Load scheduling method for zero-carbon park based on spatiotemporal coupling model and world model fusion

Abstract

The application discloses a zero-carbon park load scheduling method based on spatiotemporal coupling model and world model fusion, comprising the following steps: constructing system state variables and control variables according to the operation data of the park energy system; establishing a spatiotemporal coupling model for describing the energy distribution relationship between different nodes in the system according to the state variables and the control variables; constructing a world model for system state prediction based on the energy distribution relationship; generating a scheduling strategy according to the prediction results output by the world model and applying the scheduling strategy to the park energy system; and updating the state variables according to the system operation feedback to realize closed-loop control. Through the fusion of spatiotemporal coupling modeling and the world model, the application realizes the interpretable modeling and forward-looking optimization of the dynamic evolution of the system, and improves the low-carbon, economy and robustness of the park energy system.

Description

Technical Field

[0001] This invention belongs to the field of smart energy technology, and in particular relates to a zero-carbon park load scheduling method based on the fusion of a spatiotemporal coupling model and a world model. Background Technology

[0002] Park-level energy systems are gradually developing towards high-proportion renewable energy integration, multi-energy complementarity, and intelligent dispatch. The proportion of distributed energy sources such as photovoltaics and wind power in the park's energy structure is continuously increasing. At the same time, the entry and exit of enterprises within the park, dynamic changes in load structure, and fluctuations in electricity price signals give the park's energy system a high degree of dynamism, uncertainty, and nonlinearity. Existing park energy dispatch methods are mainly divided into two categories: one is based on mechanistic models, relying on fixed mathematical models and incorporating engineering constraints such as power balance and equipment operation, but it is difficult to accurately characterize the nonlinear evolution characteristics of the system under complex dynamic environments; the other is based on data-driven methods, using historical data to achieve state prediction, but lacking physical constraints, it is prone to producing results that do not conform to actual engineering operating laws, limiting its application in practical systems.

[0003] However, existing technologies still have the following problems: First, traditional scheduling methods typically treat load as rigid demand, lacking modeling of load-side adjustment potential and failing to fully exploit demand response capabilities, resulting in insufficient system flexibility. Second, in actual park operation, structural changes such as enterprise entry, exit, or sudden load changes can cause disturbances, but existing methods lack a modeling mechanism for the propagation of such disturbances in the system, making it difficult for scheduling strategies to dynamically adapt to changes in system structure. Third, existing methods struggle to unify the modeling of dynamic adjustments to energy allocation structures with system state evolution, and the impact of control behavior on energy flow paths lacks interpretability, resulting in insufficient foresight and robustness of scheduling strategies. Fourth, most methods employ single-step prediction or open-loop control, lacking the ability to proactively evaluate and optimize multi-step control strategies in a virtual environment, causing scheduling decisions to easily deviate from the optimal trajectory of actual operation. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a zero-carbon park load scheduling method based on the fusion of a spatiotemporal coupling model and a world model, comprising: Based on the operational data of the park's energy system, construct system state variables and control variables; Based on the system state variables and the control variables, a spatiotemporal coupling model is established to describe the energy distribution relationship between different nodes in the system; Based on the energy distribution relationship determined by the spatiotemporal coupling model, a world model for system state prediction is constructed. Based on the prediction results output by the world model, a scheduling strategy is generated and applied to the park's energy system; The system state variables are updated based on system operation feedback to achieve closed-loop control.

[0005] Preferably, the process of constructing system state variables and control variables includes: Collect operational data of the park's energy system, and perform time-scale unification and normalization processing on the operational data; Based on the processed operational data, construct the system state vector and control vector; The system state vector includes distributed energy output, load power, energy storage state of charge, electricity price, and meteorological characteristics. The control vector includes energy storage charging and discharging power, load regulation, and grid interaction power.

[0006] Preferably, the process of establishing the spatiotemporal coupling model includes: A spatiotemporal coupling matrix is ​​constructed, which is a function of system state variables and control variables, and is used to describe the energy flow path and distribution ratio between energy nodes and load nodes.

[0007] Preferably, the process of calculating the energy flow path and allocation ratio between the node and the load node based on the spatiotemporal coupling matrix includes: Based on the elements of the spatiotemporal coupling matrix and the output power of the energy node, the power allocated by the energy node to each load node is calculated, and the output power of the energy node is not less than the sum of the power allocated to all load nodes.

[0008] Preferably, the process of constructing the world model includes: A state transition model is constructed to describe the dynamic evolution process of the system under the action of control variables, and the energy flow allocation result determined by the spatiotemporal coupling model is embedded into the state transition model as an intermediate physical variable to obtain the world model.

[0009] Preferably, the process of predicting the system's operating state based on the world model includes: The system state sequence for future operation is obtained by recursively predicting the system state at multiple time steps using the state transition model.

[0010] Preferably, the process of generating a scheduling policy includes: Based on the future operating state sequence of the system, different control strategies are evaluated, and the optimal control strategy that satisfies the preset objective function is selected; the objective function includes system operating cost, carbon emissions, and operating risk indicators.

[0011] Preferably, the method further includes: A load resilience model is constructed to describe the adjustable characteristics of load power under the influence of electricity price changes and dispatch control signals. The load power is represented as the sum of the baseline load, the electricity price response term, and the dispatch command response term.

[0012] Preferably, the method further includes: A disturbance propagation model is constructed to describe the influence of local state changes on the overall system state. The disturbance propagation model includes a disturbance vector and a disturbance propagation matrix. The elements of the disturbance propagation matrix are determined based on the topological distance or association strength between nodes.

[0013] Preferably, the process of achieving closed-loop control includes: Control commands are generated according to the scheduling strategy, and then sent to the actual equipment for execution through the park's energy management system. During control execution, the system state variables are updated based on the real-time acquired system operating status, forming a closed loop of state prediction and control execution.

[0014] Compared with the prior art, the present invention has the following advantages and technical effects: First, by constructing a load resilience model, this invention transforms the load from a passive energy-consuming unit into an adjustable resource, fully tapping the demand response potential on the load side and significantly improving the flexibility and economy of system operation.

[0015] Second, by constructing a disturbance propagation model, this invention quantifies the impact of local state changes on the overall system, enabling the scheduling strategy to dynamically adapt to structural changes such as enterprise entry, exit, or load mutations, thereby enhancing the robustness and adaptability of the system.

[0016] Third, by constructing a spatiotemporal coupling model as a dynamic bridge between the system state and control variables, this invention makes the adjustment process of the energy allocation structure transparent and interpretable, and clearly expresses the impact of control behavior on the energy flow path, thereby improving the credibility and engineering practicality of the scheduling strategy.

[0017] Fourth, this invention constructs a world model with a spatiotemporal coupling model as its core, enabling continuous state projection of the system at multiple future time steps. Based on this, it conducts forward-looking evaluation and optimization of different control strategies, forming a closed-loop control mechanism. This effectively overcomes the limitations of single-step prediction and open-loop control, significantly improving the forward-looking nature and accuracy of scheduling decisions and the overall efficiency of system operation. Attached Figure Description

[0018] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention. Detailed Implementation

[0019] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0021] like Figure 1 As shown, this embodiment provides a zero-carbon park load scheduling method based on the fusion of a spatiotemporal coupling model and a world model, including: Based on the operational data of the park's energy system, construct system state variables and control variables; Based on the system state variables and control variables, a spatiotemporal coupling model is established to describe the energy distribution relationship between different nodes in the system; Based on the energy distribution relationship determined by the spatiotemporal coupling model, a world model for system state prediction is constructed. Based on the predictions output by the world model, a scheduling strategy is generated and applied to the park's energy system. The system state variables are updated based on system operation feedback to achieve closed-loop control.

[0022] Furthermore, the process of constructing system state variables and control variables includes: Collect operational data of the park's energy system and perform time-scale unification and normalization processing on the operational data; Based on the processed operational data, construct the system state vector and control vector; The system state vector includes distributed energy output, load power, energy storage state of charge, electricity price, and meteorological characteristics. The control vectors include energy storage charging and discharging power, load regulation, and grid interaction power.

[0023] Furthermore, this embodiment collects system operation data through the park's energy management system, including output data of distributed energy sources such as photovoltaics and wind power, various load data, meteorological data, electricity price information, and energy storage system operation status data. The above data undergoes time scale unification, outlier removal, missing value completion, and normalization processing to ensure the data has a unified time and numerical scale.

[0024] Furthermore, this embodiment constructs a system state vector. With control vector These are used to describe the current operating state of the system and the scheduling decision variables, respectively.

[0025] in, For discrete time steps, Photovoltaic power generation (kW). Wind power generation capacity (kW). No. Load power (kW) of each enterprise. This represents the state of charge (range 0-1) of the energy storage system. for The grid electricity price at any given time (RMB / kWh). For meteorological feature vectors, Energy storage charging power (kW). The energy storage discharge power is (kW). No. Load regulation capacity (kW) of each enterprise. This refers to the power exchanged with the power grid (kW, positive for purchasing electricity, negative for selling electricity).

[0026] The purpose of this step is to transform the park's energy system into a computable state-space representation, providing a foundation for subsequent modeling and optimization.

[0027] Furthermore, the process of establishing a spatiotemporal coupling model includes: Construct a spatiotemporal coupling matrix, which is a function of system state variables and control variables, to describe the energy flow path and distribution ratio between energy nodes and load nodes.

[0028] Furthermore, the process of calculating the energy flow path and allocation ratio between nodes and load nodes based on the spatiotemporal coupling matrix includes: Based on the elements of the spatiotemporal coupling matrix and the output power of the energy node, calculate the power allocated by the energy node to each load node, and ensure that the output power of the energy node is not less than the sum of the power allocated to all load nodes.

[0029] Furthermore, based on the system state representation, this embodiment constructs a spatiotemporal coupling model to describe the energy transmission paths and distribution relationships within the system. Unlike traditional fixed topology modeling methods, this embodiment defines the energy distribution relationship as a dynamically coupled structure driven by both the system state and control variables.

[0030] Specifically, construct the spatiotemporal coupling matrix: in, For the number of energy nodes, The number of load nodes, matrix elements Indicates the state With control variables Through the combined effect, from energy nodes To load nodes The energy distribution ratio. Based on this coupling relationship, the system energy flow can be expressed as: And satisfy the constraints: in, For energy nodes 'output power' Indicates energy node Distributed to load nodes Power. Control variables. By adjusting the energy storage charging and discharging power, load regulation, and grid interaction power, the supply and demand relationship and marginal cost of each node are changed, thereby driving the coupling matrix. Dynamic reconstruction enables adaptive adjustment of energy flow paths.

[0031] Therefore, the spatiotemporal coupling matrix not only describes the system topology but also reflects the dynamic influence of control behavior on energy allocation paths, making it a core structural variable connecting system state and scheduling decisions.

[0032] During scheduling optimization, the coupling matrix It is not treated as a fixed parameter, but rather as a control variable. Implicitly driven structural variables participate in the optimization process, and their changes reflect the adjustment of energy allocation structure under different scheduling strategies.

[0033] Furthermore, the process of constructing a world model includes: A state transition model is constructed to describe the dynamic evolution process of the system under the action of control variables. The energy flow allocation result determined by the spatiotemporal coupling model is embedded into the state transition model as an intermediate physical variable to obtain the world model.

[0034] Furthermore, the process of predicting the system's operating state based on the world model includes: By recursively predicting the system state at multiple time steps using a state transition model, a sequence of future system operating states can be obtained.

[0035] Furthermore, based on the aforementioned spatiotemporal coupling model, this embodiment further constructs a system state transition model to describe the dynamic evolution of the system under control. Unlike traditional methods that directly predict based on state and control variables, this embodiment embeds the energy flow allocation result as an intermediate physical driving variable into the state transition process. The state transition relationship is expressed as follows: in, This is a state prediction model trained based on historical data. From the coupling matrix With control variables A joint decision.

[0036] Through the above structure, the system evolution process forms the following chain relationship: This allows control variables to influence the energy flow distribution by changing the coupling structure, and ultimately determine the system state evolution.

[0037] Furthermore, this embodiment introduces energy balance constraints, energy storage operation constraints, and grid interaction constraints during model training, so that the state prediction results simultaneously meet the requirements of data fitting accuracy and engineering physical feasibility.

[0038] By explicitly embedding the spatiotemporal coupling structure into the state transition model, this embodiment achieves the unification of "structural modeling + dynamic prediction", enabling the system not only to predict future states, but also to explain the impact of control behavior on the system evolution path, thereby significantly improving the interpretability and reliability of the scheduling strategy.

[0039] Furthermore, based on the aforementioned spatiotemporal coupling model and state transition mechanism, this embodiment constructs a world model for dynamic system deduction, which is essentially a state evolution model with an energy coupling structure as its core. Unlike models used only for single-step prediction, the world model, through recursive calls to the state transition function, can continuously simulate the operating state of the system at multiple future time steps given an initial state and control strategy. Its deduction process is expressed as follows: in, This represents the multi-step recursive form of the state transition function. Indicates from time to The control sequence.

[0040] Through this mechanism, the world model can not only output the state at the next moment, but also construct the future evolution trajectory of the system under different control strategies, thereby forming a virtual simulation environment corresponding to the actual system.

[0041] Furthermore, in the process of state deduction, the state transitions involved in the world model are all calculated based on the energy distribution structure determined by the spatiotemporal coupling model, that is, the control variables first act on the coupling matrix. This, in turn, affects the distribution of energy flow. Ultimately, this drives the evolution of the system state. Therefore, the world model is essentially a system dynamic simulation model with a spatiotemporal coupled structure as its core.

[0042] Furthermore, the process of generating the scheduling policy includes: Based on the future operating state sequence of the system, different control strategies are evaluated, and the optimal control strategy that satisfies the preset objective function is selected. The objective function includes system operating cost, carbon emissions, and operating risk indicators.

[0043] Furthermore, the method also includes: A load resilience model is constructed to describe the adjustable characteristics of load power under the influence of electricity price changes and dispatch control signals. The load power is represented as the sum of the baseline load, the electricity price response term, and the dispatch command response term.

[0044] Furthermore, to enhance the load side's ability to participate in system scheduling, this embodiment constructs a load resilience model to describe the load's adjustability under external signals. The load model is expressed as: in, Indicates the first The baseline load power of each company's load. Indicates the change in electricity price. Indicates scheduling control signals, This is the load response coefficient to changes in electricity prices. This represents the load's response coefficient to scheduling commands.

[0045] Through the above modeling, the load is transformed from a traditional passive energy-consuming unit into an adjustable resource, thereby participating in the system's energy balance and optimal scheduling process and improving the overall flexibility of the system.

[0046] Furthermore, the method also includes: A disturbance propagation model is constructed to describe the influence of local state changes on the overall system state. The disturbance propagation model includes a disturbance vector and a disturbance propagation matrix. The elements of the disturbance propagation matrix are determined based on the topological distance or correlation strength between nodes.

[0047] Furthermore, in response to system structure changes caused by enterprise entry, exit, or sudden load changes during the operation of the park, this embodiment constructs a disturbance propagation model to characterize the impact of local changes on the global system.

[0048] First, define the system disturbance vector: This is used to represent changes in the state of local nodes. Further, a perturbation propagation matrix is ​​constructed: Used to describe the diffusion relationship of disturbances in a system, its propagation process is represented as: in, Represents the overall disturbance state of the system, matrix elements Indicates the disturbance originating from the node To the node The degree of influence. In a preferred embodiment, the propagation matrix can be further expressed as: in, Indicates the topological distance or association strength between nodes. This represents the coupling coefficient between nodes.

[0049] This model can quantify the impact of local disturbances on the overall system operation, providing a basis for the dynamic adjustment of scheduling strategies.

[0050] Furthermore, the process of achieving closed-loop control includes: Control commands are generated based on the scheduling strategy and then sent to the actual equipment for execution through the park's energy management system. During control execution, the system state variables are updated based on the real-time acquired system operating status, forming a closed loop of state prediction and control execution.

[0051] Furthermore, to achieve global optimization and real-time dynamic adjustment of system operation, this embodiment constructs a hierarchical optimization scheduling mechanism.

[0052] Upper-level optimization aims at improving the overall system performance, and its objective function is expressed as: in, Indicates system operating costs. Indicates carbon emissions. Indicators representing system operational risk. , , These are the weighting coefficients.

[0053] The upper-level optimization generates a global scheduling strategy and energy allocation scheme based on the future state output by the prediction model.

[0054] The lower-level optimization employs a rolling optimization strategy, utilizing a world model to prospectively evaluate candidate control strategies. This involves inputting different control variable sequences into the world model to obtain the corresponding future state sequences and performance indicators, and then selecting the optimal control strategy accordingly. The decision-making process can be represented as follows: in, Indicates the forecast time window, This represents the optimal control strategy. This is a comprehensive evaluation function based on system state.

[0055] The optimization process is carried out in a virtual simulation environment built on a world model. By performing multi-step simulations of different control strategies, a forward-looking assessment of the future operating state is achieved, thereby avoiding trial and error directly in the actual system.

[0056] Through the above two-layer optimization structure, the transformation from "state prediction" to "policy deduction" is realized, enabling the model to conduct virtual experiments and optimizations of scheduling strategies under complex operating environments, thereby significantly improving the foresight and robustness of scheduling decisions.

[0057] Furthermore, regarding grid interaction, this embodiment dynamically adjusts the power exchange range with the external grid based on the system's operating status, and the constraint relationship is expressed as follows: The upper and lower limits are dynamically determined based on the energy storage status, load level, and electricity price information, thereby improving the flexibility of system operation.

[0058] After obtaining the optimal control strategy, corresponding control commands are generated, including energy storage charging and discharging control commands, load adjustment commands, and grid interaction power commands, which are then sent to the actual equipment for execution through the park's energy management system.

[0059] During control execution, the system operates according to the following closed-loop process: Among them, the current state The control variables were obtained through optimization. Control variables affect energy flow distribution This drives the system state to evolve to .

[0060] Through continuous status feedback and strategy updates, a closed-loop control mechanism is formed, enabling the system to achieve stable, efficient, and low-carbon operation in complex operating environments.

[0061] The load optimization scheduling method provided in this embodiment, which integrates spatiotemporal coupling modeling, data-driven prediction model, and physical constraint mechanism, can be applied to park energy management systems that include distributed energy, energy storage systems, and various types of energy loads. By constructing a unified system state expression, spatiotemporal coupling model, state transition model, world model, and multi-layer optimization scheduling mechanism, dynamic modeling and closed-loop optimization control of the park energy system are realized.

[0062] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A zero-carbon industrial park load scheduling method based on the fusion of a spatiotemporal coupling model and a world model, characterized in that, include: Based on the operational data of the park's energy system, construct system state variables and control variables; Based on the system state variables and the control variables, a spatiotemporal coupling model is established to describe the energy distribution relationship between different nodes in the system; Based on the energy distribution relationship determined by the spatiotemporal coupling model, a world model for system state prediction is constructed. Based on the prediction results output by the world model, a scheduling strategy is generated and applied to the park's energy system; The system state variables are updated based on system operation feedback to achieve closed-loop control.

2. The method according to claim 1, characterized in that, The process of constructing system state variables and control variables includes: Collect operational data of the park's energy system, and perform time-scale unification and normalization processing on the operational data; Based on the processed operational data, construct the system state vector and control vector; The system state vector includes distributed energy output, load power, energy storage state of charge, electricity price, and meteorological characteristics. The control vector includes energy storage charging and discharging power, load regulation, and grid interaction power.

3. The method according to claim 1, characterized in that, The process of establishing the spatiotemporal coupling model includes: A spatiotemporal coupling matrix is ​​constructed, which is a function of system state variables and control variables, and is used to describe the energy flow path and distribution ratio between energy nodes and load nodes.

4. The method according to claim 3, characterized in that, The process of calculating the energy flow path and allocation ratio between nodes and load nodes based on the spatiotemporal coupling matrix includes: Based on the elements of the spatiotemporal coupling matrix and the output power of the energy node, the power allocated by the energy node to each load node is calculated, and the output power of the energy node is not less than the sum of the power allocated to all load nodes.

5. The method according to claim 1, characterized in that, The process of constructing the world model includes: A state transition model is constructed to describe the dynamic evolution process of the system under the action of control variables, and the energy flow allocation result determined by the spatiotemporal coupling model is embedded into the state transition model as an intermediate physical variable to obtain the world model.

6. The method according to claim 5, characterized in that, The process of predicting the system's operating state based on the world model includes: The system state sequence for future operation is obtained by recursively predicting the system state at multiple time steps using the state transition model.

7. The method according to claim 1, characterized in that, The process of generating a scheduling policy includes: Based on the future operating state sequence of the system, different control strategies are evaluated, and the optimal control strategy that satisfies the preset objective function is selected; the objective function includes system operating cost, carbon emissions, and operating risk indicators.

8. The method according to claim 1, characterized in that, The method further includes: A load resilience model is constructed to describe the adjustable characteristics of load power under the influence of electricity price changes and dispatch control signals. The load power is represented as the sum of the baseline load, the electricity price response term, and the dispatch command response term.

9. The method according to claim 1, characterized in that, The method further includes: A disturbance propagation model is constructed to describe the influence of local state changes on the overall system state. The disturbance propagation model includes a disturbance vector and a disturbance propagation matrix. The elements of the disturbance propagation matrix are determined based on the topological distance or association strength between nodes.

10. The method according to claim 1, characterized in that, The process of achieving closed-loop control includes: Control commands are generated according to the scheduling strategy, and then sent to the actual equipment for execution through the park's energy management system. During control execution, the system state variables are updated based on the real-time acquired system operating status, forming a closed loop of state prediction and control execution.

Images