Urban rail transit energy management system considering new energy and energy storage access

Abstract

The invention provides an urban rail transit energy management system considering new energy and energy storage access. The urban rail transit energy management system comprises a new energy power generation unit, an energy storage unit, a regenerative braking energy recovery unit and an energy management control platform, the new energy power generation unit and the energy storage unit are connected to a rail transit power supply system; the energy management control platform is connected with the power supply network and controls the units in a communication mode, energy flow in the rail transit power supply system is monitored, analyzed, optimized and dispatched so as to coordinate utilization of new energy power generation power, energy storage charging and discharging and regenerative braking energy, the energy utilization efficiency is improved, and the requirement for power grid outsourcing electric energy is lowered. The peak load of the system and the power grid electricity purchasing cost are effectively reduced, meanwhile, the overall energy utilization efficiency is improved, and energy conservation and consumption reduction are achieved while safe and stable operation of the rail transit system is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban rail transit, and particularly relates to an urban rail transit energy management system considering the access of new energy and energy storage. Background Art

[0002] At present, the urban rail transit energy management system mainly focuses on the monitoring of traditional power consumption and simple dispatching management. The system function is relatively single, and the comprehensive application of new energy and energy storage technologies has not been fully considered. In actual operation, due to the large fluctuations in the output of renewable energy and the limited capacity of energy storage devices, the rail transit power supply system often has difficulty in effectively balancing load changes, resulting in prominent problems such as low power supply efficiency and energy consumption waste. There are deficiencies in coordination in key links such as new energy power generation prediction, energy storage dispatching, and regenerative braking energy recovery in traditional systems, and there is a lack of an integrated control mechanism based on real-time data and intelligent algorithms, which cannot meet the requirements of safe, efficient power supply and energy conservation and consumption reduction at the same time. Summary of the Invention

[0003] The purpose of the present invention is to solve the deficiencies existing in the above background art, and provide an urban rail transit energy management system considering the access of new energy and energy storage, which can optimize the charge and discharge strategies of energy storage devices and the parameters of regenerative braking energy recovery on the basis of predicting new energy output and monitoring real-time load, effectively reduce the system peak load and the grid power purchase cost, and at the same time improve the overall energy utilization efficiency, ensuring energy conservation and consumption reduction while the rail transit system operates safely and stably.

[0004] The technical solution adopted by the present invention is: an urban rail transit energy management system considering the access of new energy and energy storage, including: a new energy power generation unit, an energy storage unit, a regenerative braking energy recovery unit, and an energy management control platform; the new energy power generation unit and the energy storage unit are connected to the rail transit power supply system; the energy management control platform is connected to the power supply system and communicates to control the above units, monitors, analyzes, and optimizes the dispatching of the energy flow in the rail transit power supply system to coordinate the utilization of new energy power generation, energy storage charge and discharge, and regenerative braking energy, improving the energy utilization efficiency and reducing the demand for externally purchased electric energy from the grid.

[0005] In the above technical solution, the data network architecture of the energy management control platform adopts a three-layer architecture design, including: The terminal perception layer collects device operation data, environmental data, and passenger flow data through various on-site intelligent meters and sensors, and realizes data monitoring; The edge computing layer, through devices such as intelligent communication gateways and switches, uploads the data collected by the terminal to the energy management control platform through a private network channel according to the standard protocol, and sends the control commands of the platform to each control unit; The system application layer, with the energy management control platform as the core, realizes the storage of all data of the subway station, the display of the equipment operation status, and the issuance of control instructions.

[0006] In the above technical solution, the control management architecture of the energy management control platform adopts a three - level hierarchical architecture, including: The field level, which collects the equipment operation conditions and environmental data through field instruments and sensors and executes local control; The station level, which monitors, statistically analyzes, and optimally controls the energy consumption within the station; The line level, which is set in the control center to centrally monitor and make dispatching decisions on the energy of the whole line, and issue control instructions to each station level to achieve the optimal coordinated control of energy across the whole line.

[0007] In the above technical solution, the energy management control platform is used to predict the output power of the new - energy power generation unit, and optimize the charge - discharge scheduling of the energy storage unit according to the prediction result to suppress the fluctuation of the new - energy power generation output.

[0008] In the above technical solution, the energy management control platform is used to reduce the peak power load of the rail transit system through demand management, thereby reducing the basic electricity charge, and adjust the charge - discharge strategy of the energy storage unit according to the time - of - use electricity price, storing electric energy during low - valley periods and releasing electric energy during peak periods to reduce the overall electricity cost.

[0009] In the above technical solution, the energy management control platform is used to dynamically adjust the working threshold of the regenerative braking energy recovery unit based on the real - time power flow condition of the traction power supply network to ensure the line energy balance and maximize the utilization rate of braking energy recovery.

[0010] In the above technical solution, the energy management control platform establishes a power flow calculation model including an AC distribution network and a DC traction network, and uses this model to conduct power flow analysis on the rail transit power supply system with new - energy and energy - storage access, providing a decision - making basis for energy dispatching; The process of the power flow analysis includes: (a) Initialization stage: Initialize the parameters of each key device according to the preset parameters, and set the initial values of the output voltage of the rectifier unit, the DC traction voltage, and the AC - side node voltage; (b) AC distribution network calculation: Form a node admittance matrix Y based on the AC distribution network parameters of the rail transit power supply system, and calculate the voltage of each AC node; (c) First traction voltage iteration: Based on the initially set output voltage of the rectifier unit at the initial moment, and using the output voltage of the rectifier unit updated by the previous round of calculation at subsequent moments, iteratively calculate the traction voltage in the DC traction network until the change in the traction voltage between two consecutive iteration results meets the preset first convergence condition, so that the DC traction network part of the power flow calculation model converges; (d) Preliminary calculation of rectifier output power: Based on the initially set output voltage of the rectifier unit at the initial moment, and using the output voltage of the rectifier unit updated by the previous round of calculation at subsequent moments, calculate the initial output power of the rectifier unit based on the power flow calculation model; (e) Second traction voltage iteration and power update: ① Update the rectifier output voltage based on the power flow calculation results of the AC-side node voltage and the DC-side voltage of the rectifier; ② Based on the updated rectifier output voltage, iteratively calculate the traction voltage in the DC traction network until the change in the traction voltage between two consecutive iteration results meets the preset convergence condition, so that the DC traction network part of the power flow calculation model converges; ③ According to the converged power flow calculation model and the updated rectifier output voltage, calculate the output power of the rectifier unit in the current round; ④ Compare the output power of the rectifier unit calculated in the current round with the initial output power recorded in step (d). If the difference meets the preset power convergence condition, it is considered that the power flow calculation at this moment is completed; otherwise, return to step (d) and repeat the above iterative process until the convergence condition is met; (f) Time-step update: When both the traction voltage and the output power of the rectifier unit at this moment reach the preset convergence conditions, use the traction voltage and AC node voltage converged at the current moment as the initial values for the power flow calculation at the next moment, and repeat steps (c) to (e) until the entire power flow analysis is completed.

[0011] In the above technical solution, the energy storage unit includes a battery energy storage sub-unit and a flywheel energy storage sub-unit; the energy management and control platform uses a proximal policy optimization-based reinforcement learning algorithm to perform intelligent scheduling control on the energy storage unit, and dynamically adjusts the charge and discharge power of the battery and the flywheel according to real-time load, new energy generation, and electricity price information, so as to maximize economic benefits and minimize the amount of electricity purchased from the power grid.

[0012] In the above technical solution, the energy management and control platform is also used to collect station environmental parameters and passenger flow information, and adjust the operation of the environmental control loads related to station air conditioning and ventilation according to the change in electricity price, so as to reduce the energy consumption of the environmental control system while ensuring passenger comfort.

[0013] In the above technical solution, the energy management platform adopts a phased optimization control strategy, including: a local optimization stage that first identifies and optimizes the energy-consuming equipment with the greatest energy-saving potential, followed by an overall optimization stage that coordinates the energy utilization efficiency of each station, and a global optimization stage that comprehensively considers multiple factors such as new energy output fluctuations, electricity price changes, passenger flow, and energy storage life; the control strategies generated by the optimization in each stage are iteratively superimposed step by step to gradually approach the optimal solution of the energy utilization efficiency of the entire system.

[0014] The beneficial effects of the present invention are as follows: By integrating a new energy power generation device, an energy storage device, and a regenerative braking energy recovery unit, and uniformly monitoring and dispatching them by an energy management control platform, the present invention realizes the collaborative optimization of each energy unit in the rail transit power supply system, thereby improving the overall energy utilization efficiency, reducing the dependence on purchasing electricity from the external power grid, and enhancing the power supply stability and economy.

[0015] Furthermore, the present invention adopts a three-layer architecture design of a terminal perception layer, an edge computing layer, and a system application layer, which makes the on-site data collection, data transmission, and central control have clear divisions of labor. This structure ensures the real-time and accurate collection and transmission of data, helps to achieve remote centralized monitoring and refined control, and further improves the response speed and system reliability of energy management.

[0016] Furthermore, the present invention adopts a three-level control architecture of a field level, a station level, and a line level, which effectively connects local control and centralized dispatching. The field level can achieve local control, the station level conducts local energy consumption optimization, and the line level makes global dispatching decisions in the control center, so as to realize the energy coordination optimization within the whole line and enhance the overall stability and energy-saving effect of the system.

[0017] Furthermore, by combining the new energy power prediction module and the energy storage dispatching module, the present invention realizes the accurate prediction of the output of new energy (such as photovoltaic) power generation, and optimizes the charge and discharge strategy of the energy storage device accordingly, thereby suppressing the fluctuations of new energy output, improving the stability and reliability of the power supply system, and ensuring the efficient matching of energy supply and demand.

[0018] Furthermore, the present invention sets up a demand management and peak-valley electricity price response function, which can monitor and control the peak electricity load of the rail transit system in real time, and use the electricity price fluctuations for dynamic adjustment, so as to reduce the basic electricity charge and the overall electricity cost and maximize the economic benefits.

[0019] Furthermore, by dynamically adjusting the charge and discharge threshold of the DC regenerative braking energy recovery device, the present invention ensures the recovery of the remaining energy generated during the braking process of the train, thereby improving the energy recovery utilization rate, reducing energy waste, and further reducing the dependence of the system on external power supply.

[0020] Furthermore, the present invention establishes an AC-DC hybrid power flow calculation model to perform real-time analysis on the power flow of the AC and DC networks within the rail transit power supply system, providing accurate data support for energy scheduling. This data-based analysis can help optimize the energy distribution in the system, improve the accuracy of scheduling decisions, and enhance the overall operation efficiency.

[0021] Furthermore, the present invention adopts a reinforcement learning scheduling algorithm based on Proximal Policy Optimization (PPO) to intelligently schedule energy storage devices, dynamically adjusting the charging and discharging power according to real-time load, new energy output, and electricity price information, thereby maximizing economic benefits, reducing the electricity purchase cost of the power grid, and enhancing the system's adaptive control ability.

[0022] Furthermore, the present invention configures an environmental control energy consumption optimization function to collect environmental parameters, passenger flow data, and equipment operation status within the station in real time, and combines electricity price information to dynamically adjust the operation strategies of environmental control equipment such as air conditioners, ventilation, and exhaust, thereby reducing the energy consumption of the environmental control system and the overall operation cost while ensuring passenger comfort and safety.

[0023] Furthermore, the present invention adopts a phased energy optimization control strategy (including three stages: local optimization, overall optimization, and global optimization), gradually iterating and optimizing the system scheduling strategy, enabling the gradual superposition of various optimization measures to approach the optimal state of the energy utilization efficiency of the entire system, thereby achieving comprehensive energy conservation and consumption reduction and improving the system economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 FIG. is the overall architecture diagram of an urban rail transit energy management system considering the access of new energy and energy storage; Figure 2 FIG. is the overall software architecture diagram of the energy management control platform; Figure 3 FIG. is the schematic diagram of the iterative process of the energy-saving control algorithm applied in a specific embodiment; Figure 4 FIG. is the schematic diagram of an urban rail transit power supply system considering the access of photovoltaic and energy storage; Figure 5a FIG. is the schematic diagram of the simplified device model of the equivalent circuit model of an urban rail DC traction power supply system in a specific embodiment; Figure 5b FIG. is the overall model schematic diagram of the equivalent circuit model of an urban rail DC traction power supply system in a specific embodiment; Figure 6 FIG. is the flowchart of the AC-DC hybrid power flow calculation for an urban rail transit considering the access of photovoltaic and energy storage; Figure 7 FIG. is the flowchart of the energy scheduling algorithm based on reinforcement learning applied in a specific embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments, which is convenient for clearly understanding the present invention, but they do not constitute a limitation to the present invention.

[0026] As Figure 1 shown, the present invention provides an urban rail transit energy management system considering the access of new energy and energy storage, including: a new energy power generation unit, an energy storage unit, a regenerative braking energy recovery unit, and an energy management control platform; the new energy power generation unit and the energy storage unit are connected to the rail transit power supply system; the energy management control platform is connected to the power supply network and communicates to control the above units, monitors, analyzes, and optimizes the dispatching of the energy flow in the rail transit power supply system to coordinate the utilization of new energy power generation, energy storage charging and discharging, and regenerative braking energy, improve the energy utilization efficiency, and reduce the demand for externally purchased electric energy from the power grid.

[0027] Specifically, the data network architecture of the energy management control platform adopts a three-layer architecture design, including: The terminal perception layer collects device operation data, environmental data, and passenger flow data through various on-site intelligent meters and sensors, and realizes data monitoring; the system mainly collects device and environmental data through various on-site intelligent meters and sensors and conducts data monitoring. By integrating a variety of sensors, the system can monitor the energy consumption, operation status, and changes in the surrounding environment of the device in real time. These sensor data will be used for subsequent analysis and diagnosis to help discover potential energy waste or equipment failures. The on-site basic data collection devices include multifunctional electricity meters, intelligent control systems for chillers, etc. The on-site environment and equipment monitoring systems, integrated automation systems, security prevention and control systems, video monitoring data, etc. in the station can also be accessed to achieve centralized management.

[0028] The edge computing layer mainly collects the data of on-site intelligent instruments and sensors through devices such as intelligent communication gateways and switches, and uploads them to the energy management system (EMS) through a private network channel according to standard protocols, and sends the control commands of the platform to each control unit.

[0029] The system application layer takes the energy management control platform as the core to realize the storage of all data of subway stations, the display of device operation status, and the issuance of control instructions. The energy management control platform conducts remote and centralized monitoring of multiple subway stations, and local duty personnel can achieve "unattended" or "few people on duty" through monitoring all operation conditions by logging in to the web page.

[0030] Specifically, the control management architecture of the energy management control platform adopts a three-level hierarchical architecture, including: The field level collects device operation conditions and environmental data through on-site instruments and sensors and executes local control; At the station level, monitor, statistically analyze, and optimize the control of the energy consumption within the station; the station-level energy management system mainly includes the BAS system, the power quality monitoring system, and the ISCS system to collect the energy and environmental data within the station, and realize the functions of online monitoring, statistics, analysis, and diagnosis at the station level.

[0031] At the line level, it is set in the control center to centrally monitor and make dispatching decisions on the energy of the whole line, and issue control instructions to each station level to achieve the optimized coordinated control of energy within the whole line range. The line level realizes data sharing among systems, that is, the data analysis and display of the energy consumption of the whole line, calculates various energy consumption indicators according to the existing index system, provides data support for the evaluation of the energy consumption level, and provides auxiliary decision-making information for train operation dispatching and general dispatching.

[0032] As Figure 1 shown, the architecture of this embodiment fully reflects the multi-layer division of labor and cooperation mode of "field - station - line / control center": the field layer is responsible for bottom-layer perception and preliminary control, the station layer realizes local energy consumption management and execution control, and the control center centrally conducts advanced data analysis and dispatching decisions. Each layer is closely connected through a dedicated subway network to form an expandable, safe, reliable, and functional rail transit energy management system, providing strong technical support for the intelligent dispatching and energy-saving strategies of the present invention, mainly including the site-level network, the subway network, and the line-level / control center network.

[0033] The site-level network includes: Field sensors, intelligent meters, and control devices are deployed inside each station, depot, or parking lot, such as multi-functional electricity meters, environmental sensors, passenger flow detectors, intelligent control systems for chillers, etc.

[0034] These field devices are interconnected through industrial switches or intelligent communication gateways to form a station-level local area network. The gateway is responsible for packing the collected data according to a unified protocol and performing edge computing (such as simple threshold judgment, fault detection, etc.) when necessary.

[0035] One or more station servers can be locally configured in the station-level network to monitor and statistically analyze the energy consumption data within the station, and can automatically adjust the adjustable loads such as air conditioners, lighting, and ventilation according to the energy-saving control instructions issued by the superior.

[0036] The subway network includes: The local area networks of each station, depot, or parking lot are all interconnected with the line-level or control center network through dedicated communication lines (i.e., the "subway network").

[0037] This subway network usually adopts the form of a dedicated network or a virtual private network (VPN), with security encryption and bandwidth guarantee to meet the requirements of the rail transit system for real-time performance and reliability.

[0038] The line-level / control center network includes: At the control center or line-level data center, a core switch and a server cluster are deployed, and a large-capacity storage device is equipped.

[0039] These servers carry the main functional modules of the energy management control platform (system application layer), including: Database and historical data storage: centrally save the real-time and historical energy consumption data, equipment status information, etc. from each station.

[0040] Visual monitoring and alarm: provide a graphical interface for operation and maintenance personnel to view the energy consumption and equipment status of the whole line in real time. If there is abnormal energy consumption or equipment failure, alarms can be automatically triggered.

[0041] Advanced scheduling and optimization algorithms: such as AC / DC hybrid power flow calculation, PPO reinforcement learning scheduling, phased optimization strategies, demand response, etc. Through the analysis of the data sent from each station, control instructions are sent down in real time.

[0042] When the control platform generates scheduling strategies (such as energy storage charge and discharge plans, adjustment of the threshold for regenerative braking energy recovery, optimization of station air conditioning and ventilation, etc.), these instructions are sent down through the subway network to the gateways or servers of each station, and then specifically executed by the station servers.

[0043] The terminal perception layer (site level) is mainly responsible for on-site data collection and preliminary monitoring, and realizes the real-time acquisition of energy consumption equipment, environmental parameters and passenger flow information through sensors and intelligent meters, and can perform local control.

[0044] The edge computing layer (station / line level) consists of intelligent gateways, switches, etc., and is responsible for data aggregation, transmission, filtering and some edge-side calculations (such as simple alarm judgments). At the same time, it can maintain two-way communication with the control center and distribute the instructions sent by the control center to each station device.

[0045] The system application layer (control center) deploys a core energy management system, including data storage, visualization, alarm management, scheduling optimization algorithms, etc., to realize comprehensive energy consumption analysis and intelligent scheduling across the whole line.

[0046] The present invention realizes basic functions, statistical analysis and energy-saving control based on the above architecture.

[0047] In terms of the basic functions, the system has functions such as data acquisition and storage, real-time monitoring, statistical analysis, query and display, report printing, user permission management, alarm indication, and software update. Through data acquisition and storage, the system can collect and store various energy data in real time; the real-time monitoring function can comprehensively track the equipment and energy usage; the statistical analysis function supports the comprehensive analysis of energy data to help users understand the energy consumption trend; the query and display function provides flexible data query and display, facilitating users to view historical and real-time energy consumption data; the report printing function can automatically generate various reports and provide necessary analysis results; user permission management ensures access control for different roles; the alarm indication issues an alarm when the energy consumption is abnormal to ensure timely response; the software update function ensures that the system remains in the latest state to adapt to changing requirements.

[0048] In terms of the statistical analysis, the system provides functions such as energy monitoring, energy supervision, energy monitoring, energy statistics, energy publicity, energy early warning and alarm, indicator management, and energy billing. The energy monitoring and supervision functions help to grasp the operation status of the system in real time and ensure the compliance of energy use. The energy monitoring supports accurate tracking of energy consumption and accurately calculates the energy consumption of each device. The energy statistics and publicity functions are convenient for displaying and analyzing energy consumption data and improving the transparency of the system. The energy early warning and alarm function can issue an alarm when the energy consumption exceeds the standard or the device fails to prevent unnecessary losses in advance. The indicator management function helps users set and monitor key energy efficiency indicators and provides data support for energy-saving goals. The energy billing module can calculate and manage electricity charges according to the energy usage to ensure the control of the system operation and maintenance costs.

[0049] In terms of the energy-saving control, the system has functions such as environmental control energy-saving, intelligent peak shaving, energy storage scheduling, intelligent lighting, expert system, load shifting, and algorithm update. The environmental control energy-saving module reduces energy consumption by optimizing the control of equipment such as air conditioners and ventilation. The intelligent peak shaving function adjusts the power consumption according to the load demand to avoid peak loads. The energy storage scheduling function improves energy efficiency by intelligently scheduling and managing the battery energy storage system. The intelligent lighting optimizes the operation of lighting equipment through intelligent control to reduce unnecessary energy consumption. The expert system provides intelligent decision-making support through big data analysis to help managers optimize energy use. The load shifting function can balance the load at different times and reduce electricity charges. The algorithm update function keeps the energy-saving algorithm continuously optimized to improve the energy-saving effect of the system.

[0050] Among them, the terminal perception layer collects the operation status of equipment, environmental parameters, and passenger flow data in real time through various intelligent meters and sensors installed on the site of the rail transit system (such as multi-functional electricity meters, temperature sensors, humidity sensors, energy consumption sensors, passenger flow counters, etc.). These devices are usually directly embedded in the field control unit or field acquisition terminal and have preliminary data monitoring and preprocessing functions.

[0051] The edge computing layer aggregates the data collected on-site through devices such as intelligent communication gateways and industrial switches according to standard communication protocols (such as MODBUS, OPCUA, etc.), and then uploads it to the central energy management and control platform via a private network channel. After the aggregated data is preliminarily processed by the edge device, if abnormal data (such as sudden increase in energy consumption or equipment failure) is detected, an early warning signal will be immediately triggered, and the alarm information will be transmitted to the system application layer.

[0052] In the system application layer at the control center, the data is stored in servers and memory libraries to form a system database, providing basic data for subsequent statistical analysis and scheduling decisions. The central monitoring software automatically generates alarm reports by comparing real-time data with preset thresholds, and displays abnormal states through a graphical interface, facilitating timely response and maintenance by operation and maintenance personnel. After data collection and storage, the energy management and control platform runs a statistical analysis module to comprehensively process historical and real-time data, generating energy consumption trend charts, comparative analysis reports, and various statistical reports. At the same time, the system provides a user-friendly query and display interface, allowing operators to query data and display results according to their needs. The report printing module can automatically output corresponding analysis reports to provide a basis for decision-making. The management software in the control center conducts hierarchical authority management for different users (such as on-site operators, dispatchers, administrators) to ensure that each role can only access and operate the data and control functions within their authority. The software update module ensures that the platform can regularly receive updates of the latest version of software and algorithms to adapt to changing operation requirements and technological progress, ensuring the long-term efficient and stable operation of the system.

[0053] As Figure 2 shown, in this embodiment, the data processing architecture can be further divided from bottom to top into: on-site equipment layer, data collection layer, data preprocessing layer, algorithm model layer, and intelligent regulation layer.

[0054] The on-site equipment layer includes: Flywheel energy storage, which is mainly used to quickly absorb or release energy, such as storing regenerative energy during train braking or releasing energy to assist power supply during peak electricity consumption.

[0055] Cooling system, which provides refrigeration function for air conditioners or other equipment with high heat dissipation requirements to maintain the safe and stable operation of the equipment.

[0056] Fresh air system, which controls the air circulation and quality in stations and tunnels, provides a comfortable environment and excludes harmful gases.

[0057] Photovoltaic. The distributed photovoltaic power generation system is a new energy power generation unit that converts solar energy into electrical energy to provide part of the power supply for the rail transit system.

[0058] Energy storage refers to the battery energy storage unit, which is used to charge during low valley electricity prices or when photovoltaic power generation is sufficient, and discharge during peak electricity prices or when the load suddenly increases, so as to smooth out power supply fluctuations.

[0059] The temperature and humidity meter detects the ambient temperature and humidity in real time, providing reference data for the regulation of air conditioning or ventilation systems.

[0060] The electricity meter monitors the power consumption or power generation at each key node, accurately measures the energy flow, and provides data support for subsequent energy consumption analysis and control strategies.

[0061] The underlying field devices are the physical basis for the system to obtain real-time data and execute control instructions. They are responsible for directly interacting with energy or the environment, providing inputs for upper-layer data collection and decision-making, and executing corresponding regulation measures.

[0062] The data collected by the data collection layer includes: Environmental data, outdoor environmental data: from temperature and humidity meters, fresh air systems, etc., reflecting the meteorological conditions inside and outside the station.

[0063] Equipment energy consumption data: collected by electricity meters, energy storage and flywheel equipment, recording equipment operating power, voltage, current, etc.

[0064] Air conditioning status data and air conditioning control data: including information such as air conditioning operation mode, temperature setting, fan speed, etc., providing key inputs for subsequent optimized scheduling.

[0065] The data collection layer collects the real-time status, energy consumption information, and environmental parameters of the underlying devices through the sensors and measuring devices of the field devices, and packages them into a standardized format, laying a data foundation for subsequent data preprocessing and algorithm analysis.

[0066] The data processing procedures processed by the data preprocessing layer include: Data cleaning: filtering out noise, duplicates, or outliers to ensure data quality.

[0067] Data conversion: converting data in different protocols or formats into a unified standard (such as unified timestamps, units, etc.) for subsequent analysis.

[0068] Detection (or "anomaly detection"): automatically identifying abnormal situations such as sudden increases in energy consumption, equipment failures, or sensor failures, and generating alarm messages.

[0069] Data completion: interpolating or inferring missing or discontinuous data to maintain data continuity as much as possible.

[0070] The data preprocessing layer is the "data quality gatekeeper". It cleans, normalizes, identifies and corrects multi-source heterogeneous data collected, providing highly reliable and consistent input data for the algorithm model, and avoiding scheduling decision-making mistakes caused by poor original data.

[0071] The algorithm models applied in the algorithm model layer include: Expert experience pool: Based on operation and maintenance experience and historical statistical information, it provides reference thresholds or parameter ranges (temperature alarm values, humidity alarm values, equipment loss changes, air-conditioning parameter adjustment ranges, etc.) for the system in boundary or special scenarios, such as upper / lower limits of air-conditioning temperature, humidity warning values, equipment life assessment, etc.

[0072] Photovoltaic power generation prediction module: Using historical power generation data, real-time meteorological information and photovoltaic power curves, it makes short-term predictions on photovoltaic power output, providing reference for energy storage and load scheduling.

[0073] Environmental control energy consumption prediction model: Combining indoor and outdoor environmental data, passenger flow changes, air-conditioning operating conditions, etc., it predicts the environmental control load requirements such as air-conditioning and ventilation in the future period, avoiding blind cooling / heating waste.

[0074] Environmental control energy consumption optimization algorithm: According to the prediction results and the set comfort range, it dynamically adjusts the operation strategies of equipment such as air-conditioning and fresh air systems, realizing the reduction of the energy consumption of the environmental control system and maintaining the comfort of passengers. Through the real-time monitoring of environmental parameters such as temperature, humidity, and air quality in the rail platform and carriage, as well as passenger flow changes, equipment operation status, and energy prices, it dynamically adjusts the operation strategies of environmental control equipment. This algorithm combines energy demand prediction and equipment performance optimization, reasonably allocates and utilizes energy resources on the premise of ensuring passenger comfort and safety, and reduces the overall energy consumption of the environmental control system. By optimizing the operation efficiency of equipment such as air-conditioning, ventilation, and exhaust, this algorithm effectively reduces the peak energy consumption during subway operation and makes full use of the energy resources during the valley period, realizing global optimal energy management, and finally achieving the goals of cost savings and operation efficiency improvement.

[0075] Energy storage intelligent peak shaving algorithm: For battery energy storage or flywheel energy storage devices, according to electricity prices, load predictions, and photovoltaic power generation prediction results, it formulates charge and discharge plans to cut the maximum load peak, profit from the peak-valley electricity price difference, and ensure system stability.

[0076] The algorithm model layer is the "core brain" of system decision-making. Through the collaborative work of various prediction models, optimization algorithms, and expert experience knowledge bases, it provides decision-making solutions for the upper-layer "intelligent regulation". This layer focuses on calculating and analyzing multi-source data to form scientific and accurate optimization strategies.

[0077] The functions realized by the intelligent regulation layer include: Air - conditioner energy saving: By combining environmental control energy consumption prediction and optimization algorithms, dynamically adjust the air - conditioner temperature setting, fan speed, etc., to achieve a balance between energy saving and comfort.

[0078] Ventilation energy saving: According to the indoor and outdoor air quality and passenger flow, control the opening degree or operation time of the fresh - air system to reduce the energy consumption of the ventilation system.

[0079] Photovoltaic power generation: Use the results of photovoltaic prediction to determine the grid - connection or energy - storage charging strategy of photovoltaic power generation, and maximize the self - use ratio of renewable energy.

[0080] Energy - storage peak shaving: Release the stored - energy electricity during peak electricity prices or peak loads to reduce the purchase of electricity from the grid; charge during off - peak or photovoltaic - surplus periods to smooth the system load fluctuations and save costs.

[0081] Flywheel energy - storage braking - energy feedback: Absorb, store and release the braking energy of the train at an appropriate time to further reduce the traction energy consumption.

[0082] The intelligent control layer implements the optimization strategies output by the algorithm model layer. Through the real - time scheduling of equipment such as air - conditioners, energy - storage devices, photovoltaics, and flywheel energy - storage, it dynamically matches the load demand and energy supply to achieve the goals of peak shaving and valley filling, energy consumption reduction, operation - cost reduction, and overall energy - efficiency improvement.

[0083] In this embodiment, the system data - processing process is from bottom to top: The bottom - layer devices operate in real - time and generate a large amount of data. The data - acquisition layer collects multi - source data. The data - preprocessing layer cleans, transforms, and performs anomaly detection on the data. The algorithm model layer conducts comprehensive analysis and decision - making based on the expert - experience pool, prediction models, and optimization algorithms. The intelligent control layer executes the scheduling strategies (such as air - conditioner, ventilation, photovoltaic, energy - storage, flywheel feedback, etc.), and real - time feedbacks the execution effects and updates the data to form a closed - loop, continuously iterating to improve the system energy efficiency. Through this five - layer structure, the present invention can achieve multi - objective collaborative optimization of renewable - energy utilization, energy - storage peak shaving, regenerative - braking energy recovery, and environmental control while ensuring the safe operation of rail transit, greatly reducing the operation energy consumption and electricity - cost, and enhancing the comprehensive energy efficiency and sustainable - development level of the urban rail - transit system.

[0084] Preferably, through the environmental control energy consumption prediction model and the environmental control energy consumption optimization algorithm, the BAS system conducts state search based on the operating status and duration of the equipment, and adjusts the control strategy according to the target state. The controller can adopt various operation modes, such as adjusting the chiller frequency, adjusting the supply water temperature, optimizing the pump frequency, etc. It judges through data analysis and preferentially executes the operations with greater potential, thereby improving the energy efficiency of the air conditioning system. The system conducts intelligent scheduling based on factors such as load data, equipment performance, and energy costs, avoids energy waste, and dynamically adjusts the operation mode of the air conditioning system according to the real-time indoor temperature and humidity data, enhances the autonomous operation ability of the equipment, and ensures the balance between energy conservation and comfort. The system can identify dangerous states, potential states, and feasible states, preferentially abandon dangerous states, concentrate computing resources on potential and feasible states, and gradually search for the optimal control path.

[0085] Preferably, the present invention uses a photovoltaic power generation system as a new energy power generation device. Multiple photovoltaic modules convert solar energy into direct current, and an inverter converts the direct current into alternating current for use in the rail transit system. The system combines a photovoltaic power generation prediction module, and estimates the future photovoltaic power generation in advance by combining meteorological data, historical power generation data, and load demand prediction. This prediction result provides support for power dispatching and energy storage management, ensuring the stability and reliability of power supply. The system is based on an energy storage intelligent peak shaving algorithm, adjusts the charge and discharge strategy of the battery energy storage system according to the change of photovoltaic power generation, stores the excess power for use during periods of insufficient sunlight, and dynamically adjusts the cooperation between the photovoltaic power generation and the energy storage system according to the real-time load demand. For example, during peak load periods, the photovoltaic system can preferentially charge the energy storage device to ensure that the energy storage system releases enough electric energy during peak power consumption periods to support the operation of the rail transit, thereby achieving load balance and optimizing power dispatching.

[0086] Preferably, the flywheel energy storage system is mainly used for energy recovery and dispatching in rail transit. The system adopts an energy storage intelligent peak shaving algorithm or a flywheel energy storage braking energy feedback algorithm. During the braking process of the train, the excess kinetic energy of the train is converted into the rotational energy of the flywheel for storage. When the train needs to accelerate or encounters a peak power demand, the flywheel energy storage system will quickly release the stored energy, reducing the dependence on the external power grid. Flywheel energy storage can provide high-power power output in a short time, has the characteristics of efficient energy storage and rapid release, and significantly improves the overall energy utilization rate of the traction system.

[0087] Preferably, the algorithm model layer can also implement the energy consumption monitoring function. Through a complete configuration of metering devices, the system comprehensively realizes the online monitoring of the energy consumption data of the entire rail transit system. The system supports accurate metering by category, item, and region, and can automatically generate energy consumption reports, providing detailed data analysis. It real-time monitors the main energy consumption links such as vehicle traction and regenerative braking systems, and conducts year-on-year, month-on-month, and horizontal comparative analyses to help management personnel understand the energy usage trends and promptly detect abnormal situations. The system has the function of automatically calculating and benchmarking energy consumption indicators, and triggers an alarm when exceeding the preset threshold to ensure the accuracy of energy efficiency management. Through the energy consumption prediction module, the system can predict future energy consumption based on historical data and real-time information, supporting the scientific formulation of quota indicators. The system also provides the function of peak-valley time-of-use metering, which can accurately distinguish the energy consumption differences in different time periods, helping to optimize energy usage and reduce costs. In addition, the system supports maximum power metering, cost accounting, cost sharing for transfer stations, and cost calculation for external units, comprehensively improving the energy management efficiency. Through intelligent data analysis and report generation, the system provides strong support for energy optimization and energy conservation and emission reduction, promoting the rail transit system to develop in a more efficient and sustainable direction.

[0088] Preferably, based on the energy storage intelligent peak shaving algorithm and the environmental control energy consumption optimization algorithm, this embodiment realizes the intelligent scheduling of the power supply system. By optimizing energy usage, it reduces the maximum power demand of the line, thereby reducing the basic electricity cost expenditure. The system comprehensively evaluates the benefits of photovoltaic power generation and the energy storage system, and reasonably arranges power generation and load through intelligent scheduling to further improve the economy of the system. The intelligent scheduling can shift part of the electricity load to the time period with a lower electricity price during the peak power demand period, thereby maximizing the electricity usage benefit and reducing the overall operation cost. The system also designs a load intelligent regulation function to real-time monitor the load conditions of each station and its three-phase power supply, ensuring reasonable load distribution and avoiding load overload. Through the dynamic detection of load imbalance, the system can actively issue adjustment suggestions to optimize power distribution, reduce the energy loss in the power supply system, and improve the power utilization rate. In addition, the power supply system is equipped with a load rate monitoring and warning function to real-time monitor the capacity utilization rate of equipment. Once it detects the operation of high-load equipment, the system will automatically issue a warning signal to help operation and maintenance personnel conduct inspections and maintenance in a timely manner, prevent equipment from overloading, and ensure the stability and reliability of the power supply system. Through these intelligent scheduling and monitoring functions, the power supply system can achieve more efficient and economic energy management.

[0089] As Figure 3 shown, this embodiment improves the energy utilization efficiency, reduces waste, and ensures the sustainable operation of the system through a phased optimization strategy. The algorithm can be divided into three development stages, gradually transitioning from local optimization to global optimization, and ultimately achieving a comprehensive improvement in the overall energy utilization rate.

[0090] In the local optimization stage, the system first collects operation data, energy consumption information, and environmental parameters (such as air-conditioning energy consumption, lighting energy consumption, equipment load, etc.) from various on-site devices and sensors. In the "Data Analysis" module, these raw data are cleaned, statistically analyzed, and visualized to help managers understand the current energy consumption profile of the system and identify high-energy-consuming devices or weak links. Based on the collected energy consumption data and equipment characteristics, digital models are established, including equipment energy consumption models, load prediction models, energy storage / new energy models, etc. In this stage, the system can use various algorithms (such as regression analysis, neural networks, expert systems, etc.) to construct a mathematical description of equipment energy consumption or performance, laying a foundation for subsequent energy-saving optimization and integrated scheduling. When the system optimizes a certain type of equipment or a certain site, it is necessary to evaluate the energy-saving effect and quantify the energy-saving benefits by comparing the energy consumption indicators before and after optimization. The evaluation results are fed back to the core algorithm layer to help update parameters or adjust strategies in the subsequent iteration process.

[0091] In the local optimization stage, for typical energy-consuming devices or links such as "lighting, chiller, composite energy storage, medium-voltage energy feed", the energy-saving algorithms are iterated respectively to find out their respective optimal energy-saving control strategies. In this link, the algorithm will conduct in-depth local optimization of specific equipment based on the aforementioned data and models (such as energy consumption prediction, equipment characteristic models).

[0092] The system will iterate the operating parameters (temperature setting, frequency adjustment, start-stop time, etc.) of the selected equipment type (such as chiller) multiple times: 1. Initial strategy: Based on expert experience or historical average strategy.

[0093] 2. Measure the energy-saving potential: Evaluate the energy consumption level of the current strategy through simulation or actual operation comparison.

[0094] 3. Update the strategy: Use digital models or heuristic algorithms to adjust control parameters. If the energy-saving effect is improved and the comfort / safety constraints are not violated, accept the new strategy.

[0095] 4. Loop iteration: Continuously repeat measurement and update to gradually approach the optimal or sub-optimal local energy consumption level.

[0096] When several equipment categories have respectively completed local energy-saving iterations and obtained relatively optimal strategies, the system will summarize these results to provide an optimized subsystem control scheme for the next layer of "comprehensive optimization", that is, global optimization.

[0097] Global optimization considers external or system-level dynamic factors such as "wind power and photovoltaic, electrochemical energy storage, real-time electricity price, meteorology, passenger flow", and focuses on system-level energy scheduling, including renewable energy access, energy storage device charge and discharge strategies, energy consumption management during passenger flow peaks, real-time electricity price response, etc.

[0098] The system combines the optimized strategies of local devices with real-time information on the external environment (such as new energy fluctuations, electricity prices, and passenger flows), and realizes the optimal allocation of overall energy through multi-objective or multi-stage scheduling algorithms. For example: Load balancing and peak shaving: Flexibly schedule energy storage devices and controllable loads at different times to avoid peak loads; Electricity price response: Minimize power purchases from the grid during high electricity price periods, or charge and store energy during low electricity price / photovoltaic surplus periods; Multi-objective optimization: Take into account energy conservation, economic benefits, equipment life, and system safety and stability.

[0099] The global optimization finally outputs a system-level energy-saving scheduling plan, covering the interactions between various devices and external energy sources: such as giving priority to supplying photovoltaic output to loads, storing the remaining power in the battery, discharging to subsidize loads when the electricity price is high, and moderately shifting the operation of air conditioners when the load is too high.

[0100] Furthermore, the first stage is the local optimization stage, which is the load selection and optimization based on the "barrel principle". The energy-saving algorithm determines the load type with the greatest energy-saving potential in the system through the "barrel principle", and preferentially selects those loads with the greatest energy-saving potential and impact for optimization, and takes local energy-saving measures. By detailed evaluation and priority ranking of load types, the system can focus on high-energy-consuming devices or weak links in the system, and adopt an iterative optimization algorithm to gradually improve local energy efficiency.

[0101] In this embodiment, the optimization goal of the first stage is to optimize the energy consumption of a single device or a single site and reduce its energy consumption. Mathematically, it can be expressed as minimizing the total energy consumption of a certain device within a given period. For example, for high-energy-consuming devices such as air conditioners, it is desired to minimize the integral of energy consumption E: ; where, where P 设备 represents the power consumption of the device, which depends on the control decision u(t) (such as the air conditioner temperature setting value, the fan speed, etc.). t represents the time, and T represents the control period. By optimizing the control decision u(t), the total energy consumption E is minimized.

[0102] In local optimization, constraint conditions such as comfort and equipment operation safety need to be satisfied, including but not limited to: Environmental comfort constraint: For example, the air conditioner temperature in the station needs to be maintained within the comfortable range for passengers. This ensures energy conservation without affecting passenger comfort.

[0103] Equipment safety constraint: Ensure that the equipment operates within the rated safety range. For example, the start-stop frequency of the air conditioner compressor cannot be too high, and the motor cannot be overloaded, to avoid shortening the equipment life or causing failures.

[0104] Therefore, the above optimization problem is essentially a constrained minimization problem. We can introduce Lagrange multipliers or penalty function methods to handle the constraints and ensure that the comfort and safety requirements are not violated during the optimization process.

[0105] In the local optimization stage, the system identifies the load type with the highest energy consumption and the greatest energy-saving potential through energy consumption data analysis and the expert experience database (this reflects the "barrel principle", giving priority to solving the short board). Once the target load is determined, such as the air conditioning system or lighting system, a dedicated optimization algorithm for that load is used for optimization.

[0106] For example, for the environmental control air conditioning system, an air conditioning dynamics model (i.e., an environmental control energy consumption prediction model) is established to predict future temperature and load. The environmental control energy consumption prediction model uses indoor and outdoor environmental data, passenger flow information, and equipment operation status, etc., to predict the demand for environmental control loads such as air conditioning and ventilation in the future for a period of time. For example, a model can be constructed: ; where T(t), H(t), Q(t), and N(t) represent temperature, humidity, air quality, and passenger flow data respectively.

[0107] The environmental control energy consumption prediction model can be a machine learning / regression model to predict the energy consumption of air conditioners, etc. according to environmental conditions. Its input variables include: historical environmental control system energy consumption data; environmental parameters (outdoor temperature and humidity, station passenger flow, etc.); current air conditioning / ventilation equipment status settings. The model fits historical data to map the relationship between environmental changes and energy consumption, providing a decision-making reference.

[0108] The environmental control energy consumption optimization algorithm then dynamically adjusts control parameters (such as chiller frequency, water supply temperature, etc.) based on the prediction results to minimize energy consumption while ensuring passenger comfort. Its optimization objective can be written as: ; The constraint conditions include: comfort (such as belonging to a specific temperature and humidity range), equipment operation safety (preventing overload and frequent start-stop), meeting the basic operation requirements of the actual load demand, air quality, and safety constraints: ventilation and air change meeting the standards.

[0109] The expert experience database provides initial strategies and adjustment rules (such as adjusting the temperature setting value and start-stop time according to historical data and expert knowledge), and then iteratively optimizes in combination with real-time data. Through multiple iterations of optimization, the energy efficiency of the equipment is gradually improved. After one equipment is optimized, the same process is continued for the next high-energy-consuming equipment to gradually reduce the energy consumption of each local load.

[0110] Taking the air conditioning system of a subway station as an example, energy consumption analysis reveals that the power consumption of the air conditioning refrigeration unit accounts for 40% of the total station energy consumption, presenting the largest energy-saving potential. Through expert experience, the comfortable temperature for passengers is determined to be around 26°C. The system adopts an environmental control energy consumption optimization algorithm to adjust the operation of the air conditioner: when the passenger flow is low, the supply air temperature is appropriately increased by 1°C (ensuring that the station temperature is between 5 - 27°C) to reduce the compressor output; before the peak passenger flow, the concourse is pre-cooled in advance to avoid excessive cooling energy consumption during the peak period. After optimization, the daily power consumption of the station's air conditioner has decreased by approximately 10%. During this process, the temperature always meets the comfort standard (for example, remaining around 26°C), and the equipment also operates within the safe range without overloading. Thus, it can be seen that the local optimization algorithm achieves energy-saving effects through precise regulation of individual equipment.

[0111] Furthermore, the second stage is overall optimization, which improves the load control strategy and enhances energy efficiency. During the continuous improvement of various load optimization algorithms, the system accumulates a large amount of data and experience, gradually improving the overall energy utilization level. Through global scheduling and resource optimization, considering the demands of different load types comprehensively, the energy use efficiency of the entire system is enhanced. The system not only optimizes the use of a single load but also can intelligently schedule the energy distribution among various loads, balancing the energy demands of different regions and equipment, thus effectively avoiding excessive energy consumption and waste and achieving more efficient energy utilization.

[0112] The main task of the overall optimization stage is to coordinately control the loads in multiple stations or regions to achieve the balance of energy resources among various loads, thereby improving the overall energy utilization efficiency of the entire network. This stage focuses on energy distribution and scheduling, ensuring that the local optimization results can be coordinated within the system, without particularly emphasizing issues such as electricity price response or peak shaving and valley filling (these contents will be introduced in the global optimization stage).

[0113] Through the photovoltaic power generation prediction model, the photovoltaic output in the future period is estimated, providing data support for the energy distribution among various loads in the region, so that the system can make full use of renewable energy when allocating energy.

[0114] The photovoltaic power generation prediction model is trained using historical power and meteorological factors (such as time series models or neural networks), and the predicted power is output by inputting future weather forecasts. Its input variables can include: historical photovoltaic power generation data; meteorological data (future solar irradiance, temperature, etc.); time (sunshine duration, season). The predicted photovoltaic power value output by the photovoltaic power generation prediction model must not be negative and shall not exceed the installed capacity; the reliability under weather uncertainty needs to be considered (to avoid overestimating power generation).

[0115] Using the data and prediction results obtained in each local optimization stage, summarize the energy consumption requirements and equipment operation status of each station. The optimization objective can be set to minimize the energy consumption fluctuation of the entire network or maximize the utilization rate of new energy. In this stage, linear programming or heuristic scheduling algorithms are mainly used to balance the energy consumption distribution in each region through resource optimization, so as to smooth the load of the entire network. Since the dynamic electricity price factor is not considered in this stage, the scheduling process is mainly based on photovoltaic power generation prediction, load prediction and historical data distribution to determine the energy allocation ratio of each region.

[0116] The inputs of the optimization objective function in this stage include: the load data after local optimization of each station, historical statistical data, and local prediction results (such as energy consumption prediction of equipment such as air conditioners, lighting, and ventilation, and photovoltaic power generation prediction).

[0117] The outputs of the solution of the objective function include: the overall network scheduling plan, that is, the energy allocation plan for each region, the smoothed load curve, and the adjustment suggestions for the control parameters of each load.

[0118] The constraint conditions include: Energy balance constraint: Ensure the balance of supply and demand at each moment in the entire network; Equipment capacity constraint: The power regulation of each equipment cannot exceed its maximum bearing range; Comfort and safety: Ensure that the load balance is carried out on the premise of not affecting the comfort of passengers and the safe operation of equipment.

[0119] Furthermore, the third stage is global optimization to achieve system performance energy management under the coupling of multiple factors. The algorithm considers more complex multi-factor coupling relationships, including the volatility of new energy power generation, the safety and life of electrochemical energy storage, real-time electricity price, real-time passenger flow, and meteorological factors, etc. The power generation of new energy has strong randomness and volatility, and the safety and service life of the electrochemical energy storage system also need to be comprehensively considered. Combining real-time electricity price, meteorological and passenger flow data, the system can intelligently adjust the energy use strategy to achieve optimal resource allocation, thereby improving the overall comprehensive utilization rate of energy and ensuring the stability and sustainability of energy supply.

[0120] The global optimization stage requires an accurate understanding of the operation status of the entire power supply network (such as the voltage, current and power distribution of each node). Figure 4The urban rail transit power supply system applied in this embodiment is shown. The subway power supply system can be divided into two parts: an AC power supply network and a DC traction network. The AC power supply network mainly includes switchgear stations, traction step-down substations, step-down substations, photovoltaic power generation, and battery energy storage. The switchgear station introduces medium-voltage AC power from the previous-stage medium-voltage AC substation. Usually, one switchgear station is arranged at each head and tail of each line. Each station is equipped with a traction step-down substation or a step-down substation. The traction step-down substation includes a traction substation and a step-down substation. The traction substation rectifies the medium-voltage AC electrical energy into DC traction network voltage through a rectifier transformer and a rectifier device, and then supplies power to the DC traction network. The step-down substation steps down the medium-voltage AC electrical energy to low-voltage AC through a transformer and supplies power to low-voltage loads such as lighting, elevators, and air conditioners. The flywheel energy storage is connected to the DC traction network. It can be seen that the urban rail transit power supply system has typical distributed characteristics.

[0121] Figures 5a - 5b This is an equivalent circuit model of a DC traction power supply system for urban rail transit provided in this embodiment. As Figure 5a shown, the rectifier unit can be equivalent to a voltage source with unidirectional current flow, where U s is the no-load voltage of the rectifier unit, i s is the output current, R s is the equivalent resistance of the rectifier unit; the flywheel energy storage is equivalent to two voltage sources with unidirectional current flow, U ce is the charging threshold, R ce is the charging resistance, U de is the discharging threshold, R de is the discharging resistance; the train can be equivalent to a current source, I t is the train current, U brk is the braking resistance action threshold, R as is the equivalent resistance of auxiliary energy consumption. Figure 5b This is the overall model of the urban rail traction power supply system. The rectifier unit is connected in parallel with the flywheel energy storage to form a simplified model of the substation. The magnitudes of the variable resistances of the up (down) line and the rail resistance depend on the distances between the up (down) line trains and the substation.

[0122] When establishing the hybrid power flow calculation model (the process shown in Figure 6 ), Figure 4 and Figure 5 provide the necessary physical parameters and structural information, thus supporting global energy scheduling decisions and ensuring the effective implementation of strategies such as energy regulation and peak shaving and valley filling across the entire network.

[0123] As Figure 6 shown, the power flow calculation method applied to the AC-DC hybrid of urban rail transit adopted in this embodiment includes the following steps: Step A: According to the port characteristics of key equipment such as rectifier units, trains, and flywheel energy storage, their equivalent circuit models are constructed respectively. Using the modular modeling method, a complete power flow calculation model for the DC traction power supply system is established for the power flow calculation model of the DC traction power supply system.

[0124] Step B: According to the AC side network of the line power supply system and specific parameters, form the nodal admittance matrix Y; assume that the initial voltage of each node is , then the real and imaginary parts of the voltage respectively satisfy: ; In the line power supply system, the photovoltaic power generation system is configured as a PQ node, and the power value of this node is determined according to the current photovoltaic power generation state; the battery energy storage system is a PQ node, and the power value of this node is confirmed according to the instruction obtained from the dispatching algorithm.

[0125] Step C: At t = 0, assume the traction voltage is , then the traction current satisfies the formula: ; where n is the number of substations.

[0126] At subsequent times, use the traction voltage obtained from the previous round of calculation as .

[0127] According to the power flow calculation model of the DC traction power supply system, based on the initially set output voltage of the rectifier unit at the initial moment, and using the output voltage of the rectifier unit updated by the previous round of calculation at subsequent times, calculate the traction voltage as , then the voltage correction amount is: .

[0128] Repeat the calculation of the traction voltage until the following formula is satisfied to make the DC part of the power flow calculation model converge:

[0129] .

[0130] Step D: According to the output voltage and output current of each rectifier unit, calculate the output power of each site's rectifier unit: ; where, at the initial moment, the output voltage of the rectifier unit adopts the corresponding initially set value, and at subsequent times, it adopts the output voltage of the rectifier unit updated by the previous round of calculation. The output current of the rectifier unit is obtained through the currently converged power flow calculation model analysis.

[0131] The AC-side output power of the rectifier unit in the DC traction substation satisfies: ; where α is the commutation angle.

[0132] Step E: According to the voltage correction equation of the Newton-Raphson method: ; P i ′ and Q i ′ usually represent the active power imbalance and reactive power imbalance of node ii.

[0133] Calculate △P AC(i) 、△Q AC(i) 、△U AC(i) where G ij is the real part of Y ij in the admittance matrix, and B ij is the imaginary part of Y ij in the admittance matrix, satisfying: Y ij =G ij +jB ij .

[0134] Calculate each element in the Jacobian matrix (H ij 、N ij 、J ij 、L ij 、R ij 、S ij ), when i≠j: .

[0135] According to the formula ; Find the voltage correction amount, and the new node voltage satisfies the formula: ; Repeat this step until the voltage correction amount satisfies the convergence formula: .

[0136] Step F: Based on the power flow calculation results of the node voltages in the AC part of the DC traction power supply system and the voltage at the DC side of the rectifier, calculate the rectifier output voltage. In this embodiment, the rectifier output voltage formula is: ; where R sw is the commutation resistance, satisfying the formula: ; where L sw is the commutation inductance.

[0137] Recalculate the traction voltage iteratively according to the output voltage of the rectifier unit and the traction current until the following convergence condition is met: .

[0138] Calculate the output power of the rectifier unit at each station at this time and the power change . The power change satisfies the formula: ; where the calculation results of step D are adopted

[0139] Repeat steps D - F until the power change satisfies the following convergence condition: .

[0140] Step G: Use the traction voltage and node voltage obtained from the power flow calculation at the previous moment as the initial values for the power flow calculation at the next time step, and repeat steps C - F to perform the power flow calculation until the calculation is completed

[0141] In the global optimization stage, based on the overall scheduling, various dynamic factors are further comprehensively considered, such as the random fluctuations of new - energy power generation, the safety and lifespan of electrochemical energy storage, real - time electricity prices, passenger flow, and meteorological conditions. Its goal is to achieve the optimal energy scheduling within the entire system through multi - factor coupling, so as to minimize the grid power purchase cost and maximize the economic benefits

[0142] In this stage, the system needs to comprehensively consider various dynamic factors (such as new - energy output fluctuations, load changes, energy - storage charge - discharge characteristics, real - time electricity prices, passenger flow, and meteorology) to formulate an optimal scheduling strategy for the entire system. The AC - DC hybrid power flow calculation method provides accurate power grid operation state data for global scheduling decisions by constructing a comprehensive model of the AC and DC traction networks and calculating the voltage, current, and power distribution of each key node in real - time. These data are important bases for ensuring that the system achieves supply - demand balance, reduces the grid power purchase cost, and guarantees the stable operation of the system during peak - shaving and valley - filling, energy - storage scheduling, and multi - factor coupling optimization. Therefore, as part of the algorithm model layer, this power flow calculation method mainly supports the global optimization stage

[0143] As Figure 7 shown, this stage adopts the Proximal Policy Optimization (PPO) scheduling algorithm (i.e., the energy - storage intelligent peak - shaving algorithm): construct a Markov decision process (MDP), define the state s t including real - time load, photovoltaic power generation, energy - storage state of charge (SOC), electricity price, passenger flow, and meteorological information; the action a t is the charge - discharge power of the energy - storage devices (batteries and flywheels). The immediate reward function is designed as: ; Where P grid (t) is the power purchased from the power grid. When it is greater than 0, it means purchasing power from the power grid, and when it is less than 0, it means selling power to the power grid. π(t) is the time-of-use electricity price. This design can ensure that the system can make corresponding decisions according to the electricity price in different situations: encourage charging (purchasing power) when the electricity price is low, and encourage discharging (selling power) when the electricity price is high. Therefore, the key lies in ensuring that the action space and load data of the system allow negative values, so that the behavior of selling power can be reflected in the mathematical model.

[0144] The goal of reinforcement learning is to maximize the cumulative discounted reward: .

[0145] The PPO algorithm updates the policy parameters through a "clip" mechanism, and its objective function is: ; Where is the policy ratio, and A is the advantage function.

[0146] In the reinforcement learning framework, the photovoltaic prediction results, load prediction, energy storage state, and external electricity price signal are used as inputs to ensure that the intelligent agent can adjust the energy storage charging and discharging strategy according to the real-time state, so as to realize peak shaving and valley filling and maximize economic benefits.

[0147] Specifically, the inputs of the reinforcement learning (PPO) scheduling algorithm include: real-time load data, photovoltaic power generation prediction, SOC of energy storage equipment, electricity price information, passenger flow data, and meteorological data. Its outputs include: the energy storage scheduling decision optimized by reinforcement learning, specifically the charging and discharging power values of the energy storage equipment at each moment, and then determine the minimized scheduling scheme for the whole network to purchase / sell electricity from the power grid. Its constraint conditions include: Energy storage equipment constraints: For example, the battery SOC needs to be kept within a certain range, and the charging and discharging power does not exceed the rated value of the equipment; Energy balance: Ensure the balance of supply and demand within each time period. Ensure that the load = grid power supply + energy storage discharge - energy storage charge + photovoltaic power generation at any moment, and peak shaving does not affect the power supply of critical loads.

[0148] Multi-objective balance: Introduce penalty terms for the energy storage life and safety in the reward function to prevent the strategy from over-consuming the energy storage equipment.

[0149] To sum up, in the local optimization stage, expert experience, environmental control energy consumption prediction, and optimization algorithms are adopted to perform local regulation on a single load (such as an air conditioning system) to reduce equipment energy consumption and meet comfort and safety constraints.

[0150] The overall optimization stage integrates the results of local optimization. Through data fusion and resource scheduling algorithms, it balances the energy consumption of each station across the entire network, achieving load balance and improving the overall energy utilization efficiency. This stage does not focus on price response or peak shaving control, but only emphasizes the optimization of energy distribution.

[0151] The global optimization stage introduces multi-factors (such as electricity price, passenger flow, meteorology, etc.) and the Proximal Policy Optimization (PPO) reinforcement learning scheduling algorithm to globally optimize the charge and discharge strategies of energy storage devices, achieving peak shaving and valley filling and minimizing the electricity purchase cost. This stage comprehensively considers energy storage safety, device lifespan, and economic benefits to construct a globally optimal scheduling strategy.

[0152] Figure 7 Further presented is the solution process of an energy storage scheduling algorithm based on the Proximal Policy Optimization (PPO) reinforcement learning framework. It takes "flywheel energy storage power" and "battery energy storage power" as the main action outputs and updates the policy through the "Actor-Critic" structure. The working principle is described in steps as follows: 1. Environment Input: System states, such as the State of Charge (SOC) of the flywheel and battery, current load demand, real-time electricity price, passenger flow information, meteorological conditions, etc.

[0153] Output: After the algorithm selects a set of energy storage charge and discharge actions, the environment will feedback the new state (updated SOC, load, etc.) and the immediate reward (based on the electricity purchase cost or economic benefit of the power grid).

[0154] The environment is extended through the "considering multi-factor characteristics (battery lifespan, real-time electricity price, meteorological information)" module to ensure that the scheduling decision comprehensively reflects external uncertainties.

[0155] 2. Actor_new and Actor_old The Actor network is responsible for outputting actions based on the current state. Here, the actions are {flywhee-power, battery-power}, that is, the charge and discharge power of the flywheel energy storage and the battery energy storage.

[0156] "Actor_new" represents the updated policy network in the current iteration, and "Actor_old" represents the policy network in the previous iteration. The PPO algorithm will first fix the old policy (Actor_old) in each training round, then interact with the environment to collect data, and finally update the new policy (Actor_new) based on these data in the optimization stage.

[0157] 3. Critic network (Critic_new) The Critic network is used to evaluate the value of the current state or state-action pair (Value Function or Q Function) to assist in calculating the advantage function (Advantage).

[0158] In PPO, the Critic network is updated based on the interaction data between the old and new policies, outputs an estimate of the environmental return, and provides an evaluation basis for the policy gradient.

[0159] 4. Sampling and Data Collection The algorithm uses Actor_old to interact with the environment for a period of time (one episode or multiple time steps) to generate a batch of quadruples of state, action, reward, and next state.

[0160] These interaction data (or "experiences") are temporarily stored and then used together to calculate the policy gradient and value error after the episode ends.

[0161] 5. Mini-batch Sampling The collected data is divided into mini-batches to facilitate the stability and efficiency of gradient calculation. Each mini-batch contains several samples of "(state, action, reward, next state)".

[0162] 6. Calculating the Advantage Function The advantage function A(s,a) reflects the quality of a certain action a relative to the average policy action, and typical calculation methods such as Generalized Advantage Estimation (GAE).

[0163] 7. Policy Update: PPO Clipping Mechanism In PPO, by comparing the probability ratio of the new policy and the old policy and restricting it within the range of [1−ϵ,1+ϵ], it is to avoid the instability caused by too large a policy update amplitude.

[0164] Loss Function: .

[0165] By performing gradient descent on this objective, the updated policy parameters θnew are obtained. The Critic network is also updated synchronously to better evaluate the value function.

[0166] 8. Updating Actor_new and Critic_new According to the calculated gradient, the network parameters of Actor_new and Critic_new are updated by gradient descent. After completion, Actor_new will replace Actor_old and become the policy network for the next round of interaction.

[0167] 9. Iterative Loop Repeat the above process: Use Actor_old to interact with the environment to collect data; Calculate advantages and value errors based on batch data; Use the PPO clipping mechanism to update Actor_new and Critic_new; Replace Actor_old with Actor_new and enter the next round of interaction.

[0168] As the number of training rounds increases, the energy storage scheduling strategy is continuously optimized. The agent gradually learns when to charge, when to discharge, and how much to discharge under different electricity prices, loads, and SOCs, so as to maximize the overall economic benefits.

[0169] Through the above three stages, the system gradually expands from local device optimization to whole-network coordination, and finally introduces dynamic external factors to achieve optimal scheduling, forming a closed-loop energy management system. The stages are interconnected: local optimization provides basic data and preliminary results for overall scheduling, overall optimization ensures the balanced energy consumption distribution in each region, and global optimization further combines external factors (such as real-time electricity prices) on this basis to achieve the optimization of economic benefits.

[0170] This phased optimization strategy not only enables the system to achieve energy conservation at the single load level, but also realizes the efficient, intelligent, and economic operation of the entire rail transit power supply system through whole-network scheduling and multi-factor coupling.

[0171] The content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

Claims

1. An urban rail transit energy management system considering the access of new energy and energy storage, characterized in that: Including: A new energy power generation unit, an energy storage unit, a regenerative braking energy recovery unit, and an energy management and control platform; the new energy power generation unit and the energy storage unit are connected to the rail transit power supply system; the energy management and control platform is connected to the power supply system and communicates to control the above units, monitors, analyzes and optimizes the scheduling of the energy flow in the rail transit power supply system to coordinate the new energy power generation, energy storage charge and discharge, and the utilization of regenerative braking energy, improve the energy utilization efficiency and reduce the demand for externally purchased electric energy from the power grid; The energy management and control platform establishes a power flow calculation model including an AC distribution network and a DC traction network, and uses this model to perform power flow analysis on the rail transit power supply system with new energy and energy storage access, providing a decision-making basis for energy scheduling; The process of the power flow analysis includes: (a) Initialize the parameters of each key device according to preset parameters, and set the initial values of the output voltage of the rectifier unit, the DC traction voltage, and the AC side node voltage; (b) Form a node admittance matrix Y based on the AC distribution network parameters of the rail transit power supply system, and calculate the AC node voltages; (c) Based on the initially set output voltage of the rectifier unit at the initial moment, and using the output voltage of the rectifier unit updated by the previous round of calculation at subsequent moments, iteratively calculate the traction voltage in the DC traction network until the change in the traction voltage between two consecutive iteration results meets a preset first convergence condition, so that the DC traction network part of the power flow calculation model converges; (d) Based on the initially set output voltage of the rectifier unit at the initial moment, and using the output voltage of the rectifier unit updated by the previous round of calculation at subsequent moments, calculate the initial output power of the rectifier unit based on the power flow calculation model; (e) Update the output voltage of the rectifier based on the power flow calculation results of the AC side node voltage and the DC side voltage of the rectifier; Based on the updated output voltage of the rectifier, perform iterative calculation on the traction voltage in the DC traction network until the change in the traction voltage between two consecutive iteration results meets the preset convergence condition, so that the DC traction network part of the power flow calculation model converges; Calculate the output power of the rectifier unit in the current round according to the converged power flow calculation model and the updated output voltage of the rectifier; Compare the output power of the rectifier unit calculated in the current round with the initial output power recorded in step (d). If the difference meets the preset power convergence condition, it is considered that the power flow calculation at this moment is completed; otherwise, return to step (d) to repeat the above iterative process until the convergence condition is met; (f) When both the traction voltage and the output power of the rectifier unit at this moment reach the preset convergence conditions, use the traction voltage and AC node voltage converged at the current moment as the initial values for the power flow calculation at the next moment, and repeat steps (c) to (e) until the entire power flow analysis is completed.

2. The system according to claim 1, wherein The data network architecture of the energy management and control platform adopts a three-layer architecture design, including: The terminal perception layer collects device operation data, environmental data, and passenger flow data through various on-site intelligent meters and sensors, and realizes data monitoring; The edge computing layer uses intelligent communication gateways and switches to send the data collected by the terminal to the energy management control platform through a dedicated network channel according to standard protocols, and sends the control commands of the platform to each control unit; The system application layer, with the energy management control platform as the core, realizes the storage of all subway station data, the display of equipment operating status, and the issuance of control instructions.

3. The system according to claim 1, characterized in that, The control management architecture of the energy management control platform adopts a three-level hierarchical architecture, including: At the field level, equipment operating conditions and environmental data are collected through field instruments and sensors, and local control is performed; At the station level, monitor, statistically analyze and optimize the energy consumption within the station; At the line level, a control center is set up to centrally monitor and make dispatch decisions on the energy of the entire line, and issue control instructions to each station level to achieve energy optimization and coordinated control across the entire line.

4. The system according to claim 1, wherein The energy management control platform is used to predict the output power of the new energy power generation unit and optimize the charging and discharging scheduling of the energy storage unit according to the prediction result to smooth the fluctuation of the new energy power generation output.

5. The system according to claim 1, characterized in that, The energy management control platform is used to reduce the peak power load of the rail transit system through demand management, thereby reducing the basic electricity fee, and adjust the charging and discharging strategy of the energy storage unit according to the time-of-use electricity price, storing electricity during off-peak hours and releasing electricity during peak hours, so as to reduce the overall electricity cost.

6. The system according to claim 1, wherein The energy management control platform is used to dynamically adjust the working threshold of the regenerative braking energy recovery unit based on the real-time power flow status of the traction power supply network to ensure line energy balance and maximize the utilization rate of braking energy recovery.

7. The system according to claim 1, wherein The energy storage unit includes a battery energy storage subunit and a flywheel energy storage subunit; the energy management control platform adopts a reinforcement learning algorithm optimized by a proximal strategy to perform intelligent scheduling and control on the energy storage unit, and dynamically adjusts the charging and discharging power of the battery and flywheel according to real-time load, new energy power generation and electricity price information to maximize economic benefits and minimize the amount of electricity purchased from the power grid.

8. The system according to claim 1, wherein The energy management control platform is also used to collect station environmental parameters and passenger flow information, and adjust the operation of station air-conditioning and ventilation related environmental control loads according to changes in electricity prices, so as to reduce the energy consumption of the environmental control system while ensuring passenger comfort.

9. The system according to claim 1, characterized in that The energy management platform adopts a phased optimization control strategy, including: a local optimization stage that prioritizes the identification and optimization of energy-consuming equipment with the greatest energy-saving potential, followed by an overall optimization stage that coordinates the energy efficiency of each station, and a global optimization stage that comprehensively considers fluctuations in new energy output, changes in electricity prices, passenger flow, and energy storage life; the control strategies generated by the optimization in each stage are iteratively superimposed step by step to gradually approach the optimal solution for the energy utilization efficiency of the entire system.

Images