A power system planning method considering extreme scenarios and seasonal energy storage coordination

Abstract

The application discloses a power system planning method considering extreme scenarios and seasonal energy storage coordination, relates to the technical field of power system planning and optimization, and comprises the following steps: extracting operation scenarios covering source- load bilateral extremes from historical operation data to construct a mixed scenario set; a hierarchical coordination model containing battery energy storage, pumped storage and hydrogen energy storage is established, and three-level energy state coupling constraints of intra-day, inter-month and inter-year are set; based on the hierarchical coordination model, a dynamic planning framework considering asset construction period, operation life and cross-year state connection is constructed to realize investment decision optimization under a full life cycle perspective; based on the mixed scenario set and the dynamic planning framework, investment decision and operation decision are respectively made with the full life cycle cost minimization as the target, and after refining the cost items and the constraint condition system, the planning of the power system is completed, and the full life cycle cost is significantly reduced under the premise of ensuring the safety of the system.

Description

Technical Field

[0001] This invention relates to the field of power system planning and optimization technology, and more specifically, to a power system planning method that considers extreme scenarios and the synergy of seasonal energy storage. Background Technology

[0002] As global efforts to address climate change deepen, building a new power system dominated by new energy sources has become an inevitable trend. The penetration rate of renewable energy sources such as wind power and solar power is increasing year by year, fundamentally changing the operating characteristics of the power system. Compared with traditional fossil fuels, renewable energy has significant volatility, intermittency, and seasonality. For example, wind power output is typically higher in spring and autumn, while solar power output peaks at midday in summer; simultaneously, influenced by temperature, electricity load also exhibits a clear "winter-summer double peak" characteristic. This seasonal mismatch between power sources and loads presents the power system with unprecedented long-term energy balance challenges. Furthermore, the frequent occurrence of extreme weather events globally in recent years (such as extreme cold, extreme heat, typhoons, and calm winds) poses a severe threat to the power supply security of systems with a high proportion of renewable energy.

[0003] Despite advancements in power system planning, existing technologies still exhibit significant shortcomings when addressing the complex characteristics of new power systems. Existing planning models often employ algorithms like K-means and hierarchical clustering to compress the annual 8760 hours of data into a few typical days. This average-based clustering method tends to retain frequently occurring common scenarios while often discarding low-probability but highly destructive "long-tail" extreme scenarios, such as several consecutive days of calm, low-light, and high-load weather. However, the adequacy of power system capacity is often determined by these extreme moments; ignoring extreme periods leads to under-planned system capacity and severe load shedding risks in actual operation. Existing models generally assume a one-year system operation cycle, requiring energy storage devices to reset their energy state at the beginning and end of the year, and decoupling operation between different years. This assumption neglects interannual climatic differences (such as wet and dry years, and years with abundant and scarce wind resources). In reality, novel long-term energy storage technologies such as hydrogen energy storage have the potential to store energy across multiple years. Existing models cannot simulate the cross-year regulation mechanism of "storing energy in bumper years and releasing it in dry years," making it difficult to cope with resource fluctuations caused by ultra-long-term climate cycles. Current energy storage planning often focuses on a single type of energy storage, mainly lithium batteries, or simply superimposing different energy storage methods. In fact, short-term energy storage (batteries), medium- and long-term energy storage (pumped hydro storage), and long-term energy storage (hydrogen storage) differ significantly in terms of response speed, capacity cost, and self-discharge rate. Existing research lacks refined hierarchical modeling of these three types of energy storage, failing to fully explore their tiered utilization value across multiple time scales, including intraday, interweekly, seasonal, and cross-year periods, resulting in high overall system investment costs. Traditional stochastic programming relies on precise probability distribution assumptions, but in the context of climate change, the statistical regularities of historical operating data may no longer apply to the future, leading to model distortion. Traditional robust optimization, based on worst-case decision-making, often results in overly conservative planning outcomes and high investment costs. How to handle the uncertainty of distribution while avoiding excessive conservatism is a major challenge for current planning technologies.

[0004] Therefore, there is an urgent need to develop a dynamic multi-stage power system planning method that can comprehensively consider extreme scenario identification, coordinated operation of multiple types of seasonal energy storage, and cross-year energy balance, so as to balance economy and system security under extreme scenarios without introducing an overly complex robust optimization architecture. Summary of the Invention

[0005] To address the aforementioned issues, the present invention aims to provide a power system planning method that considers the synergy between extreme scenarios and seasonal energy storage, in order to address the extreme weather impacts and long-term energy imbalances faced by high-proportion renewable energy systems, and to significantly reduce the total life-cycle cost while ensuring system safety.

[0006] To achieve the above technical objectives, this application provides a power system planning method that considers the synergy between extreme scenarios and seasonal energy storage, comprising the following steps: Operational scenarios covering extreme values ​​on both the source and load sides are extracted from historical operational data, and a hybrid scenario set including ordinary typical days and extreme typical days is constructed. Establish a hierarchical collaborative model that includes battery energy storage, pumped hydro storage and hydrogen energy storage, and realize a multi-type energy storage collaborative scheduling mechanism through three-level energy state coupling constraints at the intraday, inter-month and inter-year levels; Based on the hierarchical collaborative model, an investment decision optimization from a full life cycle perspective is achieved by constructing a dynamic programming framework that considers the asset construction cycle, operational life, and cross-year status transitions. Based on a hybrid scenario set and dynamic programming framework, with the goal of minimizing the total life cycle cost, investment and operational decisions are made separately, and the power system planning is completed after refining the cost items and constraint system.

[0007] Preferably, when constructing a hybrid scenario set, net load calculation and extreme value identification are performed based on historical operating data, and single-factor extreme days at the tail of the distribution are identified based on the empirical distribution of historical operating data. Considering complex extreme scenarios involving multiple coupled factors, we will identify combined extreme scenarios. Based on the identification results of combined extreme scenarios and single-factor extreme days, a hybrid scenario set is constructed.

[0008] Preferably, when obtaining single-factor extreme days, single-factor extreme days should include extreme days with high net load, low net load, high volatility, and high ramp.

[0009] Preferably, before constructing the hybrid scene set, based on the recognition results, a Wasserstein generative adversarial network with gradient penalty is used to amplify the scarce extreme samples, thereby completing the construction of the hybrid scene set.

[0010] Preferably, when obtaining coupling constraints, coupling constraints are formed based on short-term energy storage operation constraints, pumped storage operation constraints, and hydrogen energy storage operation constraints, so as to realize cross-year energy transfer modeling of hydrogen energy storage through three levels of energy state coupling constraints: intraday, inter-month, and inter-year.

[0011] Preferably, when optimizing investment decisions, the annualized investment cost is calculated based on the hierarchical collaborative model, taking into account capacity investment constraints and cross-year energy storage state transitions, thereby completing the investment decision optimization.

[0012] Preferably, the decision on new construction and decommissioning capacity is made during the investment decision-making stage.

[0013] Preferably, during the operational decision-making phase, with the goal of minimizing the total lifecycle cost, operational decisions are made based on the feasible set generated during the investment decision-making phase and the annualized investment cost, and initial state consistency constraints are incorporated during the operational decision-making process.

[0014] Preferably, when refining the cost items and constraint system, a long-term energy imbalance risk term based on CVaR is introduced into the objective function, and robust hard constraints are set for extreme typical days.

[0015] Preferably, when setting robust hard constraints for extreme typical days, the robust hard constraints include the upper limit of the load shedding ratio under extreme scenarios and the system adequacy constraints under extreme scenarios.

[0016] The present invention discloses the following technical effects: 1. This invention identifies extreme operating conditions from four dimensions: maximum and minimum net load, volatility, and ramp rate, through a multi-dimensional anomaly detection method. It also establishes a composite extreme scenario identification mechanism, such as high load-low output and low load-high output, and combines WGAN-GP data augmentation technology to enrich extreme samples. This effectively overcomes the shortcomings of traditional K-means clustering, which only retains high-frequency scenarios and smooths out extreme risks. As a result, the planning scheme can cope with extreme operating conditions with low probability of occurrence but high destructive power, such as calm and stable weather.

[0017] 2. This invention addresses the differentiated techno-economic characteristics of three types of energy storage: battery energy storage, pumped hydro storage, and hydrogen energy storage. It establishes a hierarchical and refined model and achieves true multi-timescale coordination through strict variable-level coupling of intraday SOC transfer, inter-monthly SOC transfer, year-end SOC, and the SOC of the first month of the following year. Rather than a simple summation of electricity usage constraints, this invention fundamentally ensures the consistency of the physical conservation relationship between intraday peak shaving, inter-weekly and inter-monthly regulation, and cross-year energy transfer.

[0018] 3. This invention realizes multi-scale modeling of hydrogen energy storage at the daily, monthly, annual, and trans-year scales in the planning model, allowing for a trans-year adjustment mechanism of "storing energy in good years and releasing it in bad years"; the dynamic planning framework simultaneously considers the entire life cycle factors such as asset construction cycle, operational life, and decommissioning decisions, effectively addressing the energy imbalance risks brought about by interannual climate differences and ultra-long-term uncertainties.

[0019] 4. This invention adopts a standard two-stage stochastic programming model. The first stage determines the investment variables, and the second stage independently solves the operating costs of each scenario in a mixed scenario set and obtains the expected value. The annual long-term energy gap is quantified by CVaR, so that the planning results not only focus on the expected cost, but also focus on controlling the tail risk, providing a reliable guarantee for the safe operation of a high proportion of renewable energy power system. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the extreme scene recognition and generation process described in this invention; Figure 2 This is a schematic diagram of the method described in this invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] like Figures 1-2 As shown, this invention provides a robust planning method for dynamic multi-stage power systems that considers extreme scenarios and the synergy of multiple types of seasonal energy storage, including the following steps: Step 1: Construct an extreme scene recognition and generation module based on multi-dimensional anomaly detection, accurately extract operational scenes covering extreme values ​​on both the source and load sides from historical operational data, and construct a hybrid scene set including ordinary typical days and extreme typical days; Step 2: Establish multi-type seasonal energy storage collaborative operation models, and conduct refined modeling for short-term battery energy storage, medium- and long-term pumped hydro storage and long-term hydrogen energy storage respectively. Through the three-level energy state coupling constraints of intraday, inter-month and inter-year, the energy collaborative scheduling of intraday, inter-month and inter-year is realized. Step 3: Construct a cross-year dynamic planning framework that considers the entire life cycle of assets, and introduce dynamic investment decisions, facility decommissioning mechanisms, and cross-year energy storage energy status connection constraints; Step 4: With the goal of minimizing the total life cycle cost, construct a two-stage stochastic programming model: the first stage determines the investment variables, and the second stage solves for the operating cost of each scenario in the mixed scenario set (including ordinary typical days and extreme typical days), and obtains the expected value by weighting the scenario weights. Step 5: Introduce a long-term energy imbalance risk term based on CVaR into the objective function, and set robust hard constraints (upper limit of load shedding ratio and system sufficiency constraint) for extreme typical days, so as to balance economy and safety in extreme scenarios without relying on distributed robust optimization.

[0024] In step one of the embodiments, to overcome the deficiency of the traditional K-means clustering method in ignoring extreme time periods, an extreme scene identification method integrating statistical anomaly detection and physical characteristic analysis is proposed, which includes the following steps: Step 1.1 Data Preprocessing. Collect historical operating data over many years, including hourly load curves. Wind power output coefficient Photovoltaic power output coefficient ,in For node indexing, Indexed by natural day, Intraday time period index ( To eliminate the influence of dimensions, the data is standardized: In the formula, This represents the raw data (load or output coefficient) to be standardized. and Variables The mean and standard deviation over a historical sample. This is the standardized data.

[0025] Step 1.2 Net Load Calculation and Extreme Value Identification. The system net load is defined as the original load minus the output of renewable energy: In the formula, For the first day The system net load at any given time, and These are respectively a collection of wind turbine units and a collection of photovoltaic units. , These are the power output coefficients for wind power and solar power, respectively. , These represent the rated installed capacity of wind power and solar power, respectively. Extreme value indices are calculated daily based on the net load sequence, where the daily maximum net load is... The minimum daily net load is Daily net load volatility is ,in For the first Average daily net load; Maximum daily ramp rate .

[0026] Step 1.3 Quantile Anomaly Detection. Based on the empirical distribution of historical operating data, identify single-factor extreme days at the tail of the distribution: In the formula, Representing the historical sequence of the whole year Quantiles The extreme threshold parameter is usually set to 0.05; the four sets mentioned above correspond to extreme days of high net load, low net load, high volatility, and high ramp, respectively.

[0027] Step 1.4 Combined Extreme Scenario Identification. Consider composite extreme scenarios involving multiple coupled factors, where an extreme power shortage scenario is defined as the simultaneous occurrence of high load and low renewable energy output; an extreme power curtailment scenario is defined as the simultaneous occurrence of low load and high renewable energy output, as defined below: In the formula, , The first Maximum and minimum values ​​of daily raw load For the first The daily average value of the renewable energy output coefficient of the entire system.

[0028] Step 1.5 Construction of the extreme scenario set. Merge all extreme day sets and remove duplicates, as defined below: For sets Using a hierarchical clustering algorithm, the representative day furthest from the cluster center in each cluster is selected as the extreme typical day, forming a set of extreme typical days. K-means clustering is performed on the remaining ordinary days to obtain a set of typical ordinary days. The two combine to form a hybrid typical day set. Each typical day Corresponding weight (Equal to the frequency of that type of day in the original data).

[0029] Step 1.6 Extreme Scene Data Augmentation Based on WGAN-GP. To enrich the extreme scene samples, a Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) is used to augment the scarce extreme samples. The loss functions of the generator and discriminator are defined as follows: In the formula, , These are the generator and discriminator networks, respectively. To conform to the prior distribution Latent variables, To conform to the true extreme sample distribution Historical extreme day samples, The interpolated sample is the line connecting the real sample and the generated sample. Its distribution, The gradient penalty strength coefficient is used in the third gradient penalty term to ensure that the discriminator satisfies the 1-Lipschitz constraint. To concentrate the generated samples towards extreme regions, an iterative distribution shift mechanism is introduced: In the formula, For the first The empirical distribution of samples used for discriminator training in the next iteration. The extreme degree of the sample is measured by a weighted sum of the extreme value indices defined in step 1.2. For the first The extreme threshold of the next iteration. This is an indicator function. After multiple iterations, the samples generated by the generator will be concentrated at the end of the historical running data, serving as... Supplementary samples were included in the mixed scene set.

[0030] In step two of one embodiment, based on the response characteristics and capacity features of different energy storage technologies, energy storage systems are divided into three categories (sets denoted as short-time battery energy storage). Pumped storage Hydrogen storage The three are combined into ), and establish refined operation models respectively, including the following: Step 2.1 Short-term energy storage operation constraints. For battery energy storage Its power constraints, mutual exclusion constraints, energy state transitions, and capacity constraints are defined as follows: In the formula, , Energy storage The charging and discharging power, , Its upper limit for charging and discharging power, , These are 0-1 variables representing the charging and discharging states, respectively. , These represent charge and discharge efficiencies, For energy storage charge, For the rated capacity of energy storage, , This represents the upper limit coefficient under charged state. Because batteries have a high self-discharge rate and are not suitable for long-term storage, a daily cycle constraint is set to ensure consistent operation throughout the day. ; Step 2.2 Pumped Storage Operation Constraints. For pumped storage... The pumping / power generation constraints and their mutual exclusion constraints are as follows: In the formula, , These are pumping power and power generation power, respectively. , Its upper limit, , Let be the corresponding 0-1 state variables. The reservoir energy state transition equation considering the effects of natural inflow and outflow is: In the formula, , These are pumping efficiency and power generation efficiency, respectively. Natural water flow converted into electricity This refers to the amount of water discarded.

[0031] For example, considering that pumped storage has weekly and monthly regulation capabilities, daily cycle constraints are not mandatory; by introducing coupled constraints between the end-of-day state of charge and the inter-monthly state of charge, the energy balance across days and months is explicitly established in the model. Definition For energy storage In the Year scene The Middle The charge level at the beginning of the month is as follows: In the formula, To represent the first The typical set of days in a month, For typical days In the The actual number of days represented in the month (satisfying) equal to the (Total number of days in a month) Indicates a typical day The net energy change of internal energy storage. This formula weights the charge volume at the end of each typical day by its representative day number and accumulates it in the inter-month state transition. Simultaneously, it constrains the inter-month charge volume to not exceed the limit, as follows: ; And in a typical day-first period, the transition between the day's initial load and the beginning-of-month load is forced to be consistent, as follows: ; In the formula, For typical days The month to which it belongs This represents the offset of different representative days from their state at the beginning of the month.

[0032] Step 2.3 Hydrogen storage operation constraints. For hydrogen storage... Considering the coupled operation of three stages: hydrogen production via electrolysis, hydrogen storage tanks, and fuel cells. The operating constraints and hydrogen production equations for the electrolyzer and the operating constraints and hydrogen consumption equations for the fuel cell are as follows: In the formula, To input electrical power into the electrolytic cell, For its rated power, For electrolysis efficiency, This represents the amount of hydrogen produced in the form of equivalent electrical energy. Where, To output electrical power for fuel cells, For its rated power, For fuel cell power generation efficiency, This represents the amount of hydrogen consumed.

[0033] The intraday energy state transition and capacity constraints of the hydrogen storage tank are as follows: In the formula, The hydrogen storage tank charge is measured in equivalent electrical energy (MWh). The rated capacity of the hydrogen storage tank. , These are the upper and lower limits of charge.

[0034] Step 2.5 Coupling of three-level energy states for hydrogen energy storage: intraday, intermonthly, and interannual. To achieve energy transfer across seasons and years in hydrogen energy storage, this invention constructs direct coupling between three levels of charge states: intraday, intermonthly, and interannual, to eliminate decoupling problems between different time scales.

[0035] definition For the first Year scene The Middle The hydrogen storage status at the beginning of the month, and the inter-month transfer equation based on the intraday net hydrogen quantity change defined in step 2.4: In the formula, This represents the monthly hydrogen storage self-loss rate (including losses from evaporation, leakage, and pressure maintenance). For typical days In the The number of days represented by a month. This formula aggregates the net hydrogen production per day into the energy change between months by weighting the number of days it represents, thus strictly coupling the intraday operation with the intermonthly state at the variable level.

[0036] To ensure that the charge level at the beginning of a typical day is consistent with the hydrogen storage status at the beginning of the month, a month-end transition constraint is introduced: In the formula, To represent the first The formula guarantees that the first day of a typical month is the first day in the sequence. The amount of hydrogen stored at the beginning of the first typical day of the month represents the state at the beginning of that month.

[0037] definition For the first Hydrogen storage status at the end of the year (i.e., the end of December), and its connection with the inter-month status: ; The formula adds up the daily net hydrogen production in the 12th month to the state at the beginning of the 12th month to obtain the hydrogen storage state at the end of the year.

[0038] The key innovation of this invention lies in allowing hydrogen energy storage to transfer energy between adjacent years, i.e., the first The state at the beginning of the first month of the year is equal to the first... The year-end status, after deducting annual self-losses, is as follows: In the formula, This represents the annual hydrogen storage self-loss rate. This constraint, together with the aforementioned intraday and intermonthly coupled constraints, constitutes a complete constraint system from intraday to intermonthly to annual to cross-year, thus truly realizing a cross-year regulation mechanism for hydrogen energy storage: "storing energy in bumper years and releasing it in lean years." To avoid artificially injected energy at the boundary, a consistency constraint is set between the state at the beginning of the first month of the first year and the state at the end of the planning period: In the formula, This represents the initial hydrogen storage level during the planning period.

[0039] Step 2.6 Multi-type Energy Storage Cooperative Scheduling Mechanism. Based on the response characteristics of different energy storage types, priority scheduling constraints are established to ensure that fast-response tasks are preferentially undertaken by short-term energy storage, as follows: In the formula, This represents the net system load in the corresponding scenario. The sharing ratio factor for short-term energy storage to bear the rapid net load increase ( ).

[0040] The coordinated scheduling of the above three types of energy storage is reflected in the power balance constraints described in step 5.3.

[0041] In step three of one implementation, to optimize investment decisions from a full life-cycle perspective, a dynamic programming framework is constructed that considers the asset construction cycle, operational lifespan, and cross-year status transitions. This framework includes the following: Step 3.1 Capacity Investment Constraints. The capacity accumulation equation considering the construction lag effect is as follows: In the formula, Index of power generation / energy storage technologies, among which This is a collection of power generation technologies, representing thermal power, hydropower, wind power, and photovoltaic power. For the first Year technology Available capacity For the first New capacity in 2018 For the first Annual decommissioning capacity For technology The construction period is measured in years. For In this case, take Meanwhile, the retirement constraint considering the equipment's lifespan is as follows: In the formula, For technology The operational lifespan, that is, the lifespan after commissioning. It must be decommissioned within one year. The upper and lower capacity limits are: In the formula, , technology The upper and lower limits of installed capacity are determined by resource endowment or policy constraints.

[0042] Step 3.2 Cross-Year Energy Storage Status Transfer. The cross-year transition of seasonal energy storage (mainly hydrogen storage) has already been addressed in Step 2.5. To ensure consistency in year-end planning decisions across different scenarios, the same initial energy storage state is set for all scenarios: Step 3.3 Dynamic Investment Cost Calculation. The annualized investment cost considering the time value of money is: In the formula, For technology The unit investment cost For the set of candidate transmission lines, For the first Annual candidate routes 0-1 investment decision variables For the line The investment cost, This is the capital recovery factor. is the discount rate. The net present value of the total investment during the planning period is: In the formula, This is for the planned annual total.

[0043] In step four of one implementation, a two-stage stochastic programming model is constructed with the objective of minimizing the total lifecycle cost. The first stage is the investment decision stage, which determines the capacity to be newly built or decommissioned; the second stage is the operation decision stage, which solves for the operation cost of each scenario under the mixed typical daily scenario set constructed in step one, and obtains the expected value according to the scenario probability weights. Specifically, it includes the following: Step 4.1 Scene Set and Probability Weights. Define a complete scene consisting of scene year. With typical day Together they constitute the probability weights as follows: In the formula, For the scene year The probability, For typical days The weight.

[0044] Step 4.2 Two-stage stochastic programming objective function. The model aims to minimize the sum of the expected values ​​of investment cost and operating cost across all scenarios: In the formula, This forms the feasible set for the first phase of investment decisions. The annualized investment cost, In investment decision With Scene Year The optimal operating cost for the second stage is as follows: In the formula, For the scene year The second stage of the operating variable set (including conventional power output) Energy storage charging and discharging power Pumped storage Hydrogen energy storage load shedding abandoned electricity With the trend ); , These are short-term and long-term operating costs, respectively. These are the weighting coefficients for short-term and long-term costs. Furthermore, for multi-year hydrogen energy storage energy status, to avoid "energy cheating" caused by inconsistent initial values ​​during the planning period under different scenarios, an explicit initial state consistency constraint is added, as follows: In the formula, This represents the initial hydrogen storage level during the planning period.

[0045] In step five of one embodiment, under the two-stage stochastic programming objective framework constructed in step 4.2, the cost items and constraint system are further refined, including the following: Step 5.1 Specific expression of the three types of cost items. The annualized investment cost, For intraday scale in the scene The short-term operating costs, This refers to the long-term operating costs, including CVaR risk items, on a monthly / annual scale.

[0046] In the formula, For technology Fixed operation and maintenance costs; For power generation technology Variable operating costs, To reduce the cost of load shedding, To account for the cost of power curtailment penalties, , , Scenes The power generation output, load shedding, and abandoned power during that period. , The probability weights for scenario year and typical day are: Constitutes the standard expectation operator; For the long-term variable cost of thermal power, For the first Year Monthly thermal power generation For the annual long-term energy deficit, For nodes In the Monthly long-term load shedding, CVaR confidence level.

[0047] Step 5.2 CVaR Risk Constraints. Introduce auxiliary variables. and CVaR can be expressed as: In the formula, For the scene probability weights, For the scene Below The part.

[0048] Step 5.3 Short-term energy balance and network constraints.

[0049] The upper limit constraint for renewable energy output is: In the formula, For the scene Renewable Energy The output coefficient (given by historical samples or WGAN-GP generated samples). Available capacity determined for investment decisions.

[0050] The node power balance constraint is: In the formula, , Access nodes A collection of power generation / energy storage, , These are the inflow and outflow nodes, respectively. A collection of routes, For the line The trend.

[0051] The power flow constraints of transmission lines are simplified to use a DC power flow model, as follows: In the formula, For the line susceptivity, The phase angle of the node. This refers to the line capacity.

[0052] The investment logic constraints for the candidate routes are: Step 5.4 Long-term energy balance constraints. Monthly energy balance constraints are: In the formula, For the unit In the Total monthly electricity generation , For nodes No. Total monthly load electricity, For the first Monthly routes Net power transmission; This represents the monthly long-term load shedding. The annual electricity balance is directly calculated from the aforementioned monthly equations. Sum and export.

[0053] Step 5.5 Renewable Energy Penetration Constraints. To promote the achievement of new energy development goals, a lower limit for renewable energy electricity penetration is set: In the formula, This represents the lower limit for renewable energy electricity penetration.

[0054] Step 5.6 Robust Constraints for Extreme Scenarios. For extreme typical days... On top of the goal of minimizing expected costs, hard robust constraints are superimposed to ensure that the system still has sufficient safety margins in extreme scenarios.

[0055] The maximum load shedding ratio under extreme scenarios is: In the formula, This represents the maximum allowable load shedding percentage under extreme scenarios (typically 5%). This represents the load under this extreme typical day.

[0056] The system sufficiency constraint under extreme scenarios is: In the formula, For reserve rate in extreme scenarios, For the first Maximum load under an extreme typical day set in a year.

[0057] In summary, this invention discloses a dynamic multi-stage robust planning method for power systems that considers extreme scenarios and the synergy of multiple types of seasonal energy storage. First, it extracts extreme operating scenarios through multi-dimensional anomaly detection, constructing a hybrid scenario set including both ordinary and extreme typical days to address the problem of smoothing extreme risks using traditional clustering. Second, it establishes a hierarchical collaborative model encompassing battery energy storage, pumped hydro storage, and hydrogen energy storage, and focuses on modeling cross-year energy transfer for hydrogen energy storage through three levels of energy state coupling constraints: intraday, inter-month, and inter-year. Subsequently, it constructs a cross-year dynamic planning framework in conjunction with asset lifecycle management. Finally, with the goal of minimizing the lifecycle cost, it constructs a two-stage stochastic programming model. The first stage determines investment variables, and the second stage calculates the expected operating cost for each scenario in the hybrid scenario set, incorporating a long-term energy imbalance risk measure based on Conditional Value at Risk (CVaR). Additional load shedding ratio caps and system adequacy hard constraints are imposed on extreme typical days, thereby providing deterministic robust guarantees under extreme scenarios. This invention can effectively address the challenges of extreme weather and long-term energy imbalances faced by high-proportion renewable energy systems, and significantly reduce the total life cycle cost while ensuring system safety.

[0058] This invention integrates multi-dimensional anomaly characteristics such as net load extremes, volatility, and ramp rate with physical mechanism analysis, and combines WGAN-GP to identify and enhance extreme scenarios, constructing a hybrid scenario set that simultaneously covers both typical daily and source-load combined extremes. Based on this, a hierarchical collaborative model is established for short-term battery energy storage, medium-to-long-term pumped hydro storage, and long-term hydrogen energy storage. In particular, a three-level energy state coupling constraint is constructed for hydrogen energy storage at the intraday, intermonthly, and interannual levels, achieving physical consistency modeling for cross-year energy transfer. Furthermore, a robust planning architecture is proposed, and CVaR risk measurement is embedded into the full lifecycle cost minimization model, thereby simultaneously improving system safety, long-term energy balance capability, and planning economy under extreme scenarios without excessively increasing optimization complexity.

[0059] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0060] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0061] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A power system planning method considering the synergy between extreme scenarios and seasonal energy storage, characterized in that, Includes the following steps: Operational scenarios covering extreme values ​​on both the source and load sides are extracted from historical operational data, and a hybrid scenario set including ordinary typical days and extreme typical days is constructed. Establish a hierarchical collaborative model that includes battery energy storage, pumped hydro storage and hydrogen energy storage, and realize a multi-type energy storage collaborative scheduling mechanism through three-level energy state coupling constraints at the intraday, inter-month and inter-year levels; Based on the aforementioned hierarchical collaborative model, an investment decision optimization from a full life-cycle perspective is achieved by constructing a dynamic programming framework that considers asset construction cycle, operational lifespan, and cross-year status transitions. Based on the hybrid scenario set and the dynamic programming framework, with the goal of minimizing the total life cycle cost, investment and operational decisions are made separately, and the cost items and constraint system are refined to complete the planning of the power system.

2. The power system planning method considering extreme scenarios and seasonal energy storage synergy as described in claim 1, characterized in that: When constructing a hybrid scenario set, net load calculation and extreme value identification are performed based on historical operating data, and single-factor extreme days at the tail of the distribution are identified based on the empirical distribution of historical operating data. Considering complex extreme scenarios involving multiple coupled factors, we will identify combined extreme scenarios. Based on the identification results of the combined extreme scenarios and the single-factor extreme days, the hybrid scenario set is constructed.

3. The power system planning method considering extreme scenarios and seasonal energy storage synergy as described in claim 2, characterized in that: When obtaining single-factor extreme days, the single-factor extreme days include extreme days with high net load, low net load, high volatility, and high ramp.

4. The power system planning method considering extreme scenarios and seasonal energy storage coordination according to claim 2, characterized in that: Before constructing the hybrid scene set, based on the recognition results, a Wasserstein generative adversarial network with gradient penalty is used to amplify the scarce extreme samples, thus completing the construction of the hybrid scene set.

5. The power system planning method considering extreme scenarios and seasonal energy storage synergy according to claim 1, characterized in that: When obtaining coupling constraints, coupling constraints are formed based on short-term energy storage operation constraints, pumped storage operation constraints, and hydrogen energy storage operation constraints. Through the three-level energy state coupling constraints of intraday, intermonthly, and interannual, the cross-year energy transfer model of hydrogen energy storage can be realized.

6. The power system planning method considering extreme scenarios and seasonal energy storage synergy as described in claim 1, characterized in that: When optimizing investment decisions, based on the aforementioned hierarchical collaborative model, considering capacity investment constraints and cross-year energy storage state transitions, the annualized investment cost taking into account the time value of money is calculated to complete the investment decision optimization.

7. The power system planning method considering extreme scenarios and seasonal energy storage synergy as described in claim 1, characterized in that: During the investment decision-making stage, decisions are made regarding the construction and decommissioning of capacity.

8. The power system planning method considering extreme scenarios and seasonal energy storage synergy according to claim 7, characterized in that: During the operational decision-making phase, with the goal of minimizing the total lifecycle cost, operational decisions are made based on the feasible set generated during the investment decision-making phase and the annualized investment cost, and initial state consistency constraints are added during the operational decision-making process.

9. The power system planning method considering extreme scenarios and seasonal energy storage synergy according to claim 1, characterized in that: When refining the cost items and constraint system, a long-term energy imbalance risk term based on CVaR is introduced into the objective function, and robust hard constraints are set for extreme typical days.

10. The power system planning method considering extreme scenarios and seasonal energy storage synergy according to claim 9, characterized in that: When setting robust hard constraints for extreme typical days, the robust hard constraints include the upper limit of the load shedding ratio under extreme scenarios and the system adequacy constraints under extreme scenarios.

Images