Urban rail energy management and control system and method

Abstract

The present application belongs to the technical field of urban rail transit power supply system, and particularly relates to a kind of urban rail energy management and control system and method. Including system sensing layer, platform control layer and edge control layer. System sensing layer collects source-net-load-storage each node operation data;Platform control layer includes data acquisition unit, prediction unit, digital twin unit and optimization scheduling unit, data acquisition unit outputs full system data flow under unified time section after time synchronization and alignment of operation data, prediction unit generates prediction value according to full system data flow, digital twin unit constructs twin data model according to full system data flow and prediction value, optimization scheduling unit generates optimization control instruction according to full system data flow, prediction value and twin data model;Edge control layer receives optimization control instruction and adjusts controllable power electronic device power according to real-time state of substation. The present application breaks through source-net-load-storage data link, and realizes multi-energy collaborative optimization scheduling.

Description

Technical Field

[0001] This invention relates to the field of urban rail transit power supply system technology, specifically to an urban rail energy management system and method. Background Technology

[0002] Traction energy consumption in urban rail transit accounts for more than 50% of the total system energy consumption, and power supply costs constitute a significant proportion of total operating costs. Improving the energy utilization efficiency of urban rail power supply systems has become an important issue that urgently needs to be addressed in this field.

[0003] Traditional urban rail traction power supply systems mostly use diode uncontrolled rectifier units, where energy flows unidirectionally between the AC grid and the DC traction grid. This has the following technical defects: the energy generated by the train's regenerative braking cannot be fed back to the grid, and the surplus energy that is not absorbed by neighboring trains can only be dissipated as heat through the braking resistor; the DC traction grid voltage fluctuates greatly due to changes in train operating conditions; and the centralized power supply architecture relies on the main substation, which can easily lead to large-scale power outages if a fault occurs.

[0004] To address these issues, several improved technologies have emerged in this field in recent years. On the equipment side, bidirectional converters are gradually replacing uncontrolled rectifier units, enabling bidirectional energy flow. On the source side, distributed energy sources such as photovoltaics are beginning to be integrated into urban rail power supply systems. Some solutions also introduce energy storage devices to mitigate power fluctuations. However, these improvements still have significant shortcomings: the photovoltaic system, traction power supply system, power and lighting system, and energy storage system operate independently, with data scattered across different monitoring platforms, lacking a unified data acquisition and time synchronization mechanism, thus hindering global collaborative optimization. Urban rail traction loads are characterized by strong impact and high volatility, making it difficult for conventional scheduling methods to simultaneously consider daily economic planning and second-level dynamic response. When photovoltaic output is excessive within a power supply zone, existing technologies struggle to flexibly transfer surplus power to other power supply zones, limiting the absorption rate of distributed photovoltaic energy.

[0005] Therefore, there is an urgent need for a systematic urban rail energy management and control solution that can connect the data links of source, grid, load and storage and achieve multi-timescale collaborative optimization. Summary of the Invention

[0006] This invention addresses at least some of the aforementioned technical problems, providing an urban rail transit energy management system and method. Through a three-layer distributed architecture—sensing layer, platform management layer, and edge control layer—it achieves a collaborative optimization of all elements—source, grid, load, and storage—in its energy management system and method.

[0007] Embodiments of this disclosure provide an urban rail transit energy management system, comprising: The system perception layer is communicatively connected to the source-side equipment, grid-side equipment, load-side equipment, and storage-side equipment in the urban rail power supply system. It is configured to collect and output the operating data of each node in the source-grid-load-storage system, and the operating data includes timestamps. The platform control layer includes: The data acquisition unit, connected to the output of the system perception layer, is configured to receive the operation data of each node of the power supply system (source-grid-load-storage), synchronize and align the operation data under a unified time reference according to the timestamp, and output the whole system data stream under a unified time section. The prediction unit, connected to the output of the data acquisition unit, is configured to receive the entire system data stream and generate a prediction value based on the entire system data stream through a prediction model. A digital twin unit, connected to the output of the data acquisition unit and the output of the prediction unit, is configured to receive the full system data stream and the predicted value, and to construct and output a twin data model based on the full system data stream and the predicted value; The optimization scheduling unit, connected to the output of the data acquisition unit, the output of the prediction unit, and the output of the digital twin unit, is configured to generate optimization control commands based on the whole system data flow, the predicted values, and the twin data model. The edge control layer, deployed in the traction substation of the urban rail power supply system, includes a bidirectional converter controller, an energy routing device controller, a flexible interconnection device controller, an SVG controller, and an energy storage device controller. Each controller is connected to a corresponding controllable power electronic device. The edge control layer is connected to the output of the optimization scheduling unit and is configured to receive the optimization control command and adjust the power of each controllable power electronic device according to the optimization control command and the real-time status of the substation.

[0008] The technical solution provided in this application brings at least the following benefits: it realizes the interconnection of data of all elements of source-network-load-storage through a three-layer distributed architecture, provides a data foundation for upper-level optimization control, and realizes closed-loop collaborative control of perception-decision-execution.

[0009] In other embodiments of this application, the edge control layer further includes a local sensing interface, which is connected to sensing devices within the traction substation and configured to switch to an independent operation mode when the platform control layer fails. In the independent operation mode, the edge control layer obtains the real-time status of the traction substation through the local sensing interface and performs normal energy distribution within the coverage area of ​​the traction substation based on locally preset control logic.

[0010] The technical solution provided in this application brings at least the following benefits: the edge control layer has a local perception interface, which can independently collect local data and operate autonomously when the platform management layer fails, ensuring uninterrupted power supply in extreme scenarios.

[0011] In other embodiments of this application, the source-side equipment includes a main substation and a photovoltaic power generation device; the grid-side equipment includes a bidirectional converter, an energy routing device, a flexible interconnection device, and an SVG; the load-side equipment includes a train and a power and lighting distribution system; and the storage-side equipment includes an energy storage device. The operating data on the source side includes: main substation voltage and power, photovoltaic output voltage, photovoltaic output power, photovoltaic capacity, irradiance, and temperature; The grid-side operating data includes: DC-side voltage, DC-side power, AC-side voltage, AC-side current, operating status and remaining power capacity of each grid-side device; The operational data on the load side includes: train timetable, train position, train speed, train power, train voltage, and power and lighting system power. The operating data on the energy storage side includes: energy storage device voltage, energy storage device current, and energy storage device SOC.

[0012] The technical solution provided in this application brings at least the following benefits: it clarifies the data collection content of various equipment in the source-grid-load-storage system, providing complete and accurate input parameters for prediction models and digital twin models.

[0013] In other embodiments of this application, the prediction unit includes a photovoltaic power output prediction subunit, a traction load prediction subunit, a power and lighting load prediction subunit, and a rolling correction subunit; The photovoltaic power output prediction subunit is configured to construct a physical model based on solar elevation angle, photovoltaic module parameters and irradiance data to perform basic power output estimation, and to perform time series correction on the residual sequence output by the physical model and multi-source meteorological time series data through the Transformer model to output the photovoltaic power output prediction value. The traction load prediction subunit is configured to construct a physical traction calculation model based on train timetable, headway, passenger capacity, line gradient, station spacing and vehicle traction characteristics to estimate the basic load, and to perform time-series correction on the residual sequence of the physical traction calculation model through an LSTM model to output the traction load prediction value. The power lighting load prediction subunit is configured to use passenger flow, indoor and outdoor temperature, equipment operation plan, holiday information and energy-saving control strategy as input features, and perform power lighting load prediction through Transformer multi-feature fusion time series model.

[0014] The technical solution of this application adopts a hybrid architecture of "physical model + AI time series correction" to predict photovoltaic power output and multiple types of loads respectively, taking into account both physical interpretability and nonlinear capture capability, thereby improving prediction accuracy.

[0015] In other embodiments of this application, the prediction unit further includes a rolling correction subunit, which is connected to the photovoltaic power output prediction subunit, the traction load prediction subunit, and the power and lighting load prediction subunit, and is configured to collect actual power output data and actual load data, and correct the model parameters of each prediction subunit based on the actual power output data and actual load data.

[0016] The technical solution of this application uses a rolling correction mechanism to enable the prediction model to continuously track equipment aging and environmental changes, maintaining the prediction accuracy throughout the entire life cycle.

[0017] In other embodiments of this application, the digital twin unit includes a mechanism model subunit, a data-driven correction subunit, and a closed-loop feedback subunit. The mechanism model subunit is connected to the output end of the data acquisition unit and is configured to receive the data stream of the entire system. Based on the operating status of the train traction power supply system, the characteristics of photovoltaic distributed energy generation, the charging and discharging characteristics of the energy storage device and the train operating status, it establishes the system power flow equation and outputs the system power flow reference value. The data-driven correction subunit is connected to the output of the mechanism model subunit and the output of the data acquisition unit. It is configured to receive the system power flow reference value and the whole system data flow, and generate the parameter correction amount of the mechanism model subunit through a machine learning model based on the residual between the whole system data flow and the system power flow reference value. The input terminal of the closed-loop feedback subunit is connected to the output terminal of the data-driven correction subunit and is configured to receive the parameter correction amount; the first feedback output terminal of the closed-loop feedback subunit is connected to the input terminal of the mechanism model subunit and is configured to feed back the parameter correction amount to the mechanism model subunit to correct the parameters of the mechanism model; the second feedback output terminal of the closed-loop feedback subunit is connected to the input terminal of the data-driven correction subunit and is configured to feed back the residual between the system power flow reference value output by the mechanism model subunit based on the corrected parameters and the actual system data to the data-driven correction subunit.

[0018] The technical solution of this application constructs a self-calibrating digital twin model through "mechanism model + data-driven model + closed-loop feedback" to achieve high-precision real-time mapping of the physical system.

[0019] In other embodiments of this application, the optimized scheduling unit includes a daily-level optimization subunit, a scheduling coordination subunit, and a real-time optimization subunit; The first input terminal of the daily optimization subunit is connected to the output terminal of the photovoltaic power output prediction subunit and the output terminal of the traction load prediction subunit, and is configured to receive the photovoltaic power output prediction value and the traction load prediction value; the second input terminal of the daily optimization subunit is connected to the output terminal of the digital twin unit, and is configured to receive the twin model data. The daily optimization subunit is configured to generate a daily scheduling plan based on the photovoltaic output forecast, the traction load forecast, and the twin model data. The input end of the scheduling coordination subunit is connected to the output end of the daily optimization subunit, and is configured to receive the daily scheduling plan and convert the daily scheduling plan into boundary constraints for real-time optimization. The input end of the real-time optimization subunit is connected to the output end of the scheduling and coordination subunit and the output end of the data acquisition unit. It is configured to receive the boundary constraints and the whole system data stream, iteratively calculate the network space matrix expression of the traction network according to the boundary constraints and the whole system data stream, optimize through intelligent algorithms under the objective function constraints, generate optimization control instructions for each controllable device and send them to the edge control layer.

[0020] The technical solution proposed in this application achieves optimal overall energy efficiency across all scenarios by combining "daily economic planning + real-time power allocation" with dual-layer optimization, taking into account both system economy and dynamic response capability.

[0021] In other embodiments of this application, the optimized scheduling unit further includes a supply-demand bidirectional matching subunit; The input of the supply-demand bidirectional matching subunit is connected to the output of the real-time optimization subunit, and is configured to schedule energy supply sources according to a first preset priority rule and allocate energy demand according to a second preset priority rule. The first preset priority rule includes: First priority: Train regenerative braking energy; Second priority: Distributed photovoltaic energy; Third priority: Energy storage and discharge; Fourth priority: Purchasing electricity from the power grid; The second preset priority rule includes: First priority: Train traction load; Second priority: Power lighting load; Third priority: Energy storage and charging needs; Fourth priority: Grid feedback.

[0022] The technical solution proposed in this application adopts a two-way priority matching strategy between supply and demand, giving priority to the consumption of zero-cost energy, ensuring power supply to core loads, and maximizing energy utilization efficiency.

[0023] In other embodiments of this application, the optimized scheduling unit further includes a cross-partition scheduling subunit; The first input terminal of the cross-regional scheduling subunit is connected to the output terminal of the digital twin unit and is configured to obtain the power flow distribution status of the entire system. The second input terminal is connected to the data output terminal of the platform control layer and is configured to obtain the power output of the main substation and vehicle load demand of each power supply zone. The output terminal is connected to the flexible interconnection device controller of the edge control layer and is configured to drive the energy flow of this zone to other power supply zones through the flexible interconnection device when the photovoltaic output in the power supply zone cannot be fully absorbed by this zone, based on the power flow distribution status and the power output of the main substation and vehicle load demand of each power supply zone.

[0024] The technical solution of this application realizes cross-regional energy flow through flexible interconnection devices, and transmits surplus photovoltaic power to other regions when the local area cannot absorb it, thereby improving the renewable energy absorption rate.

[0025] A second aspect of this application provides a method for urban rail transit energy management, implemented based on the urban rail transit energy management system described in the first aspect of this application. The method includes: Step S1: Collect operational data from each node of the source-grid-load-storage system, wherein the operational data includes timestamps; Step S2: Synchronize and align the running data under a unified time reference to obtain the full system data stream under a unified time profile; Step S3: Receive the entire system data stream and generate predicted values ​​based on the entire system data stream using a prediction model; Step S4: Receive the entire system data stream and the predicted value, and construct and output a twin data model based on the entire system data stream and the predicted value; Step S5: Generate optimized control instructions based on the entire system data stream, the predicted values, and the twin data model; Step S6: Receive the optimized control command, and adjust the power of each controllable power electronic device according to the optimized control command and the real-time status of the substation.

[0026] Step S7: After the real-time optimization in step S6 is executed, the actual running data is fed back to steps S3 and S4 for rolling updates of the prediction model parameters and digital twin model parameters.

[0027] Compared with the prior art, the present invention has the following beneficial effects: (1) Through the three-layer distributed architecture of perception layer, platform management and control layer and edge control layer and the data closed loop between layers, the connection of all elements of source, network, load and storage is realized, and a complete closed-loop control link of perception, cognition, decision-making, execution and feedback is constructed, which solves the problem of data dispersion and information silos in multiple subsystems and provides a data foundation for global collaborative optimization.

[0028] (2) Through the dual-layer optimization and scheduling synergy of daily economic planning and real-time power allocation, the daily level utilizes peak-valley electricity prices to reduce costs, and the real-time second-level dynamic power adjustment ensures stability, taking into account both economic efficiency and dynamic response.

[0029] (3) By matching supply and demand in two-way priority and cross-regional scheduling, energy is supplied according to the priority of regenerative braking > photovoltaic > energy storage > power grid, and surplus photovoltaic is transmitted across regions through flexible interconnection to improve the new energy consumption rate.

[0030] The above description is merely an overview of the technical solution disclosed herein. In order to better understand the technical means of this disclosure and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this disclosure more apparent and understandable, specific embodiments of this disclosure are described below. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a topology diagram of the urban rail energy management system according to an embodiment of this application; Figure 2 This is a structural diagram of the integrated urban rail transit source-grid-load-storage system according to an embodiment of this application; Figure 3 This is a flowchart of the urban rail energy management method according to an embodiment of this application. Detailed Implementation

[0033] To make the technical problems, technical solutions, and beneficial effects to be solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of this application.

[0034] The prefixes such as "first" and "second" used in this application embodiment are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this application embodiment does not constitute a limitation on the described objects. The description of the described objects is given in the claims or the context of the embodiments, and should not constitute unnecessary restrictions due to the use of such prefixes. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0035] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. In the description of the embodiments of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B; the term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone.

[0036] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0037] In this application, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0038] Existing energy management systems are mostly limited to energy consumption monitoring and data display, failing to establish a complete data link across all elements of "source-grid-load-storage". The published patent CN119852964A only achieves basic control of photovoltaic and energy storage access to the DC traction network, without considering the coordinated scheduling of source, grid, load, and storage; CN111823876A achieves integrated vehicle-to-ground energy regulation, but does not cover multi-source coordinated power supply scenarios. On the other hand, general-purpose grid distributed energy regulation systems (such as CN120638631A) are not adapted to the special voltage levels and application scenarios of urban rail DC traction networks.

[0039] In addition, the "Technical Specifications for Operation of Power Supply System of Urban Rail Transit (Trial)" issued by the Ministry of Transport in 2025 clearly requires that newly built lines have the ability to access distributed energy and the function of coordinated energy regulation, but existing technologies have not yet provided a systematic solution that can meet the above requirements.

[0040] To address the above issues, this application provides an intelligent energy management and control system for urban rail transit that can establish data links between energy sources, grids, loads, and storage, and achieve coordinated and optimized scheduling of multiple energy sources. The specific implementation methods of this application will be described in detail below with reference to the accompanying drawings.

[0041] The first aspect of this disclosure discloses an urban rail energy management system, such as... Figure 1 As shown, it includes a system perception layer, a platform management and control layer, and an edge control layer.

[0042] The system perception layer communicates with the source-side equipment, grid-side equipment, load-side equipment, and storage-side equipment in the urban rail power supply system. It is used to collect and output the operation data of each node in the source-grid-load-storage system. The operation data includes timestamps.

[0043] The platform control layer includes a data acquisition unit, a prediction unit, a digital twin unit, and an optimization scheduling unit.

[0044] The data acquisition unit connects to the output of the system's perception layer and is used to receive the operating data of each node in the power supply system, from source to grid to load to storage. Based on the timestamp, the operating data is synchronized and aligned under a unified time reference, and the entire system data stream is output under a unified time profile.

[0045] The prediction unit is connected to the output of the data acquisition unit and is used to receive the data stream of the entire system and generate predicted values ​​based on the data stream through the prediction model.

[0046] The digital twin unit connects the output of the data acquisition unit and the output of the prediction unit. It is used to receive the data stream and predicted values ​​of the entire system, and to construct and output the twin data model based on the data stream and predicted values ​​of the entire system.

[0047] The optimized scheduling unit connects to the outputs of the data acquisition unit, the prediction unit, and the digital twin unit, and is used to generate optimized control commands based on the system's data flow, predicted values, and the digital twin data model.

[0048] The edge control layer is deployed in the traction substation of the urban rail power supply system. It includes bidirectional converter controller, energy routing device controller, flexible interconnection device controller, SVG controller and energy storage device controller. Each controller is connected to the corresponding controllable power electronic device. The edge control layer is connected to the output of the optimization scheduling unit to receive optimization control commands and adjust the power of each controllable power electronic device according to the optimization control commands and the real-time status of the substation.

[0049] Specifically, the system perception layer is deployed at each substation and equipment site of the urban rail power supply system. For example... Figure 2 As shown, the system perception layer establishes communication connections with source-side equipment (main substations, photovoltaic power generation devices), grid-side equipment (bidirectional converters, energy routing devices, flexible interconnection devices, SVG), load-side equipment (trains, power and lighting distribution systems), and storage-side equipment (energy storage devices) through communication protocols such as Industrial Ethernet. The system perception layer collects operational data from each node, including voltage, current, power, equipment operating status, and energy storage SOC, and adds a timestamp to each frame of data before packaging and uploading it.

[0050] The platform control layer is deployed at the line control center and includes data acquisition units, prediction units, digital twin units, and optimization scheduling units. Each unit can be deployed as a server or server cluster, interconnected via a high-speed local area network. The platform control layer receives operational data uploaded from the system perception layer and performs data aggregation, time synchronization and alignment, prediction, digital twin modeling, and optimization decision-making.

[0051] The data acquisition unit connects to the output of the system's perception layer. After receiving timestamped operational data, it compares the timestamps from different data sources and uses an interpolation synchronization algorithm to synchronize and align the data under a unified time reference, outputting a system-wide data stream at a unified time profile. The prediction unit receives the system-wide data stream and uses a prediction model to generate predicted photovoltaic power output and predicted values ​​for various load types. The digital twin unit receives the system-wide data stream and predicted values, constructs a digital twin model of the traction power supply system, and outputs the twin data model. The optimization and scheduling unit receives the system-wide data stream, predicted values, and the twin data model, performs multi-objective optimization calculations, and generates optimized control commands.

[0052] The edge control layer is deployed in each traction substation and includes bidirectional converter controllers, energy routing controllers, flexible interconnection controllers, SVG controllers, and energy storage controllers. Each controller is connected to its corresponding controllable power electronic device via a fieldbus. After receiving optimization control commands from the optimization scheduling unit, the edge control layer adjusts the power of each controllable power electronic device according to the real-time operating status of the substation.

[0053] In traditional urban rail power supply systems, data collected by the sensing layer is directly uploaded to various independent subsystems (power monitoring, vehicle monitoring, etc.), lacking unified aggregation and time alignment. This results in upper-level optimization being unable to obtain synchronized system cross-sectional data. The technical solution in this application collects data uniformly at the sensing layer and adds timestamps, then synchronizes and aligns the data through data acquisition units, outputting a unified system-wide data stream at a unified time cross-section, providing a data foundation for prediction, twinning, and optimization. In the three-layer architecture, the sensing layer addresses the issue of data availability, the platform layer addresses data usability and decision-making, and the edge layer addresses execution, forming a complete link from sensing to decision-making to execution. This solves the industry pain points of traditional systems, such as scattered data, delayed decision-making, and passive execution.

[0054] In other embodiments of this application, the edge control layer further includes a local sensing interface, which is connected to the sensing devices in the traction substation. This interface is used to switch to an independent operation mode when the platform control layer fails. In the independent operation mode, the edge control layer obtains the real-time status of the traction substation through the local sensing interface and performs normal energy distribution within the coverage area of ​​the traction substation based on the locally preset control logic.

[0055] The local sensing interface is directly connected to the sensing devices in the traction substation via hardwiring or fieldbus.

[0056] During normal operation, the edge control layer adjusts power through optimized control commands issued by the platform management layer, while continuously monitoring the communication status with the platform management layer. When the edge control layer detects a communication interruption with the platform management layer, a device failure in the platform management layer, or a timeout of received commands (for example, no valid commands are received for three consecutive cycles), it automatically switches to independent operation mode.

[0057] In independent operation mode, the edge control layer acquires real-time status data such as voltage, current, and energy storage SOC within the traction substation through a local sensing interface, no longer relying on data distribution from the platform control layer. The edge control layer invokes locally pre-set control logic (e.g., voltage regulation control strategy, overvoltage protection strategy, and energy storage emergency charging / discharging strategy), autonomously makes decisions and controls various controllable power electronic devices based on the local real-time status, ensuring normal energy distribution and power supply continuity within the substation's coverage area.

[0058] Once the edge control layer detects that communication with the platform management layer has been restored, it automatically switches back from independent operation mode to normal mode and resumes receiving optimized control commands from the platform management layer.

[0059] Urban rail power supply systems have extremely high reliability requirements; any power outage can affect train operation safety. In traditional centralized architectures, if the control center or communication network fails, all substations along the line will lose control. The technical solution in this application configures local sensing interfaces and independent control logic for the edge control layer, enabling each substation to remain autonomous in place during network outages, ensuring uninterrupted power supply to critical loads and greatly improving system robustness.

[0060] In other embodiments of this application, the source-side equipment includes a main substation and a photovoltaic power generation device; the grid-side equipment includes a bidirectional converter, an energy routing device, a flexible interconnection device, and an SVG; the load-side equipment includes a train and a power and lighting distribution system; and the storage-side equipment includes an energy storage device. The operating data on the source side includes: main substation voltage and power, photovoltaic output voltage, photovoltaic output power, photovoltaic capacity, irradiance, and temperature; The grid-side operating data includes: DC side voltage, DC side power, AC side voltage, AC side current, operating status and remaining power capacity of each grid-side device; The operational data on the load side includes: train timetable, train position, train speed, train power, train voltage, and power and lighting system power; The operating data on the energy storage side includes: energy storage device voltage, energy storage device current, and energy storage device SOC.

[0061] As shown in Table 1, the data requirements of the urban rail energy management and control system include the source-side operation data collected by the system's perception layer, such as: main substation voltage and power (reflecting the power supply capacity of the external power grid), photovoltaic output voltage and output power (reflecting the real-time photovoltaic power generation capacity), photovoltaic capacity (reflecting the photovoltaic installed capacity), irradiance (reflecting the lighting conditions), and temperature (an environmental factor that affects the photovoltaic power generation efficiency).

[0062] Grid-side equipment includes bidirectional converters, energy routing devices, flexible interconnection devices, and SVG (Static Var Generator). The grid-side operating data collected by the system sensing layer includes: DC-side voltage and DC-side power (reflecting the operating status of the DC traction network), AC-side voltage and AC-side current (reflecting the operating status of the AC distribution network), the operating status of each grid-side device (such as operation / shutdown / fault), and remaining power capacity (reflecting the adjustability margin of the equipment).

[0063] The load-side equipment includes trains and power and lighting distribution systems. The load-side operation data collected by the system sensing layer includes: train timetable (reflecting train dispatching plans), train position, train speed, train power (reflecting real-time traction load), train voltage (reflecting train power receiving status), and power and lighting system power (reflecting the power and lighting load of stations and sections).

[0064] Storage-side equipment includes energy storage devices. The storage-side operating data collected by the system sensing layer includes: energy storage device voltage, energy storage device current, and energy storage device SOC (State of Charge, reflecting the remaining energy storage capacity).

[0065] Table 1 Data Requirements for Urban Rail Energy Management System

[0066] The accuracy of a prediction model depends on the completeness and accuracy of the input data. This application's technical solution clarifies the specific content of data collection from four dimensions: source, grid, load, and storage. It covers key parameters such as irradiance and temperature required for photovoltaic power output prediction, train timetables and location speeds required for traction load prediction, and State of Charge (SOC) required for energy storage scheduling, providing a complete data foundation for subsequent hybrid prediction and digital twin modeling.

[0067] In other embodiments of this application, the prediction unit includes a photovoltaic power output prediction subunit, a traction load prediction subunit, and a power and lighting load prediction subunit.

[0068] The photovoltaic power output prediction subunit is used to build a physical model based on solar elevation angle, photovoltaic module parameters and irradiance data to estimate the basic power output. It also uses the Transformer model to perform time-series correction on the residual sequence output by the physical model and multi-source meteorological time-series data to output the photovoltaic power output prediction value.

[0069] Multi-source meteorological time-series data includes time series data of various meteorological elements such as irradiance, temperature, humidity, and wind speed. Among them, irradiance and temperature directly affect the physical model calculation of photovoltaic power output, while humidity and wind speed are used as auxiliary features to correct for the impact of nonlinear factors such as abrupt weather changes on power output.

[0070] Specifically, the workflow of the photovoltaic power output prediction subunit is as follows: First, based on the solar elevation angle, photovoltaic module parameters (photovoltaic conversion efficiency η, effective area A, temperature coefficient β) and irradiance data G(t), a physical output model is constructed to estimate the basic output.

[0071] The physical model formula is:

[0072] in,P PV (t) Characterizes the photovoltaic output (kW) at time t; η For the photoelectric conversion efficiency of the component; A The effective area of ​​the photovoltaic array (m²) 2 ); G(t) The actual irradiance at time t; β Temperature coefficient; Tc(t) Let t be the temperature of the solar panel (°C). T ref This is the standard test temperature.

[0073] Then, a Transformer model is introduced to perform time-series correction on the residual sequence of the above physical model. The input of the Transformer model is the joint feature vector of historical power output sequence and multi-source meteorological time-series data, including irradiance, temperature, humidity, wind speed, and historical power output. The Transformer encoder calculates the attention weights between each time step through a multi-head self-attention mechanism, capturing the impact of nonlinear factors such as weather abrupt changes and cloud cover on power output, and outputs the predicted value of photovoltaic power output. The physical model provides the basic values, and the Transformer model compensates for nonlinear biases that are difficult for the physical model to characterize. The two are integrated to form a complete hybrid prediction architecture.

[0074] The traction load prediction subunit is used to construct a physical traction calculation model based on train timetables, headway, passenger capacity, track gradient, station spacing and vehicle traction characteristics to estimate the basic load. It also uses an LSTM model to perform time-series correction on the residual sequence of the physical traction calculation model and outputs the predicted traction load value.

[0075] Specifically, the workflow of the traction load prediction subunit is as follows: First, based on train timetables, headway, passenger capacity, track gradient, station spacing, and vehicle traction characteristics, a physical traction calculation model is constructed to estimate the basic load. The physical traction calculation model calculates the train's traction force, operating speed, and auxiliary system power at each time point according to the train timetable, using formulas... Estimate the foundation traction load.

[0076] Where F(t) is the train traction force, and v(t) is the running speed. For motor efficiency, This is for auxiliary system power.

[0077] Then, using historical vehicle operation data as input, an LSTM model is employed to perform time-series correction on the residual sequence of the physical model. By learning the time dependencies of the physical model residuals, the LSTM model compensates for actual operational deviations not captured by the physical model and outputs predicted traction load values.

[0078] The power and lighting load prediction subunit is used to predict power and lighting loads using a Transformer multi-feature fusion time series model, with passenger flow, indoor and outdoor temperature, equipment operation plan, holiday information and energy-saving control strategy as input features.

[0079] Specifically, the workflow of the power lighting load prediction subunit is as follows: using passenger flow, indoor and outdoor temperatures, equipment operation plans, holiday information, and energy-saving control strategies as input features, a joint feature vector is constructed, and power lighting load is predicted through a Transformer multi-feature fusion time series model. The Transformer's multi-head self-attention mechanism automatically learns the nonlinear influence weights of each feature on the load and outputs the predicted power lighting load value for future time periods.

[0080] The output of urban rail photovoltaic systems is greatly affected by sudden weather changes. Traction load varies periodically with train schedules but fluctuates randomly due to factors such as driving habits. Neither a single physical model nor a pure data model can achieve high-precision predictions. Pure physical models are based on idealized assumptions and cannot adapt to component aging and sudden weather changes; pure data models (such as using Transformer or LSTM alone) lack physical constraints and may output unreasonable results under extreme conditions. The technical solution in this application adopts a hybrid architecture where a physical model provides the baseline values ​​and an AI model compensates for the residuals. This allows the interpretability of the physical model to complement the nonlinear fitting capability of the AI ​​model, maintaining high prediction accuracy across different time scales (long-term planning, medium-term scheduling, and ultra-short-term real-time), providing reliable boundary conditions for optimized scheduling.

[0081] In other embodiments of this application, the prediction unit further includes a rolling correction subunit, which is connected to the photovoltaic power output prediction subunit, the traction load prediction subunit, and the power and lighting load prediction subunit. The rolling correction subunit is used to collect actual power output data and actual load data, and to correct the model parameters of each prediction subunit based on the actual power output data and actual load data.

[0082] The rolling correction subunit collects actual output data and actual load data (i.e., actual photovoltaic power generation, actual train traction power, and actual power consumption for power and lighting). The rolling correction subunit compares the predicted values ​​output by each prediction subunit with their corresponding actual values ​​and calculates the prediction residuals. Then, a sliding window online update strategy is adopted: a fixed-length time window of length W is maintained, containing the prediction residual data from the most recent W time steps. The gradient direction of the model parameters is calculated using the residual data within the window, and the prediction model parameters are fine-tuned using an online learning rate α.

[0083] The parameter update formula is: , Where L is the loss function, , These are the parameter values ​​before the update, where α is the online learning rate. This is the gradient of the loss function.

[0084] In this way, the rolling correction subunit continuously corrects the model parameters of each prediction subunit using the measured data of the most recent W time steps, so that the prediction model can track dynamic factors such as the efficiency degradation of photovoltaic modules, train schedule adjustments, and load characteristic shifts caused by seasonal changes.

[0085] The efficiency of photovoltaic modules declines year by year with age, train schedules are adjusted regularly according to operational needs, and passenger flow varies systematically due to seasonality and holidays. If the prediction model parameters remain fixed, the prediction accuracy will gradually decrease over time. The technical solution in this application continuously collects actual data and corrects model parameters online through a rolling correction subunit, enabling the prediction model to have adaptive capabilities and maintain high prediction accuracy throughout its entire life cycle.

[0086] In other embodiments of this application, the digital twin unit includes a mechanism model subunit, a data-driven correction subunit, and a closed-loop feedback subunit.

[0087] The mechanism model subunit is connected to the output of the data acquisition unit to receive the data stream of the entire system. Based on the operating status of the train traction power supply system, the characteristics of photovoltaic distributed energy generation, the charging and discharging characteristics of the energy storage device, and the train operating status, it establishes the system power flow equation and outputs the system power flow reference value.

[0088] Specifically, the system power flow equations are as follows:

[0089] in, For traction load, For power lighting load, Contribute to photovoltaic power This refers to the energy storage charging and discharging power (discharging is positive).

[0090] The data-driven correction subunit connects the output of the mechanistic model subunit and the output of the data acquisition unit. It receives the system power flow reference value and the entire system data flow. Based on the residual between the entire system data flow and the system power flow reference value, it generates the parameter correction amount for the mechanistic model subunit through a machine learning model. θ.

[0091] The input of the closed-loop feedback subunit is connected to the output of the data-driven correction subunit to receive parameter correction values. θ; The first feedback output of the closed-loop feedback subunit is connected to the input of the mechanism model subunit, and is used to convert the parameter correction amount. θ is fed back to the sub-units of the mechanism model to correct the parameters of the mechanism model.

[0092] The second feedback output of the closed-loop feedback subunit is connected to the input of the data-driven correction subunit. This feedback feeds back the residual between the system power flow baseline value output by the mechanistic model subunit based on the corrected parameters and the actual system data to the data-driven correction subunit. The data-driven correction subunit then uses this residual to continuously update the parameters of the machine learning model, making the correction amount in the next prediction more accurate.

[0093] Traditional digital twin models are mostly pure mechanistic models with fixed parameters, making them unable to adapt to long-term drift caused by equipment aging and environmental changes, resulting in a decrease in mapping accuracy over time. While pure data-driven models can fit complex nonlinear relationships, they lack physical constraints and may output unreasonable results under extreme conditions. The technical solution in this application employs a hybrid driving approach that combines physical constraints and a baseline framework provided by a mechanistic model with online parameter correction by a data-driven model. This ensures that the digital twin model maintains physical rationality while continuously tracking changes in the physical system. The closed-loop feedback mechanism creates a positive feedback optimization cycle of "mechanistic model prediction → data-driven correction → feedback update," continuously improving the accuracy of the twin model over time and providing a high-fidelity system state mirror for optimized scheduling.

[0094] In other embodiments of this application, the optimized scheduling unit includes a daily-level optimization subunit, a scheduling coordination subunit, and a real-time optimization subunit.

[0095] The first input terminal of the daily-level optimization subunit is connected to the output terminals of the photovoltaic power output prediction subunit and the traction load prediction subunit, and is used to receive the photovoltaic power output prediction value and the traction load prediction value; the second input terminal of the daily-level optimization subunit is connected to the output terminal of the digital twin unit, and is used to receive the twin model data.

[0096] The daily optimization subunit is used to generate a daily scheduling plan based on the predicted photovoltaic output, the predicted traction load, and twin model data.

[0097] Specifically, the daily optimization subunit operates on a 24-hour scheduling cycle, using photovoltaic (PV) output forecasts, electricity pricing policies (time-of-use pricing curves), and train schedules as inputs, and energy storage SOC upper and lower limits (exemplarily 20%~90%) and PV installed capacity as constraints. The daily optimization subunit constructs an objective function: minimizing the sum of grid power purchase cost + energy storage loss cost + PV curtailment penalty cost. Specifically, the grid power purchase cost is calculated as the product of the purchased electricity volume and electricity price for each time period based on the predicted load and energy storage charging / discharging plan; the energy storage loss cost is calculated based on the charging / discharging depth and cycle count; and the PV curtailment penalty cost is calculated based on the difference between the predicted PV output and the actual consumption. By solving this optimization problem, the daily optimization subunit outputs a daily scheduling plan, including the energy storage charging / discharging power plan for each time period and the target SOC trajectory for energy storage.

[0098] The input of the scheduling coordination subunit is connected to the output of the daily optimization subunit. It is used to receive the daily scheduling plan and convert the daily scheduling plan into boundary constraints for real-time optimization.

[0099] Specifically, the scheduling and coordination subunit converts the energy storage SOC target value and charging / discharging power plan in the daily scheduling plan into the boundary constraints required for real-time optimization, including: the energy storage SOC target range (for example, the target value ±5%), the upper and lower limits of power for each substation, and the allowable fluctuation range of grid voltage.

[0100] The input end of the real-time optimization subunit is connected to the output end of the scheduling and coordination subunit and the output end of the data acquisition unit. It is used to receive boundary constraints and the data flow of the entire system. Based on the boundary constraints and the data flow of the entire system, it iteratively calculates the network space matrix expression of the traction network. Under the constraints of the objective function, it optimizes through intelligent algorithms, generates optimized control commands for each controllable device, and sends them to the edge control layer.

[0101] Specifically, the real-time optimization subunit constructs the traction network node admittance matrix and power flow equations based on the location of each substation, the boundary of the power supply zone, the current position of the train, and the power demand. Under the constraints of minimizing dispatching losses, ensuring that substation power does not exceed limits, keeping grid voltage fluctuations within allowable ranges, meeting harmonic content standards, and meeting power factor standards, the real-time optimization subunit uses intelligent algorithms such as the interior-point method or genetic algorithms to find the optimal power output command for each controllable device, generates optimized control commands, and sends them to the edge control layer.

[0102] Urban rail traction power supply systems exhibit both daily regularity (train schedules are basically fixed daily, and electricity prices vary according to peak and off-peak hours) and second-level volatility (train starting / braking causes drastic power changes). Daily planning alone cannot address second-level load fluctuations; real-time control alone cannot utilize peak-valley electricity prices for economic optimization. This application's technical solution employs a two-tiered optimization approach: daily economic planning and real-time power allocation. At the daily level, energy storage charging and discharging strategies are planned from a global perspective to reduce electricity costs. At the real-time level, power allocation is dynamically adjusted based on second-level state changes to ensure stable operation. These two approaches are coupled through a scheduling coordination subunit (daily output provides boundary constraints for real-time operation), achieving a balance between economic efficiency and dynamic response.

[0103] In other embodiments of this application, the optimized scheduling unit further includes a supply-demand bidirectional matching subunit.

[0104] The input of the supply-demand bidirectional matching subunit is connected to the output of the real-time optimization subunit, and is used to schedule energy supply sources according to a first preset priority rule and allocate energy demand according to a second preset priority rule. The first preset priority rule includes: First priority: Train regenerative braking energy; the regenerative energy generated during train braking is a zero-cost energy source, and if it is not consumed in time, it will cause the traction network voltage to rise, triggering the braking resistor to be put into operation and causing energy waste. Therefore, it must be consumed first.

[0105] Second priority: Photovoltaic distributed energy; Photovoltaic power generation is also a zero-cost energy source and should be given priority to reduce the purchase of electricity from the grid.

[0106] Third priority: Energy storage discharge; the cost of the electrical energy stored in the energy storage device is lower than that of the grid-purchased electricity (derived from off-peak electricity pricing or regenerative braking energy storage), and it is released first when photovoltaic and regenerative braking energy are insufficient.

[0107] Fourth priority: purchasing electricity from the grid; electricity purchased from the external public grid has the highest cost and should only be used when the above three types of energy cannot meet the load demand.

[0108] The second preset priority rule includes: First priority: Train traction load; normal train operation is the core function of urban rail transit, and traction load must be guaranteed first.

[0109] Second priority: power and lighting loads; the power supply reliability requirements for loads such as station lighting, air conditioning, and escalators are lower than those for traction loads, but still need to be guaranteed.

[0110] Third priority: Energy storage charging demand. The energy storage device's need to replenish energy can be arranged after the first two types of loads are met.

[0111] Fourth priority: Grid feedback; When the energy supply in this power supply zone exceeds the demand and the energy storage is full, the excess energy will be inverted and fed back to the AC ring network. This demand has the lowest priority.

[0112] The supply-demand bidirectional matching subunit embeds the above bidirectional matching strategy into the constraints of the real-time optimization subunit to ensure that the optimization results meet the priority ranking under any operating condition.

[0113] In traditional energy dispatching, the dispatching priorities of various energy sources are unclear or ranked solely based on cost, failing to fully consider the urgency of absorbing regenerative braking energy and the importance of power supply to various loads. The technical solution in this application ensures the priority absorption of zero-cost regenerative braking energy and photovoltaic energy through clear two-way priority matching of supply and demand, reducing energy waste and electricity purchase costs; simultaneously, it guarantees the power supply reliability of traction loads, a core load, achieving a balance between energy efficiency and safety.

[0114] In other embodiments of this application, the optimized scheduling unit further includes a cross-partition scheduling subunit.

[0115] The first input of the cross-regional scheduling subunit is connected to the output of the digital twin unit to obtain the power flow distribution status of the entire system. The second input is connected to the data output of the platform control layer to obtain the power output of the main substation and vehicle load demand of each power supply zone. The output is connected to the flexible interconnection device controller of the edge control layer to drive the energy flow of this zone to other power supply zones through the flexible interconnection device when the photovoltaic output in the power supply zone cannot be fully absorbed by this zone, based on the power flow distribution status and the power output of the main substation and vehicle load demand of each power supply zone.

[0116] Specifically, the cross-regional dispatching subunit continuously monitors the balance between photovoltaic output and local load in each power supply zone. When the photovoltaic output in a power supply zone exceeds the local load demand (traction load + power and lighting load + energy storage charging demand) and the energy storage device has reached its charging limit, the cross-regional dispatching subunit determines that there is surplus photovoltaic power in that zone.

[0117] At this point, the cross-regional scheduling subunit, based on the overall system power flow distribution provided by the digital twin unit, assesses the load demand and receiving capacity of adjacent power supply zones and selects the optimal power transmission path. The output of the cross-regional scheduling subunit is connected to the flexible interconnection device controller in the edge control layer, sending cross-regional scheduling commands to the flexible interconnection device controller to control the flexible interconnection device to transmit surplus photovoltaic power from its own power supply zone to other power supply zones, thus realizing cross-regional energy flow.

[0118] Urban rail power supply systems are typically divided into multiple power supply zones. Power flow between these zones is constrained by the open / closed state of traditional circuit breakers, hindering flexible energy dispatch. When a zone has a large installed photovoltaic capacity but a low local load, surplus photovoltaic power cannot be absorbed by other zones, resulting in wasted solar power. This embodiment connects different power supply zones through flexible interconnection devices. Combined with the system-wide power flow status and real-time load data provided by digital twins, it enables on-demand cross-zone energy flow, increasing the proportion of renewable energy consumption and reducing the overall system's electricity purchase demand.

[0119] The second aspect of this application provides a method for urban rail energy management, implemented based on the urban rail energy management system described in the first aspect of this application, such as... Figure 3 As shown, urban rail transit energy management methods include: Step S1: Collect operational data from each node of the source-grid-load-storage system. The operational data includes timestamps.

[0120] Specifically, the system perception layer collects operational data from each node of the source-grid-load-storage system, including voltage, current, power, equipment status, and energy storage SOC, and adds a timestamp to each frame of data collected.

[0121] Step S2: Synchronize and align the running data under a unified time base to obtain the whole system data stream under a unified time section.

[0122] The data acquisition unit of the platform control layer receives the running data collected in step S1, compares the timestamps of different data sources, and uses an interpolation synchronization algorithm to synchronize and align the running data under a unified time reference. This obtains the whole system data stream under a unified time section, ensuring that the data used for subsequent prediction, twinning, and optimization comes from the same moment, thus avoiding decision-making biases caused by data asynchrony.

[0123] Step S3: Receive the entire system data stream and generate predicted values ​​based on the prediction model using the entire system data stream.

[0124] Specifically, the prediction unit receives the system-wide data stream output in step S2 and generates predicted values ​​based on the data stream using a prediction model. These predicted values ​​include photovoltaic power output predictions, traction load predictions, and power and lighting load predictions.

[0125] Step S4: Receive the entire system data stream and predicted values, and construct and output the twin data model based on the entire system data stream and predicted values.

[0126] Specifically, the digital twin unit receives the full system data stream output in step S2 and the predicted value output in step S3. Based on the real-time running data and predicted value in the full system data stream (as feedforward boundary conditions), it constructs a digital twin model driven by a hybrid mechanism model and a data-driven model, and outputs a twin data model.

[0127] Step S5: Generate optimized control instructions based on the whole system data flow, predicted values, and twin data model.

[0128] The optimization scheduling unit receives the system-wide data stream output from step S2, the predicted values ​​output from step S3, and the twin data model output from step S4, performs two-level optimization calculations, and generates optimized control commands. Specifically, the daily optimization subunit performs daily economic scheduling using the predicted values ​​as input, while the real-time optimization subunit performs real-time power allocation using both the system-wide data stream and the twin data model as input.

[0129] Step S6: Receive the optimization control command and adjust the power of each controllable power electronic device according to the optimization control command and the real-time status of the substation.

[0130] Specifically, the edge control layer receives the optimized control commands generated in step S5 and adjusts the power of each controllable power electronic device based on the optimized control commands and the real-time status of each traction substation. In addition, the edge control layer simultaneously monitors the communication status with the platform management layer and switches to independent operation mode in the event of a platform management layer failure.

[0131] Step S7: After the real-time optimization in step S6, the actual running data is fed back to steps S3 and S4 for rolling updates of the prediction model parameters and digital twin model parameters.

[0132] Specifically, the actual operating data is fed back to the rolling correction subunit of the prediction unit to update the model parameters of each prediction subunit; at the same time, it is fed back to the closed-loop feedback subunit of the digital twin unit to update the machine learning model parameters of the data-driven correction subunit.

[0133] Those skilled in the art will recognize that the modules, units, and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0134] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. An urban rail transit energy management and control system, characterized in that, include: The system perception layer is communicatively connected to the source-side equipment, grid-side equipment, load-side equipment, and storage-side equipment in the urban rail power supply system. It is configured to collect and output the operating data of each node in the source-grid-load-storage system, and the operating data includes timestamps. The platform control layer includes: The data acquisition unit, connected to the output of the system perception layer, is configured to receive the operation data of each node of the power supply system (source-grid-load-storage), synchronize and align the operation data under a unified time reference according to the timestamp, and output the whole system data stream under a unified time section. The prediction unit, connected to the output of the data acquisition unit, is configured to receive the entire system data stream and generate a prediction value based on the entire system data stream through a prediction model. A digital twin unit, connected to the output of the data acquisition unit and the output of the prediction unit, is configured to receive the full system data stream and the predicted value, and to construct and output a twin data model based on the full system data stream and the predicted value; The optimization scheduling unit, connected to the output of the data acquisition unit, the output of the prediction unit, and the output of the digital twin unit, is configured to generate optimization control commands based on the whole system data flow, the predicted values, and the twin data model. The edge control layer, deployed in the traction substation of the urban rail power supply system, includes a bidirectional converter controller, an energy routing device controller, a flexible interconnection device controller, an SVG controller, and an energy storage device controller. Each controller is connected to a corresponding controllable power electronic device. The edge control layer is connected to the output of the optimization scheduling unit and is configured to receive the optimization control command and adjust the power of each controllable power electronic device according to the optimization control command and the real-time status of the substation.

2. The urban rail energy management and control system according to claim 1, characterized in that, The edge control layer also includes a local sensing interface, which is connected to sensing devices within the traction substation and configured to switch to an independent operation mode when the platform control layer fails. In the independent operation mode, the edge control layer obtains the real-time status of the traction substation through the local sensing interface and performs normal energy distribution within the coverage area of ​​the traction substation based on locally preset control logic.

3. The urban rail energy management and control system according to claim 1, characterized in that, The source-side equipment includes a main substation and a photovoltaic power generation device; the grid-side equipment includes a bidirectional converter, an energy routing device, a flexible interconnection device, and an SVG; the load-side equipment includes a train and a power and lighting distribution system; and the storage-side equipment includes an energy storage device. The operating data on the source side includes: main substation voltage and power, photovoltaic output voltage, photovoltaic output power, photovoltaic capacity, irradiance, and temperature; The grid-side operating data includes: DC-side voltage, DC-side power, AC-side voltage, AC-side current, operating status and remaining power capacity of each grid-side device; The operational data on the load side includes: train timetable, train position, train speed, train power, train voltage, and power and lighting system power. The operating data on the energy storage side includes: energy storage device voltage, energy storage device current, and energy storage device SOC.

4. The urban rail energy management and control system according to claim 1, characterized in that, The prediction unit includes a photovoltaic power output prediction subunit, a traction load prediction subunit, and a power and lighting load prediction subunit; The photovoltaic power output prediction subunit is configured to construct a physical model based on solar elevation angle, photovoltaic module parameters and irradiance data to perform basic power output estimation, and to perform time series correction on the residual sequence output by the physical model and multi-source meteorological time series data through the Transformer model to output the photovoltaic power output prediction value. The traction load prediction subunit is configured to construct a physical traction calculation model based on train timetable, headway, passenger capacity, line gradient, station spacing and vehicle traction characteristics to estimate the basic load, and to perform time-series correction on the residual sequence of the physical traction calculation model through an LSTM model to output the traction load prediction value. The power lighting load prediction subunit is configured to use passenger flow, indoor and outdoor temperature, equipment operation plan, holiday information and energy-saving control strategy as input features, and perform power lighting load prediction through Transformer multi-feature fusion time series model.

5. The urban rail energy management and control system according to claim 4, characterized in that, The prediction unit also includes a rolling correction subunit, which is connected to the photovoltaic power output prediction subunit, the traction load prediction subunit, and the power and lighting load prediction subunit. It is configured to collect actual power output data and actual load data, and correct the model parameters of each prediction subunit based on the actual power output data and actual load data.

6. The urban rail energy management and control system according to claim 1, characterized in that, The digital twin unit includes a mechanism model subunit, a data-driven correction subunit, and a closed-loop feedback subunit. The mechanism model subunit is connected to the output end of the data acquisition unit and is configured to receive the data stream of the entire system. Based on the operating status of the train traction power supply system, the characteristics of photovoltaic distributed energy generation, the charging and discharging characteristics of the energy storage device and the train operating status, it establishes the system power flow equation and outputs the system power flow reference value. The data-driven correction subunit is connected to the output of the mechanism model subunit and the output of the data acquisition unit. It is configured to receive the system power flow reference value and the whole system data flow, and generate the parameter correction amount of the mechanism model subunit through a machine learning model based on the residual between the whole system data flow and the system power flow reference value. The input terminal of the closed-loop feedback subunit is connected to the output terminal of the data-driven correction subunit and is configured to receive the parameter correction amount; the first feedback output terminal of the closed-loop feedback subunit is connected to the input terminal of the mechanism model subunit and is configured to feed the parameter correction amount back to the mechanism model subunit to correct the parameters of the mechanism model. The second feedback output terminal of the closed-loop feedback subunit is connected to the input terminal of the data-driven correction subunit and is configured to feed back the residual between the system power flow reference value output by the mechanism model subunit based on the corrected parameters and the actual system data to the data-driven correction subunit.

7. The urban rail energy management and control system according to claim 4, characterized in that, The optimized scheduling unit includes a daily-level optimization subunit, a scheduling coordination subunit, and a real-time optimization subunit; The first input terminal of the daily-level optimization subunit is connected to the output terminal of the photovoltaic power output prediction subunit and the output terminal of the traction load prediction subunit, and is configured to receive the photovoltaic power output prediction value and the traction load prediction value; The second input terminal of the daily-level optimization subunit is connected to the output terminal of the digital twin unit and is configured to receive the twin model data; The daily optimization subunit is configured to generate a daily scheduling plan based on the photovoltaic output forecast, the traction load forecast, and the twin model data. The input end of the scheduling coordination subunit is connected to the output end of the daily optimization subunit, and is configured to receive the daily scheduling plan and convert the daily scheduling plan into boundary constraints for real-time optimization. The input end of the real-time optimization subunit is connected to the output end of the scheduling and coordination subunit and the output end of the data acquisition unit. It is configured to receive the boundary constraints and the whole system data stream, iteratively calculate the network space matrix expression of the traction network according to the boundary constraints and the whole system data stream, optimize through intelligent algorithms under the objective function constraints, generate optimization control instructions for each controllable device and send them to the edge control layer.

8. The urban rail energy management and control system according to claim 1, characterized in that, The optimized scheduling unit also includes a supply-demand bidirectional matching subunit; The input of the supply-demand bidirectional matching subunit is connected to the output of the real-time optimization subunit, and is configured to schedule energy supply sources according to a first preset priority rule and allocate energy demand according to a second preset priority rule. The first preset priority rule includes: First priority: Train regenerative braking energy; Second priority: Distributed photovoltaic energy; Third priority: Energy storage and discharge; Fourth priority: Purchasing electricity from the power grid; The second preset priority rule includes: First priority: Train traction load; Second priority: Power lighting load; Third priority: Energy storage and charging needs; Fourth priority: Grid feedback.

9. The urban rail energy management and control system according to claim 1, characterized in that, The optimized scheduling unit also includes a cross-regional scheduling subunit; The first input terminal of the cross-regional scheduling subunit is connected to the output terminal of the digital twin unit and is configured to obtain the power flow distribution status of the entire system. The second input terminal is connected to the data output terminal of the platform control layer and is configured to obtain the power output of the main substation and vehicle load demand of each power supply zone. The output terminal is connected to the flexible interconnection device controller of the edge control layer and is configured to drive the energy flow of this zone to other power supply zones through the flexible interconnection device when the photovoltaic output in the power supply zone cannot be fully absorbed by this zone, based on the power flow distribution status and the power output of the main substation and vehicle load demand of each power supply zone.

10. A method for managing energy consumption in urban rail transit, characterized in that, Based on the urban rail energy management system according to any one of claims 1-9, the method includes: Step S1: Collect operational data from each node of the source-grid-load-storage system, wherein the operational data includes timestamps; Step S2: Synchronize and align the running data under a unified time reference to obtain the full system data stream under a unified time profile; Step S3: Receive the entire system data stream and generate predicted values ​​based on the entire system data stream using a prediction model; Step S4: Receive the entire system data stream and the predicted value, and construct and output a twin data model based on the entire system data stream and the predicted value; Step S5: Generate optimized control instructions based on the entire system data stream, the predicted values, and the twin data model; Step S6: Receive the optimized control command, and adjust the power of each controllable power electronic device according to the optimized control command and the real-time status of the substation; Step S7: After the real-time optimization in step S6 is executed, the actual running data is fed back to steps S3 and S4 for rolling updates of the prediction model parameters and digital twin model parameters.

Images