A zero-carbon park dispatching system and method thereof

Abstract

This invention relates to the field of zero-carbon park technology, and more particularly to a zero-carbon park scheduling system and method. The system includes a physical park layer, a digital twin layer, a carbon flow calculation engine, a scheduling strategy generator, and an edge controller. The method includes the following steps: real-time mirroring, carbon potential calculation, twin inference, strategy generation and execution, and feedback correction. This invention achieves real-time synchronization and high-fidelity mapping of the status of distributed energy devices, energy storage devices, and load devices, enabling the tracing of the source of each kilowatt-hour of electricity. It can generate a spatiotemporal carbon potential map with time and location tags, and simultaneously achieve synergistic optimization of economic operation and low-carbon emission reduction. It solves the problems of long computation time and difficulty in coping with rapidly changing scenarios in traditional mathematical optimization methods. It can continuously adapt to environmental changes and equipment performance degradation, accurately reflect the true state of the park, and provide technical assurance for long-term reliable operation.

Description

Technical Field

[0001] This invention relates to the field of zero-carbon park technology, specifically to a zero-carbon park scheduling system and method. Background Technology

[0002] Industrial parks, as important carriers of regional economic development and core spaces for industrial agglomeration, are not only concentrated areas of energy consumption but also major sources of carbon emissions. Statistics show that various industrial parks account for a significant proportion of the nation's total carbon emissions, making the promotion of low-carbon transformation in these parks a primary objective.

[0003] In recent years, the construction of zero-carbon industrial parks has received high attention from national and local governments. Multiple departments have jointly issued documents to promote pilot projects for zero-carbon industrial parks, and various provinces and cities have successively introduced construction plans and application guidelines, clearly requiring parks to establish accurate carbon emission statistical accounting systems, increase the proportion of clean energy consumption, and build an integrated "source-grid-load-storage" collaborative mechanism. However, the transformation from traditional high-carbon industrial parks to zero-carbon industrial parks still faces a series of challenges, as follows: (1) The foundation of carbon accounting is weak, and there is a lack of dynamic and accurate carbon footprint tracking capabilities. Currently, carbon emission calculations in the park mostly adopt macro-statistical methods based on primary energy consumption, which are difficult to reveal the carbon emission characteristics of energy systems with prominent network attributes. The carbon flow process of energy in the production, transmission, conversion, storage and consumption stages within the park is unclear, making it impossible to achieve the accounting requirements of "clear baseline and accurate data". Existing accounting methods are mostly ex-post statistics, lacking the ability to perceive instantaneous carbon potential and predicted carbon potential in real time, making it difficult to provide a refined decision-making basis for low-carbon scheduling.

[0004] (2) Inefficient energy management and insufficient multi-system collaborative scheduling capabilities. Many industrial parks suffer from "information silos" where HVAC, security, lighting, and power distribution systems are independent and data is fragmented, lacking a unified centralized management platform and cross-system linkage and collaboration mechanisms. Faced with the collaborative scheduling needs of various types of resources such as photovoltaics, energy storage, and controllable loads, traditional rule-based control strategies are difficult to adapt to complex scenarios with multiple time scales and multiple objectives.

[0005] (3) Carbon emissions and energy dispatch are disconnected, lacking a coordinated optimization mechanism for "electricity-carbon". Existing dispatch strategies often take minimizing economic operating costs as the sole objective, failing to incorporate carbon emissions as an optimizable decision variable into the dispatch model. Although the park has introduced low-carbon equipment such as distributed photovoltaics and energy storage, due to a lack of deep understanding of the flow patterns of carbon emissions, it is difficult to achieve priority dispatch of low-carbon resources and global optimization of carbon-energy coupling. How to integrate carbon footprint information into real-time dispatch decisions so that the consumption of each kilowatt-hour of electricity can be associated with its carbon emission equivalent has become a technical problem that zero-carbon parks urgently need to solve.

[0006] (4) Closed-loop optimization between digital twins and physical systems has not yet been formed. Although digital twin technology has been initially applied in the field of smart parks and has realized the digital mapping of the physical world, most applications are still at the level of "visual and observable". They lack real-time two-way interaction with edge control systems, the digital model has insufficient reverse control capability over the physical system, and the "perception-decision-control-feedback" closed loop between the twin and the entity has not yet been established, making it difficult to achieve dynamic optimization and online correction of scheduling strategies.

[0007] To address the aforementioned issues, scholars at Tsinghua University proposed the theory of carbon emission flows in power systems. This theory combines carbon emission analysis with power flow calculations, defining core indicators such as branch carbon flow rate and node carbon potential, thus laying the theoretical foundation for dynamic tracking of carbon footprints. Subsequently, this theory was extended to multi-energy systems, establishing a multi-period carbon emission flow model that considers the coupling characteristics of energy storage periods and the transmission process in the pipeline network. At the application level, existing research has explored using carbon emission flows for distribution network optimization and demand response guidance. Meanwhile, some enterprises have attempted to build an integrated system of "source-grid-load-storage + AI regulation" to achieve peak shaving and valley filling and green electricity consumption through intelligent scheduling. However, most solutions still focus primarily on economic optimization, with insufficient consideration for real-time embedding of carbon potential indicators and multi-objective synergy.

[0008] No solutions have yet been proposed for the relevant technical issues. Summary of the Invention

[0009] To address the problems in related technologies, this invention proposes a zero-carbon park scheduling system and method to overcome the aforementioned technical issues in existing technologies. The purpose of this invention is to achieve real-time synchronization and high-fidelity mapping of the status of distributed energy devices, energy storage devices, and load devices, enabling the tracing of the source of each kilowatt-hour of electricity, generating a spatiotemporal carbon potential map with time and location tags, and simultaneously achieving synergistic optimization of economic operation and low-carbon emission reduction. It solves the problems of long calculation time and difficulty in coping with rapidly changing scenarios in traditional mathematical optimization methods, and can continuously adapt to environmental changes and equipment performance degradation, accurately reflecting the real status of the park and providing technical support for long-term reliable operation.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a zero-carbon park scheduling system, comprising: The physical park layer includes distributed energy equipment, energy storage equipment, controllable load equipment, and data acquisition terminals installed on-site within the park. A digital twin layer is connected to the physical park layer. The digital twin layer receives real-time data from the data acquisition terminal and simultaneously constructs a digital twin model reflecting the real-time status of the physical park. The digital twin model includes an equipment mechanism model, an energy topology model, and a carbon footprint dynamic model. A carbon flow calculation engine is set in the digital twin layer. Based on the real-time data and the carbon footprint dynamic model, the carbon flow calculation engine calculates the instantaneous carbon potential index and predicted carbon potential index of each node and branch in the park. The instantaneous carbon potential index and predicted carbon potential index are used to characterize the carbon emission equivalent corresponding to a unit of energy consumption. A scheduling strategy generator is used to perform simulation and deduction based on the digital twin layer. With the goals of minimizing the overall operating cost of the park and minimizing carbon emissions, the generator combines the instantaneous carbon potential index and the predicted carbon potential index to generate a multi-objective scheduling strategy. An edge controller is located at the physical campus layer. The edge controller receives the scheduling strategy and converts it into control commands to adjust the distributed energy equipment, energy storage equipment and controllable load equipment in real time. The state changes of the physical park layer are fed back to the digital twin layer via the data acquisition terminal to update the digital twin model and form a closed-loop optimization scheduling loop.

[0011] Preferably, the carbon flow computing engine includes: The carbon flow tracking module is used to dynamically calculate the carbon footprint of energy during transmission and conversion based on the carbon emission flow theory, according to the source of energy injection and carbon emission factors, and generate a spatiotemporal carbon potential map with time stamps and spatial locations.

[0012] Preferably, the scheduling policy generator includes: The behavior cloning reinforcement learning module is pre-trained offline using historical best scheduling data to clone the scheduling behavior of traditional optimization algorithms. During the online scheduling phase, based on the current park status, instantaneous carbon potential index, and predicted carbon potential index, it uses a deep deterministic policy gradient algorithm to output scheduling actions in real time.

[0013] Preferably, the digital twin layer includes: The virtual carbon trading market module is used to access real-time market data from external carbon trading markets and conduct simulated carbon trading in the digital twin model to evaluate the gains and losses of carbon assets under different scheduling strategies. The gains and losses of carbon assets are then input as constraints into the scheduling strategy generator.

[0014] To achieve the above objectives, the present invention also provides the following technical solution: A zero-carbon industrial park scheduling method includes the following steps: Step S1, Real-time Mirroring: Acquire multi-source heterogeneous data through data acquisition terminals set up in the physical park, and reconstruct the operating status of the physical park in real time in the digital twin; Step S2, Carbon Potential Calculation: Based on the operating state, calculate the spatiotemporal carbon potential index of key nodes in the park's energy network at the current moment and within the preset future time domain using carbon emission flow theory. Step S3, Twin Simulation: In the digital twin, with the optimization objectives of minimizing operating costs and minimizing carbon emissions, and in conjunction with the spatiotemporal carbon potential index, simulations are performed on multiple candidate scheduling strategies to predict the park's operating trajectory and carbon emission effects under different strategies. Step S4, Strategy Generation and Execution: Select the optimal scheduling strategy based on the simulation results and send it to the edge controller, which then coordinates the control of the underlying devices in the physical park. Step S5, Feedback Correction: Collect the physical park response data after the control command is executed, compare the deviation between the prediction result and the actual result of the digital twin, and use the deviation to correct the digital twin model online.

[0015] Preferably, step S2 specifically includes: Based on the topology of the park's energy network, different basic carbon emission factors are assigned to externally purchased electricity, renewable energy power generation, and gas turbine output. At the energy conversion equipment node, the carbon flow transfer relationship between input energy and output energy is calculated based on the conversion efficiency. At the energy storage device node, the time-accumulated carbon potential of the stored energy is dynamically calculated based on the charging and discharging status.

[0016] Preferably, step S3 further includes: By utilizing a trained behavior cloning reinforcement learning model, scheduling action sequences for multiple time scales can be generated quickly. The digital twin simulates in parallel the comprehensive impact of the scheduling action sequence on the future state of the park, including its impact on energy supply and demand balance, equipment lifespan reduction, and the overall zero-carbon attributes of the park.

[0017] Preferably, the cooperative control in step S4 includes: Prioritize scheduling equipment with low-carbon and negative-carbon capabilities; The charging and discharging strategies of energy storage devices are dynamically adjusted based on the high and low levels of the spatiotemporal carbon potential index.

[0018] Preferably, in step S5, the parameters of the equipment efficiency degradation model and carbon flow tracking model in the digital twin are adaptively corrected by comparing the error between the actual carbon flow and the simulated carbon flow.

[0019] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention is a zero-carbon park scheduling system and method. By constructing a digital twin of the physical park, it realizes real-time synchronization and high-fidelity mapping of the status of distributed energy equipment, energy storage equipment and load equipment, providing a reliable data foundation for accurate decision-making. By setting a carbon flow calculation engine in the digital twin, the instantaneous carbon potential and predicted carbon potential of each node and branch are calculated based on the carbon emission flow theory. It can not only trace the source of each kilowatt-hour of electricity, but also generate a spatiotemporal carbon potential map with time and location labels, transforming abstract carbon emissions into spatiotemporal indicators, and solving the problems of strong lag and coarse granularity in traditional carbon accounting. (2) This invention is a zero-carbon park scheduling system and method. The scheduling strategy generator takes minimizing the overall operating cost and minimizing carbon emissions as dual objectives. It combines real-time carbon potential indicators for simulation and deduction, so that the scheduling decision can dynamically capture the changes in low-carbon and high-carbon periods. It charges / uses electricity when the electricity price is low and the carbon potential is low, and discharges / uses energy when the potential is high, so as to achieve synergistic optimization of economic operation and low-carbon emission reduction. The virtual carbon trading market module introduced by the digital twin layer can simulate the carbon asset gains and losses under different scheduling strategies before decision-making, so that park managers can predict the financial impact of scheduling behavior in the carbon market in advance, and use carbon cost or carbon revenue as decision constraints, thereby improving the park's economic adaptability and risk management capabilities in the context of the carbon market. (3) This invention is a zero-carbon park scheduling system and method. Through the behavior cloning reinforcement learning module, the scheduling experience of traditional optimization algorithms is pre-learned and cloned. When running online, it can output high-dimensional scheduling actions in real time according to the current state and carbon potential index using the deep deterministic strategy gradient algorithm. This solves the problems of long calculation time and difficulty in dealing with rapidly changing scenarios in traditional mathematical optimization methods, and achieves millisecond-level intelligent response. The generated scheduling strategy is converted into control instructions by the edge controller to accurately adjust the underlying equipment, reducing communication delay and dependence on the center. At the same time, the state changes of the physical park after execution are fed back to the digital twin in real time, forming a closed-loop optimization loop of perception-decision-control-feedback, ensuring that the system can continuously adapt to environmental changes and equipment performance degradation. (4) The present invention is a zero-carbon park scheduling system and method. In the feedback correction step, by comparing the deviation between the actual response data and the twin simulation results, the key parameters in the digital twin model are adaptively corrected online, ensuring that the digital twin will not be distorted due to the long-term operation and aging of physical equipment, and can always accurately reflect the real state of the park, providing technical guarantee for long-term reliable operation. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the framework structure of the system of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0022] Example Please see Figure 1 This invention proposes a technical solution for a zero-carbon park scheduling system and method: A zero-carbon park scheduling system includes: The physical park layer includes distributed energy equipment, energy storage equipment, controllable load equipment, and data acquisition terminals installed on-site within the park. The digital twin layer is connected to the physical park layer. The digital twin layer receives real-time data from the data acquisition terminal and simultaneously constructs a digital twin model that reflects the real-time status of the physical park. The digital twin model includes an equipment mechanism model, an energy topology model, and a dynamic carbon footprint model. The carbon flow calculation engine is set in the digital twin layer. Based on real-time data and a dynamic carbon footprint model, the carbon flow calculation engine calculates the instantaneous carbon potential index and predicted carbon potential index of each node and branch in the park. The instantaneous carbon potential index and predicted carbon potential index are used to characterize the carbon emission equivalent corresponding to a unit of energy consumption. The scheduling strategy generator is based on the digital twin layer for simulation and deduction, and aims to minimize the overall operating cost of the park and minimize carbon emissions. It combines instantaneous carbon potential indicators and predicted carbon potential indicators to generate multi-objective scheduling strategies. The edge controller is located at the physical campus layer. It receives scheduling policies and converts them into control commands to adjust distributed energy devices, energy storage devices and controllable load devices in real time. The state changes of the physical park layer are fed back to the digital twin layer through the data acquisition terminal to update the digital twin model and form a closed-loop optimization scheduling loop.

[0023] In this embodiment, the data acquisition terminal not only collects the status data of the equipment (such as power and voltage), but also collects the energy flow data and equipment efficiency data (such as the gas consumption and power generation of the gas turbine). These data are the basis for building the carbon footprint model. Equipment mechanism models describe the physical characteristics of individual devices, such as the light-power curve of photovoltaic panels, the charge-discharge efficiency and aging model of energy storage batteries; energy topology models describe the connection relationships between devices and the paths of energy flow (power lines, heat pipes, etc.); the carbon footprint dynamic model is a dynamic model that can calculate the "carbon content" of each kilowatt-hour of electricity and each cubic meter of heat in real time based on the energy source, conversion path and storage process.

[0024] Instantaneous carbon potential tells staff whether the electricity used by a particular piece of equipment is "green" or "gray" electricity. For example, during periods of high solar power generation, the carbon potential of the park's busbar is low; when the proportion of electricity purchased from the external grid is high, the carbon potential is high.

[0025] Carbon potential is predicted based on weather forecasts (affecting photovoltaics), load forecasts, and external power grid carbon emission factor forecasts, allowing for advance prediction of carbon potential trends over a future period.

[0026] The role of the scheduling strategy generator is to find the optimal balance between the lowest cost and the lowest carbon emissions (i.e., the Pareto optimal solution).

[0027] Furthermore, the carbon flow computing engine includes: The carbon flow tracing module is used to dynamically calculate the carbon footprint of energy during transmission and conversion based on carbon emission flow theory, according to the source of energy injection and carbon emission factors, and generate a spatiotemporal carbon potential map with time stamps and spatial location.

[0028] In this embodiment, the carbon emission flow theory refers to the fact that starting from the source (such as electricity purchased from the grid or photovoltaic power generation), each source is given an initial "carbon concentration" (i.e., carbon emission factor). When energy flows through the line, through the transformer, and into the energy storage or load, the carbon flow also "flows", "mixes", or is "stored".

[0029] The spatiotemporal carbon potential map visualizes the calculation results. On the digital twin model, different colors are used to mark the carbon potential at different locations and times in the park, making the abstract carbon emissions intuitive and visible.

[0030] Furthermore, the scheduling policy generator includes: The Behavior Cloning Reinforcement Learning Module is pre-trained offline using historical best scheduling data to clone the scheduling behavior of traditional optimization algorithms. During the online scheduling phase, based on the current park status, instantaneous carbon potential index, and predicted carbon potential index, it uses a deep deterministic policy gradient algorithm to output scheduling actions in real time.

[0031] In this embodiment, the behavior clone is first pre-trained using historical best scheduling data (such as the result calculated by traditional mathematical optimization algorithms) to allow the model to learn "what is good scheduling behavior", which can greatly accelerate the learning speed.

[0032] When reinforcement learning is running online, in the face of complex and uncertain environments (such as load fluctuations and photovoltaic forecasting deviations), the model continuously optimizes its strategy through continuous trial and error interaction with the environment (digital twin) and outputs real-time, adaptive scheduling actions.

[0033] Furthermore, the digital twin layer includes: The virtual carbon trading market module is used to access real-time market data from external carbon trading markets and simulate carbon trading in a digital twin model to evaluate the gains and losses of carbon assets under different scheduling strategies. The gains and losses of carbon assets are then used as constraints to input into the scheduling strategy generator.

[0034] In this embodiment, the virtual carbon trading market module can add an economic dimension to the scheduling strategy. It can purchase more electricity when carbon prices are low, and when carbon prices are high, the carbon emissions saved through the scheduling strategy can be traded in the virtual market to assess its potential benefits.

[0035] A zero-carbon industrial park scheduling method includes the following steps: Step S1, Real-time Mirroring: Acquire multi-source heterogeneous data through data acquisition terminals set up in the physical park, and reconstruct the operating status of the physical park in real time in the digital twin; Step S2, Carbon Potential Calculation: Based on the operating status, calculate the spatiotemporal carbon potential index of key nodes in the park's energy network at the current moment and within the preset future time domain using carbon emission flow theory; Step S3, Twin Simulation: In the digital twin, with the optimization objectives of minimizing operating costs and minimizing carbon emissions, and combined with spatiotemporal carbon potential indicators, simulations are performed on various candidate scheduling strategies to predict the park's operating trajectory and carbon emission effects under different strategies. Step S4, Strategy Generation and Execution: Select the optimal scheduling strategy based on the simulation results and send it to the edge controller, which then coordinates the control of the underlying devices in the physical park. Step S5, Feedback Correction: Collect the physical park response data after the control command is executed, compare the deviation between the prediction results of the digital twin and the actual results, and use the deviation to correct the digital twin model online.

[0036] Furthermore, step S2 specifically includes: Based on the topology of the park's energy network, different basic carbon emission factors are assigned to externally purchased electricity, renewable energy power generation, and gas turbine output. At the energy conversion equipment node, the carbon flow transfer relationship between input energy and output energy is calculated based on the conversion efficiency. At the energy storage device node, the time-accumulated carbon potential of the stored energy is dynamically calculated based on the charging and discharging status.

[0037] Furthermore, step S3 further includes: By utilizing a trained behavior cloning reinforcement learning model, scheduling action sequences for multiple time scales can be generated quickly. The parallel simulation of the scheduling action sequence in a digital twin will comprehensively impact the future state of the park, including its effects on energy supply and demand balance, equipment lifespan reduction, and the overall zero-carbon attributes of the park.

[0038] Furthermore, the coordinated control in step S4 includes: Prioritize scheduling equipment with low-carbon and negative-carbon capabilities; The charging and discharging strategies of energy storage devices are dynamically adjusted based on the high and low levels of spatiotemporal carbon potential indicators.

[0039] Furthermore, in step S5, by comparing the error between the actual carbon flow and the simulated carbon flow, the parameters of the equipment efficiency degradation model and the carbon flow tracking model in the digital twin are adaptively corrected.

[0040] Scene setting A high-tech industrial park (a zero-carbon pilot park) includes: rooftop photovoltaics, a gas turbine (CCHP, combined cooling, heating and power), an energy storage system, and several controllable industrial loads (such as charging piles with adjustable production times and temperature-adjustable air conditioning systems). The park is connected to the external power grid.

[0041] Objective: To achieve the lowest possible operating cost and carbon emissions throughout the day while meeting the energy needs of production and daily life.

[0042] Phase 1: Early Morning (6:00 - 8:00) Physical Park Level: As the sun rises, photovoltaic output gradually increases. Park load (lighting, office equipment) begins to rise.

[0043] Digital twin layer (twin mirror): Real-time synchronization of data between digital twins.

[0044] Carbon Stream Calculation Engine: The carbon stream calculation engine performs calculations based on real-time data.

[0045] At this time, the carbon emission factor of electricity purchased from the external power grid is relatively high (assuming it is mainly thermal power).

[0046] The output of photovoltaic power is tracked, and its carbon emission factor is 0.

[0047] The instantaneous carbon potential of the park's busbar was calculated to be at a moderate level.

[0048] Carbon potential forecast: Based on the weather forecast (sunny today), photovoltaic power is expected to increase significantly in the next few hours, and the carbon potential of the busbar will decrease significantly.

[0049] Strategy Generation and Derivation: The scheduling strategy generator (pre-trained) begins operation. It predicts a sharp increase in load after 8 PM, but photovoltaic load is also increasing synchronously. It derives multiple scenarios in the twin: Option A: Now discharge the energy storage to replenish the load.

[0050] Option B: Keep the energy storage stationary for now, or even charge it using off-peak electricity (assuming electricity prices are low in the morning).

[0051] The analysis concluded that Option B was the superior option. This is because, based on predicted carbon potential, when solar power generation surges in the future, using "zero-carbon" solar energy to charge energy storage will result in lower overall carbon emissions and lower costs compared to currently using "high-carbon" grid power to supply the load (by utilizing future free solar energy).

[0052] Strategy execution: Upon receiving the instruction, the edge controller remains silent on energy storage and prioritizes the absorption of photovoltaic power.

[0053] Phase Two: Morning (10:00 - 12:00) Physical park level: Photovoltaics reach their peak, and the load is also at a high level. The instantaneous carbon potential becomes very low (close to zero carbon) due to the addition of photovoltaics.

[0054] Strategy generation and deduction: The system observed that the carbon potential was extremely low and predicted that there might be cloud cover in the afternoon, which would reduce photovoltaic power.

[0055] Carbon potential map: This shows that the carbon potential of the energy storage node is very low at this time because it is charged with photovoltaic power.

[0056] Strategy: The system decides to activate carbon capture or power-to-gas (PTO) devices (if available in the park) at this time because they consume a large amount of electricity, which is "zero-carbon" at this moment, making the overall carbon balance most favorable. Simultaneously, the scheduling strategy generator instructs energy storage devices to charge at maximum power to store this inexpensive zero-carbon electricity.

[0057] Strategy execution: The edge controller executes the command, and the energy storage system enters fast charging mode.

[0058] Phase Three: Evening (18:00 - 20:00) Physical park level: Photovoltaic output drops to zero, and the park experiences evening peak traffic. Carbon emission factors from purchasing electricity from the external grid remain high.

[0059] Carbon Flow Calculation Engine: At this moment, the bus's instantaneous carbon potential rises sharply because the main energy source comes from high-carbon mains electricity and gas turbines. However, the carbon flow tracking module shows that some of the energy supplied to the load is "zero-carbon electricity" stored in the afternoon's energy storage system.

[0060] Strategy generation and deduction: The system faces a choice: to discharge energy from the energy storage system or to start the gas turbine? The carbon flow calculation engine provides key data: the current energy storage has a very low "time-cumulative carbon potential" (because it is charged with photovoltaic power), while the instantaneous carbon potential of gas turbine power generation is higher.

[0061] Simulation in a virtual twin: Prioritizing energy storage discharge can smooth load, reduce electricity purchase costs, and lower overall carbon emissions. Simultaneously, the system calculates the carbon emission savings from prioritizing energy storage discharge based on the real-time carbon price in the virtual carbon trading market (assuming a high price at this time). If traded on the market, this could generate substantial revenue.

[0062] Strategy execution: The edge controller prioritizes energy storage discharge and only starts the gas turbine to supplement when the energy storage SOC (state of charge) is lower than the safety threshold, thereby achieving dual optimization of cost and carbon emissions.

[0063] Phase Four: Nighttime Closed-Loop Feedback (After 22:00) Feedback correction: Compare the actual daily operating data (actual photovoltaic power generation, load fluctuation, energy storage charging and discharging efficiency, actual electricity purchase cost, and actual total carbon emissions) with the predicted data from the digital twin on the morning of the same day.

[0064] Model Correction: The actual energy storage charging and discharging efficiency was found to be 2% lower than the model prediction. This deviation was automatically input into the device mechanism model of the digital twin, adaptively correcting the efficiency degradation parameters of the energy storage device. After correction, tomorrow's predictions will be more accurate, thus forming a closed loop of continuous optimization.

[0065] In summary, by constructing a mirror image using digital twins, tracking carbon traces using a carbon flow computing engine, generating strategies using reinforcement learning, introducing economic leverage through a virtual market, and executing instructions using an edge controller, a complete closed loop of intelligent scheduling for zero-carbon industrial parks can be achieved, encompassing perception, analysis, decision-making, execution, and feedback.

[0066] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A zero-carbon industrial park scheduling system, characterized in that, include: The physical park layer includes distributed energy equipment, energy storage equipment, controllable load equipment, and data acquisition terminals installed on-site within the park. A digital twin layer is connected to the physical park layer. The digital twin layer receives real-time data from the data acquisition terminal and simultaneously constructs a digital twin model reflecting the real-time status of the physical park. The digital twin model includes an equipment mechanism model, an energy topology model, and a carbon footprint dynamic model. A carbon flow calculation engine is set in the digital twin layer. Based on the real-time data and the carbon footprint dynamic model, the carbon flow calculation engine calculates the instantaneous carbon potential index and predicted carbon potential index of each node and branch in the park. The instantaneous carbon potential index and predicted carbon potential index are used to characterize the carbon emission equivalent corresponding to a unit of energy consumption. A scheduling strategy generator is used to perform simulation and deduction based on the digital twin layer. With the goals of minimizing the overall operating cost of the park and minimizing carbon emissions, the generator combines the instantaneous carbon potential index and the predicted carbon potential index to generate a multi-objective scheduling strategy. An edge controller is located at the physical campus layer. The edge controller receives the scheduling strategy and converts it into control commands to adjust the distributed energy equipment, energy storage equipment and controllable load equipment in real time. The state changes of the physical park layer are fed back to the digital twin layer via the data acquisition terminal to update the digital twin model and form a closed-loop optimization scheduling loop.

2. The zero-carbon park scheduling system according to claim 1, characterized in that, The carbon stream computing engine includes: The carbon flow tracking module is used to dynamically calculate the carbon footprint of energy during transmission and conversion based on the carbon emission flow theory, according to the source of energy injection and carbon emission factors, and generate a spatiotemporal carbon potential map with time stamps and spatial locations.

3. The zero-carbon park scheduling system according to claim 1, characterized in that, The scheduling policy generator includes: The behavior cloning reinforcement learning module is pre-trained offline using historical best scheduling data to clone the scheduling behavior of traditional optimization algorithms. During the online scheduling phase, based on the current park status, instantaneous carbon potential index, and predicted carbon potential index, it uses a deep deterministic policy gradient algorithm to output scheduling actions in real time.

4. A zero-carbon industrial park scheduling system according to claim 1, characterized in that, The digital twin layer includes: The virtual carbon trading market module is used to access real-time market data from external carbon trading markets and conduct simulated carbon trading in the digital twin model to evaluate the gains and losses of carbon assets under different scheduling strategies. The gains and losses of carbon assets are then input as constraints into the scheduling strategy generator.

5. A zero-carbon park scheduling method for a zero-carbon park scheduling system as described in any one of claims 1 to 4, characterized in that, Includes the following steps: Step S1, Real-time Mirroring: Acquire multi-source heterogeneous data through data acquisition terminals set up in the physical park, and reconstruct the operating status of the physical park in real time in the digital twin; Step S2, Carbon Potential Calculation: Based on the operating state, calculate the spatiotemporal carbon potential index of key nodes in the park's energy network at the current moment and within the preset future time domain using carbon emission flow theory. Step S3, Twin Simulation: In the digital twin, with the optimization objectives of minimizing operating costs and minimizing carbon emissions, and in conjunction with the spatiotemporal carbon potential index, simulations are performed on multiple candidate scheduling strategies to predict the park's operating trajectory and carbon emission effects under different strategies. Step S4, Strategy Generation and Execution: Select the optimal scheduling strategy based on the simulation results and send it to the edge controller, which then coordinates the control of the underlying devices in the physical park. Step S5, Feedback Correction: Collect the physical park response data after the control command is executed, compare the deviation between the prediction result and the actual result of the digital twin, and use the deviation to correct the digital twin model online.

6. A zero-carbon industrial park scheduling method according to claim 5, characterized in that, Step S2 specifically includes: Based on the topology of the park's energy network, different basic carbon emission factors are assigned to externally purchased electricity, renewable energy power generation, and gas turbine output. At the energy conversion equipment node, the carbon flow transfer relationship between input energy and output energy is calculated based on the conversion efficiency. At the energy storage device node, the time-accumulated carbon potential of the stored energy is dynamically calculated based on the charging and discharging status.

7. A zero-carbon industrial park scheduling method according to claim 5, characterized in that, Step S3 further includes: By utilizing a trained behavior cloning reinforcement learning model, scheduling action sequences for multiple time scales can be generated quickly. The digital twin simulates in parallel the comprehensive impact of the scheduling action sequence on the future state of the park, including its impact on energy supply and demand balance, equipment lifespan reduction, and the overall zero-carbon attributes of the park.

8. A zero-carbon industrial park scheduling method according to claim 5, characterized in that, The coordinated control in step S4 includes: Prioritize scheduling equipment with low-carbon and negative-carbon capabilities; The charging and discharging strategies of energy storage devices are dynamically adjusted based on the high and low levels of the spatiotemporal carbon potential index.

9. A zero-carbon industrial park scheduling method according to claim 5, characterized in that, In step S5, the parameters of the equipment efficiency degradation model and carbon flow tracking model in the digital twin are adaptively corrected by comparing the error between the actual carbon flow and the simulated carbon flow.

Images