A wind-solar-hydrogen storage micro-grid optimal scheduling method suitable for a traction power supply system in a plateau region

Abstract

This invention discloses an optimized scheduling method for wind-solar-hydrogen-storage microgrids suitable for traction power supply systems in plateau regions. The method includes: establishing a wind-solar-hydrogen-storage microgrid model; establishing a grid interaction model; establishing a traction load model for plateau regions; in renewable energy-rich sections, calculating the output of renewable energy sources using the wind-solar-hydrogen-storage microgrid model, thereby utilizing the microgrid to supply power to the traction loads in plateau regions; in renewable energy-scarce sections, prioritizing the use of the wind-solar-hydrogen-storage microgrid for power supply, with the remaining power calculated using the grid interaction model to supplement the traction loads; determining the optimized scheduling objective of the wind-solar-hydrogen-storage microgrid; establishing and optimizing a multi-stage microgrid optimized scheduling framework to determine the optimal optimized scheduling scheme for the wind-solar-hydrogen-storage microgrid. This invention can formulate differentiated operation plans based on the resource characteristics and environmental conditions of different sections along railway lines in plateau regions.

Description

Technical Field

[0001] This invention belongs to the field of power system dispatching and control technology, specifically relating to an optimized dispatching method for wind-solar-hydrogen-storage microgrids suitable for traction power supply systems in plateau regions. Background Technology

[0002] The high altitude, low air pressure, strong radiation, large diurnal temperature range, and weak power grid structure of plateau regions pose severe challenges to the construction and operation of power systems. On the one hand, traditional centralized power grids cannot fully cover remote plateau areas, and transmission line construction is costly, loss-prone, and difficult to maintain. On the other hand, traction loads in plateau regions are intermittent and impulsive, making it difficult to balance power supply and demand in the context of large-scale renewable energy power supply. Microgrid systems centered on wind, solar, hydrogen, and energy storage, due to their flexible networking, local consumption, and seamless switching between grid connection and off-grid operation, have become a key technological path to improve energy self-sufficiency in plateau regions, promote the efficient use of clean energy, and ensure reliable power supply to critical loads.

[0003] In the field of traction power supply in high-altitude areas, traction loads such as electrified railways, highway tunnel ventilation, and important military facilities exhibit significant impact, volatility, and high reliability requirements. These loads experience drastic power fluctuations and are extremely sensitive to voltage and frequency stability, placing immense pressure on the stable operation of microgrids. Especially in scenarios with highly random wind and solar power output, complex dynamic responses of hydrogen energy storage systems, and numerous system safety constraints, designing optimized scheduling strategies that balance economy, reliability, and cleanliness has become a critical issue that urgently needs to be addressed in current research and practice.

[0004] Currently, research on microgrid optimal dispatch has made some progress. Existing technologies can be mainly divided into three categories: First, deterministic optimization methods based on traditional mathematical programming, such as mixed integer linear programming (MILP) and dynamic programming, which focus on economic dispatch at a single time scale, either day-ahead or intraday; second, uncertain optimization methods based on robust optimization or stochastic programming, which improve the robustness of dispatch schemes by considering uncertain factors such as wind and solar forecasting errors and load fluctuations; and third, intelligent optimization methods based on artificial intelligence, such as reinforcement learning and deep learning, which attempt to learn the optimal dispatch strategy in complex environments from data.

[0005] Existing research mainly focuses on forecasting wind and solar power output and load demand under conventional climate conditions. With the optimization objective of system operation economy or renewable energy absorption rate, it uses methods such as mixed integer linear programming, stochastic programming, robust optimization or model predictive control to construct microgrid day-ahead or real-time dispatch models. Through coordinated control of energy storage system charging and discharging, controllable unit output and grid interaction power, microgrid operation optimization is achieved.

[0006] The existing technology has the following shortcomings:

[0007] (1) Lack of segment-differentiated operation strategy: Existing research focuses on plains or relatively stable environments. However, the geographical and climatic conditions along the railway in plateau areas vary greatly. The wind energy, photovoltaic resources and equipment operating environments of different segments are different. Existing research lacks differentiated operation plans based on the resource characteristics and environmental conditions of each segment. (2) Neglecting the spatial non-uniform distribution characteristics: In the microgrid dispatching of traction power supply systems in plateau areas, existing technologies usually adopt centralized or single-segment independent optimization models, which fail to fully consider the spatial non-uniform distribution characteristics of renewable energy installed capacity and load demand caused by geographical and resource endowment differences among multiple segments along the traction power supply network. (3) High-dimensional strong coupling real-time decision bottleneck: Plateau traction power supply scenarios have high-dimensional, continuous, non-convex, and strong coupling decision characteristics. When dealing with such complex problems, traditional optimization algorithms or basic reinforcement learning algorithms may fall into the dimensionality curse, converge to suboptimal solutions, or have unstable training. Therefore, the dispatching system needs to have real-time response capabilities and be able to flexibly adjust equipment output according to actual conditions to ensure that it can cope with the fluctuations of renewable energy and sudden load demands. Summary of the Invention

[0008] Purpose of the invention: The purpose of this invention is to provide an optimized scheduling method for wind-solar-hydrogen-storage microgrids suitable for traction power supply systems in plateau areas, which can formulate differentiated operation plans based on the resource characteristics and environmental conditions of different sections along railway lines in plateau areas.

[0009] Technical solution: The present invention provides an optimized scheduling method for wind-solar-hydrogen-storage microgrids suitable for traction power supply systems in plateau areas, comprising:

[0010] Establish wind power output model, photovoltaic power output model, hydrogen energy storage system model, and battery energy storage system model. The wind power output model, photovoltaic power output model, hydrogen energy storage system model, and battery energy storage system model together constitute a wind-solar-hydrogen-storage microgrid model.

[0011] A grid interaction model is established by combining the wind-solar-hydrogen-storage microgrid model;

[0012] A traction load model for plateau regions was established, and the power required for traction load was calculated using this model.

[0013] Based on the calculated power required for the traction load, in renewable energy-rich areas, the output of renewable energy is directly calculated using a wind-solar-hydrogen-storage microgrid model, thereby using the wind-solar-hydrogen-storage microgrid to supply power to the traction load in the plateau region; in renewable energy-scarce areas, the wind-solar-hydrogen-storage microgrid is given priority for power supply, and the insufficient part is calculated using a grid interaction model to obtain the power that can be obtained from the external grid to supplement the power supply to the traction load.

[0014] Under the goal of optimizing the dispatch of wind-solar-hydrogen-storage microgrids, the optimization dispatch target of wind-solar-hydrogen-storage microgrids is determined by combining the microgrid model, the grid interaction model, and the traction load model in plateau areas.

[0015] A multi-stage microgrid optimization scheduling framework is established by combining the wind-solar-hydrogen-storage microgrid model, the grid interaction model, and the plateau region traction load model. Through the multi-stage microgrid optimization scheduling framework, the optimal scheduling scheme of the wind-solar-hydrogen-storage microgrid can be determined with the minimum total system operating cost.

[0016] The multi-stage microgrid optimal scheduling framework is optimized by adopting a course-based, dual-delay, deep deterministic strategy gradient optimization method, resulting in the final optimized scheduling scheme.

[0017] Furthermore, a traction load model for plateau regions is established, and this model is used to calculate the power required for the traction load. The expression for the traction load model for plateau regions is as follows:

[0018] ;

[0019] in, for Total traction load power at all times; for Base load at any time; for Constant impact load components; This is the load random fluctuation coefficient; For the first Each impact load amplitude; For the first The moment when the impact load occurs; It is a unit impulse function.

[0020] Furthermore, the multi-stage microgrid optimization scheduling framework includes day-ahead scheduling plans, real-time security monitoring, and intraday optimization scheduling plans, as detailed below:

[0021] Day-ahead scheduling plan: Based on day-ahead forecast data of wind, solar and load, with the goal of minimizing the total system operating cost, a mixed integer linear programming problem is solved using a MILP solver to determine the start-up and shutdown status of hydrogen energy storage, the basic charge and discharge plan of the battery, and the initial transaction curve with the power grid for each time period of the next day.

[0022] Real-time safety monitoring: During daily real-time operation, the deviation between the actual operating status of the system and the daily plan is monitored at a high frequency, and the system safety risks are assessed; real-time safety margin indicators are calculated, and when it is found that the system margin is insufficient to cope with the impact load of the next stage, the daily optimized scheduling plan is immediately triggered.

[0023] Intraday Optimized Scheduling Plan: An intraday opportunity-constrained scheduling model is established and transformed into a Markov decision process under a reinforcement learning framework. A training mechanism for the correction network is proposed, which adopts opportunity-constrained programming to allow the violation of certain non-critical constraints with extremely low probability, adjusts the battery output and fuel cell power, corrects the deviation of the day-ahead plan, and ensures that the wind-solar-hydrogen-storage microgrid does not disconnect from the grid or lose power in the complex environment of the plateau.

[0024] Furthermore, the calculation of the real-time safety margin index, when it is found that the system margin is insufficient to cope with the impact load of the next stage, immediately triggers the intraday optimized scheduling plan, including:

[0025] A data-driven approach is used to establish a safety margin assessment network:

[0026] ;

[0027] ;

[0028] in, for Safety margin indicators for real-time systems; For neural network mapping functions; These are network weight parameters; For monitoring state vectors; for Actual wind power output at any given moment; for Real-time photovoltaic output; for The hydrogen storage tank must be kept in constant hydrogen-load status. for Constant battery state of charge; for Total traction load power at all times; for Power exchanged with the external power grid at all times; The real-time electricity price of the external power grid;

[0029] Warning conditions: If And continue During a specific time period, the intraday optimized scheduling plan will be triggered, in which... This represents the system's safety margin threshold.

[0030] Furthermore, the transformation of the intraday opportunity-constrained scheduling model into a Markov decision process within a reinforcement learning framework includes:

[0031] State Space: The complete state space of the wind-solar-hydrogen-storage microgrid system for traction power supply in plateau areas consists of the day-ahead planned state. Real-time status within the day Together they constitute the basis. During the intraday rolling correction phase, the decision relies on the fully observable real-time state, whose mathematical expression is as follows:

[0032] ;

[0033] in, For the future Short-term forecasted output of wind power, short-term forecasted output of photovoltaic power, and short-term forecasted total power of traction load for each time period; for Actual wind power output at any given moment; for Real-time photovoltaic output; for Total traction load power at all times; for Constant battery state of charge; for The hydrogen storage tank must be kept in constant hydrogen-load status. for Power exchanged with the external power grid at all times; The real-time electricity price of the external power grid;

[0034] Define the modified action space:

[0035] ;

[0036] in, To correct the motion space; for The amount of power adjustment for the electrolytic cell at any given time; for Real-time fuel cell power adjustment; for Adjustment of battery energy storage power at all times; for Real-time grid interaction power adjustment;

[0037] Define the reward function:

[0038] ;

[0039] Define economic rewards:

[0040] ;

[0041] Define security penalties:

[0042] ;

[0043] in, As an economic reward; As a security penalty; To terminate the reward, additional rewards or penalties will be given based on the final energy storage status; This is a weighting coefficient for economic efficiency; Weighting coefficients for safety; The unit penalty cost coefficient for adjusting equipment power; Let be the Manhattan norm of the action vector; These are the weighting coefficients for safety penalties, corresponding to safety limit exceedance penalties for battery energy storage, hydrogen energy storage, and grid interaction power, respectively. This is an indicator function that takes the value 1 when the condition is true and 0 otherwise. for Constant battery state of charge; These are the minimum and maximum states of charge; for The hydrogen storage tank must be kept in constant hydrogen-load status. The minimum and maximum hydrogen-charged states; for The power exchange between the wind, solar, hydrogen, and energy storage microgrid and the external power grid at all times; This represents the upper limit of the absolute value of the allowed grid interaction power.

[0044] Define the optimization objective:

[0045] ;

[0046] in, It is the maximum value; This is the end time of a complete scheduling cycle; For policy functions; Discount factor; For strategy The expected value.

[0047] Furthermore, the process of establishing the wind power output model is as follows:

[0048] Using historical meteorological data from the plateau region and combining it with the Weibull distribution to characterize the uncertainty of wind speed, the air density in the formula is corrected based on the thin air characteristics of the plateau. Furthermore, by introducing an uncertainty coefficient, random disturbances caused by plateau gusts are simulated. The predicted and actual power outputs are modeled as follows:

[0049] ;

[0050] ;

[0051] in, The current state of the wind-solar-hydrogen-storage microgrid system for traction power supply in plateau regions; for Real-time wind power output forecast; air density; The area swept by the wind turbine; for Wind speed at all times; The wind energy utilization coefficient is the ratio of the blade tip speed to the blade tip speed. and pitch angle related; for Actual wind power output at any given moment; The uncertainty factor for wind power output; For wind power random disturbance coefficient. .

[0052] Furthermore, the process of establishing the photovoltaic power output model is as follows:

[0053] A power temperature coefficient is introduced to monitor the deviation between the photovoltaic panel temperature and the standard temperature in real time, and to correct the photovoltaic output; a photovoltaic output model with random perturbations is established.

[0054] ;

[0055]

[0056] in, for Real-time photovoltaic power output forecast; Rated photovoltaic power under standard test conditions; for Light intensity at any given time; Standard test for light intensity; The power temperature coefficient; for Constant photovoltaic panel temperature; Standard test temperature; for Real-time photovoltaic output; The uncertainty factor for photovoltaic power output; This represents the photovoltaic random disturbance coefficient.

[0057] Furthermore, the process of establishing the battery energy storage system model is as follows:

[0058] During charging, electrical energy is converted into chemical energy for storage; during discharging, chemical energy is converted back into electrical energy and output to the grid or load. The entire energy conversion process is monitored and protected by the battery management system, and bidirectional conversion between DC and AC is completed by the power conversion system. Finally, under the scheduling of the energy management system, it collaboratively completes multiple functions such as peak shaving and valley filling, frequency and voltage regulation, reserve capacity, and microgrid support. The expression of the battery energy storage system model is as follows:

[0059] ;

[0060] in, for Constant battery state of charge; For charging and discharging efficiency; for Constant charging and discharging power; This refers to the battery's rated capacity. The length of the scheduling period; These are the minimum and maximum states of charge; This represents the maximum charging and discharging power.

[0061] Furthermore, the process of establishing the power grid interaction model is as follows:

[0062] Establish a grid interaction model between the wind-solar-hydrogen-storage microgrid and the external power grid:

[0063] ;

[0064] in, for The power exchange between the wind-solar-hydrogen-storage microgrid and the external power grid at any time is represented by positive values ​​for purchasing electricity and negative values ​​for selling electricity. for Total load at any given time; for Actual wind power output at any given moment; for Real-time photovoltaic output; for Real-time fuel cell power generation capacity; For the net output of the battery, , for Constant charging and discharging power; Maximum power purchase capacity; This represents the maximum power output.

[0065] Furthermore, the determination of the optimal scheduling objective of the wind-solar-hydrogen-storage microgrid by combining the wind-solar-hydrogen-storage microgrid model, the grid interaction model, and the traction load model for plateau regions includes:

[0066] The objective function takes minimizing the total operating cost of the system as its core optimization objective, comprehensively considering all core expenses of the wind-solar-hydrogen-storage microgrid during the dispatch cycle. The expression of the objective function is as follows:

[0067] ;

[0068] ;

[0069] ;

[0070] ;

[0071] in, Cost of purchasing / selling electricity; For equipment operation and maintenance costs; For the cost of hydrogen production; The real-time electricity price of the external power grid; for Power exchanged with the external power grid at all times; for Hydrogen production at any given time; For equipment The unit power operation and maintenance cost coefficient, For equipment The power.

[0072] Beneficial effects: Compared with the prior art, the significant technical effects of the present invention are as follows: (1) In view of the shortcomings of the existing scheduling methods in failing to fully characterize the differences in resources and environment in different sections of the plateau, the present invention proposes a differentiated section energy balance and coordinated scheduling mechanism. Specifically, by constructing a multi-stage microgrid optimization scheduling framework that can characterize the spatiotemporal transfer characteristics of traction load and the dynamic energy flow between sections, the present invention realizes the targeted support of surplus energy in resource-rich sections to scarce sections, maximizes the local consumption and optimized allocation of renewable energy along the line, and solves the problem of energy mutual assistance caused by uneven distribution of resources and load mobility along the plateau; (2) The present invention proposes a multi-stage microgrid optimization scheduling framework of "day-ahead scheduling planning + real-time safety monitoring + intraday optimization correction"; through Deterministic scheduling is carried out recently. Based on the prediction of renewable energy and load, a preliminary operation plan is formulated. At the same time, the system operation status is monitored in real time and the safety margin is assessed. When the system safety margin is found to be insufficient, the correction mechanism is quickly triggered. By flexibly adjusting the equipment output within the day, the energy configuration is optimized to ensure the reliability and economy of the system. (3) The present invention adopts a dual-delay deep deterministic strategy gradient optimization method based on course learning to optimize the multi-stage microgrid optimization scheduling framework. The algorithm can gradually cope with the uncertainty and complexity in the environment through phased training. Through reinforcement learning and deep strategy search, the scheduling strategy is optimized to cope with the rapid changes in traction load and the random fluctuations in wind and solar power output in real time, thereby improving the system's adaptability and response speed. Attached Figure Description

[0073] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0074] The technical solution of the present invention will now be described in detail with reference to specific embodiments and accompanying drawings.

[0075] like Figure 1As shown, the present invention provides an optimized scheduling method for a wind-solar-hydrogen-storage microgrid suitable for traction power supply systems in plateau areas, comprising the following steps:

[0076] S1. Establish wind power output model, photovoltaic power output model, hydrogen energy storage system model, and battery energy storage system model. The wind power output model, photovoltaic power output model, hydrogen energy storage system model, and battery energy storage system model together constitute the wind-solar-hydrogen-storage microgrid model.

[0077] The specific implementation process of step S1 is as follows:

[0078] S1.1 Establish a wind power output model, as follows:

[0079] Using historical meteorological data from the plateau region and combining it with the Weibull distribution to characterize the uncertainty of wind speed, the air density in the formula is corrected based on the thin air characteristics of the plateau. Furthermore, by introducing an uncertainty coefficient, random disturbances caused by plateau gusts are simulated. The predicted and actual power outputs are modeled as follows:

[0080]

[0081]

[0082] in, The current state of the wind-solar-hydrogen-storage microgrid system for traction power supply in plateau regions; for The wind power output (kW) at any given moment can be obtained by short-term wind speed prediction through time series analysis based on historical meteorological data. air density ( ); The swept area of ​​the wind turbine ( ); for Wind speed at any time ( ); The wind energy utilization coefficient is the ratio of the blade tip speed to the blade tip speed. and pitch angle related; for Actual wind power output at any given moment (kW); The uncertainty factor for wind power output; For wind power random disturbance coefficient. .

[0083] S1.2 Establish a photovoltaic power output model, as follows:

[0084] A power temperature coefficient is introduced to monitor the deviation between the photovoltaic panel temperature and the standard temperature in real time, and to correct the photovoltaic output. Considering the potential changes in module efficiency caused by strong ultraviolet radiation and low ambient temperature at high altitudes, a photovoltaic output model with random perturbations is established to ensure prediction accuracy under extreme weather conditions, as detailed below:

[0085]

[0086]

[0087] in, for The instantaneous photovoltaic power output (kW) can be obtained by short-term irradiance prediction based on historical meteorological data and numerical weather forecasting. The rated power of photovoltaic power (kW) under standard test conditions; for Light intensity at any given time (W / m²); For standard testing of light intensity ( ); The power temperature coefficient ( ); for Photovoltaic panel temperature at any time ( ); For standard test temperature ( ); for Real-time photovoltaic output; The uncertainty factor for photovoltaic power output; For photovoltaic random disturbance coefficients, This indicates that the photovoltaic random disturbance coefficient follows a normal distribution with a mean of 0 and a standard deviation of 1 (standard normal distribution).

[0088] S1.3 Establish a hydrogen energy storage system model. When wind and solar resources are sufficient to supply traction load, the hydrogen energy storage system is responsible for storing surplus electricity. When wind and solar resources are insufficient, the hydrogen energy storage system serves as a supplement. In the energy conversion chain of the hydrogen energy storage system, the electrolyzer, hydrogen storage tank, and fuel cell form a coupled energy-matter cycle through gaseous hydrogen medium. The electrolyzer, as the energy input unit, produces hydrogen by electrolyzing water when there is excess electricity. The hydrogen storage tank, as the static storage unit, is responsible for safely and efficiently storing the hydrogen produced by the electrolyzer. When there is a power shortage, the fuel cell converts the chemical energy of the stored hydrogen back into electrical energy.

[0089] The specific implementation process of step S1.3 is as follows:

[0090] S1.3.1 Electrolyzer Model (Hydrogen Production): The electrolyzer forcibly breaks down water molecules through electrochemical reactions, utilizing excess electrical energy to produce hydrogen. The actual power consumption and hydrogen production of the electrolyzer are modeled as follows:

[0091]

[0092]

[0093] in, for Actual power consumption of the electrolytic cell at any given time (kW); For electrolytic cell efficiency; for The constant input of the electrolytic cell's control power (kW); for Hydrogen production at any given time (kg). The low calorific value of hydrogen (kWh / kg); For hydrogen production efficiency.

[0094] S1.3.2 Hydrogen Storage Tank Model: The hydrogen storage tank enables the temporal and spatial transfer of energy. Low-pressure hydrogen gas from the electrolyzer must be compressed to 350-700 times atmospheric pressure to be stored in a specially designed high-pressure tank with a reasonable energy density. The compression process consumes additional electrical energy, but this allows the energy to be safely stored for days or even months with almost zero decay. When needed, the stored chemical energy can be released at any time via a valve to provide hydrogen. The hydrogen state-of-the-art is modeled as follows:

[0095]

[0096] in, for The hydrogen storage tank is in constant hydrogen charge status ( ); Rated capacity of the hydrogen storage tank (kg); for The amount of hydrogen produced and consumed (kg) at any given time, used for fuel cell power generation; The minimum and maximum hydrogen-charged states; For the maximum hydrogen production and consumption rate ( ).

[0097] S1.3.3 Fuel Cell Model (Power Generation): The power generation process of a fuel cell is the reverse of that of an electrolyzer. Hydrogen is catalytically decomposed into protons and electrons at the anode. Electrons generate current through an external circuit, doing work, while protons pass through the membrane to the cathode and combine with oxygen to form water. Chemical energy is thus directly converted into direct current (DC) electrical energy. The output DC power needs to be converted to alternating current (AC) by an inverter, which can then be connected to the power grid or used to power loads. The power generation model of the fuel cell is as follows:

[0098]

[0099] in, for Fuel cell power generation at any time (kW); For fuel cell power generation efficiency; This represents the maximum output power (kW) of the fuel cell.

[0100] S1.4 Establish a battery energy storage system model, as follows:

[0101] During charging, electrical energy is converted into chemical energy for storage; during discharging, chemical energy is converted back into electrical energy and output to the grid or load. The entire energy conversion process is monitored and protected by the Battery Management System (BMS), and bidirectional conversion between DC and AC power is achieved through the Power Conversion System (PCS). Finally, under the scheduling of the Energy Management System (EMS), it collaboratively performs multiple functions such as peak shaving and valley filling, frequency and voltage regulation, reserve capacity, and microgrid support. The expression for the battery energy storage system model is as follows:

[0102]

[0103] in, for Battery state of charge at all times ( ); For charging and discharging efficiency; for Real-time charging and discharging power (kW); The rated capacity of the battery is (kWh). The scheduling period length (h); These are the minimum and maximum states of charge; Maximum charging and discharging power (kW).

[0104] S2. Based on the wind-solar-hydrogen-storage microgrid model, establish a grid interaction model; details are as follows:

[0105] In renewable energy-rich areas, when wind and solar power generation is sufficient to cover local loads and energy storage systems are fully charged, the system will maximize energy self-sufficiency and clean energy consumption. In renewable energy-scarce areas, when wind and solar output is insufficient and energy storage systems cannot meet load demands, the system will actively draw power from the external grid to ensure reliable power supply to traction loads. A grid interaction model will be established between the wind-solar-hydrogen-storage microgrid and the external grid:

[0106]

[0107] in, for The power (kW) of the wind-solar-hydrogen-storage microgrid interacting with the external power grid at any time is represented by positive values ​​for purchasing electricity and negative values ​​for selling electricity. for Total load at any given time (kW), including traction load and other auxiliary loads; for Actual wind power output at any given moment; for Real-time photovoltaic output; for Real-time fuel cell power generation capacity; Net battery output (kW). , for Constant charging and discharging power; Maximum power purchase capacity (kW); Maximum power output (kW).

[0108] S3. Establish a traction load model for plateau regions and use the plateau region traction load model to calculate the power required for traction load.

[0109] The traction load of railways in plateau regions differs from traditional industrial and civilian loads; it is essentially a mobile, high-power, and impulsive AC load. In this embodiment, the expression for the traction load model in plateau regions is as follows:

[0110]

[0111] in, for Total traction load power (kW) at any given time; for The base load (kW) at any given time can be obtained by short-term load forecasting based on historical traction operation data using time series methods; for Impact load component at any moment (kW); For load random fluctuation coefficient, ; For the first Each impact load amplitude (kW); For the first The moment when the impact load occurs; It is a unit impulse function.

[0112] This invention divides the traction load model for high-altitude and cold regions into basic load and impact load components, and introduces a random fluctuation factor to overcome the difficulty in characterizing the large impact and fluctuation of traction load.

[0113] S4. Based on the calculated power required for the traction load, in renewable energy-rich areas, the output of renewable energy is directly calculated using the wind-solar-hydrogen-storage microgrid model, thereby using the wind-solar-hydrogen-storage microgrid to supply power to the traction load in the plateau region; in renewable energy-scarce areas, the wind-solar-hydrogen-storage microgrid is used first for power supply, and the insufficient part is calculated using the grid interaction model to obtain the power that can be obtained from the external grid to supplement the power supply to the traction load.

[0114] S5. The optimal scheduling objective of the wind-solar-hydrogen-storage microgrid is determined by combining the wind-solar-hydrogen-storage microgrid model, the grid interaction model, and the traction load model for plateau regions; specifically as follows:

[0115] The objective function aims to minimize the total operating cost of the system, comprehensively considering all core expenses of the wind-solar-hydrogen-storage microgrid during the dispatch cycle. Specifically, the objective function considers external grid interaction costs (electricity purchase or sale fees based on real-time electricity price fluctuations), internal equipment operation and maintenance costs (covering operating losses of wind turbines, photovoltaics, fuel cells, electrolyzers, and batteries), and hydrogen production costs. This objective function aims to guide the system to achieve the most economically efficient dispatch decision while satisfying all safety constraints. In this embodiment, the expression of the objective function is as follows:

[0116]

[0117]

[0118]

[0119]

[0120] in, Cost of electricity purchase / sale (RMB); Equipment maintenance cost (RMB); The cost of hydrogen production (in yuan); The real-time electricity price of the external power grid; for Power exchanged with the external power grid at all times; for Hydrogen production at any given time; For equipment The unit power operation and maintenance cost coefficient, For equipment The power.

[0121] Safety constraints include: power balance constraints (first row of formula (10)), equipment operation constraints (second row of formula (7), second row of formula (8) and the two rows below formula (9)), network security constraints (the voltage of each node and the current of each branch are within the safe operating range), etc.

[0122] S6. Under the objective of optimizing the scheduling of wind-solar-hydrogen-storage microgrids, a multi-stage microgrid optimization scheduling framework is established by combining the wind-solar-hydrogen-storage microgrid model, the grid interaction model, and the traction load model in plateau areas. Through the multi-stage microgrid optimization scheduling framework, the optimal scheduling scheme of the wind-solar-hydrogen-storage microgrid can be determined with the minimum total system operating cost.

[0123] In the aforementioned steps, wind power output model (step S1), photovoltaic power output model (step S1), hydrogen energy storage system model (step S1), battery energy storage system model (step S1), plateau traction load model (step S3), and grid interaction model (step S2) have been established respectively. These models together constitute the dynamic operation description and energy balance basis of wind-solar-hydrogen-storage microgrid. Among them, the wind power output model provides a characterization of the uncertainty of wind energy input for the dispatch system by comparing the predicted and actual wind power outputs; the photovoltaic output model reflects the time-varying characteristics of solar resources by calculating the predicted and actual photovoltaic outputs; the hydrogen energy storage system model realizes cross-time energy transfer between electricity, hydrogen, and electricity through the power consumption of the electrolyzer, the hydrogen state of charge of the hydrogen storage tank, and the power generation of the fuel cell; the battery energy storage system model provides rapid power regulation capabilities by comparing the state of charge and charging / discharging power of the battery energy storage system; the traction load model for plateau regions calculates the total traction load power (including base load and impact load components) to characterize the spatiotemporal fluctuation characteristics of the load; and the grid interaction model calculates the grid interaction power as a supplement and regulation means for system power balance. Based on the above models and their state variables, this invention constructs the following multi-stage microgrid optimization dispatch framework, which includes day-ahead dispatch plan, real-time safety monitoring, and intraday optimization dispatch plan, enabling full-process optimization dispatch from day-ahead planning and real-time monitoring to intraday correction:

[0124] S6.1 Day-ahead scheduling plan: Based on day-ahead forecast data of wind, solar and load, with the goal of minimizing the total system operating cost, a mixed integer linear programming problem is solved using a MILP solver to determine the start-up and shutdown status of hydrogen storage, the basic charge and discharge plan of the battery, and the initial transaction curve with the main grid for each period of the next day.

[0125] S6.2 Real-time Safety Monitoring: During intraday real-time operation, monitor the deviation between the actual system operating status and the daily plan at a high frequency (seconds / minutes), and assess system safety risks, such as power limit violations, voltage limit violations, and energy storage status limit violations. Calculate real-time safety margin indicators, and immediately trigger the intraday optimized scheduling plan when it is found that the system margin is insufficient to cope with the impact load of the next stage;

[0126] In this embodiment, a data-driven approach is used to establish a safety margin assessment network:

[0127]

[0128]

[0129] in, for Safety margin indicators for real-time systems; For neural network mapping functions; These are network weight parameters; This is the monitoring state vector.

[0130] It is worth noting that the safety margin assessment network is a neural network model used to assess in real time the distance between the current operating state of the power system and the safety boundary, and to quantify the system's reserve capacity to cope with future disturbances.

[0131] Warning conditions: If And continue During a specific time period, the intraday optimized scheduling plan will be triggered, in which... This represents the system's safety margin threshold.

[0132] S6.3 Intraday Optimized Scheduling Plan: First, an intraday opportunity-constrained scheduling model is established. Then, this model is transformed into a Markov decision process within a reinforcement learning framework. A training mechanism for the calibration network is proposed, employing opportunity-constrained programming, which allows for the violation of certain non-critical constraints with extremely low probability to achieve better economic efficiency. To address wind and solar forecasting errors and unexpected load pulses, battery output and fuel cell power are adjusted to correct day-ahead plan deviations, ensuring the wind-solar-hydrogen-storage microgrid remains connected and powered even in the complex high-altitude environment. The Markov decision process consists of four parts: state space, action space, state transition probabilities, and reward function. The intraday opportunity-constrained scheduling model is transformed into a Markov decision process within a reinforcement learning framework, as detailed below:

[0133] In step S6.3, the intraday opportunity-constrained scheduling model, within the intraday rolling optimization framework, allows for the violation of certain non-critical constraints with a specified low probability, thereby improving operational economy and flexibility while ensuring basic system safety. The correction network is a neural network model used to correct deviations in the daily plan during real-time intraday operation.

[0134] With a very low probability, say 10%, certain non-critical constraints are those that, if violated, will lead to a decrease in economic efficiency or a deviation from optimal operating conditions, but will not immediately cause a safety incident.

[0135] State Space: The complete state space of the wind-solar-hydrogen-storage microgrid system for traction power supply in plateau areas consists of the day-ahead planned state. Real-time status within the day These are the components. However, the precise baseline plan established in the day-ahead phase cannot predict some key state information in advance when facing ultra-short-term uncertainties. Therefore, this state component is unobservable information at the time of real-time decision-making. In the intraday rolling correction phase, the decision relies on the fully observable real-time state, the mathematical expression of which is as follows:

[0136]

[0137] in, For the future Short-term forecasted output of wind power, short-term forecasted output of photovoltaic power, and short-term forecasted total power of traction load for each time period; for Actual wind power output at any given moment; for Real-time photovoltaic output; for Total traction load power at all times; for Constant battery state of charge; for The hydrogen storage tank must be kept in constant hydrogen-load status. for Power exchanged with the external power grid at all times; This refers to the real-time electricity price of the external power grid.

[0138] Define the modified action space:

[0139]

[0140] in, To correct the motion space; for The amount of power adjustment for the electrolytic cell at any given time; for Real-time fuel cell power adjustment; for Adjustment of battery energy storage power at all times; for The amount of power adjustment between the power grid at any given time.

[0141] Define the reward function:

[0142]

[0143] Define economic rewards:

[0144]

[0145] Define security penalties:

[0146]

[0147] in, As an economic reward; As a security penalty; To terminate the reward, additional rewards or penalties will be given based on the final energy storage status; The weighting factor for economic efficiency (usually taken as 0.5 ~ 2); The weighting factor for security (usually taken as 1 to 5); The unit penalty cost coefficient for adjusting equipment power (usually taken as 0.01 ~ 0.1 yuan / kW). The Manhattan norm of the action vector (the sum of the absolute values ​​of each adjustment). The weighting coefficients for safety penalties correspond to safety limit violations for battery energy storage, hydrogen energy storage, and grid interaction power, respectively, and are typically set to large positive numbers, such as 10 to 100. This is an indicator function that takes the value 1 when the condition is true and 0 otherwise. for Constant battery state of charge; These are the minimum and maximum states of charge; for The hydrogen storage tank must be kept in constant hydrogen-load status. The minimum and maximum hydrogen-charged states; for The power exchange between the wind, solar, hydrogen, and energy storage microgrid and the external power grid at all times; This represents the upper limit of the absolute value of the allowed grid interaction power.

[0148] Define the optimization objective:

[0149]

[0150] in, It is the maximum value; This is the end time of a complete scheduling cycle; The policy function maps states to actions; This is a discount factor used to calculate cumulative rewards; it is an action-value function. For strategy The expected value.

[0151] This invention employs a three-stage scheduling framework—day-ahead deterministic scheduling, real-time safety monitoring, and intraday optimization and correction—to dynamically adjust the system's scheduling plan while ensuring safety margins. This maximizes the economic efficiency and clean energy utilization of wind-solar-hydrogen-storage microgrids suitable for traction power supply systems in plateau regions.

[0152] S7. The multi-stage microgrid optimal scheduling framework is optimized by adopting a course-based, dual-delay, deep deterministic strategy gradient optimization method to obtain the final optimized scheduling scheme.

[0153] Among them, the dual-delay deep deterministic strategy gradient optimization method based on course learning is the solution algorithm for the "intraday optimization correction" stage in the multi-stage microgrid optimization scheduling framework of "day-ahead scheduling planning + real-time safety monitoring + intraday optimization correction".

[0154] Specific environmental parameters for the plateau (such as air pressure and extreme temperature differences) are set, and measured wind, solar, and traction load data are imported into a wind-solar-hydrogen-storage microgrid model for simulation. A dual-delay deep deterministic policy gradient optimization method based on course learning is used to solve the aforementioned Markov decision process. First, the basic policy is trained in an environment with smooth wind, solar, and load fluctuations and small prediction errors. Then, typical traction load impact characteristics are introduced to increase the fluctuation amplitude. Finally, highly random wind and solar fluctuations under extreme weather conditions in the plateau are simulated, incorporating extreme scenarios such as partial equipment failure and communication delays to improve the policy's generalization and fault tolerance. The agent learns progressively from easy to difficult, ultimately obtaining a scheduling policy suitable for the complex plateau environment. After training, the obtained optimized scheduling policy... This strategy can be deployed in actual plateau-area wind-solar-hydrogen-storage microgrids via an energy management system (EMS). Using state observation data as input (including wind power predicted and actual output, photovoltaic power predicted and actual output, current operating status of each device in the hydrogen storage system, total traction load power, grid interaction power, real-time electricity price, and system operating constraints), the system outputs real-time scheduling decisions for each part of the microgrid (including power allocation instructions or start / stop instructions for the hydrogen storage system, power purchase and sale decisions, and instructions for switching operating modes of each device).

[0155] This invention gradually increases the uncertainty of the training environment through a course learning mechanism, enabling the agent to gradually adapt to the complex environment of the plateau, and ultimately obtains a wind-solar-hydrogen-storage microgrid dispatching strategy suitable for traction power supply systems in high-altitude and cold regions.

[0156] This invention overcomes the shortcomings of existing technologies in the dispatching of wind-solar-hydrogen-storage microgrids in plateau regions, and solves the multi-objective collaborative optimization problem between the strong randomness of wind and solar power output, the strong impact of traction loads, and the high safety requirements of system operation. Under the premise of strictly ensuring reliable power supply to plateau traction loads, it significantly improves the economic efficiency of microgrid system operation and the proportion of clean energy such as wind and solar power consumption, providing key technical support for green, low-carbon, safe, and efficient energy supply in plateau regions. This invention is applicable to the economical, reliable, and clean dispatching of microgrids in wind-solar-hydrogen-storage joint operation scenarios.

Claims

1. A method for optimized scheduling of wind-solar-hydrogen-storage microgrids suitable for traction power supply systems in plateau areas, characterized in that, include: Establish wind power output model, photovoltaic power output model, hydrogen energy storage system model, and battery energy storage system model. The wind power output model, photovoltaic power output model, hydrogen energy storage system model, and battery energy storage system model together constitute a wind-solar-hydrogen-storage microgrid model. A grid interaction model is established by combining the wind-solar-hydrogen-storage microgrid model; A traction load model for plateau regions was established, and the power required for traction load was calculated using this model. Based on the calculated power required for the traction load, in renewable energy-rich areas, the output of renewable energy is directly calculated using a wind-solar-hydrogen-storage microgrid model, thereby using the wind-solar-hydrogen-storage microgrid to supply power to the traction load in the plateau region; in renewable energy-scarce areas, the wind-solar-hydrogen-storage microgrid is given priority for power supply, and the insufficient part is calculated using a grid interaction model to obtain the power that can be obtained from the external grid to supplement the power supply to the traction load. The optimal scheduling objective of the wind-solar-hydrogen-storage microgrid is determined by combining the wind-solar-hydrogen-storage microgrid model, the grid interaction model, and the traction load model in the plateau region. Under the goal of optimizing the scheduling of wind-solar-hydrogen-storage microgrids, a multi-stage microgrid optimization scheduling framework is established by combining the wind-solar-hydrogen-storage microgrid model, the grid interaction model, and the traction load model in plateau areas. Through the multi-stage microgrid optimization scheduling framework, the optimal scheduling scheme of wind-solar-hydrogen-storage microgrids can be determined with the minimum total system operating cost. The multi-stage microgrid optimal scheduling framework is optimized by adopting a course-based, dual-delay, deep deterministic strategy gradient optimization method, resulting in the final optimized scheduling scheme.

2. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 1, characterized in that, The process involves establishing a traction load model for plateau regions and using this model to calculate the power required for traction load. The expression for the plateau region traction load model is as follows: ； in, for Total traction load power at all times; for Base load at any time; for Constant impact load components; This is the load random fluctuation coefficient; For the first Each impact load amplitude; For the first The moment when the impact load occurs; It is a unit impulse function.

3. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 1, characterized in that, The multi-stage microgrid optimal scheduling framework includes day-ahead scheduling plan, real-time security monitoring, and intraday optimal scheduling plan, as detailed below: Day-ahead scheduling plan: Based on day-ahead forecast data of wind, solar and load, with the goal of minimizing the total system operating cost, a mixed integer linear programming problem is solved using a MILP solver to determine the start-up and shutdown status of hydrogen energy storage, the basic charge and discharge plan of the battery, and the initial transaction curve with the power grid for each time period of the next day. Real-time safety monitoring: During daily real-time operation, the deviation between the actual operating status of the system and the daily plan is monitored at a high frequency, and the system safety risks are assessed; real-time safety margin indicators are calculated, and when it is found that the system margin is insufficient to cope with the impact load of the next stage, the daily optimized scheduling plan is immediately triggered. Intraday Optimized Scheduling Plan: An intraday opportunity-constrained scheduling model is established and transformed into a Markov decision process under a reinforcement learning framework. A training mechanism for the correction network is proposed, which adopts opportunity-constrained programming to allow the violation of certain non-critical constraints with extremely low probability, adjusts the battery output and fuel cell power, corrects the deviation of the day-ahead plan, and ensures that the wind-solar-hydrogen-storage microgrid does not disconnect from the grid or lose power in the complex environment of the plateau.

4. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 3, characterized in that, The calculation of real-time safety margin indicators, when it is found that the system margin is insufficient to cope with the impact load of the next stage, immediately triggers the intraday optimized scheduling plan, including: A data-driven approach is used to establish a safety margin assessment network: ； ； in, for Safety margin indicators for real-time systems; For neural network mapping functions; These are network weight parameters; For monitoring state vectors; for Actual wind power output at any given moment; for Real-time photovoltaic output; for The hydrogen storage tank must be kept in constant hydrogen-load status. for Constant battery state of charge; for Total traction load power at all times; for Power exchanged with the external power grid at all times; The real-time electricity price of the external power grid; Warning conditions: If And continue During a specific time period, the intraday optimized scheduling plan will be triggered, in which... This represents the system's safety margin threshold.

5. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 3, characterized in that, The transformation of the intraday opportunity-constrained scheduling model into a Markov decision process within a reinforcement learning framework includes: State Space: The complete state space of the wind-solar-hydrogen-storage microgrid system for traction power supply in plateau areas consists of the day-ahead planned state. Real-time status within the day Together they constitute the basis. During the intraday rolling correction phase, the decision relies on the fully observable real-time state, whose mathematical expression is as follows: ； in, For the future Short-term forecasted output of wind power, short-term forecasted output of photovoltaic power, and short-term forecasted total power of traction load for each time period; for Actual wind power output at any given moment; for Real-time photovoltaic output; for Total traction load power at all times; for Constant battery state of charge; for The hydrogen storage tank must be kept in constant hydrogen-load status. for Power exchanged with the external power grid at all times; The real-time electricity price of the external power grid; Define the modified action space: ； in, To correct the motion space; for The amount of power adjustment for the electrolytic cell at any given time; for Real-time fuel cell power adjustment; for Adjustment of battery energy storage power at all times; for Real-time grid interaction power adjustment; Define the reward function: ； Define economic rewards: ； Define security penalties: ； in, As an economic reward; As a security penalty; To terminate the reward, additional rewards or penalties will be given based on the final energy storage status; This is a weighting coefficient for economic efficiency; Weighting coefficients for safety; The unit penalty cost coefficient for adjusting equipment power; Let be the Manhattan norm of the action vector; These are the weighting coefficients for safety penalties, corresponding to safety limit exceedance penalties for battery energy storage, hydrogen energy storage, and grid interaction power, respectively. This is an indicator function that takes the value 1 when the condition is true and 0 otherwise. for Constant battery state of charge; These are the minimum and maximum states of charge; for The hydrogen storage tank must be kept in constant hydrogen-load status. The minimum and maximum hydrogen-charged states; for The power exchange between the wind, solar, hydrogen, and energy storage microgrid and the external power grid at all times; This represents the upper limit of the absolute value of the allowed grid interaction power. Define the optimization objective: ； in, It is the maximum value; This is the end time of a complete scheduling cycle; For policy functions; Discount factor; For strategy The expected value.

6. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 1, characterized in that, The process of establishing the wind power output model is as follows: Using historical meteorological data from the plateau region and combining it with the Weibull distribution to characterize the uncertainty of wind speed, the air density in the formula is corrected based on the thin air characteristics of the plateau. Furthermore, by introducing an uncertainty coefficient, random disturbances caused by plateau gusts are simulated. The predicted and actual power outputs are modeled as follows: ； ； in, The current state of the wind-solar-hydrogen-storage microgrid system for traction power supply in plateau regions; for Real-time wind power output forecast; air density; The area swept by the wind turbine; for Wind speed at all times; The wind energy utilization coefficient is the ratio of the blade tip speed to the blade tip speed. and pitch angle related; for Actual wind power output at any given moment; The uncertainty factor for wind power output; For wind power random disturbance coefficient. .

7. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 1, characterized in that, The process of establishing the photovoltaic power output model is as follows: A power temperature coefficient is introduced to monitor the deviation between the photovoltaic panel temperature and the standard temperature in real time, and to correct the photovoltaic output; a photovoltaic output model with random perturbations is established. ； in, for Real-time photovoltaic power output forecast; Rated photovoltaic power under standard test conditions; for Light intensity at any given time; Standard test for light intensity; The power temperature coefficient; for Constant photovoltaic panel temperature; Standard test temperature; for Real-time photovoltaic output; The uncertainty factor for photovoltaic power output; This represents the photovoltaic random disturbance coefficient.

8. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 1, characterized in that, The process of establishing the battery energy storage system model is as follows: During charging, electrical energy is converted into chemical energy for storage; during discharging, chemical energy is converted back into electrical energy and output to the grid or load. The entire energy conversion process is monitored and protected by the battery management system, and bidirectional conversion between DC and AC is completed by the power conversion system. Finally, under the scheduling of the energy management system, it collaboratively completes multiple functions such as peak shaving and valley filling, frequency and voltage regulation, reserve capacity, and microgrid support. The expression of the battery energy storage system model is as follows: ； in, for Constant battery state of charge; For charging and discharging efficiency; for Constant charging and discharging power; This refers to the battery's rated capacity. The length of the scheduling period; These are the minimum and maximum states of charge; This represents the maximum charging and discharging power.

9. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 1, characterized in that, The process of establishing the power grid interaction model is as follows: Establish a grid interaction model between the wind-solar-hydrogen-storage microgrid and the external power grid: ； in, for The power exchange between the wind, solar, hydrogen, and energy storage microgrid and the external power grid at any time is represented by positive values ​​for purchasing electricity and negative values ​​for selling electricity. for Total load at any given time; for Actual wind power output at any given moment; for Real-time photovoltaic output; for Real-time fuel cell power generation capacity; For the net output of the battery, , for Constant charging and discharging power; Maximum power purchase capacity; This represents the maximum power output.

10. The optimized scheduling method for wind-solar-hydrogen-storage microgrids applicable to traction power supply systems in plateau areas according to claim 1, characterized in that, The method of combining the wind-solar-hydrogen-storage microgrid model, the grid interaction model, and the traction load model for plateau regions to determine the optimal scheduling objective of the wind-solar-hydrogen-storage microgrid includes: The objective function takes minimizing the total operating cost of the system as its core optimization objective, comprehensively considering all core expenses of the wind-solar-hydrogen-storage microgrid during the dispatch cycle. The expression of the objective function is as follows: ； ； ； ； in, Cost of purchasing / selling electricity; For equipment operation and maintenance costs; For the cost of hydrogen production; The real-time electricity price of the external power grid; for Power exchanged with the external power grid at all times; for Hydrogen production at any given time; For equipment The unit power operation and maintenance cost coefficient, For equipment The power.

Images