin model. These data are used to predict the energy demand for the next 24 hours or longer, generate a prediction curve, and predict the electricity demand in different time periods. Generate energy demand forecast data, record the predicted demand every hour or every minute, and use it for supply and demand balance calculation. Daily demand data extracted from the digital twin model, such as the average daily load and daily peak time period in the past 30 days. Based on these historical data, the model predicts the demand for the next 24 hours. For example: 00:00-06:00: low demand, average 4kWh, 06:00-18:00: gradually rising, peak at 12:00, demand 10kWh, 18:00-24:00: demand falls, average 7kWh, smooth and trend the historical data to obtain the future demand curve. Generate energy demand forecast data, record the expected demand in different time periods, and use it for subsequent supply and demand balance calculation.

[0185] S42. Evaluate the energy production capacity of the energy system digital twin model to obtain energy production capacity evaluation data;

[0186] Specifically, based on the digital twin model of the energy system, the maximum production capacity of different power generation equipment and the output limit under environmental conditions are evaluated. Check the current status (such as power output, equipment health status) and output upper limit of photovoltaic, wind power, energy storage and other equipment. Combine meteorological data (such as light, wind speed) to evaluate the production capacity of photovoltaic and wind power systems under different conditions. For example, the maximum photovoltaic power generation on a sunny day is 8kW, and the maximum on a cloudy day is 5kW. Generate energy production capacity evaluation data to record the available production capacity under different conditions. Combine the equipment information (such as photovoltaic, wind power, energy storage system) and environmental conditions (such as meteorological data) of the digital twin model to evaluate future energy production capacity. Input the rated power and environmental data (such as expected sunshine and wind speed) of photovoltaic, wind power and energy storage systems. It is expected that the sunshine will be good, and the power generation of the photovoltaic system will fluctuate between 5-8kW in the next 24 hours. The wind speed is 4-6m / s, and the power generation of the wind power system is between 3-4kW. The current storage is 50kWh, and the maximum charge and discharge rate of the energy storage system is 3kW / h. Combine the sunshine and wind speed conditions of photovoltaic and wind power systems to calculate the hourly power generation. Generate energy production capacity assessment data and record the power generation range of each device as the basis for supply and demand balance calculation.

[0187] S43, performing energy supply and demand balance calculation based on energy demand forecast data and energy production capacity assessment data to obtain energy supply and demand balance table data;

[0188] Specifically, based on energy demand forecast data and energy production capacity assessment data, calculate the energy supply and demand balance and generate a supply and demand balance sheet. Calculate the difference between supply and demand hourly to identify power supply gaps or surpluses, such as less than 2kW of power supply during peak load periods and 1.5kW of power supply surplus during off-peak load periods. Record the supply and demand difference data as supply and demand balance sheet data, including supply, demand and difference for each time period. Generate energy supply and demand balance sheet data to provide a basis for scheduling optimization strategies.

[0189] Specifically, based on the energy demand forecast data and energy production capacity assessment data, calculate the energy supply and demand balance for each time period. 06:00-12:00: The forecast demand is 8kWh, and the production capacity is 6kWh (the difference is -2kWh, energy storage supplement), 12:00-18:00: The forecast demand is 10kWh, and the production capacity is 11kWh (the difference is +1kWh, energy storage charging), 18:00-24:00: The forecast demand is 7kWh, and the production capacity is 5kWh (the difference is -2kWh, energy storage supplement). Record the supply, demand and difference of each time period, and use the energy storage system to supplement during the power shortage period. Generate energy supply and demand balance table data, list the supply and demand differences every hour, and provide a basis for the next step of scheduling optimization strategy.

[0190] S44, generating an energy dispatch optimization strategy based on the energy supply and demand balance table data to obtain energy dispatch optimization strategy data;

[0191] Specifically, based on the data of the supply and demand balance table, the dispatch optimization strategy of the energy system is generated to ensure the balance of supply and demand under load fluctuations. The energy storage system is enabled during the period of insufficient power supply, and discharge is prioritized to supplement the load; the energy storage system is switched to charging mode during the period of excess power supply. The specific time nodes and power output of charging and discharging are set for the energy storage equipment, and the output arrangement of the power generation equipment is optimized. Energy dispatch optimization strategy data is generated to record the discharge / charging arrangement of the energy storage system and the power setting of the power generation equipment. Dispatching optimization strategies are generated based on the data of the supply and demand balance table to maintain the supply and demand balance of the system during peak and valley periods. Strategy settings such as 06:00-12:00: the energy storage system discharges 2kW to ensure peak demand, 12:00-18:00: the energy storage system charges 1kW to absorb excess power generation, and 18:00-24:00: the energy storage system discharges 2kW again to balance demand. Specific charging and discharging time and power are set for the energy storage system, and the output arrangement of the power generation equipment is optimized. Energy dispatch optimization strategy data is generated to record the charging and discharging arrangement and the power setting of the power generation equipment.

[0192] S45, performing simulation verification according to the energy dispatch optimization strategy data to obtain dispatch verification data;

[0193] Specifically, the scheduling strategy is simulated and verified in the digital twin model to evaluate its effectiveness in actual applications. Run the scheduling strategy in the digital twin model to observe the load response, the use effect of the energy storage system, and whether the supply and demand differences can be eliminated. Record the supply and demand differences in each time period to see whether the scheduling strategy achieves the goal of supply and demand balance. Generate scheduling verification data to evaluate the effectiveness of the scheduling strategy for the next step of correction. Simulate the actual effect of the scheduling strategy in the digital twin model to verify the effectiveness of the strategy. Run the scheduling strategy in the simulation scenario to observe whether the system can achieve supply and demand balance in each time period. 06:00-12:00 Discharge support is effective, 12:00-18:00 Charging is smooth, and 18:00-24:00 Power supply is balanced. Record the supply and demand differences in each time period during the simulation and analyze whether the scheduling strategy effectively balances supply and demand. Generate scheduling verification data for the next step of strategy correction.

[0194] S46, correcting the energy dispatch optimization strategy data according to the dispatch verification data to obtain energy system energy dispatch data;

[0195] Specifically, the dispatch strategy is adjusted according to the simulation verification results to ensure its applicability under different load conditions. Compare the simulation results with the actual demand and analyze the supply and demand deviation in each period. Increase the discharge of the energy storage system or adjust the output of the power generation equipment during the insufficient period, and increase the charging amount or reduce the output power of the power generation equipment during the surplus period. Generate corrected dispatch data for the energy system to ensure that the dispatch strategy can achieve the expected effect in a variety of scenarios. According to the dispatch verification data, optimize the dispatch strategy to ensure the effectiveness of dispatch under different conditions. Adjustment examples such as 06:00-12:00: insufficient discharge in the verification, increase the energy storage discharge to 2.5kW, 12:00-18:00: verification found excessive charging, reduce the charging power to 0.8kW. Adjust the output of the energy storage and power generation systems according to the verification results of each period to achieve a more accurate supply and demand match. Generate corrected energy dispatch data for the energy system to ensure that the strategy is effective in different scenarios.

[0196] S47. Send the energy system energy scheduling data to the edge device to perform energy system scheduling optimization operations.

[0197] Specifically, the generated energy dispatch data is sent to the edge device to perform the actual dispatch optimization of the energy system. The energy dispatch data is packaged and transmitted to the edge to ensure data integrity and real-time performance. After receiving the dispatch data, the edge device adjusts the output of each device and the charging and discharging operation of the energy storage system according to the instructions to achieve optimized dispatch. The energy system achieves balanced and optimized operation under load fluctuations.

[0198] Therefore, from any point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0199] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for constructing a digital twin service platform based on cloud-edge collaboration, characterized in that: The method comprises: S1. Obtain basic data of energy system through edge devices; S2, using the basic data of the energy system to perform preliminary state modeling at the edge device and construct a virtual energy system model for transmission to the cloud server; S3, controlling the cloud server to receive the virtual energy system model uploaded by the edge device, and constructing a digital twin model of the virtual energy system model to obtain a digital twin model of the energy system; S4. Optimize energy scheduling based on the digital twin model of the energy system to obtain energy scheduling data of the energy system and send it to the edge device for energy system scheduling optimization.

2. The method according to claim 1, characterized in that S1 includes: Through the sensors deployed on photovoltaic inverter equipment, basic data of photovoltaic energy system is collected in real time; Through the sensors deployed on wind turbine equipment, basic data of wind power energy system is collected in real time; Through the sensors deployed on the energy storage battery equipment, the basic data of the energy storage energy system is collected in real time; Benchmark the photovoltaic energy system basic data, the wind power energy system basic data and the energy storage energy system basic data to obtain energy system benchmark verification data; Time synchronization is performed based on the energy system benchmark verification data to obtain the energy system basic data.

3. The method according to claim 1, characterized in that S2 include: The edge device uses the basic data of the energy system to perform daily energy trend analysis to obtain daily energy trend data; Perform power generation and power consumption matching analysis based on daily energy trend data to obtain power generation supply and demand deviation data; Classify energy status according to power generation supply and demand deviation data to obtain energy status classification data; Analyze the status of the energy storage system based on the basic data of the energy system to obtain the estimated data of the energy storage capacity; Edge state simulation is performed based on the basic data of the energy system, energy state classification data, and energy storage capacity estimation data to obtain a virtual energy system model for transmission to the cloud server.

4. The method according to claim 1, characterized in that The analysis of power generation and power consumption matching includes: Conduct demand response elasticity analysis based on daily energy trend data to obtain response elasticity score data; The flexibility score of power generation equipment is scored based on daily energy trend data and basic energy system data to obtain the flexibility score data of power generation equipment; Perform multi-scenario supply and demand matching simulation based on daily energy trend data, response elasticity score data, and power generation equipment flexibility score data to obtain multi-scenario supply and demand simulation data; Generate a short-term balance plan based on multi-scenario supply and demand simulation data to obtain short-term balance plan data; Make real-time deviation adjustments based on short-term balance plan data to obtain real-time deviation adjustment numbers; Supply and demand stability assessment is performed based on multi-scenario supply and demand simulation data and real-time deviation adjustment numbers to obtain supply and demand stability assessment data; The power generation supply and demand deviation is generated according to the response elasticity score data, the power generation equipment flexibility score data and the supply and demand stability assessment data to obtain the power generation supply and demand deviation data.

5. The method according to claim 1, characterized in that S3 include: Control the cloud server to receive the virtual energy system model uploaded by the edge device; Initializing the digital twin model according to the virtual energy system model to obtain an initialized digital twin model; Performing digital twin model optimization on the initialized digital twin model to obtain a digital twin optimization model; Multi-source data fusion is performed according to the digital twin optimization model to obtain a digital twin fusion model; Perform scene simulation based on the digital twin fusion model to obtain scene simulation data; The digital twin fusion model is modified according to the scenario simulation data to obtain the digital twin model of the energy system.

6. The method according to claim 5, characterized in that The digital twin model optimization includes: Performing a first digital twin model optimization on the initialized digital twin model to obtain a first digital twin optimization model; Performing a second digital twin model optimization on the initialized digital twin model to obtain a second digital twin optimized model; Cross-validation and error correction are performed based on the first digital twin optimization model and the second digital twin optimization model to obtain a digital twin optimization model, wherein the first digital twin model optimization and the second digital twin model optimization are different digital twin model optimization methods.

7. The method according to claim 6, characterized in that The first digital twin model optimization is as follows: Select the parameter space to be optimized for the initialized digital twin model to obtain the parameter space data to be optimized; Selecting an optimization target based on the parameter space data to be optimized to obtain optimization target selection data; Select data according to the optimization target to generate evaluation criteria, and obtain optimized evaluation standard data; Perform Gaussian process regression calculation based on the initialized digital twin model to obtain the initial proxy model; Perform Bayesian iterative optimization based on the initial proxy model to obtain an updated proxy model; Convergence judgment is performed based on the update agent model to obtain the optimal parameter data of the digital twin; Optimization is performed based on the digital twin optimal parameter data and the initialized digital twin model to obtain the first digital twin optimization model.

8. The method according to claim 6, characterized in that The second digital twin model optimization is specifically as follows: Perform baseline performance evaluation on the initialized digital twin model to obtain model baseline performance data; Select the maximum deviation optimization target according to the model baseline performance data to obtain the maximum deviation optimization target data; Adjust the control parameters of the initialized digital twin model according to the maximum deviation optimization target data to obtain control parameter adjustment data, wherein the control parameter adjustment includes model learning rate adjustment, model smoothing coefficient adjustment, and model loss coefficient adjustment; According to the control parameters, the data is adjusted and the digital twin model is initialized to perform optimization iteration and model adjustment to obtain the second digital twin optimization model.

9. The method according to claim 1, characterized in that S4 include: Energy demand forecasting is performed based on the energy system digital twin model to obtain energy demand forecast data; Evaluate the energy production capacity of the energy system digital twin model to obtain energy production capacity evaluation data; Calculate energy supply and demand balance based on energy demand forecast data and energy production capacity assessment data to obtain energy supply and demand balance table data; Generate energy dispatch optimization strategy based on energy supply and demand balance table data to obtain energy dispatch optimization strategy data; Conduct simulation verification based on energy dispatch optimization strategy data to obtain dispatch verification data; The energy dispatch optimization strategy data is modified according to the dispatch verification data to obtain the energy dispatch data of the energy system; The energy system energy scheduling data is sent to the edge device to perform energy system scheduling optimization operations.

Images