Power system flexible resource planning method, system and device considering multi-time scale extreme weather, and storage medium

Abstract

The invention discloses an electric power system flexible resource planning method, system and device considering multi-time scale extreme weather and a storage medium, and relates to the technical field of electric power system planning, and the method comprises the steps: generating a multi-time scale extreme weather scene based on a conditional diffusion model; constructing a flexible resource planning model of the mixed time scale power system; and calling a Gurobi solver by using a Yalmip toolkit of Matlab to solve the flexible resource planning model of the mixed time scale power system, and outputting an optimal planning scheme of flexible resources of the power system. According to the method, the multi-day extreme weather scene is generated based on the conditional diffusion model, spatial-temporal characteristics of multi-day extreme weather such as typhoon and high temperature can be accurately described, and compared with other methods, a diversified scene set with more statistical coverage can be reconstructed under limited historical extreme data; and a multi-time-scale operation scene basis with typicality and extreme property is provided for power system planning.

Description

Technical Field

[0001] This invention relates to the field of power system planning technology, specifically to a flexible resource planning method, system, device, and storage medium for power systems that takes into account extreme weather at multiple time scales. Background Technology

[0002] In recent years, the power system has faced a dual challenge: on the one hand, industrial transformation and rising living standards have driven significant seasonal variations in electricity demand, with air conditioning load continuing to climb. On the other hand, frequent extreme weather events have exacerbated the supply-demand imbalance. Meanwhile, the high proportion of renewable energy connected to the grid has further amplified system volatility. Against this backdrop, tapping into the potential of flexible resources on the demand side and optimizing the synergistic effects of various types of flexible resources have become key pathways to addressing the insufficient flexibility of the power system.

[0003] Source-load-storage coordinated planning considering flexibility requirements has become a research hotspot for high-proportion renewable energy power systems. Flexibility planning needs to be based on the coordination of the entire "source-grid-load-storage" process, taking into account technical, economic, and safety constraints, and promoting the complementarity of multiple resource types. Existing technologies have proposed methods for coordinated planning of power system source, load, and storage resources considering flexibility requirements, achieving better economic and environmental benefits while meeting flexibility needs. They have studied the sources and balancing measures of multi-timescale flexibility requirements and proposed an iterative model for alternating planning and operation based on multi-timescale flexibility evaluation. Existing technologies consider the coordination of daily and multi-day timescale operating scenarios in the planning problem of high-proportion renewable energy power systems. However, the construction of extreme scenario sets often adopts model-driven methods, and the multi-day timescale extreme scenarios generated under conditions of limited historical samples and high feature dimensions differ significantly from real-world scenarios. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that existing power system resource planning methods have the problem that the multi-day timescale extreme scenarios generated by them have a large gap with the real-world scenarios when there are few historical samples and high feature dimensions.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a flexible resource planning method for power systems that considers extreme weather at multiple time scales, comprising generating extreme weather scenarios at multiple time scales based on a conditional diffusion model; constructing a flexible resource planning model for power systems at mixed time scales; using the Yalmip toolkit in Matlab to call the Gurobi solver to solve the flexible resource planning model for power systems at mixed time scales, and outputting the optimal planning scheme for flexible resources of the power system.

[0007] As a preferred embodiment of the flexible resource planning method for power systems considering extreme weather at multiple time scales as described in this invention, the generation of multi-time scale extreme weather scenarios based on the conditional diffusion model includes generating dimensionality-reduced meteorological feature identification and conditional sequences for different extreme weather conditions, extracting features such as daily maximum temperature, daily average precipitation, daily maximum wind speed, daily maximum wind speed fluctuation, daily maximum solar intensity, season, month, and holidays, and concatenating the multi-day features according to the number of days. The multi-day feature sequence is represented as follows: , , , , , , in, For feature l, a multi-day feature sequence, , , , , , , , These represent the daily maximum temperature, daily average precipitation, daily maximum wind speed, daily maximum wind speed fluctuation, daily maximum solar radiation intensity, seasonal characteristics, monthly characteristics, and holiday characteristics for day d, respectively. , , , These represent the normalized temperature, wind speed, light intensity, and precipitation at time t, respectively. , , , , , Let d represent 0-1 variables indicating whether day d falls in spring, summer, autumn, winter, weekend, or other statutory holidays. The total number of time periods on day d. The month in which day d falls. This refers to the number of days included in a multi-day extreme weather event. Extreme weather events that affect the power system are divided into two categories: extreme disaster scenarios and extreme stable weather. For extreme weather events, extract the characteristic rate of change within different time windows; For extreme stable weather, extract the maximum and minimum values ​​of features and the corresponding number of days, and extract extreme scenarios with a probability less than a threshold. For weather conditions with accurate dates and definitions, extreme scenarios can be extracted directly. After extracting extreme scenarios, different multi-day meteorological feature sequences within the same scenario are normalized and concatenated into a curve. The multi-day meteorological feature sequences are then clustered using the k-means clustering method. The multi-day feature sequences at the cluster centers are taken as the conditional sequences of the extreme weather scenarios. The conditional sequences formed by the k-th cluster centers and the scenario probability are represented as follows: , in, Let k be the conditional sequence for the k-th scenario. Let be the probability of the k-th scenario. Let K be the total number of scenarios within the k-th scenario category. This represents the total number of conditional sequence scenarios. Based on the correlation between the target object and different features, differentiated conditional sequences are extracted for different target objects. The features corresponding to the target object; The characteristics of the conditions for generating wind power output are: daily maximum temperature, daily maximum wind speed, daily maximum wind speed fluctuation, daily average precipitation, and seasonal characteristics. The conditional sequence characteristics for generating photovoltaic power output are daily maximum temperature, daily maximum solar irradiance, daily average precipitation, and seasonal characteristics; the conditional sequence characteristics for generating load are daily maximum temperature, daily average precipitation, monthly characteristics, and holiday characteristics.

[0008] As a preferred embodiment of the flexible resource planning method for power systems that takes into account extreme weather at multiple time scales as described in this invention, the step of generating extreme weather scenarios at multiple time scales based on the condition diffusion model includes obtaining the condition sequence corresponding to the extreme weather and then generating multi-day wind power, photovoltaic, and load scenarios respectively through the condition diffusion model. The diffusion model consists of two inverse processes: forward diffusion and reverse denoising. During the forward diffusion phase, the multi-day historical scenes in the original scene dataset are transformed into noisy scenes conforming to the prior distribution after n noise superpositions, which is represented by an n-order Markov chain: , , in, This is a multi-day historical scene from the original wind power, solar power, and load data sets. The scene sequence after n noise addition processes. Let n be the scaling factor for step n. Noise injected at each step; The inverse denoising process aims to use conditional sequences Guided by prior distribution, the probability distribution of the target scene is reconstructed from the prior distribution. Conditional probabilities are obtained through a learnable denoising network. And hope the posterior distribution With prior distribution Similarly, the inverse denoising process can also be represented as an nth-order Markov process: , , , , in, The conditional sequence input for the scene. Denoising noise mapped to the denoising network; , , All are scale adjustment coefficients during the noise addition process. The parameters to be calculated; The goal is to ensure that the generated scene follows the same distribution as the original historical scene, with the training objective being to minimize the posterior distribution. and prior distribution The KL divergence, which is equivalent to the variational lower bound of the likelihood function, is expressed as the loss function of the conditional diffusion model: , in, Let be the loss function of the conditional diffusion model. Let be the expectation of the logarithmic joint probability; The input condition sequence is established through the denoising process in the diffusion model. The mapping to the target sequence P is achieved by training multi-day extreme weather scenario condition diffusion models for wind power, photovoltaics, and load respectively, and then inputting condition sequences under different extreme scenarios. After the reverse denoising process, the target sequence generated in the k-th clustering scenario is output. The probability of the scenario is , as input for multi-day timescale operation scenarios; The daily timescale operation scenario is generated using conventional methods such as Monte Carlo sampling, or by using a diffusion model with a day count of 1. The scenario is generated in a model with no conditional sequence input and then reduced using the k-means clustering method. The final set of operating scenarios for the flexible resource planning model of the power system consists of a multi-day timescale operating scenario set and a daily timescale operating scenario set.

[0009] As a preferred embodiment of the flexible resource planning method for power systems considering extreme weather at multiple time scales as described in this invention, the construction of the flexible resource planning model for a power system with mixed time scales includes: the objective function of the multi-stage flexible resource planning model for power systems considering mixed time scale operating scenarios includes the comprehensive cost of flexible resource investment costs and power system operating costs; the flexible resource investment costs include multi-stage energy storage construction costs, demand response resource modification costs, and equipment residual value; and the power system operating costs under each scenario include multi-type resource operating costs, maintenance costs, carbon emission costs, and risk costs, expressed as: , , , , , , , , in, , , , , , These are the flexible resource investment costs, operating costs, risk costs, carbon emission costs, maintenance costs, and equipment residual value of the power system. The number of typical operating scenarios, The total number of scheduling periods for scenario s. Let be the probability of scenario s occurring. and These represent the investment costs for energy storage and demand-side resources, respectively. The system operating cost for period t in scenario s of year y is... For the unit Unit coal consumption cost Coal consumption, For unit start-up and shutdown costs, , These are the unit shutdown and startup status variables, respectively. The operating cost per unit power of energy storage, , For energy storage The amount of charge and discharge, This is the collection of all system resources, including all traditional units. renewable energy Energy storage and demand-side resources ; and The unit cost of upper and lower standby capacity for resource i, respectively. and These are the upper and lower reserve capacities of resource i, respectively. The cost per unit capacity of carbon emission rights. Let represent the carbon emissions of resource i during time period t in scenario s in year y. This represents the maximum permissible carbon emissions for year y. It refers to the total number of years the power generation resources operate, starting from the investment year y until the end of the planning period. It refers to the lifespan of power generation resources. It is the net residual value rate. It is the cost of load loss per unit of electricity. For the load shedding during period t under scenario s in year y, It is the cost of curtailing wind and solar power per unit of electricity. For the curtailed wind and solar power volume during time period t under scenario s in year y; and These represent the total number of traditional generating units and energy storage resources, respectively.

[0010] As a preferred embodiment of the flexible resource planning method for power systems that takes into account extreme weather at multiple time scales as described in this invention, the construction of the flexible resource planning model for power systems at mixed time scales includes the following constraints: energy storage resource constraints, demand response resource constraints, traditional unit operation constraints, system power balance constraints, and system flexibility constraints. Demand response resource constraints are represented as follows: , , , , in, The investment cost of demand response resources in year y. For the types and quantities of resources required to respond to demand, Let i be the unit investment cost of the i-th type of demand response resource in year y. The transformation ratio of the i-th type of demand response resources in year y. The maximum transformation capacity of the resource for responding to the i-th type of demand. y is the discount rate, and y is the planning year index. Let i be the operating cost of the resource for responding to the i-th type of demand in the t-th time period of the y-th year under scenario s. Let i be the unit capacity cost of the demand response resource of type i in the t-th time period of year y under scenario s. Let i be the actual response volume of the resource for responding to the i-th type of demand in the t-th time period of the y-th year under scenario s. and These are the quadratic and primary coefficients of the unit electricity cost of the i-th type of demand response resource in the y-th year and t-th time period under scenario s. This represents the maximum response ratio of resources for the i-th type of demand in the t-th time period of the y-th year under scenario s. It is the total number of years in the planning period. This represents the total number of resources available for demand response.

[0011] Considering the learning and fatigue effects of demand response resources, the cost coefficient and maximum response ratio Both are functions of construction time and continuous load gap. Proportional to unit cost The proportion of the maximum response to the total capacity is expressed as follows: , , , , in, For the first Cost change coefficient of the i-th type of demand response in the t-th period of the year. For the first The maximum response ratio change coefficient of the i-th type of demand response resource in the t-th time period of the year is represented by the learning-fatigue effect function. and These are the initial values ​​of the resource learning-fatigue effect function for the i-th type of demand response. and Let be the intensity coefficient of the resource learning effect of the i-th type of demand response. Let be the learning rate coefficient of the resource learning effect of the i-th type of demand response. and Let be the intensity coefficient of the resource fatigue effect in the i-th type of demand response. The baseline time at which the fatigue effect begins is used; the coefficients are obtained through regression analysis based on historical response data. The basic unit operating cost coefficient for the i-th type of demand response resource in year y. This represents the baseline maximum response ratio for the i-th type of demand. Energy storage resource constraints are represented as follows: , , , in, For energy storage The construction capacity in year y For energy storage The indicator of whether construction will commence in year y is a 0-1 variable. and For energy storage The maximum and minimum investment capacity per transaction, For energy storage Unit capacity cost in year y; Energy storage resource constraints also include: , , , , , , , in, , These are the upper limits of the initial charge and discharge power of the energy storage, respectively. , The maximum charging and discharging power per unit capacity of energy storage to be constructed. For energy storage capacity, , These are the upper and lower limits of the initial energy storage capacity, respectively. , The upper and lower limits of energy storage capacity for the unit capacity of energy storage to be constructed; This is the self-loss coefficient for energy storage discharge. , These are the charge and discharge coefficients of the energy storage, respectively.

[0012] As a preferred embodiment of the flexible resource planning method for power systems that considers extreme weather at multiple time scales as described in this invention, the construction of the flexible resource planning model for a hybrid time scale power system includes the following: traditional unit operation constraints are expressed as follows: , , , , , , , in, Indicates traditional units The running status during scheduling period t, and These represent traditional generating units. The maximum value of landslide and uphill power, For traditional generator sets The duration of power-on, For traditional generator sets The duration of the downtime, and For traditional units Minimum start-up and shutdown times, It is a constant; The power balance constraint is expressed as: , in, For external power input.

[0013] As a preferred embodiment of the flexible resource planning method for power systems that considers extreme weather at multiple time scales as described in this invention, the construction of the flexible resource planning model for a hybrid time scale power system includes the following: the up- and down-reserve requirements of the power system are affected by renewable energy and load fluctuations, and the system flexibility balance constraint is expressed as: , , in, For renewable energy In planning the year's scenario, the effort put in at the moment t, To plan the system load at time t in scenario y of year y, , , Resources on the power generation side Demand-side resources and energy storage resources The backup capacity at time t in the planning of scenario y in year y , , Resources on the power generation side Demand-side resources and energy storage resources Under the planning of the backup capacity in scenario y at time t. and These are the upstream and downstream reserve demand coefficients caused by fluctuations in renewable energy, respectively. and These are the upper and lower reserve requirement coefficients caused by load fluctuations and system failures, respectively. This represents the total amount of renewable energy resources.

[0014] Another objective of this invention is to provide a flexible resource planning system for power systems that takes into account extreme weather at multiple time scales. This system can generate extreme weather scenarios at multiple time scales and construct a flexible resource planning model for power systems at mixed time scales by using a conditional diffusion model. This solves the problem that current power system resource planning methods generate multi-day extreme weather scenarios at multiple time scales that differ significantly from real-world scenarios when there are few historical samples and high feature dimensions.

[0015] As a preferred embodiment of the flexible resource planning system for power systems that considers extreme weather at multiple time scales as described in this invention, the system includes: a multi-time-scale extreme weather scenario construction module, a mixed-time-scale power system flexible resource planning model construction module, and an optimal planning scheme output module. The multi-time-scale extreme weather scenario construction module generates multi-time-scale extreme weather scenarios based on a conditional diffusion model, including dimensionality reduction meteorological feature identification and conditional sequence generation for different extreme weather conditions, as well as the construction of a conditional diffusion model for multi-time-scale extreme weather scenarios. The mixed-time-scale power system flexible resource planning model construction module constructs a mixed-time-scale power system flexible resource planning model and applies constraints. The optimal planning scheme output module outputs the planning model to obtain the optimal planning scheme for the flexible resources of the power system.

[0016] Another object of the present invention is to provide a flexible resource planning device for power systems that takes into account extreme weather at multiple time scales, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales.

[0017] Another object of the present invention is to provide a power system flexible resource planning storage medium that takes into account extreme weather at multiple time scales, wherein a computer program is stored thereon, and when the computer program is executed by a processor, the steps of the power system flexible resource planning method that takes into account extreme weather at multiple time scales are implemented.

[0018] The beneficial effects of this invention are as follows: The flexible resource planning method for power systems that takes into account extreme weather at multiple time scales provided by this invention generates multi-day extreme weather scenarios based on a conditional diffusion model. This method can accurately characterize the spatiotemporal features of multi-day extreme weather events such as typhoons and high temperatures. Compared with other methods, it can reconstruct a more statistically comprehensive and diverse set of scenarios under limited historical extreme data, providing a foundation for power system planning with multi-time scale operating scenarios that are both typical and extreme. The diversified flexible resource allocation strategy effectively balances the system's economy and sufficiency. Compared with other planning models, it can reduce the total cost of the power system during the planning period, ensuring the system's ability to cope with extreme weather while achieving reasonable allocation and time-series optimization of flexible resources. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of 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.

[0020] Figure 1 The above is an overall flowchart of a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, as provided in Embodiment 1 of the present invention.

[0021] Figure 2 This is a multi-day extreme weather scenario generation result diagram provided by Embodiment 2 of the present invention, which is a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales.

[0022] Figure 3 This is a real-world scenario diagram of a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, as provided in Embodiment 2 of the present invention.

[0023] Figure 4 This is a diagram showing the planning results of a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, as provided in Embodiment 2 of the present invention.

[0024] Figure 5 The figure shows the simulation results of optimized scheduling in a continuous multi-day high temperature extreme weather scenario, which is a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, provided in Embodiment 2 of the present invention.

[0025] Figure 6 This is a cumulative new capacity diagram of different resources during the planning period under various planning methods of a flexible resource planning system for power systems that takes into account extreme weather at multiple time scales, as provided in Embodiment 2 of the present invention. Detailed Implementation

[0026] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0027] Example 1, referring to Figure 1 As an embodiment of the present invention, a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales is provided, comprising: S1: Generate extreme weather scenarios at multiple time scales based on conditional diffusion models.

[0028] Furthermore, based on the conditional diffusion model, multi-timescale extreme weather scenarios are generated, including dimensionality reduction meteorological feature identification and conditional sequence generation for different extreme weather events, and the construction of a multi-timescale extreme weather scenario conditional diffusion model. The process is as follows: For dimensionality reduction meteorological feature identification and conditional sequence generation for different extreme weather conditions, features such as daily maximum temperature, daily average precipitation, daily maximum wind speed, daily maximum wind speed fluctuation, daily maximum solar intensity, season, month, and holidays are extracted. Then, the multi-day features are concatenated according to the number of days in the multi-day sequence. Therefore, the multi-day feature sequence can be represented as: , , , , , , in, Let l be a multi-day feature sequence, where , , , , , , , These represent the daily maximum temperature, daily average precipitation, daily maximum wind speed, daily maximum wind speed fluctuation, daily maximum solar intensity, seasonal characteristics, monthly characteristics, and holiday characteristics for day d, respectively. , , , These are the normalized temperature, wind speed, light intensity, and precipitation at time t, respectively. , , , , , These are 0-1 variables representing whether day d falls in spring, summer, autumn, winter, weekend, or other statutory holidays; The total number of time periods on day d. The month in which day d falls. This refers to the number of days included in a multi-day extreme weather event.

[0029] Extreme weather events with significant impact on the power system are categorized into two types: extreme disaster scenarios (such as typhoons) and extreme stable weather (such as heat waves and prolonged periods of calm). For extreme disaster weather, due to the drastic environmental changes during these periods, the rate of change of features (such as daily average precipitation) within different time windows can be extracted. For extreme stable weather, the maximum and minimum values ​​of features (such as daily maximum temperature) and their corresponding durations are extracted, and then extreme scenarios with probabilities less than a certain threshold are extracted. For weather events with precise dates and definitions, such as typhoons, extreme scenarios can also be directly extracted. After extracting the extreme scenarios, different multi-day meteorological feature sequences within the same scenario are normalized and concatenated into a curve. The multi-day meteorological feature sequences are then clustered using the k-means clustering method, and the optimal number of clusters can be determined using the elbow method and the silhouette coefficient method. Finally, the multi-day feature sequences of the cluster centers are taken as the conditional sequences of the extreme weather scenarios. The conditional sequence formed by the k-th cluster centers and the scenario probability can be expressed as: , in, Let k be the conditional sequence for the k-th scenario. The probability of the k-th scenario; Let K be the total number of scenarios within the k-th scenario category. This represents the total number of conditional sequence scenarios; it allows for the extraction of differentiated conditional sequences for different target objects based on the correlation between the target object and different features. For the characteristics corresponding to the target object, the present invention takes the following conditional sequence characteristics for generating wind power output as maximum daily temperature, maximum daily wind speed, maximum daily wind speed fluctuation, average daily precipitation, and seasonal characteristics; the following conditional sequence characteristics for generating photovoltaic power output as maximum daily temperature, maximum daily irradiance, average daily precipitation, and seasonal characteristics; and the following conditional sequence characteristics for generating load as maximum daily temperature, average daily precipitation, monthly characteristics, and holiday characteristics.

[0030] It should be noted that the conditional diffusion model for multi-timescale extreme weather scenarios is constructed by obtaining the conditional sequences corresponding to extreme weather events. The model is then used to generate multi-day wind power, photovoltaic, and load scenarios. The diffusion model consists of two inverse processes: forward diffusion and reverse denoising. The forward process gradually injects noise to transform the original scenario data into a priori distribution, while the reverse process recovers the target scenario from the noise distribution based on the conditional sequences. The conditional diffusion model takes the conditional sequences as input and embeds them into each step of the noise addition and denoising process. By establishing a mapping between the conditional sequences and the target sequences, the target sequence can be generated in a controllable manner. Because the conditional diffusion model can generate differentiated scenarios with the same distribution as the original scenario from random noise and has the ability to extract features from small samples, it can reconstruct a diverse set of scenarios with statistical coverage under limited historical extreme data, providing power system planning with operational scenario samples that are both typical and extreme.

[0031] During the forward diffusion phase, the multi-day historical scenes in the original scene dataset are transformed into noisy scenes conforming to the prior distribution after n noise superpositions. This process is achieved through an n-order Markov chain: , , in, This is a multi-day historical scene from the original wind power, solar power, and load data sets. The scene sequence after n noise addition processes. Let n be the scaling factor for step n. Noise injected at each step.

[0032] When the number of iterations N is large enough It will converge to a predefined prior distribution (choosing a normal distribution). As a prior distribution, and since the same noise is added at each step, denoising can be achieved through a reverse process. The reverse denoising process aims to denoise through conditional sequences. Guided by prior distribution, the probability distribution of the target scene is reconstructed from the prior distribution. This process obtains conditional probabilities through a learnable denoising network. And hope the posterior distribution With prior distribution Similarly, the inverse denoising process can also be represented as an nth-order Markov process: , , , , in, The conditional sequence input for the scene. Denoising noise mapped to the denoising network; , , All are scale adjustment coefficients during the noise addition process. The calculated parameters, when and When the size is small, the forward diffusion and reverse denoising processes have the same functional form.

[0033] To ensure that the generated scene distribution is consistent with the original historical scene distribution, the training objective is to minimize the posterior distribution. and prior distribution The KL (Kullback-Leibler) divergence, which can be equivalent to the variational lower bound (VLB) of the likelihood function, can be expressed as: , in, Let be the loss function of the conditional diffusion model. Let be the expectation of the logarithmic joint probability; The core UNet structure in the conditional diffusion model denoising process includes an encoding-decoding process, capable of processing information at multiple scales. It can capture local details of new energy output and load scenarios while maintaining global consistency across multiple days of extreme weather. The convolutional layers of the UNet structure proposed in this invention can extract temporal features, the attention mechanism module can capture long-distance dependencies in the sequence, the downsampling layer can reduce feature dimensionality and extract high-level features, the upsampling layer restores feature dimensionality, and the skip connections directly pass encoder features to the decoder, enabling the fusion of features from different levels.

[0034] Therefore, the input condition sequence is established through the denoising process in the diffusion model. The mapping to the target sequence P is achieved by training multi-day extreme weather scenario condition diffusion models for wind power, photovoltaics, and load respectively, and then inputting condition sequences under different extreme scenarios. After the reverse denoising process, the target sequence generated in the k-th clustering scenario can be obtained. The probability of the scenario is The daily timescale operating scenarios serve as inputs for the multi-day timescale operating scenarios. Due to the large number of samples, daily timescale operating scenarios can be generated using conventional methods such as Monte Carlo sampling, or a diffusion model can be used, setting the number of days to 1. This generates a large number of scenarios from the unconditional sequence input model, which are then reduced using k-means clustering. Ultimately, the operating scenario set of the power system flexible resource planning model consists of both the multi-day timescale operating scenario set and the daily timescale operating scenario set.

[0035] S2: Construct a flexible resource planning model for a hybrid time-scale power system.

[0036] Furthermore, a flexible resource planning model for a power system with a hybrid time scale is constructed, including the objective function and constraints, as follows: The objective function of the multi-stage planning model for flexible resources in power systems considering mixed time-scale operation scenarios is a comprehensive cost including flexible resource investment costs and power system operation costs. Flexible resource investment costs include multi-stage energy storage construction costs, demand response resource modification costs, and equipment residual value. Power system operation costs in each scenario include multi-type resource operation costs, maintenance costs, carbon emission costs, and risk costs, which can be expressed as: , , , , , , , , in, , , , , , These are the flexible resource investment costs, operating costs, risk costs, carbon emission costs, maintenance costs, and equipment residual value of the power system. Number of typical operating scenarios; The total number of scheduling periods for scenario s; Let s be the probability of scenario s occurring. and These are the investment costs for energy storage and demand-side resources, respectively. The system operating cost for period t in scenario s of year y; For the unit Unit coal consumption cost Coal consumption; For unit start-up and shutdown costs, , These are the unit shutdown and startup status variables, respectively; The operating cost per unit power of energy storage, , For energy storage The charge and discharge amounts; to simplify the formula, It is the collection of all system resources, including all traditional units. renewable energy Energy storage and demand-side resources ; and The unit cost of upper and lower standby capacity for resource i, respectively; and These are the upper and lower reserve capacities of resource i, respectively; The cost per unit capacity of carbon emission rights; Let i be the carbon emissions of resource i in scenario s during period t in year y; This represents the maximum permissible carbon emissions for year y. It is the total number of years the power generation resources operate, starting from the investment year y until the end of the planning period. It refers to the lifespan of power generation resources; It is the net residual value rate; It is the cost of load loss per unit of electricity; The power loss during time period t under scenario s in year y; It is the cost of curtailing wind and solar power per unit of electricity; For the curtailed wind and solar power volume during time period t under scenario s in year y; and These represent the total number of traditional generating units and energy storage resources, respectively.

[0037] The constraints of the power system flexible resource multi-stage planning model considering mixed time-scale operation scenarios include energy storage resource constraints, demand response resource constraints, traditional unit operation constraints, system power balance constraints, and system flexibility constraints.

[0038] Demand response resource constraints are represented as follows: , , , , in, The investment cost of demand response resources in year y. For the types and quantities of resources required to respond to demand, Let i be the unit investment cost of the i-th type of demand response resource in year y. The transformation ratio of the i-th type of demand response resources in year y. The maximum transformation capacity of the resource for responding to the i-th type of demand. y is the discount rate, and y is the index of the planning year. Let i be the operating cost of the resource for responding to the i-th type of demand in the t-th time period of the y-th year under scenario s. Let i be the unit capacity cost of the demand response resource of type i in the t-th time period of year y under scenario s. Let i be the actual response volume of the resource for responding to the i-th type of demand in the t-th time period of the y-th year under scenario s. and These are the quadratic and primary coefficients of the unit electricity cost of the i-th type of demand response resource in the y-th year and t-th time period under scenario s; The maximum response ratio of the resource for responding to the i-th type of demand in the t-th time period of the y-th year under scenario s; It is the total number of years in the planning period. This represents the total number of demand response resources. The cost coefficient is calculated by considering the learning and fatigue effects of these resources. and maximum response ratio Both are functions of construction time and continuous load gap. The larger the value, the higher the unit cost. This represents the proportion of the maximum response to the total capacity. A larger value indicates lower uncertainty. The specific values ​​are as follows: , , , , in, For the first Cost change coefficient of the i-th type of demand response in the t-th period of the year. For the first The maximum response ratio change coefficient of the i-th type of demand response resource in the t-th time period of the year is represented by the learning-fatigue effect function, where... and These are the initial values ​​of the resource learning-fatigue effect function for the i-th type of demand response. and is the intensity coefficient of the learning effect of the resource response to the i-th type of demand; the larger the value, the greater the impact of the learning effect. is the learning rate coefficient of the learning effect of the resource response to the i-th type of demand; the larger the value, the faster the learning speed. and Let be the intensity coefficient of the resource fatigue effect in the i-th type of demand response; the larger the value, the greater the impact of the fatigue effect. The reference time for the onset of fatigue effect is given, and the coefficients are obtained by regression based on historical response data. The basic unit operating cost coefficient for the i-th type of demand response resource in year y. This represents the baseline maximum response ratio for the i-th type of demand.

[0039] Energy storage resource constraints are expressed as: , , , in, For energy storage The construction capacity in year y; For energy storage The indicator of whether construction will commence in year y is a 0-1 variable; and For energy storage The maximum and minimum investment capacity per transaction; For energy storage Unit capacity cost in year y.

[0040] Energy storage resource constraints also include: , , , , , , , in, , These are the upper limits of the initial charging and discharging power of the energy storage; , The maximum charging and discharging power per unit capacity of energy storage to be constructed; For energy storage capacity, , These are the upper and lower limits of the initial energy storage capacity, respectively. , The upper and lower limits of energy storage capacity for the unit capacity of energy storage to be constructed; This is the self-loss coefficient for energy storage discharge; , These are the charge and discharge coefficients of the energy storage, respectively.

[0041] Traditional unit operating constraints are expressed as follows: , , , , , , , in, Indicates traditional units The running status during scheduling period t; and These represent traditional generating units. The maximum values ​​of landslide and climbing power; For traditional generator sets The duration of power-on, For traditional generator sets The duration of the downtime; and For traditional units Minimum start-up and shutdown times; It is a very large constant.

[0042] The power balance constraint is expressed as: , in, For external power input.

[0043] Flexibility balance constraints include the requirement that power system flexibility constraints ensure sufficient power system reserves. The upstream and downstream reserve requirements of the power system are affected by renewable energy and load fluctuations. The system flexibility balance constraints can be expressed as: , , in, For renewable energy The contribution of planning the year's scenario at time t; To plan the system load at time t in scenario y of year y; , , Resources on the power generation side Demand-side resources and energy storage resources Backup capacity at time t in scenario y of the planned year; , , Resources on the power generation side Demand-side resources and energy storage resources Backup capacity under the planning year y scenario s time t; and These are the upstream and downstream reserve demand coefficients caused by fluctuations in renewable energy; and These are the upper and lower reserve demand coefficients caused by load fluctuations and system accidents, respectively. This represents the total amount of renewable energy resources.

[0044] S3: Use the Yalmip toolkit in Matlab to call the Gurobi solver to solve the flexible resource planning model of the power system with mixed time scales, and output the optimal planning scheme for the flexible resources of the power system.

[0045] Furthermore, by using the Yalmip toolkit in Matlab to call Gurobi to solve the above planning model, the optimal planning scheme for the flexible resources of the power system is obtained, which not only ensures the power system's ability to cope with extreme weather, but also realizes the rational allocation and timing optimization of flexible resources.

[0046] Example 2, refer to Figures 2-6 As an embodiment of the present invention, a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales is provided. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0047] First, a case study analysis was conducted using a regional power system with a planning period of 5 years. The installed capacity and load growth of various types of generating units during the planning period are shown in Table 1. Among the flexible resources, energy storage resources consider energy-type energy storage (such as pumped storage) and capacity-type energy storage (such as electrochemical energy storage). Demand-side resources are divided into two categories: active and controllable. The parameter settings for flexible resources are shown in Table 2.

[0048] Table 1. Installation and load growth of various types of generating units during the planning period. Resources / MW y1 y2 y3 y4 y5 Coal-fired power units 67997 67997 67997 67997 67997 Gas turbine unit 1450 1450 1450 1450 1450 Photovoltaic units 47270 51181 55416 60002 64967 wind turbine 6490 7124 7757 8391 9024 nuclear power units 9170 10936 12702 14468 16234 Hydropower units 13880 13880 13880 13880 13880 peak load 123000 128535 134319 140363 146680

[0049] Table 2 Flexible Resource Planning Parameters Flexible resource types Energy storage Capacity-type energy storage Active DR Controllable DR Initial capacity 44460MWh / 9880MW 6003MWh / 2058MW 6301MW 3038MW Initial unit investment cost / 10,000 RMB·MW⁻¹ 440.0 208.0 1.0 2.0 Unit maintenance rate % 2.5 0.5 0.5 0.5 Initial value of learning-fatigue effect \ \ 0.9 1 Learning effect strength \ \ 1 0.5 Learning rate \ \ 0.9 1.7 Fatigue effect intensity \ \ 0.001 0.001

[0050] Power system operation scenarios under extreme weather conditions at multiple time scales were generated. One scenario each from the generated scenarios for typhoon weather and sustained high-temperature weather were compared with the actual scenarios. The generated scenarios and actual scenarios for multi-day extreme weather are shown below. Figure 2 and Figure 3As shown in the figure, the multi-day extreme weather scenarios generated by the conditional diffusion model successfully reproduced the daily and multi-day continuous characteristics of wind power, photovoltaic power, and load under different meteorological conditions. In the typhoon scenario, wind power output exhibited drastic fluctuations, photovoltaic power output decreased significantly due to cloud cover, while the power load remained at a high and stable level of 0.7 to 0.9 of the maximum load. The model also captured the overall decline during typhoon landfall and the overall rebound after landfall. These characteristics are highly consistent with actual typhoon scenarios, proving that the model accurately captures the dynamic characteristics of various elements during typhoons. In the sustained high-temperature scenario, the wind power output generated by the model was generally low and stable, consistent with the characteristics of a high-temperature, low-wind environment. Photovoltaic power generation showed a clear daily cycle but a low peak value (approximately 0.5 of the installed capacity), and the load curve remained high with daily fluctuations. Comparison with actual high-temperature scenarios shows that the conditional diffusion model can successfully capture the typical patterns of various elements under sustained high-temperature weather. Therefore, the conditional diffusion model can accurately capture the characteristics of renewable energy output and load curves under different extreme weather conditions over multiple days with small sample sizes, accurately reflect the complex nonlinear relationship between meteorological conditions and power system elements, and at the same time ensure the diversity of generated scenarios. In power planning models, it can more comprehensively assess the operational risks under extreme weather conditions.

[0051] The planning results of various flexible resources are obtained, such as Figure 4 As shown, the optimized scheduling simulation results for the first and fifth planning years under the scenario of continuous high temperatures for multiple days are as follows: Figure 5 As shown.

[0052] To better demonstrate the superiority of this invention, the multi-year planning results optimized by the proposed method (denoted as M-SD) are compared with those optimized by a planning method that does not consider changes in the long-term and short-term behavior characteristics of demand (denoted as M-ND), a planning method for extreme weather generation based on deep clustering algorithms (denoted as M-DC), and a planning method for extreme weather generation based on conditional generative adversarial models (denoted as M-CG). The cumulative new capacity of different resources under each planning method in five planning years is shown in the figure. Figure 6 As shown in Table 3, the planning and operating costs under different planning methods are as follows.

[0053] Table 3 Comparison of planning and operating costs of different planning methods Planning Methods Total cost Planning costs Operating costs The M-SD proposed in this invention 168.87 23.39 145.48 M-ND 177.00 22.88 154.12 M-DC 174.82 31.87 142.95 M-CG 170.73 25.04 145.69

[0054] As shown in the table, the total planning-operation cost of the flexible resource planning method proposed in this invention, which considers extreme weather risks and the long- and short-term behavior characteristics of the demand side, is RMB 16.887 billion. This is 4.59%, 3.40%, and 1.09% lower than the three comparative methods, M-ND, M-DC, and M-CG, respectively, thus improving the economic efficiency of flexible resource investment and operation in the power system. As shown in the figure, regarding DR resource investment, the planning method considering changes in the long- and short-term behavior characteristics of the demand side made large-scale investments in the early planning stages (the first and second planning years), while the M-ND method postponed DR resource investment to the fifth planning year. Although both methods achieved a total planned capacity of approximately 16,000 MW, the available DR capacity of the M-ND method was only 53.9% of that of the proposed method. In terms of energy storage resource allocation, the M-SD, M-ND, and M-CG methods are more cautious in their investment in capacity-type energy storage, significantly lower than the investment capacity of the M-DC method, thus avoiding over-investment. For energy-type energy storage, the M-SD method adopts a gradual investment strategy, reaching 10860MW by the fifth planning year, similar to the final investment amount of other methods but with a more reasonable investment pace. Comprehensive cost analysis shows that although the proposed M-SD method has a slightly higher planning cost than the M-ND method, its operating cost is reduced by 5.61%, resulting in a 4.59% reduction in total cost. Compared to the M-DC and M-CG methods, the proposed M-SD method reduces planning costs, indicating that by moderately increasing investment in the early stages and rationally allocating multiple types of flexible resources, the M-SD method effectively balances investment and operating costs. The conditional diffusion model can better capture extreme weather characteristics, generating more representative extreme weather scenarios and avoiding over-investment. Compared to the three comparative methods, the proposed planning method combining the conditional diffusion model and dynamic evolution of user behavior saves an average of 3.03% of the total cost, achieving the optimal balance between the economy and reliability of flexible resource planning.

[0055] In summary, in response to the combined uncertainty challenges brought to the power system by extreme weather and user behavior, this invention proposes a flexible resource collaborative planning method for the power system that takes into account multi-scale extreme weather risks and demand-side behavioral characteristics. First, a method for generating multi-day extreme weather scenarios based on a conditional diffusion model is proposed. This method can accurately characterize the spatiotemporal features of multi-day extreme weather events such as typhoons and high temperatures. Compared with other methods, it can reconstruct a more statistically comprehensive and diverse set of scenarios with limited historical extreme data, providing a foundation for multi-timescale operation scenarios that are both typical and extreme for power system planning. Second, a demand-side behavior characteristic model considering learning and fatigue effects is established, revealing the dynamic evolution of demand response resources across multiple timescales. Finally, the effectiveness and superiority of the proposed method are verified using actual meteorological and power data. Simulation results show that this method reduces the total cost by 4.59%, 3.40%, and 1.09% compared to the M-ND, M-DC, and M-CG planning models, respectively. The diversified and flexible resource allocation strategy effectively balances the system's economy and sufficiency. Early demand-side resource investment cultivates user participation and reduces long-term response uncertainty. In the mid-to-late stages, investment in multiple types of energy storage is increased, forming a resource combination that is synergistic and complementary between demand response and energy storage. This ensures the system's ability to cope with extreme weather while achieving reasonable allocation and time-series optimization of investment costs.

[0056] Example 3, one embodiment of the present invention, provides a flexible resource planning system for power systems that takes into account extreme weather at multiple time scales, including a time-scale extreme weather scenario construction module, a hybrid time-scale power system flexible resource planning model construction module, and an optimal planning scheme output module.

[0057] The multi-timescale extreme weather scenario construction module is used to generate multi-timescale extreme weather scenarios based on the conditional diffusion model. This includes dimensionality reduction meteorological feature identification and conditional sequence generation for different extreme weather conditions, as well as the construction of a conditional diffusion model for multi-timescale extreme weather scenarios. The mixed-timescale power system flexible resource planning model construction module is used to construct a mixed-timescale power system flexible resource planning model and apply constraints. The optimal planning scheme output module is used to output the planning model and obtain the optimal planning scheme for the flexible resources of the power system.

[0058] This embodiment also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, as proposed in the above embodiment.

[0059] This embodiment also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements a flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, as proposed in the above embodiment.

[0060] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0061] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0062] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0063] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0064] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, characterized in that, include: Generating extreme weather scenarios across multiple time scales based on conditional diffusion models; Constructing a flexible resource planning model for a hybrid time-scale power system; The Yalmip toolkit in Matlab is used to call the Gurobi solver to solve the flexible resource planning model of the power system with mixed time scales, and the optimal planning scheme for the flexible resources of the power system is output.

2. The flexible resource planning method for power systems considering extreme weather at multiple time scales as described in claim 1, characterized in that: The multi-timescale extreme weather scenarios generated based on the conditional diffusion model include... Dimensionally reduced meteorological feature identification and conditional sequences for different extreme weather conditions are generated. Features such as daily maximum temperature, daily average precipitation, daily maximum wind speed, daily maximum wind speed fluctuation, daily maximum solar intensity, season, month, and holidays are extracted. These multi-day features are concatenated based on the number of days in the multi-day sequence, which is represented as follows: ， ， ， ， ， ， in, For feature l, a multi-day feature sequence, , , , , , , , These represent the daily maximum temperature, daily average precipitation, daily maximum wind speed, daily maximum wind speed fluctuation, daily maximum solar radiation intensity, seasonal characteristics, monthly characteristics, and holiday characteristics for day d, respectively. , , , These represent the normalized temperature, wind speed, light intensity, and precipitation at time t, respectively. , , , , , Let d represent 0-1 variables indicating whether day d falls in spring, summer, autumn, winter, weekend, or other statutory holidays. The total number of time periods on day d. The month in which day d falls. This refers to the number of days included in a multi-day extreme weather event. Extreme weather events that affect the power system are divided into two categories: extreme disaster scenarios and extreme stable weather. For extreme weather events, extract the characteristic rate of change within different time windows; For extreme stable weather, extract the maximum and minimum values ​​of features and the corresponding number of days, and extract extreme scenarios with a probability less than a threshold. For weather conditions with accurate dates and definitions, extreme scenarios can be extracted directly. After extracting extreme scenarios, different multi-day meteorological feature sequences within the same scenario are normalized and concatenated into a curve. The multi-day meteorological feature sequences are then clustered using the k-means clustering method. The multi-day feature sequences at the cluster centers are taken as the conditional sequences of the extreme weather scenarios. The conditional sequences formed by the k-th cluster centers and the scenario probability are represented as follows: ， in, Let k be the conditional sequence for the k-th scenario. Let be the probability of the k-th scenario. Let K be the total number of scenarios within the k-th scenario category. This represents the total number of conditional sequence scenarios. Based on the correlation between the target object and different features, differentiated conditional sequences are extracted for different target objects. The features corresponding to the target object; The characteristics of the conditions for generating wind power output are: daily maximum temperature, daily maximum wind speed, daily maximum wind speed fluctuation, daily average precipitation, and seasonal characteristics. The conditional sequence characteristics for generating photovoltaic power output are daily maximum temperature, daily maximum solar irradiance, daily average precipitation, and seasonal characteristics; the conditional sequence characteristics for generating load are daily maximum temperature, daily average precipitation, monthly characteristics, and holiday characteristics.

3. The flexible resource planning method for power systems considering extreme weather at multiple time scales as described in claim 1 or 2, characterized in that: The multi-timescale extreme weather scenarios generated based on the conditional diffusion model include... After obtaining the condition sequence corresponding to extreme weather, the multi-day wind power, photovoltaic, and load scenarios are generated respectively using the condition diffusion model; The diffusion model consists of two inverse processes: forward diffusion and reverse denoising. During the forward diffusion phase, the multi-day historical scenes in the original scene dataset are transformed into noisy scenes conforming to the prior distribution after n noise superpositions, which is represented by an n-order Markov chain: ， ， in, This is a multi-day historical scene from the original wind power, solar power, and load data sets. The scene sequence after n noise addition processes. Let n be the scaling factor for step n. Noise injected at each step; The inverse denoising process aims to use conditional sequences Guided by prior distribution, the probability distribution of the target scene is reconstructed from the prior distribution. Conditional probabilities are obtained through a learnable denoising network. And hope the posterior distribution With prior distribution Similarly, the inverse denoising process can also be represented as an nth-order Markov process: ， ， ， ， in, The conditional sequence input for the scene. Denoising noise mapped to the denoising network; , , All are scale adjustment coefficients generated during the noise addition process. The parameters to be calculated; The goal is to ensure that the generated scene follows the same distribution as the original historical scene, with the training objective being to minimize the posterior distribution. and prior distribution The KL divergence, which is equivalent to the variational lower bound of the likelihood function, is expressed as the loss function of the conditional diffusion model: ， in, Let be the loss function of the conditional diffusion model. Let be the expectation of the logarithmic joint probability; The input condition sequence is established through the denoising process in the diffusion model. The mapping to the target sequence P is achieved by training multi-day extreme weather scenario condition diffusion models for wind power, photovoltaics, and load respectively, and then inputting condition sequences under different extreme scenarios. After the reverse denoising process, the target sequence generated in the k-th clustering scenario is output. The probability of the scenario is , as input for multi-day timescale operation scenarios; The daily timescale operation scenario is generated using conventional methods such as Monte Carlo sampling, or by using a diffusion model with a day count of 1. The scenario is generated in a model with no conditional sequence input and then reduced using the k-means clustering method. The final set of operating scenarios for the flexible resource planning model of the power system consists of a multi-day timescale operating scenario set and a daily timescale operating scenario set.

4. The flexible resource planning method for power systems considering extreme weather at multiple time scales as described in claim 3, characterized in that: The construction of a flexible resource planning model for a hybrid time-scale power system includes... The objective function of the multi-stage planning model for flexible resources in a power system considering mixed time-scale operation scenarios includes the combined cost of flexible resource investment and power system operation costs. Flexible resource investment costs include multi-stage energy storage construction costs, demand response resource modification costs, and equipment residual value. Power system operation costs under each scenario include multi-type resource operation costs, maintenance costs, carbon emission costs, and risk costs, expressed as: ， ， ， ， ， ， ， ， in, , , , , , These are the flexible resource investment costs, operating costs, risk costs, carbon emission costs, maintenance costs, and equipment residual value of the power system. The number of typical operating scenarios, The total number of scheduling periods for scenario s. Let be the probability of scenario s occurring. and These represent the investment costs for energy storage and demand-side resources, respectively. The system operating cost for period t in scenario s of year y is... For the unit Unit coal consumption cost Coal consumption, For unit start-up and shutdown costs, , These are the unit shutdown and startup status variables, respectively. The operating cost per unit power of energy storage , For energy storage The amount of charge and discharge, This is the collection of all system resources, including all traditional units. renewable energy Energy storage and demand-side resources ; and The unit cost of upper and lower standby capacity for resource i, respectively. and These are the upper and lower reserve capacities of resource i, respectively. The cost per unit capacity of carbon emission rights. Let represent the carbon emissions of resource i in scenario s during period t in year y. This represents the maximum permissible carbon emissions for year y. It refers to the total number of years the power generation resources operate, starting from the investment year y until the end of the planning period. It refers to the lifespan of power generation resources. It is the net residual value rate. It is the cost of load loss per unit of electricity. For the load shedding during period t under scenario s in year y, It is the cost of curtailing wind and solar power per unit of electricity. For the curtailed wind and solar power volume during time period t under scenario s in year y; and These represent the total number of traditional generating units and energy storage resources, respectively.

5. The flexible resource planning method for power systems considering extreme weather at multiple time scales as described in claim 4, characterized in that: The construction of a flexible resource planning model for a hybrid time-scale power system includes... The constraints of the multi-stage planning model for flexible resources of power systems considering mixed time-scale operation scenarios include energy storage resource constraints, demand response resource constraints, traditional unit operation constraints, system power balance constraints, and system flexibility constraints. Demand response resource constraints are represented as follows: ， ， ， ， in, The investment cost of demand response resources in year y. For the types and quantities of resources required to respond to demand, Let i be the unit investment cost of the i-th type of demand response resource in year y. The transformation ratio of the i-th type of demand response resources in year y. The maximum transformation capacity of the resource for responding to the i-th type of demand. Here, y represents the discount rate, and y represents the planning year index. Let i be the operating cost of the resource for responding to the i-th type of demand in the t-th time period of the y-th year under scenario s. Let i be the unit capacity cost of the demand response resource of type i in the t-th time period of year y under scenario s. Let i be the actual response volume of the resource for responding to the i-th type of demand in the t-th time period of the y-th year under scenario s. and These are the quadratic and primary coefficients of the unit electricity cost of the i-th type of demand response resource in the y-th year and t-th time period under scenario s. This represents the maximum response ratio of resources for the i-th type of demand in the t-th time period of the y-th year under scenario s. It is the total number of years in the planning period. This represents the total number of resources available for demand response. Considering the learning and fatigue effects of demand response resources, the cost coefficient and maximum response ratio Both are functions of construction time and continuous load gap. Proportional to unit cost The proportion of the maximum response to the total capacity is expressed as follows: ， ， ， ， in, For the first Cost change coefficient of the i-th type of demand response in the t-th period of the year. For the first The maximum response ratio change coefficient of the i-th type of demand response resource in the t-th time period of the year is represented by the learning-fatigue effect function. and These are the initial values ​​of the resource learning-fatigue effect function for the i-th type of demand response. and Let be the intensity coefficient of the resource learning effect of the i-th type of demand response. Let be the learning rate coefficient of the resource learning effect of the i-th type of demand response. and Let be the intensity coefficient of the resource fatigue effect in the i-th type of demand response. The baseline time at which the fatigue effect begins is used; the coefficients are obtained through regression analysis based on historical response data. The basic unit operating cost coefficient for the i-th type of demand response resource in year y. This represents the baseline maximum response ratio for the i-th type of demand. Energy storage resource constraints are represented as follows: ，， ， ， in, For energy storage The construction capacity in year y For energy storage The indicator of whether construction will commence in year y is a 0-1 variable. and For energy storage The maximum and minimum investment capacity per transaction, For energy storage Unit capacity cost in year y; Energy storage resource constraints also include: ， ， ， ， ， ， ， in, , These are the upper limits of the initial charge and discharge power of the energy storage, respectively. , The maximum charging and discharging power per unit capacity of energy storage to be constructed. For energy storage capacity, , These are the upper and lower limits of the initial energy storage capacity, respectively. , The upper and lower limits of energy storage capacity for the unit capacity of energy storage to be constructed; This is the self-loss coefficient for energy storage discharge. , These are the charge and discharge coefficients of the energy storage, respectively.

6. The flexible resource planning method for power systems considering extreme weather at multiple time scales as described in claim 5, characterized in that: The construction of a flexible resource planning model for a hybrid time-scale power system includes... Traditional unit operating constraints are expressed as follows: ， ， ， ， ， ， ， in, Indicates traditional units The running status during scheduling period t, and These represent traditional generating units. The maximum value of landslide and uphill power, For traditional generator sets The duration of power-on, For traditional generator sets The duration of the downtime, and For traditional units Minimum start-up and shutdown times, It is a constant; The power balance constraint is expressed as: ， in, For external power input.

7. The flexible resource planning method for power systems considering extreme weather at multiple time scales as described in claim 6, characterized in that: The construction of a flexible resource planning model for a hybrid time-scale power system includes... The power system's uplink and downlink reserve requirements are affected by renewable energy and load fluctuations. The system flexibility balance constraint is expressed as: ， ， in, For renewable energy In planning the year's scenario, the contribution of time t, To plan the system load at time t in scenario y of year y, , , Resources on the power generation side Demand-side resources and energy storage resources The backup capacity at time t in the planning of scenario y in year t. , , Resources on the power generation side Demand-side resources and energy storage resources The backup capacity is planned for scenario y at time t. and These are the upstream and downstream reserve demand coefficients caused by fluctuations in renewable energy, respectively. and These are the upper and lower reserve requirement coefficients caused by load fluctuations and system failures, respectively. This represents the total amount of renewable energy resources.

8. A flexible resource planning system for power systems that considers extreme weather at multiple time scales, employing the flexible resource planning method for power systems that considers extreme weather at multiple time scales as described in any one of claims 1 to 7, characterized in that: It includes a multi-timescale extreme weather scenario construction module, a hybrid timescale power system flexible resource planning model construction module, and an optimal planning scheme output module; The multi-timescale extreme weather scenario construction module is used to generate multi-timescale extreme weather scenarios based on the conditional diffusion model, including dimensionality reduction meteorological feature identification and conditional sequence generation for different extreme weather events, as well as the construction of a conditional diffusion model for multi-timescale extreme weather scenarios. The hybrid time-scale power system flexible resource planning model construction module is used to construct a hybrid time-scale power system flexible resource planning model and apply constraints. The optimal planning scheme output module is used to output the planning model and obtain the optimal planning scheme for the flexible resources of the power system.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the flexible resource planning method for power systems that takes into account extreme weather at multiple time scales, as described in any one of claims 1 to 7.

Images