Multi-source renewable energy optimization scheduling method and related device

Abstract

The invention belongs to the technical field of intelligent power grid optimal scheduling and low-carbon operation, and discloses a multi-source renewable energy optimal scheduling method and a related device. The multi-source renewable energy source optimal scheduling method comprises the steps that on the basis of to-be-processed multi-source data at a selected moment, an optimal scheduling model is used for prediction, and an optimal scheduling strategy is obtained; wherein the optimal scheduling model adopts a physical information neural network, and a composite loss function adopted during training of the optimal scheduling model comprises a renewable energy output loss item, a power flow loss item, a voltage loss item and a carbon footprint loss item. The technical scheme disclosed by the invention specifically relates to a multi-source renewable energy optimization scheduling scheme based on a physical information neural network, which can effectively fuse multi-source data, and can effectively reduce the carbon footprint of an active power distribution network on the premise of ensuring the operation safety of a power grid.

Description

Technical Field

[0001] This invention belongs to the field of smart grid optimization scheduling and low-carbon operation technology, and specifically relates to a multi-source renewable energy optimization scheduling method and related devices. Background Technology

[0002] As the penetration rate of renewable energy in the power distribution network continues to increase, its intermittency and uncertainty bring new challenges to the operation of the power grid.

[0003] Faced with these new challenges, existing traditional optimization methods (such as linear programming and nonlinear programming) rely on accurate models and are difficult to effectively cope with real-time changing multi-source data. In addition, existing machine learning methods (such as recurrent neural networks and long short-term memory networks) can process time-series data, but lack the embedding of physical constraints, which can easily lead to infeasible solutions. Furthermore, although physical information neural networks can embed physical equations, they are mostly limited to single-source data and are difficult to handle the fusion and optimization of multi-source heterogeneous data in distribution networks, especially the optimization that comprehensively considers carbon emission flows.

[0004] In summary, given the following problems still existing in the multi-source renewable energy optimization scheduling scheme: insufficient multi-source data fusion, lack of explicit embedding of grid physical constraints, and difficulty in effectively balancing carbon footprint optimization and grid operation safety, it is urgent to explore new multi-source renewable energy optimization scheduling schemes. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-source renewable energy optimization scheduling method and related apparatus to address the technical challenges of insufficient multi-source data fusion, lack of explicit embedding of power grid physical constraints, and difficulty in effectively balancing carbon footprint optimization and power grid operation safety in existing technologies. Specifically, the technical solution disclosed in this invention is a multi-source renewable energy optimization scheduling scheme based on a Physical Information Neural Network (PINN), which can effectively fuse multi-source data and, while ensuring power grid operation safety, effectively reduce the carbon footprint of the Active Distribution Network (ADN).

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for optimal scheduling of multi-source renewable energy, comprising the following steps: Based on the target active distribution network, acquire multi-source data to be processed at selected times; Based on the multi-source data to be processed, an optimized scheduling strategy is obtained by using an optimized scheduling model for prediction. In the training and acquisition process of the optimized scheduling model, historical multi-source data is first acquired based on the target active distribution network. Then, based on the historical multi-source data, the physical information neural network is trained by minimizing the composite loss function, and the trained physical information neural network is used as the optimized scheduling model. The multi-source data to be processed and the historical multi-source data both include renewable energy power generation output, load demand, grid topology, and carbon emission factors. The composite loss function includes renewable energy output loss, power flow loss, voltage loss, and carbon footprint loss.

[0007] A further improvement to the technical solution of this invention lies in that the step of acquiring historical multi-source data based on the target active distribution network includes: Based on the target active distribution network, acquire raw multi-source data for a selected historical time period; Based on the original multi-source data, the historical multi-source data is obtained by standardizing the data and constructing a spatiotemporally aligned data input format.

[0008] A further improvement of the technical solution of the present invention is that the composite loss function is expressed as: ; In the formula, This is the total composite loss function; For carbon footprint loss items, For tidal loss items, For voltage stabilization loss items, Losses due to renewable energy output; , , In order, they are respectively , , Weighting coefficients; The renewable energy output loss term represents the mean square error between the predicted renewable energy power generation and the observed data.

[0009] A further improvement to the technical solution of this invention is that, in the composite loss function, the carbon footprint loss term is expressed as: ; In the formula, and They represent the positions respectively. ,time Active power generation of non-renewable and renewable energy sources under certain conditions; N This represents the total number of node samples.

[0010] A further improvement of the technical solution of the present invention is that, in the composite loss function, the power flow loss term is expressed as: ; In the formula, M This indicates the number of neighboring node samples of the selected node; , Indicates the location and time Active power and reactive power values ​​in actual power flow constraints; , Indicates the location and time The active power and reactive power values ​​predicted by the physical information neural network.

[0011] A further improvement of the technical solution of the present invention is that, in the composite loss function, the voltage regulation loss term is expressed as: ; In the formula, N This represents the total number of node samples. The voltage at the selected node; These are the lower and upper limits of the voltage, respectively.

[0012] A further improvement of the technical solution of the present invention is that the optimized scheduling strategy includes the power generation allocation of each energy source in the target active distribution network and the power transmission allocation scheme.

[0013] In a second aspect, the present invention provides a multi-source renewable energy optimized scheduling system, comprising: The data acquisition unit is used to acquire multi-source data to be processed at a selected time based on the target active distribution network. The strategy prediction unit is used to make predictions based on the multi-source data to be processed using an optimized scheduling model to obtain an optimized scheduling strategy. In the training and acquisition process of the optimized scheduling model, historical multi-source data is first acquired based on the target active distribution network. Then, based on the historical multi-source data, the physical information neural network is trained by minimizing the composite loss function, and the trained physical information neural network is used as the optimized scheduling model. The multi-source data to be processed and the historical multi-source data both include renewable energy power generation output, load demand, grid topology, and carbon emission factors. The composite loss function includes renewable energy output loss, power flow loss, voltage loss, and carbon footprint loss.

[0014] A further improvement to the technical solution of this invention lies in that the step of acquiring historical multi-source data based on the target active distribution network includes: Based on the target active distribution network, acquire raw multi-source data for a selected historical time period; Based on the original multi-source data, the historical multi-source data is obtained by standardizing the data and constructing a spatiotemporally aligned data input format.

[0015] A further improvement of the technical solution of the present invention is that the composite loss function is expressed as: ; In the formula, This is the total composite loss function; For carbon footprint loss items, For tidal loss items, For voltage stabilization loss items, Losses due to renewable energy output; , , In order, they are respectively , , Weighting coefficients; The renewable energy output loss term represents the mean square error between the predicted renewable energy power generation and the observed data.

[0016] A further improvement to the technical solution of this invention is that, in the composite loss function, the carbon footprint loss term is expressed as: ; In the formula, and They represent the positions respectively. ,time Active power generation of non-renewable and renewable energy sources under certain conditions; N This represents the total number of node samples.

[0017] A further improvement of the technical solution of the present invention is that, in the composite loss function, the power flow loss term is expressed as: ; In the formula, M This indicates the number of neighboring node samples of the selected node; , Indicates the location and time Active power and reactive power values ​​in actual power flow constraints; , Indicates the location and time The active power and reactive power values ​​predicted by the physical information neural network.

[0018] A further improvement of the technical solution of the present invention is that, in the composite loss function, the voltage regulation loss term is expressed as: ; In the formula, N This represents the total number of node samples. The voltage at the selected node; These are the lower and upper limits of the voltage, respectively.

[0019] A further improvement of the technical solution of the present invention is that the optimized scheduling strategy includes the power generation allocation of each energy source in the target active distribution network and the power transmission allocation scheme.

[0020] In a third aspect, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the multi-source renewable energy optimization scheduling method as described in any one of the first aspects of the present invention.

[0021] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the multi-source renewable energy optimization scheduling method as described in any one of the first aspects of the present invention.

[0022] In a fifth aspect, the present invention provides a computer program product comprising computer instructions which, when executed by a processor, implement the steps of the multi-source renewable energy optimization scheduling method as described in any one of the first aspects of the present invention.

[0023] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a multi-source renewable energy optimal scheduling scheme that fully considers and integrates multi-source data during model training and prediction, allowing the information contained in different data to complement and corroborate each other. Furthermore, by explicitly embedding grid physical constraints as loss terms into a composite loss function, the physical information neural network continuously optimizes itself during training, ensuring that the prediction results and generated optimal scheduling strategies not only meet data accuracy requirements but also strictly adhere to the physical operation laws of the grid. Moreover, since the composite loss function comprehensively considers carbon footprint loss, renewable energy output loss, power flow loss, and voltage loss, the physical information neural network comprehensively considers these factors during training to generate a comprehensively optimal scheduling strategy. This invention effectively solves the technical problems of insufficient multi-source data integration, lack of explicit embedding of grid physical constraints, and difficulty in effectively balancing carbon footprint optimization and grid operation safety in existing technologies by comprehensively acquiring and integrating multi-source data, explicitly embedding grid physical constraints, and comprehensively considering carbon footprint optimization and grid operation safety. Thus, it achieves the technical effect of effectively integrating multi-source data and effectively reducing the carbon footprint of active distribution networks while ensuring grid operation safety. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0025] Figure 1 This is a flowchart illustrating a multi-source renewable energy optimization scheduling method in an embodiment of the present invention; Figure 2 This is a schematic diagram of a multi-source renewable energy optimization scheduling system in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention; obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0027] Based on the technical solutions disclosed in the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0028] Please see Figure 1 The present invention provides a method for optimizing the scheduling of multi-source renewable energy, comprising the following steps: Step 1: Based on the target active distribution network, acquire the multi-source data to be processed at the selected time. Step 2: Based on the multi-source data to be processed obtained in Step 1, use the optimized scheduling model to make predictions and obtain the optimized scheduling strategy; In the training process of the optimized scheduling model, historical multi-source data is first obtained based on the target active distribution network; then, based on the historical multi-source data, the physical information neural network is trained by minimizing the composite loss function, and the trained physical information neural network is used as the optimized scheduling model; the multi-source data to be processed and the historical multi-source data both include: renewable energy power generation output, load demand, grid topology and carbon emission factor; the composite loss function includes four components: renewable energy output loss term, power flow loss term, voltage loss term and carbon footprint loss term.

[0029] The technical solution provided in this embodiment of the invention clarifies in step 1 that multi-source data to be processed at a selected time is obtained based on the target active distribution network, and in the process of training the optimization scheduling model in step 2, historical multi-source data is also obtained based on the target active distribution network. The multi-source data here covers renewable energy power generation output, load demand, grid topology, and carbon emission factors. By comprehensively acquiring multi-source data in this way, the foundation for multi-source data fusion is laid from the data collection level. When training the optimization scheduling model, these rich multi-source data are used as input and processed using a physical information neural network, so that different types of data can be correlated and work together during the model training process, thereby achieving deep fusion of multi-source data. For example, combining renewable energy generation output data with load demand data can more accurately grasp the dynamic changes in energy supply and demand; grid topology data provides a physical framework for energy transmission and distribution, and when integrated with generation output and load demand data, it can provide a more comprehensive understanding of the grid's operating status. This comprehensive data fusion enables optimized scheduling strategies to be formulated based on more complete and accurate information, thereby effectively solving the problem of insufficient multi-source data fusion in existing technologies, providing reliable data support for subsequent effective optimized scheduling, and thus helping to reduce the carbon footprint while ensuring the safe operation of the grid.

[0030] Furthermore, during the training of the optimized scheduling model, a method based on historical multi-source data is adopted to train the physical information neural network by minimizing a composite loss function. This composite loss function comprises four components: renewable energy output loss, power flow loss, voltage loss, and carbon footprint loss. Among these, the power flow loss and voltage loss are directly related to the physical characteristics of the power grid. Power flow reflects the flow of electricity within the grid, while voltage is a crucial parameter for grid operation. The setting of these two loss terms ensures that the model adheres to the physical laws of the power grid during training, explicitly embedding the grid's physical constraints. During training, the physical information neural network continuously adjusts its parameters based on these loss terms to ensure that the generated optimized scheduling strategy conforms to the physical constraints of the power grid. For example, when considering power flow loss, the model ensures that the distribution and flow of energy in the power grid conforms to the actual physical conditions, avoiding unreasonable power flow distribution; the voltage loss term ensures that the voltage of each node in the power grid is within a reasonable range, thereby ensuring the safe operation of the power grid; this explicit embedding of power grid physical constraints allows the optimization scheduling scheme to fully consider the physical characteristics of the power grid when it is formulated, effectively solving the problem of the lack of explicit embedding of power grid physical constraints in existing technologies, and thus providing a safe and reliable operating environment for achieving the goal of reducing carbon footprint.

[0031] Furthermore, the composite loss function of the technical solution in this embodiment of the invention also includes a carbon footprint loss term, considering carbon footprint optimization as an important objective during model training. Simultaneously, as mentioned earlier, power flow loss and voltage loss terms ensure the safe operation of the power grid. When training the physical information neural network, the model needs to minimize these four loss terms simultaneously. In finding the optimal scheduling strategy, a balance must be struck between reducing the carbon footprint and ensuring the safe operation of the power grid. The physical information neural network continuously adjusts its parameters to try to find an optimal scheduling scheme that can effectively reduce the carbon footprint while ensuring reasonable power flow and stable voltage. For example, in reducing the carbon footprint, the model may prioritize the use of renewable energy, but at the same time, it will ensure through power flow loss and voltage loss terms that this energy allocation method will not threaten the safe operation of the power grid. This training method, which comprehensively considers multiple objectives, enables the generated optimal scheduling scheme to effectively reduce the carbon footprint of the active distribution network while ensuring the safe operation of the power grid, successfully solving the technical problem in the prior art where carbon footprint optimization and power grid operation safety are difficult to effectively balance.

[0032] In a specific exemplary technical solution of this invention, a method for optimizing carbon emissions from multiple sources of renewable energy based on a physical information neural network is provided, comprising the following steps: Data construction: Collect multi-source data from the distribution network, including renewable energy photovoltaic and wind power output, load demand, grid topology, carbon emission factors, etc.; in the preferred technical solution, the data can be standardized and a spatiotemporally aligned data input format can be constructed.

[0033] PINN modeling: A composite loss function is used to guide neural network learning; the composite loss function includes renewable energy output loss, power flow loss, voltage loss and carbon footprint loss.

[0034] PINN Training and Optimization: PINN is trained using the backpropagation algorithm and the Adam optimizer. The neural network is trained by minimizing the composite loss function to output the optimal scheduling policy. Real-time scheduling and control: Deploy the trained PINN model into the actual system to ultimately achieve low-carbon, safe, and economical real-time scheduling.

[0035] In the technical solution of this invention, a multi-source data fusion mechanism is adopted to input multi-source data such as power generation, load, topology, and carbon emissions into the neural network model in a unified manner; a PINN network structure is adopted to embed the loss function design with physical constraints; in real-time scheduling, through data acquisition, PINN inference, control command generation and execution, the carbon footprint of the active distribution network can be effectively reduced while ensuring the safe operation of the power grid.

[0036] In a specific exemplary technical solution of this invention, a method for optimal scheduling of multi-source renewable energy based on a physical information neural network is provided, comprising the following steps: Step 1: Multi-source data fusion and preprocessing.

[0037] In this step, multi-source data such as renewable energy output, load demand, topology, and carbon emission factors in the distribution network are collected and processed for subsequent neural network input. In a specific exemplary technical solution, the following process can be adopted: collect multi-source data in the distribution network, including renewable energy power generation output (such as photovoltaic and wind power), load demand, grid topology, carbon emission factors, etc.; standardize the data and construct a spatiotemporally aligned data input format.

[0038] Step 2: Use a composite loss function to guide neural network learning.

[0039] In this step, the composite loss function comprises four components: renewable energy output loss, power flow loss, voltage loss, and carbon footprint loss. The formula is as follows: ; In the formula, This is the total composite loss function; For carbon footprint loss items; For tidal loss; This is the voltage stabilization loss item; Losses due to renewable energy output; , , In order, they are respectively , , The weighting coefficients.

[0040] In this embodiment of the invention, the explanation of the carbon footprint loss term is as follows: A carbon footprint is a measure of the total amount of carbon dioxide and other greenhouse gases emitted into the atmosphere by human activities (interpretably, especially energy production, transportation, and industrial processes). For the power system, the carbon footprint primarily stems from the combustion of fossil fuels for power generation. Furthermore, the carbon footprint of a power system can be quantified by calculating the amount of carbon dioxide emitted per unit of electricity consumed or produced, typically expressed as grams of carbon dioxide per kilowatt-hour (gCO2 / kWh). To calculate the carbon footprint of a power system, carbon intensity (CI) is commonly used; carbon intensity refers to the amount of carbon dioxide emitted per unit of electricity consumed, calculated using the following formula: ; In the formula, Carbon strength; It comes from energy sourcesi Electricity generated (e.g., by solar, wind, coal, etc.); Indicates energy source i The carbon dioxide emission factor (in gCO2 / kWh) is the total energy demand or total electricity consumption of the power grid. n The total number of energy sources; by reducing carbon intensity and increasing the proportion of renewable energy in the energy mix, the overall carbon footprint can be reduced.

[0041] As shown above, the carbon footprint loss item encourages the use of renewable energy while minimizing the use of non-renewable resources. Therefore, the carbon footprint of ADN can be calculated as follows: ; In the formula, , These are the carbon footprint factors for non-renewable energy and renewable energy, respectively. These factors should be updated in a timely manner with reference to the calculation results and methods published by relevant national departments, and will not be elaborated here. and These represent the active power generation of non-renewable energy and renewable energy sources, respectively.

[0042] In summary, the carbon footprint loss term can be expressed as: ; In the formula, and They represent the positions respectively ,time Active power generation of non-renewable and renewable energy sources under certain conditions; Interpretationally, this incentive model prioritizes renewable energy generation to reduce reliance on non-renewable energy sources.

[0043] In this embodiment of the invention, the explanation of the power flow loss term is as follows: The power flow equations in ADN can be described using AC power flow equations derived from Kirchhoff's Current Law (KCL) and Voltage Law (KVL). The aim is to optimize renewable energy generation while ensuring that the power flow constraints of the grid are met. For nodes in the network... i The AC power flow equations for active and reactive power are as follows: ; ; in, For nodes i , j Voltage amplitude at the location; For nodes i , j The electrical conductance and susceptance between them; For nodes i , j The phase angle difference between them; For nodes i The set of adjacent nodes; , For nodes i Active power and reactive power.

[0044] As shown above, the tidal current loss term can be expressed as: ; In the formula, M Indicates the number of nodes; , Represents the power flow constraints of a node. and This is the predicted value for PINN.

[0045] In this embodiment of the invention, the explanation of the voltage regulation loss term is as follows: Voltage regulation ensures the voltage of each node in the network. Staying within acceptable limits is expressed as: ; in, These are the lower and upper limits of the voltage, respectively.

[0046] The voltage regulation loss term ensures that the predicted voltage is within these limits, and it can be expressed as: ; In the formula, N The number of voltage measurement points; interpretably, this term penalizes neural networks for violating voltage constraints.

[0047] In this embodiment of the invention, the explanation of the renewable energy output loss item is as follows: Ensure that the neural network's predictions align with the renewable power generation data observed by the ADN. Consistent with the measured system states (such as voltage and current), this term is typically expressed as the mean square error (MSE) between the predicted renewable energy generation and the observed data: ; in, For the first i Predicted renewable energy generation at various spatiotemporal points; It is the observed renewable energy power generation.

[0048] The model training and optimization process in this embodiment of the invention is represented as follows: ; Where W represents the weights of the neural network, and W* represents the optimal set of weights that minimizes the loss function.

[0049] Step 3, PINN Training and Optimization: PINN is trained using the backpropagation algorithm and the Adam optimizer. The neural network is trained by minimizing the composite loss function to output the optimal scheduling strategy. In an exemplary optional technical solution, the carbon emission factor may include carbon emission factor data corresponding to renewable energy power generation and traditional energy power generation, respectively. Step 4: Real-time Scheduling and Control: Deploy the trained model into the actual system to achieve low-carbon, safe, and economical real-time scheduling. Specifically, the model's output is the optimal scheduling strategy, which guides the power generation plans and power allocation of various energy sources in the power system to achieve the goals of low-carbon, safe, and economical operation. In the example technical solution, the optimal scheduling strategy may include the following: Power generation allocation of various energy sources: Determine the power generation capacity of renewable energy (photovoltaic, wind power) and traditional energy at different times to fully utilize renewable energy and reduce dependence on traditional high-carbon energy; Power transmission and distribution scheme: Based on the grid topology and load demand, rationally arrange the power transmission path and distribution ratio to ensure that power can be stably and efficiently delivered to each load node.

[0050] In principle, the Physical Information Neural Network (PINN) is a type of machine learning model that directly incorporates physical laws into the network's training process. PINNs are specifically designed to solve complex problems governed by partial differential equations (PDEs) or ordinary differential equations (ODEs). By embedding the governing equations into the network architecture, PINNs can learn solutions that not only conform to existing data but also comply with fundamental physical laws. The standard form of PINN involves a neural network that approximates the solution to a system described by differential equations. Assume the system is governed by PDEs or ODEs: ; In the formula, It is a differential operator representing the laws of physics. It is the unknown function that the network is trying to approximate. Represents the spatial domain, Indicates time.

[0051] Composite loss function in PINN It is usually defined as: ; In the formula, It is the renewable energy output loss term, which measures the error between the predicted solution and the observed data; It is a physics-based loss term that ensures the predicted solution conforms to the physical laws governing the system. It is a regularization parameter used to balance the relative importance of data accuracy and adherence to physical constraints.

[0052] Further interpreting, the renewable energy output loss term quantifies the difference between neural network predictions and available observation data; it is typically represented as the network prediction. With corresponding real data The mean square error (MSE) between them is expressed as: ; In the formula, N This represents the total number of data points; , These represent the predicted and observed values ​​of the unknown function, respectively.

[0053] Further explanation is needed, as a physics-based loss term ensures that the neural network's predictions conform to the physical laws governing the system. This is typically achieved by incorporating the governing differential equations into the loss function. The physics-based loss term can be expressed as: ; In the formula, Describing differential operators The residuals applied to the network prediction ensure the predicted solution. Satisfies the governing differential equation; M This indicates the number of configuration points used to assess physical losses.

[0054] In the specific exemplary power distribution system test case of this invention, the IEEE-33 node and IEEE-123 node test systems from MATPOWER are used; the hardware environment includes an Intel(R) Core(TM) i7-6700HQ processor, an NVIDIA GeForce GTX 1660 graphics processor, a system equipped with 8GB of memory, a computer running the Windows 11 operating system, and the Python 3.8 programming language.

[0055] In this case study, the PINN-based optimization method proposed in this invention is compared with two existing benchmark methods. The existing RNN methods employ a model-based approach to reduce the carbon footprint of the power grid (for explanation, specific technical solutions can be found in H. Zhang, H. Lei, H. Chi, C. Zou, T. Liu, S. Li, Z. Zhou, Y. Zheng, and M. Dong, “Predicting and applying the electricity-related carbonemission coefficient based on recurrent neural network: A decision-making reference for carbon emission policies,” in Proc. 2024 Int. Conf. PowerElectron. Artif. Intell. (PEAI), ser. PEAI '24, Jul. 2024, pp. 620–623). Additionally, the M-PINN method applies PINN to optimize the CO2 methanation process in a catalytic reactor (for explanation, specific technical solutions can be found in SI Ngo and Y.-I. Lim, “Forward physics-informed neural networks for catalytic CO2methanation via…”). “Isothermal fixed-bed reactor,” in Proc. 14thInt. Symp. Process Syst. Eng., ser. Comput. Aided Chem. Eng., vol. 49, pp.1675–1680, Jul. 2022. Explainedly, these benchmark methods were chosen because they all focus on carbon reduction in energy systems, providing a solid foundation for assessing the effectiveness of PINN in optimizing renewable energy output and reducing the carbon footprint of ADN.

[0056] To comprehensively evaluate the performance of the multi-source PINN-based optimization framework proposed in this invention, four key metrics were employed. These metrics reflect both environmental impact and system operational quality. The aim of these metrics is to assess the effectiveness of each method in minimizing carbon emissions, maximizing renewable energy utilization, and maintaining power system stability.

[0057] Specifically, this includes: Carbon footprint (kilograms of CO2): This metric measures the total carbon emissions generated by the entire power distribution network during energy production. The lower the value, the more environmentally friendly and sustainable the solution.

[0058] Renewable Energy Output (MWh): This metric measures the total energy provided by renewable energy sources such as solar and wind power. A higher output value indicates better integration and utilization of clean energy resources.

[0059] Power flow stability (%): Defined as the proportion of time over which the system can maintain a feasible and balanced power flow solution. This reflects the method's ability to maintain the integrity of system operation under dynamic conditions.

[0060] Voltage Deviation (%): This indicator reflects the average deviation of the node voltage amplitude from the nominal value. The lower the deviation, the better the voltage regulation effect and the higher the power quality.

[0061] The carbon footprint reduction analysis of the IEEE-33 node system is as follows: The purpose of this evaluation is to benchmark the model's ability to minimize carbon emissions, maximize renewable energy utilization, and maintain system-level feasibility (particularly in terms of power flow stability and voltage regulation). From a computational perspective, these standards reflect a general interest in constrained deep learning, multi-source representation learning, and generalization under physical constraints. Experimental results are shown in Table 1.

[0062] Table 1. Experimental results of the IEEE-33 node system

[0063] Based on the experimental results in Table 1 above, it can be seen that the method proposed in this invention outperforms the two existing benchmark methods in all evaluation metrics. This invention achieves the lowest carbon footprint (139.33 kg CO2) and the highest renewable energy output (33.85 MWh), demonstrating excellent control strategy learning capabilities in clean energy integration. Furthermore, the model of this invention exhibits strong generalization ability under system constraints, reflected in high power flow stability (98.55%) and low voltage deviation (3.01%). In summary, the experimental results highlight the ability of the model proposed in this invention to ensure feasibility through physics-based loss design, while effectively utilizing multi-source input data. In contrast, while existing RNN methods have predictive capabilities, they cannot comply with system constraints under variability, resulting in the highest emissions and the largest voltage deviation. The M-PINN model benefits from physical embedding but lacks the multi-source structure of the method proposed in this invention, and its performance is not as stable as the model proposed in this invention, especially in maximizing renewable energy.

[0064] The carbon footprint reduction analysis of the IEEE-123 node system is as follows: To further evaluate the scalability and robustness of the proposed multi-source PINN framework, additional experiments were conducted on the IEEE-123 node distribution network. Compared to the IEEE-33 system, the IEEE-123 network introduces significantly higher complexity in terms of system topology and operational variability, thus becoming a more challenging benchmark for evaluating the performance of learning-based optimization under structural constraints. The experimental design is the same as that of IEEE-33, focusing on three main objectives: maximizing renewable energy output, minimizing carbon emissions, and maintaining physical feasibility in terms of power flow solvability and voltage regulation. This setup can test the model's ability to generalize to more complex system states while maintaining optimization quality and constraint satisfaction. The experimental results are shown in Table 2.

[0065] Table 2. Experimental results of the IEEE-123 node system

[0066] The experimental results in Table 2 demonstrate that the proposed method outperforms both RNN and M-PINN methods in terms of carbon footprint and renewable energy output. It achieves the lowest carbon footprint (454.53 kg CO2) and the highest renewable energy output (93.14 MWh), showcasing its effectiveness in reducing carbon footprint and optimizing renewable energy integration. Furthermore, the proposed method ensures stable power flow (98.06%) and voltage regulation (2.79%), highlighting its equivalent performance in maintaining grid stability. These experimental results underscore the potential of PINN in improving sustainable energy management within an ADN framework. While the M-PINN method also incorporates physical constraints, its application scope is narrower, and it lacks the comprehensive multi-source fusion employed in the model of this invention. The RNN baseline exhibits significant performance degradation in environmental and operational metrics, highlighting its limitations in generalizing to large-scale systems without a clear structure-aware learning mechanism.

[0067] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not disclosed in the apparatus embodiments, please refer to the embodiments of the method of the present invention.

[0068] Please see Figure 2 In this embodiment of the invention, a multi-source renewable energy optimized scheduling system is provided, comprising: The data acquisition unit is used to acquire multi-source data to be processed at a selected time based on the target active distribution network. The strategy prediction unit is used to make predictions based on the multi-source data to be processed using an optimized scheduling model to obtain an optimized scheduling strategy. In the training and acquisition process of the optimized scheduling model, historical multi-source data is first acquired based on the target active distribution network. Then, based on the historical multi-source data, the physical information neural network is trained by minimizing the composite loss function, and the trained physical information neural network is used as the optimized scheduling model. The multi-source data to be processed and the historical multi-source data both include renewable energy power generation output, load demand, grid topology, and carbon emission factors. The composite loss function includes renewable energy output loss, power flow loss, voltage loss, and carbon footprint loss.

[0069] In one embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used to execute the operation of a multi-source renewable energy optimization scheduling method.

[0070] In one embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the terminal. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM (Random Access Memory) or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the multi-source renewable energy optimal scheduling method in the above embodiments.

[0071] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0072] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0073] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0074] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for optimal scheduling of multi-source renewable energy, characterized in that, Includes the following steps: Based on the target active distribution network, acquire multi-source data to be processed at selected times; Based on the multi-source data to be processed, an optimized scheduling strategy is obtained by using an optimized scheduling model for prediction. In the training and acquisition process of the optimized scheduling model, historical multi-source data is first acquired based on the target active distribution network. Then, based on the historical multi-source data, the physical information neural network is trained by minimizing the composite loss function, and the trained physical information neural network is used as the optimized scheduling model. The multi-source data to be processed and the historical multi-source data both include renewable energy power generation output, load demand, grid topology, and carbon emission factors. The composite loss function includes renewable energy output loss, power flow loss, voltage loss, and carbon footprint loss.

2. The multi-source renewable energy optimal scheduling method according to claim 1, characterized in that, The steps for obtaining historical multi-source data based on a target active distribution network include: Based on the target active distribution network, acquire raw multi-source data for a selected historical time period; Based on the original multi-source data, the historical multi-source data is obtained by standardizing the data and constructing a spatiotemporally aligned data input format.

3. The multi-source renewable energy optimal scheduling method according to claim 1, characterized in that, The composite loss function is expressed as follows: ； In the formula, This is the total composite loss function; For carbon footprint loss items, For tidal loss items, For voltage stabilization loss items, Losses due to renewable energy output; , , In order, they are respectively , , Weighting coefficients; The renewable energy output loss term represents the mean square error between the predicted renewable energy power generation and the observed data.

4. The multi-source renewable energy optimal scheduling method according to claim 3, characterized in that, In the composite loss function, the carbon footprint loss term is expressed as: ； In the formula, and They represent the positions respectively. ,time Active power generation of non-renewable and renewable energy sources under certain conditions; N This represents the total number of node samples.

5. The multi-source renewable energy optimal scheduling method according to claim 3, characterized in that, In the composite loss function, the power flow loss term is expressed as: ； In the formula, M This indicates the number of neighboring node samples of the selected node; , Indicates the location and time Active power and reactive power values ​​in actual power flow constraints; , Indicates the location and time The active power and reactive power values ​​predicted by the physical information neural network.

6. The multi-source renewable energy optimal scheduling method according to claim 3, characterized in that, In the composite loss function, the voltage regulation loss term is expressed as: ； In the formula, N This represents the total number of node samples. The voltage at the selected node; These are the lower and upper limits of the voltage, respectively.

7. The multi-source renewable energy optimal scheduling method according to claim 1, characterized in that, The optimized scheduling strategy includes the power generation allocation of each energy source in the target active distribution network and the power transmission allocation scheme.

8. A multi-source renewable energy optimized scheduling system, characterized in that, include: The data acquisition unit is used to acquire multi-source data to be processed at a selected time based on the target active distribution network. The strategy prediction unit is used to make predictions based on the multi-source data to be processed using an optimized scheduling model to obtain an optimized scheduling strategy. In the training and acquisition process of the optimized scheduling model, historical multi-source data is first acquired based on the target active distribution network. Then, based on the historical multi-source data, the physical information neural network is trained by minimizing the composite loss function, and the trained physical information neural network is used as the optimized scheduling model. The multi-source data to be processed and the historical multi-source data both include renewable energy power generation output, load demand, grid topology, and carbon emission factors. The composite loss function includes renewable energy output loss, power flow loss, voltage loss, and carbon footprint loss.

9. A multi-source renewable energy optimized scheduling system according to claim 8, characterized in that, The steps for obtaining historical multi-source data based on a target active distribution network include: Based on the target active distribution network, acquire raw multi-source data for a selected historical time period; Based on the original multi-source data, the historical multi-source data is obtained by standardizing the data and constructing a spatiotemporally aligned data input format.

10. A multi-source renewable energy optimized scheduling system according to claim 8, characterized in that, The composite loss function is expressed as follows: ； In the formula, This is the total composite loss function; For carbon footprint loss items, For tidal loss items, For voltage stabilization loss items, Losses due to renewable energy output; , , In order, they are respectively , , Weighting coefficients; The renewable energy output loss term represents the mean square error between the predicted renewable energy power generation and the observed data.

11. A multi-source renewable energy optimized scheduling system according to claim 10, characterized in that, In the composite loss function, the carbon footprint loss term is expressed as: ； In the formula, and They represent the positions respectively. ,time Active power generation of non-renewable and renewable energy sources under certain conditions; N This represents the total number of node samples.

12. A multi-source renewable energy optimized scheduling system according to claim 10, characterized in that, In the composite loss function, the power flow loss term is expressed as: ； In the formula, M This indicates the number of neighboring node samples of the selected node; , Indicates the location and time Active power and reactive power values ​​in actual power flow constraints; , Indicates the location and time The active power and reactive power values ​​predicted by the physical information neural network.

13. A multi-source renewable energy optimized scheduling system according to claim 10, characterized in that, In the composite loss function, the voltage regulation loss term is expressed as: ； In the formula, N This represents the total number of node samples. The voltage at the selected node; These are the lower and upper limits of the voltage, respectively.

14. A multi-source renewable energy optimized scheduling system according to claim 8, characterized in that, The optimized scheduling strategy includes the power generation allocation of each energy source in the target active distribution network and the power transmission allocation scheme.

15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the multi-source renewable energy optimization scheduling method as described in any one of claims 1 to 7.

16. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-source renewable energy optimization scheduling method as described in any one of claims 1 to 7.

17. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the multi-source renewable energy optimization scheduling method as described in any one of claims 1 to 7.

Images