A Method and Device for Constructing New Energy Scenario Chains Based on Extreme Daily Anchoring and Order-Preserving Clustering

By calculating hourly joint extreme scores to filter extreme natural days and their adjacent days, and performing order-preserving clustering on ordinary days, a scenario chain sequence is constructed. This solves the stability and interface problems of time-series compression in the power planning of new energy bases and improves the fidelity of planning results.

CN123117573APending Publication Date: 2026-11-17STATE GRID ECONOMIC TECH RES INST CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610871731.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-06-16
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies in power planning for new energy bases struggle to stably cover various extreme risk windows while compressing time-series scale, and lack reliable scenario chain output interfaces that directly provide downstream processes, resulting in capacity configuration results deviating from the true boundaries.

Method used

By acquiring hourly new energy resource data and power transmission demand data throughout the year, hourly joint extreme scores are calculated, extreme natural days and their adjacent days are selected and incorporated into the anchored sample set, order-preserving clustering is performed on ordinary days, the real representative days are extracted, and the scenarios are merged and sorted to construct a sequence of scenarios, forming a structured interface.

Benefits of technology

It achieves the goal of compressing the annual time series scale while stably preserving extreme risk windows and time series dependencies, improving the fidelity of the time series compression results to the real planning boundary, and providing an interface that can be directly called by downstream programs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117573A_ABST
    Figure CN123117573A_ABST
Patent Text Reader

Abstract

This specification provides a method and apparatus for constructing new energy scenario chains using extreme day anchoring and order-preserving clustering. The method includes acquiring hourly new energy resource data and power transmission demand data throughout the year to calculate hourly joint extreme scores; based on the hourly joint extreme scores, selecting extreme natural days from the natural days throughout the year, selectively incorporating each extreme natural day and its adjacent natural days into the anchoring sample set, and treating the remaining natural days as ordinary days; performing order-preserving clustering on the ordinary days to obtain multiple ordinary day clusters, and extracting real natural days from each ordinary day cluster as representative days; merging the natural days in the anchoring sample set with the extracted representative days, and sorting them according to the original time order to construct a scenario chain sequence and the corresponding structured interface, thereby improving the fidelity of the actual planning boundaries of new energy bases while compressing the time series scale.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments in this specification relate to the field of new energy power system planning technology, and in particular to a method and apparatus for constructing new energy scenario chains based on extreme day anchoring and order-preserving clustering. Background Technology

[0002] In the power planning of new energy bases, it is usually necessary to input the annual wind and solar resource time-series data, power transmission demand, and various regulation capacity constraints into the planning model. Due to the huge scale of the 8760-hour time-series sample throughout the year, existing engineering practices generally adopt time-series compression methods, such as selecting representative days, typical weeks, or performing cluster analysis, to reduce the computational complexity of the subsequent planning model.

[0003] The existing technologies closest to this invention mainly fall into two categories. The first category revolves around representative day compression, order-preserving clustering, or representative time-period splicing. Its core idea is to reduce the scale of the annual time series while preserving the chronological order as much as possible, thereby reducing the underestimation of the value of energy storage's cross-day state transfer. The second category focuses on multi-day energy storage modeling and extreme risk preservation, emphasizing that if continuous low output, continuous high net load, and multi-day energy transfer processes are not preserved, subsequent capacity configuration results will systematically deviate from the true boundaries.

[0004] However, most existing technologies still remain at the general modeling level of "minimizing average error" or "preserving some time sequence as much as possible." Their main drawbacks are: First, the traditional representative day method cannot guarantee from an algorithmic perspective that extreme net load, extreme ramp-up, extreme curtailment risk, and continuous low output windows will necessarily be included in the compression results; Second, even if some time sequence is preserved, there is generally a lack of stable rules for "extreme day anchoring + adjacent day completion," making it difficult to reflect continuous low output and cross-day state transmission; Third, existing time series compression methods mostly serve general power system modeling and have not formed a scenario chain output interface that can directly provide input for new energy base capacity planning, energy storage planning, and robust verification programs.

[0005] Therefore, how to compress the time series scale while stably covering various extreme risk windows, and form an interface that can be directly called by downstream programs, thereby improving the fidelity of the time series compression results to the real planning boundary, is a technical problem that urgently needs to be solved. Summary of the Invention

[0006] In view of this, embodiments of this specification provide a method for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering. One or more embodiments of this specification also relate to an apparatus for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering, a computing device, a computer-readable storage medium, and a computer program, to address the technical deficiencies existing in the prior art.

[0007] According to a first aspect of the embodiments of this specification, a method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering is provided, comprising: Obtain hourly data on new energy resources and power transmission demand throughout the year; Based on the aforementioned new energy resource data and power transmission demand data, calculate the hourly joint extreme score; Based on the hourly joint extreme scores, extreme days are selected from the natural days of the whole year, and each extreme day and its adjacent days are selectively incorporated into the anchored sample set, and the remaining natural days are treated as ordinary days. Perform order-preserving clustering on the ordinary days to obtain multiple ordinary day clusters, and extract the real natural day as the representative day from each ordinary day cluster; The natural days in the anchored sample set are merged with the extracted representative days and sorted according to the original time order to construct a scene chain sequence and the structured interface corresponding to the scene chain sequence.

[0008] According to a second aspect of the embodiments of this specification, a new energy scenario chain construction device based on extreme daily anchoring and order-preserving clustering is provided, comprising: The acquisition module is used to acquire hourly data on new energy resources and power transmission demand throughout the year; The calculation module is used to calculate the hourly joint extreme score based on the new energy resource data and power transmission demand data; The filtering module is used to filter out extreme natural days from the natural days of the whole year based on the hourly joint extreme score, and selectively incorporate each extreme natural day and its adjacent natural days into the anchored sample set, and treat the remaining natural days as ordinary days. The clustering module is used to perform order-preserving clustering on the ordinary days to obtain multiple ordinary day clusters, and extract the real natural day as the representative day from each ordinary day cluster; The module is used to merge the natural days in the anchored sample set with the extracted representative days, sort them according to the original time order, and construct the scene chain sequence and the structured interface corresponding to the scene chain sequence.

[0009] According to a third aspect of the embodiments of this specification, a computing device is provided, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the above-mentioned method for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering.

[0010] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the above-described method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering.

[0011] According to a fifth aspect of the embodiments of this specification, a computer program is provided, wherein when the computer program is executed in a computer, the computer is instructed to perform the steps of the above-described method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering.

[0012] One embodiment of this specification obtains hourly new energy resource data and power transmission demand data throughout the year, calculates multiple risk characteristics and merges them into an hourly joint extreme score, then anchors high-risk extreme natural days and their adjacent days based on the score, performs order-preserving clustering on the remaining ordinary days to maintain the original time order and extracts the true representative days, and finally merges and sorts the anchored samples and representative days to generate a structured interface containing hourly time periods, scene weights and adjacency relationship sets. This achieves the goal of compressing the annual time series scale while stably preserving the extreme risk windows and time series dependencies that play a decisive role in the planning boundary of new energy bases, and forming a scene link interface that can be directly called by downstream programs, thereby improving the fidelity of the time series compression results to the real planning boundary. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating a method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering, as provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating the processing steps of another method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering, provided in an embodiment of the present invention. Figure 3 Select daily distribution maps for extreme scenario chains; Figure 4 A comparison chart of extreme feature capture capabilities; Figure 5 This is a schematic diagram of the structure of the new energy scenario chain construction system based on extreme daily anchoring and order-preserving clustering provided in this embodiment of the invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0014] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0015] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0016] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0017] This specification provides a method for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering. This specification also relates to a device for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering, a computing device, and a computer-readable storage medium, which will be described in detail in the following embodiments.

[0018] See Figure 1 , Figure 1 This is a flowchart illustrating a method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering, as provided in an embodiment of the present invention. The method specifically includes the following steps.

[0019] Step 101: Obtain hourly data on new energy resources and power transmission demand throughout the year; Step 102: Calculate the hourly joint extreme score based on the new energy resource data and power transmission demand data; Step 103: Based on the hourly joint extreme score, select extreme natural days from the natural days of the whole year, and selectively incorporate each extreme natural day and its adjacent natural days before and after into the anchored sample set, and treat the remaining natural days as ordinary days; Step 104: Perform order-preserving clustering on the ordinary days to obtain multiple ordinary day clusters, and extract the real natural day as the representative day from each ordinary day cluster; Step 105: Merge the natural days in the anchored sample set with the extracted representative days, sort them according to the original time order, and construct the scene chain sequence and the structured interface corresponding to the scene chain sequence.

[0020] In a specific implementation process, energy resource data refers to time-series data describing wind and solar power generation capacity, typically including wind power resource sequences (such as hourly wind power capacity coefficients) and solar power resource sequences (such as hourly solar power capacity coefficients). Power transmission demand data refers to time-series data of the target power that new energy bases need to transmit to the external power grid. Hourly joint extreme score is a comprehensive quantitative indicator used to assess the degree of extremeness of each hour across multiple risk dimensions (such as net load levels, power ramp-up speed, curtailment risk, and continuous low output). The higher the score, the more extreme the risk during that period. Selective inclusion does not simply add all adjacent days before and after extreme days to the anchored sample set, but rather adds them selectively based on preset rules (such as whether adjacent days exceed the annual range or are already covered by other extreme days) to control the sample set size. Order-preserving clustering is a special clustering algorithm whose core constraint is to maintain the original chronological order of data points (here, natural days), meaning that earlier days cannot be assigned to later clusters to ensure that the compressed time series still reflects the original seasonal and trend changes. The scenario chain sequence is an ordered sequence of natural days, composed of natural days from the anchor sample set and representative days extracted from the ordinary day cluster, arranged in the order in which they actually appear throughout the year. The structured interface is a pre-packaged data output format that contains all the information needed for subsequent planning programs to run, such as specific hourly time periods, the weight of each time period, and the connections between scenario days, facilitating direct program calls.

[0021] In a specific implementation process, two core data sequences can be read from the database: the first is new energy resource data, which includes hourly wind power resource sequences and photovoltaic resource sequences throughout the year; the second is power transmission demand data, which defines the power that the new energy base needs to transmit to the target grid every hour. To uniformly describe the input data, the hourly sample set throughout the year is defined as shown in the following formula (1): (1); in, For the first Equivalent load or sending-end demand for a given time period For wind power resource capacity coefficient, The photovoltaic resource capacity coefficient, For the target power supply, The total number of time periods throughout the year. The mathematical meaning of Equation (1) is to define a unified input domain for subsequent risk identification and scenario construction; its engineering significance is to enable the fluctuation of wind and solar resources and the boundary of power transmission demand to be expressed in the same hourly coordinate system.

[0022] After acquiring the data, the total wind and solar power output for each hour can be calculated based on the input wind and solar resource data and the preset installed capacity of the base. Then, combined with power transmission demand, several risk characteristics reflecting the system's operational pressure are calculated, such as net load and power ramp-up intensity. These risk characteristics with different dimensions can be normalized, and a feature weight can be assigned to each feature. Finally, a weighted sum is calculated to obtain the "hourly joint extreme score" for each hour throughout the year. This score intuitively reflects the comprehensive extreme degree of each hour across multiple risk dimensions.

[0023] After obtaining the hourly joint extreme scores, the joint extreme scores for each of the 24 hours within a day are compared, and the maximum value is taken as the "comprehensive extreme score" for that natural day. Then, the scores are sorted from highest to lowest, and the top-ranked natural days are selected as "extreme natural days." For each selected extreme natural day, a "selective inclusion" operation is performed: it checks whether the day before and after the extreme day exists and whether it is covered by the neighborhood of other extreme days. Only if the adjacent days exist and are not covered will they be added to the "anchored sample set"; otherwise, only the extreme day itself is added. Natural days throughout the year that are not selected into the anchored sample set are marked as "ordinary days."

[0024] To compress the scale of ordinary days without losing the chronological information of the entire year, an "order-preserving clustering" algorithm can be used. This algorithm divides all ordinary days into several consecutive "ordinary day clusters" according to chronological order, with days within each cluster exhibiting similarity in risk characteristics. For each cluster, the system selects the day closest to the cluster center from the cluster's real historical days as the "true representative day" of that cluster.

[0025] Furthermore, all natural days in the "anchored sample set" are merged with all "real representative days" extracted from the ordinary day cluster, and these days are strictly sorted according to their chronological order in the original year-round time series to obtain the final "scene chain sequence". Based on this sequence, the system constructs a "structured interface", which includes an hourly scene time period set, a scene weight set, and a set of chronological relationships between adjacent scene days, for direct use by downstream planning programs.

[0026] This embodiment acquires hourly new energy resource data and power transmission demand data throughout the year, calculates multiple risk characteristics, and merges them into an hourly joint extreme score. Then, based on this score, it anchors high-risk extreme natural days and their adjacent days before and after. Next, it performs order-preserving clustering on the remaining ordinary days to maintain the original time order and extracts the true representative days. Finally, it merges and sorts the anchored samples and representative days to generate a structured interface containing hourly time periods, scene weights, and a set of adjacent relationships. This achieves the goal of compressing the annual time series scale while stably preserving the extreme risk windows and time series dependencies that play a decisive role in the planning boundary of new energy bases, and forming a scene link interface that can be directly called by downstream programs, thereby improving the fidelity of the time series compression results to the real planning boundary.

[0027] In some embodiments, the new energy resource data includes wind power resource sequences and photovoltaic resource sequences, and the power transmission demand data includes power transmission demand sequences; based on the new energy resource data and power transmission demand data, an hourly joint extreme score is calculated, including: Calculate hourly renewable power output based on the wind power resource sequence, photovoltaic resource sequence, and preset wind power and photovoltaic installed capacity; Based on the hourly renewable output and the power supply demand sequence, calculate multiple hourly risk characteristics; The risk characteristics are normalized, and the time-by-time joint extreme score is calculated based on each normalized risk characteristic and its corresponding feature weight.

[0028] In a specific implementation process, wind power resource sequence refers to time-series data describing the potential of wind power generation, usually the wind power capacity coefficient, which is the ratio of actual output to installed capacity. Photovoltaic resource sequence refers to time-series data describing the potential of photovoltaic power generation, usually the photovoltaic capacity coefficient. Renewable output refers to the total power that wind and photovoltaic power can actually generate at a given installed capacity at a certain moment. It is calculated by multiplying the wind power installed capacity by the wind power resource coefficient, and adding the photovoltaic installed capacity multiplied by the photovoltaic resource coefficient. Risk characteristics refer to specific quantitative indicators that characterize the operational risk of the system, such as net load, ramp-up intensity, curtailment sensitivity, and continuous low output intensity.

[0029] In detail, the hourly renewable power output is calculated as follows: based on the wind power resource sequence, the photovoltaic resource sequence, the preset wind power installed capacity, and the preset photovoltaic installed capacity, the product of the value of the wind power resource sequence and the wind power installed capacity is added to the product of the value of the photovoltaic resource sequence and the photovoltaic installed capacity to obtain the hourly renewable power output. Specifically, for a given time period (Typically one hour), read the wind power resource capacity coefficient for that time period from the wind power resource sequence. Read the photovoltaic resource capacity coefficient for that period from the photovoltaic resource sequence. Simultaneously, the preset wind power installed capacity is obtained from the system parameters. and photovoltaic installed capacity The renewable energy output for time period t is obtained by the following equation (2): (2); in, The combined renewable power output of wind and solar power in time period t. This represents the available wind power output during time period t. The available photovoltaic output for time period t is represented. The mathematical meaning of equation (2) is to map the resource capacity coefficient to the hourly output related to the scheme; its engineering meaning is to provide the basic supply for subsequent calculation of net load and curtailment sensitivity.

[0030] The hourly net load is calculated as follows: based on the target power supply in the power supply demand sequence and the hourly renewable output, the target power supply is subtracted from the renewable output to obtain the hourly net load. Specifically, after obtaining the renewable power output, the target power transmission capacity for that period is read from the power transmission demand sequence. Net load Defined as the portion of power supplied by conventional power sources (such as coal-fired power, energy storage discharge, etc.) after deducting renewable power output, it is calculated using the following formula (3): (3); The mathematical meaning of Equation (3) is to map the power supply demand and renewable energy output into a single net demand; its engineering meaning is to explicitly quantify the real capacity pressure of the new energy base.

[0031] The hourly climbing intensity is calculated as follows: based on the hourly net load, the absolute value of the difference between the net load of the current time period and the net load of the previous time period is calculated to obtain the hourly climbing intensity. Specifically, the climbing intensity is used to measure the degree of drastic change in net load per unit time, and is calculated by the following formula (4): (4); in, The climbing intensity is the absolute value of the change in net load between adjacent time periods. The meaning is the same as above. This represents the net load of the previous period. If the underlying data is an hourly series, then... This represents the net load for the first hour. In the boundary case, when t=1, it can be agreed in the engineering implementation to start the calculation from t=2, or the ramp intensity of the first period can be set to 0. The mathematical meaning of equation (4) is to transform the hourly net load change into a comparable ramp intensity index; its engineering meaning is to identify the period most sensitive to the ramp of coal-fired power plants and the rapid response of energy storage.

[0032] The hourly power curtailment sensitivity is calculated as follows: based on the renewable power output and the target power transmission, the maximum value between the difference between the renewable power output and the target power transmission and zero is taken to obtain the hourly power curtailment sensitivity. Specifically, the power curtailment sensitivity is used to quantify the potential risk of power curtailment, and is calculated using the following formula (5): (5); in, For time period The sensitivity to power curtailment. The mathematical meaning of Equation (5) is to extract surplus when the renewable output is higher than the target power transmission capacity; its engineering meaning is to identify the period when the transmission end is most likely to be under pressure and power curtailment risk.

[0033] The calculation method for the hourly continuous low power output intensity is as follows: within a preset review window, determine the proportion of time periods during which the renewable power output is continuously lower than a preset low power output threshold, and use this proportion of time periods as the hourly continuous low power output intensity.

[0034] Specifically, continuous low power output intensity is used to identify events where renewable power output remains consistently low. Its calculation requires presetting two parameters: the lookback window length. (e.g., 24 hours) and low output threshold (For example, 20% of installed capacity). For time periods Calculated using the following formula (6): (6); in, For time period Continuous low output strength, To start from the current time period Looking back Total renewable energy output over a given period To achieve a low output threshold, To check the window length, This is an indicator function (value 1 when the condition in parentheses is true, otherwise 0). Mathematically, this formula represents the percentage of periods below a threshold in the most recent H time periods; its engineering significance is to identify consecutive low-output events that are most sensitive to long-term energy storage configuration and continuous supply capacity.

[0035] In a specific implementation process: superscript Indicates renewable energy, i.e., the total output of renewable energy; subscript Indicates from the current time period Looking back A time period. When "Time" corresponds to the current time period; when The time corresponds to the previous time period; and so on, until the previous time period. Therefore, Equation (6) does not consider the low output in a single hour, but rather measures the persistence of low output within a rolling window.

[0036] In a specific implementation process, multiple sets of candidate feature weights can be set based on the above judgments and the calculated normalized risk characteristics. Before setting the weights, it is necessary to normalize each risk characteristic to eliminate the dimensional differences between different indicators. Normalization is accomplished by the following formula (7): (7); in, The normalized index value, index During the period The original value, and These are the minimum and maximum values ​​of the corresponding indicators in the annual sample, respectively, and ε is a very small positive number to prevent the denominator from being zero. The mathematical meaning of equation (7) is to compress different risk characteristics to a unified scale; its engineering meaning is to avoid a certain indicator from unreasonably dominating extreme scores due to its large numerical dimension.

[0037] Based on normalization, the joint extreme score is calculated using the following formula (8): (8); in, The joint extreme score for time period t, , , and These are the normalized values ​​of net load, ramp intensity, power curtailment sensitivity, and continuous low output intensity, respectively. α, β, γ, and δ are the corresponding weights, and the sum of the weights is 1, with each weight not less than 0. The mathematical meaning of Equation (8) is to convert multiple types of risks into a unified ranking index; its engineering meaning is to avoid missing high-risk windows that truly determine the planning boundary by selecting samples based solely on a single extreme condition.

[0038] This embodiment integrates multiple risk characteristics such as net load, ramp-up, power curtailment, and continuous low output, providing a comprehensive and configurable quantitative basis for extreme day screening, thereby improving the comprehensiveness and scientific nature of risk assessment.

[0039] In some embodiments, the step of obtaining feature weights includes: Based on the key focus of the plan, determine the types of extreme risks that the new energy base should focus on; Based on the extreme risk types and normalized risk characteristics, multiple sets of candidate feature weights are set; each set of candidate feature weights corresponds to a different risk preference. Based on the weights of each set of candidate features, the joint extreme scores are calculated sequentially over time to obtain the candidate scene chain sequence corresponding to each set of candidate feature weights. Calculate the average extreme feature capture rate for each candidate scene chain sequence and count the total number of natural days contained in each candidate scene chain sequence; Each candidate scenario chain sequence is input into the downstream capacity planning, energy storage configuration, and robustness verification program, and the fluctuation range of the output results corresponding to each candidate scenario chain sequence is recorded. The set of candidate feature weights with the highest average extreme feature capture rate, the total number of natural days in the scene chain does not exceed the preset scale limit, and the smallest fluctuation in downstream output results are determined as the feature weights to be used in the official application.

[0040] In this embodiment, the planning focus refers to the main contradictions or risks most concerning in the planning of new energy bases, such as supply capacity, regulation capacity, or curtailment issues. Candidate feature weights refer to multiple sets of different weight coefficients set to find the optimal risk preference. A candidate scenario chain sequence refers to a scenario chain result obtained after fully running the method described in Example 1 using each set of candidate weights. The average extreme feature capture rate is a comprehensive indicator used to quantitatively evaluate the scenario chain's ability to retain extreme risks. Downstream capacity planning, energy storage configuration, and robustness verification procedures refer to the core calculation procedures commonly used in new energy base planning. The fluctuation range of the output results refers to the degree of difference in planning results obtained by inputting different candidate scenario chain sequences into the same downstream planning program.

[0041] In a specific implementation process, different renewable energy bases may face different primary challenges: some bases may face significant supply gaps, thus focusing more on peak net load and continuous low-output events; some bases may have weak grid regulation capabilities, thus focusing more on sharp net load ramp-ups; and some bases may have limited absorption capacity, thus focusing more on the risk of curtailment. Planners need to determine the type of extreme risk that is of greatest concern to the current base based on the actual situation.

[0042] Based on the above judgment and the calculated normalized risk characteristics, multiple sets of candidate feature weights can be set. For example, for formula (8), multiple sets of candidate feature weights can be set. For example, the following five typical schemes can be set: Scheme 1 is the equal weight scheme, that is... Option 2 is a net load bias scheme, with the net load weight set to 0.5 and the other three weights equally weighted; Option 3 is a ramp bias scheme, with the ramp weight set to 0.5 and the other three weights equally weighted; Option 4 is a curtailment bias scheme, with the curtailment weight set to 0.5 and the other three weights equally weighted; Option 5 is a continuous low output bias scheme, with the continuous low output weight set to 0.5 and the other three weights equally weighted. Furthermore, fine-tuning can be made to the above schemes to generate more candidate groups.

[0043] For each set of candidate weights, the entire process from calculating the hourly joint extreme score to generating the final scene chain sequence is repeated. Each set of weights will generate a corresponding candidate scene chain sequence.

[0044] Furthermore, these candidate solutions can be quantitatively evaluated. First, calculate the average extreme feature capture rate for each candidate scenario chain sequence. Second, count the total number of natural days included in each candidate scenario chain sequence; this metric reflects the compressed computational scale. Third, input each candidate scenario chain sequence into downstream capacity planning, energy storage configuration, and robustness verification programs, run these programs, and record the fluctuation range of their output results (such as optimal installed capacity, energy storage duration, system cost, etc.). The smaller the fluctuation of the output results, the more stable and reliable the scenario chain sequence is as input.

[0045] After evaluating candidate solutions, a comprehensive optimization can be performed. The following three conditions are used as selection criteria: first, the highest average extreme feature capture rate to ensure fidelity to the planning boundary; second, the total number of natural days in the scenario chain does not exceed a preset upper limit (e.g., no more than 90 days) to ensure computational efficiency; and third, the smallest fluctuation in the output results of downstream programs to ensure the stability of the planning scheme. The set of candidate feature weights that simultaneously meets these three conditions or has the highest comprehensive score will be determined as the final officially used feature weights. Through the optimization process in this embodiment, the optimal weight configuration can be adaptively found for different new energy bases and application scenarios.

[0046] In some embodiments, the average extreme feature capture rate for each group of candidate scene chain sequences is calculated, including: Select at least one target risk feature from multiple risk features to calculate the average extreme feature capture rate; For any target risk feature, extract the first feature statistical value of the target risk feature in the original sequence of the whole year, and extract the second feature statistical value of the target risk feature in the candidate scenario chain sequence; wherein, the original sequence of the whole year includes the hourly wind power resource sequence, the hourly photovoltaic resource sequence and the hourly power transmission demand sequence of the whole year; Divide the second feature statistical value by the first feature statistical value to obtain the single feature capture rate of the target risk feature; The average extreme feature capture rate of each candidate scenario chain sequence is obtained by summing the single feature capture rates of all target risk features and then dividing by the total number of target risk features.

[0047] In this embodiment, the target risk feature refers to a feature specifically selected from multiple risk features to evaluate the fidelity of the scenario chain, such as peak net load and maximum ramp. The first feature statistic refers to the statistic calculated for a specific target risk feature on raw, uncompressed time-series data throughout the year. The second feature statistic refers to the statistic calculated for the same target risk feature within the generated candidate scenario chain sequence. The single feature capture rate is the ratio of the second feature statistic to the first feature statistic. The average extreme feature capture rate is the arithmetic mean of all single feature capture rates.

[0048] In a specific implementation process, several risk characteristics can be selected from the obtained risk features (such as net load, ramp intensity, curtailment sensitivity, and continuous low output intensity) as target risk features for calculating the average capture rate. These features should comprehensively reflect the core risks that the new energy base planning is concerned with. For example, the annual peak net load, maximum ramp intensity, net load value corresponding to the high-risk power shortage quantile, and the coverage of continuous low output events can be selected.

[0049] For each selected target risk characteristic, corresponding key indicators can be set. This requires two extraction operations. The first operation is to extract the characteristic statistics of this key indicator from the raw sequence of the entire year (i.e., 8760 hours of data without any compression), denoted as the first characteristic statistics. For example, regarding the characteristic of "peak net load", This refers to the maximum value in the annual net load sequence. The second operation is to extract statistical values ​​of the same feature from the candidate scenario chain sequence, denoted as the second feature statistical value. For example, in the candidate scenario chain sequence, find the maximum value among all the net load values ​​of the retained hours, as... .

[0050] For each target risk feature corresponding to the key indicator x, the single feature capture rate is calculated using the following formula (9). : (9); The mathematical meaning of equation (9) is to characterize the fidelity of the compression result for the extreme features of the whole year using a relative ratio; its engineering meaning is to provide a unified evaluation scale for comparing the effects of different scene construction methods. If the capture rate is close to 1, it means that the scene chain has preserved the extreme feature well; if the capture rate is greater than 1, it means that the scene chain exposes the extreme feature to a degree that even exceeds that of the original sequence (which is useful when evaluating conservatism); if the capture rate is much less than 1, it means that the extreme feature was lost during the compression process.

[0051] Furthermore, the single-feature capture rates of all selected target risk features are summed, and then divided by the number of target risk features to obtain the average extreme feature capture rate, which is calculated using the following formula (10): (10); in, The average extreme feature capture rate, This is the set of key indicators for evaluation. The mathematical meaning of Equation (10) is to average the capture rates of multiple key extreme features; its engineering meaning is to output a comprehensive index that directly compares the overall fidelity of the "scene chain method" and the "traditional representative day method." This average value comprehensively reflects the overall fidelity capability of the scene chain across multiple key extreme dimensions, and is a comprehensive index used to compare the merits of different compression methods or different weight configurations. Through the method defined in this embodiment, the quality of scene chains generated by different candidate weight recombinations can be objectively and quantitatively evaluated.

[0052] It should be noted that the reason why both equation (9) and equation (10) are used in this embodiment is that they have different functions. Equation (9) mainly refers to "whether a certain type of extreme feature is retained", which is a single diagnostic indicator; equation (10) mainly refers to "which method is better overall", which is a comprehensive ranking indicator. If only equation (9) is used, although it can diagnose item by item, it is difficult to give an overall judgment of superiority or inferiority; if only equation (10) is used, although it can make an overall comparison, it is not clear which type of extreme feature was missed.

[0053] Furthermore, the term "average" in Equation (10) does not contradict the criticism in the background technique regarding "minimizing the average error." The background technique reflects that averaging the point-by-point error or overall morphological error of the entire 8760-hour sample throughout the year can easily overwhelm the extreme windows that truly determine the planning boundary. However, the average in Equation (10) is not an average of the point-by-point error throughout the year, but an average of the capture rates of several key extreme indicators. In other words, what is averaged here is "fidelity across multiple extreme risk dimensions," not "average error across the entire time series sample." Therefore, Equations (9) and (10) do not return to the old approach of prioritizing the average morphological aspect, but instead establish a two-tiered evaluation system of "single-item diagnosis + comprehensive evaluation" under the premise of "prioritizing extreme features." If the capture rate of a certain indicator is greater than 1, it should be understood that the exposure of the scenario chain to that extreme indicator is more sensitive than the original sequence throughout the year, rather than mechanically interpreted as a probability that must fall between 0 and 1.

[0054] In some embodiments, selecting a preset number of extreme natural days from the annual natural days based on the hourly joint extreme score includes: For each natural day, the maximum value of the hourly joint extreme scores across all time periods within that natural day is taken as the comprehensive extreme score for that natural day; Based on the overall extreme scores from highest to lowest, the top K natural days are selected as extreme natural days, where K is an integer greater than or equal to 1; when the overall extreme scores of multiple natural days are tied, the natural day with the earlier time is retained, or the tie is broken by using the single risk feature with preset priority and then the extreme natural day is added.

[0055] In this embodiment, the comprehensive extreme score for a calendar day is a value obtained by aggregating the combined extreme scores of all 24 hours within a calendar day. The aggregation method uses the "maximum value" approach. The preset priority refers to a pre-defined comparison rule to break up ties when multiple calendar days have the same comprehensive score, such as comparing the peak net load first, then comparing the maximum ramp intensity, etc.

[0056] In a specific implementation, after obtaining the hourly scores, they need to be aggregated to the day level. For each day of the year, the maximum value of the joint extreme scores within its 24 hours is taken as the comprehensive extreme score for that day. Then, all days are sorted from highest to lowest according to the comprehensive extreme scores, and the top K days are selected as extreme days (K is a preset positive integer, such as 15, 30, etc.). For each selected extreme day, not only is it retained, but its preceding and following days are also added to an anchored sample set specifically used to store key dates. In this way, it is ensured that high-risk periods and their transition processes are not discarded. The remaining days of the year that are not selected into the anchored sample set are marked as ordinary days.

[0057] In a specific implementation process, after calculating the joint extreme score for each hour throughout the year, the system needs to aggregate these hourly scores to the day level for daily filtering. For a 24-hour day $d$, all time periods within it belong to the set. The comprehensive extreme score of this day is defined as follows: take the maximum value among all hourly scores within the day, as shown in equation (11): (11); in, For the first The overall extreme score for each natural day. Belongs to the calendar day The set of time periods. The mathematical meaning of Equation (11) is to use the maximum extreme score within the day to represent the risk level of that natural day; its engineering meaning is to ensure that natural days with truly high-risk periods are not diluted by averaging during daily screening.

[0058] This aggregation method, which takes the maximum value, ensures that if a calendar day has a high-risk period, it can still be identified as an extreme calendar day even if the other 23 hours have low risk. Conversely, if an average method is used, the risk contribution of the high-risk period is diluted by the other low-risk periods, potentially causing truly dangerous dates to be missed. Therefore, using maximum value mapping can preserve information about extreme events to the greatest extent possible.

[0059] After obtaining the comprehensive extreme scores for all natural days throughout the year, the system sorts these scores in descending order, with the natural day with the highest score ranked first. Then, based on a pre-set parameter K (K is an integer greater than or equal to 1), the top K natural days are selected and marked as extreme natural days, as shown in equation (12) below: (12); in, For the collection of extreme natural days, Score for natural days The descending order of the dates is given, where K is the number of extreme natural days to be anchored. The mathematical meaning of Equation (12) is to select a high-risk subset from the set of natural days throughout the year; its engineering meaning is to prioritize the retention of key natural days that truly determine the planning boundary. The value of K can be set by the user according to actual needs. For example, if the planner wants to focus on the 15 days with the highest risk throughout the year, K can be set to 15; if they want to cover a wider range of risk scenarios, K can be set to 30 or greater. From an algorithmic perspective, as long as K≥1, it is guaranteed that at least one extreme natural day can be selected, and there will be no situation where the selection result is an empty set. Through the selection rules of this embodiment, the system can stably and controllably determine the set of high-risk dates that need to be prioritized.

[0060] In a specific implementation process, when selecting the top K natural days as extreme natural days, if the comprehensive extreme scores of multiple natural days are tied, the natural day with the earlier time is retained, or the tie is broken by using a single risk feature. Among these, breaking down parallel structures by utilizing individual risk characteristics includes: For natural days with the same comprehensive extreme scores, first compare the corresponding values ​​according to the single risk feature with the highest priority, and natural days with larger values ​​are ranked first; if the values ​​corresponding to the single risk feature with the highest priority are the same, then compare the single risk features with the next lower priority, and so on; natural days that are ranked first after the second sort are added to the extreme natural day set until K are satisfied.

[0061] In detail, in actual data, it is common to see multiple days with the same or very similar extreme scores. Since it is necessary to accurately select the top K days, additional rules need to be introduced to break up ties, ensuring the uniqueness and rationality of the selection results.

[0062] The first tie-breaking rule is to retain the earlier calendar day. This is the simplest and most direct method. When multiple calendar days have the same score, the system compares their original date numbers and selects the earlier (i.e., more recent) calendar day for priority inclusion in the extreme calendar day set. This rule is simple to implement and does not rely on the calculation results of other risk characteristics.

[0063] The second approach to breaking the parallel rule involves secondary prioritization using individual risk characteristics. This method is more refined and scientific. First, a priority order needs to be pre-defined for each individual risk characteristic (such as net load, ramp-up intensity, curtailment sensitivity, and continuous low output intensity). This priority order should align with the planning focus. For example, if the planning prioritizes supply gaps, net load can be set as the highest priority; if it prioritizes regulation capacity pressure, ramp-up intensity can be set as the highest priority.

[0064] When multiple calendar days have the same composite extreme score, the system first compares the values ​​of these calendar days on the highest priority single risk feature (e.g., comparing the net load peak within each day). The calendar day with the larger value is ranked higher. If all the calendar days with the same value on this feature are still the same, the system continues to compare the next priority single risk feature, such as comparing the maximum ramp intensity, and so on, until the order can be distinguished. In this way, the system performs a secondary sorting of all the calendar days with the same score, and then adds the calendar days at the top of the secondary sorting to the extreme calendar day set until K extreme calendar days are reached. The tie-breaking rule provided in this embodiment ensures the determinism and repeatability of the extreme calendar day screening process, while guiding the screening results towards the risk types that planners are more concerned with.

[0065] In some embodiments, each extreme natural day and its preceding and following natural days are selectively incorporated into the anchored sample set, including: If the extreme natural day and its adjacent natural days do not exceed the annual boundary and are not covered by the neighborhood of other extreme natural days, then the extreme natural day and its adjacent natural days are included in the anchored sample set. If the extreme natural day and its adjacent natural days are covered by the neighborhood of other extreme natural days or exceed the annual boundary, then only the extreme natural day itself is included in the anchored sample set. The annual boundary refers to the valid range of natural days, from the first day of the year to the 365th day (or the 366th day in a leap year). Neighborhood coverage refers to whether a specific natural day has already been added to the anchored sample set because it is adjacent to another extreme natural day.

[0066] In a specific implementation, the preset merging rules include: If the preceding adjacent day of any extreme natural day does not exceed the boundary range of natural days throughout the year and is not covered by the neighborhood of other extreme natural days, then that extreme natural day and its preceding adjacent day are incorporated into the anchored sample set. If the preceding adjacent day of any extreme day does not exceed the boundary of the annual days, but is covered by the neighborhood of other extreme days, then that extreme day will be incorporated into the anchored sample set. If the preceding day of any extreme day exceeds the boundary of the annual number of days, the extreme day will be included in the anchored sample set. If the next adjacent day of any extreme natural day does not exceed the boundary of the natural days of the whole year and is not covered by the neighborhood of other extreme natural days, then the extreme natural day and its next adjacent day are incorporated into the anchored sample set. If the next adjacent day of any extreme day does not exceed the boundary of the annual natural days, but is covered by the neighborhood of other extreme days, then that extreme day is incorporated into the anchored sample set. If the next adjacent day of any extreme natural day exceeds the boundary range of natural days in the whole year, the extreme natural day will be incorporated into the anchored sample set.

[0067] In detail, based on the previously selected set of extreme natural days, a set containing the extreme days and their adjacent days needs to be constructed, as shown in equation (13): (13); in, An anchored sample set consisting of extreme days and their adjacent days before and after them. For day numbers that cross the annual boundary, they can be cropped into the effective natural day interval. The mathematical meaning of Equation (13) is to incorporate the local time neighborhood of extreme days into the same set; its engineering meaning is to explicitly preserve cross-day energy transfer, continuous low-output continuation, and transition processes before and after extreme events.

[0068] However, directly adding all d-1 and d+1 values ​​of all extreme days could result in an excessively large set, especially when multiple extreme days are close together, leading to significant overlap in their neighborhoods. Therefore, the following approach can be used: For each selected extreme natural day d, the system needs to process its preceding adjacent day d-1 and its following adjacent day d+1 respectively. The core principle of the processing is to ensure that the extreme day itself is preserved, while preserving as many days before and after it as possible. However, if the adjacent day has already been covered by the neighborhood of other extreme days, it will not be added again, in order to control the total size of the anchored sample set.

[0069] Specifically, there are three scenarios for handling the preceding adjacent natural day d-1: Scenario 1: If d-1 exists (i.e., d is not the first day of the year) and d-1 is not covered by the neighborhood of any other extreme natural day (i.e., it is not the day before or after any other extreme day), then the system will include the extreme day d and its previous day d-1 into the anchored sample set.

[0070] Scenario 2: If d-1 exists, but d-1 has been covered by the neighborhood of other extreme natural days (for example, the day after another extreme day is exactly d-1), then in order to avoid repeated anchoring, the system only includes the extreme day d itself into the anchoring sample set.

[0071] Scenario 3: If d-1 does not exist (i.e. d is the first day of the year), then only the extreme day d itself is included in the anchored sample set.

[0072] For the processing of the next adjacent natural day d+1, there are also three similar scenarios: Scenario 1: If d+1 exists (i.e., d is not the last day of the year) and d+1 is not covered by the neighborhood of any other extreme natural day, then the system will include the extreme day d and its next day d+1 into the anchored sample set.

[0073] Scenario 2: If d+1 exists, but d+1 has been covered by the neighborhood of other extreme natural days, then the system will only include the extreme day d itself into the anchored sample set.

[0074] Scenario 3: If d+1 does not exist (i.e. d is the last day of the year), then the system will only include the extreme day d itself in the anchored sample set.

[0075] After completing the extreme day anchoring, the remaining set of ordinary days can be defined as shown in equation (14): (14); in, For the remaining ordinary days, The total number of natural days in the year is denoted as . The mathematical meaning of Equation (14) is to divide the natural days of the year into an anchor set and an ordinary day set; its engineering significance is to provide a clear boundary for subsequent clustering compression to be performed only on ordinary days.

[0076] This embodiment ensures the integrity of each high-risk extreme day and its surrounding transition process (by adding uncovered adjacent days), and controls the size of the set through a deduplication mechanism, avoiding excessive expansion of the anchor sample set due to the clustering of extreme days.

[0077] In some embodiments, order-preserving clustering is performed on the ordinary days to obtain multiple ordinary day clusters, and a real natural day is extracted from each ordinary day cluster as a representative day, including: Each typical day's configuration includes a feature vector comprising average net load, maximum ramp intensity, average power curtailment sensitivity, and maximum continuous low output intensity. With the clustering objective of minimizing the sum of squared Euclidean distances from each ordinary day to the center of its cluster, the feature vectors of each ordinary day are divided into multiple ordinary day clusters by order-preserving clustering. For each ordinary day cluster, the ordinary day with the smallest Euclidean distance from its internal feature vector to the cluster center is selected as the real natural day, thus obtaining the representative day of that ordinary day cluster.

[0078] The feature vector is a multi-dimensional numerical vector used to quantitatively describe the comprehensive characteristics of an ordinary day across multiple key risk dimensions. This feature vector contains four dimensions: Euclidean distance: the straight-line distance between two points in multi-dimensional space, used to measure the similarity between the feature vector of an ordinary day and the feature vector of a cluster center. Order-preserving clustering partitioning: a clustering algorithm with time-order constraints, requiring that each cluster must be a continuous segment on the time axis, without temporal overlap.

[0079] In a specific implementation, a feature vector can be constructed for each natural day marked as an "ordinary day". This feature vector is a multi-dimensional numerical array used to quantitatively describe the overall characteristics of the natural day across multiple risk dimensions. In this embodiment, the feature vector consists of four components, as shown in equation (15) below: (15); in, Let d be the feature vector of the d-th ordinary day. The average net load for that day is calculated by summing the net load values ​​for the 24 hours of the day and dividing by 24. This represents the maximum climbing intensity for that day (the maximum value is taken from all adjacent hourly climbing intensities within the day). This represents the average daily power curtailment sensitivity. The maximum continuous low power output intensity of the day (the maximum value is taken from the continuous low power output intensity index of all time periods within the day). The mathematical meaning of Equation (15) is to map an ordinary day to a unified multidimensional feature space; its engineering meaning is to make clustering not only consider the average shape, but also the extreme attributes that really affect the planning boundary.

[0080] Furthermore, order-preserving clustering can be performed, which means that the original temporal order of data points (ordinary days in this case) must be maintained during the clustering process. Specifically, if natural day A is earlier than natural day B in time, then they cannot be assigned to different clusters unless all natural days between A and B also belong to either A's cluster or B's cluster. In other words, each cluster must be a continuous interval on the time axis. The optimization objective of clustering is to minimize the sum of squares of the Euclidean distances from the feature vectors of all ordinary days to the central feature vectors of their respective clusters, as shown in Equation (16): (16); in, The clustering objective function is... This represents the number of clusters in a typical daily cluster. For the k-th ordinary daily cluster, Let be the feature vector of the center of the k-th ordinary daily cluster. (Symbol) Denotes the 2-norm (Euclidean norm), its square... This is equivalent to the sum of squares of the differences in each dimension. The mathematical meaning of Equation (16) is to minimize the squared distance from the ordinary day samples to their respective cluster centers; its engineering meaning is to preserve the overall distribution pattern of ordinary days as much as possible while controlling the compression scale. The smaller the sum of squares of the Euclidean distances, the more compact the days within the cluster are in the feature space, that is, they have similar risk characteristics. By solving this optimization problem, all ordinary days can be divided into several time-continuous ordinary day clusters with similar internal characteristics.

[0081] Furthermore, the true representative day can be extracted from each ordinary day cluster. For the k-th cluster, the Euclidean distance from each ordinary day in the cluster to the cluster center needs to be calculated first, as shown in equation (17): (17); Where q(d, k) is the Euclidean distance from ordinary day d to the kth cluster center. The mathematical meaning of Equation (17) is to measure the consistency between the real day and the cluster center; its engineering meaning is to ensure that the output scene chain is composed of real natural days, rather than abstract virtual center days.

[0082] Next, the ordinary day with the smallest Euclidean distance can be selected as the true representative day of the cluster, as shown in the following equation (18): (18); in, This is the set of truly representative days selected from each ordinary day cluster. Let the day be the true representative day with the smallest distance from the cluster center in the k-th ordinary day cluster. For the k-th ordinary daily cluster, Let be the number of clusters for ordinary days. The mathematical meaning of Equation (18) is to select the real sample day with the smallest distance from the cluster center for each cluster; its engineering meaning is to retain real and traceable time-series samples while compressing the number of ordinary days. Although this representative day is selected from real historical data, it represents the average behavioral characteristics of all days in the entire cluster. Using real natural days instead of artificially synthesized days can avoid generating physically impossible output curves, ensuring the credibility and traceability of the scenario chain.

[0083] In some embodiments, constructing a structured interface corresponding to a scene chain sequence includes: Each scene day in the scene chain sequence is expanded according to the annual hourly index correspondence, and then merged to obtain an hourly scene time period set; The scene weights are assigned based on the day of the scene: for natural days from the anchored sample set, the weight is set to 1; for representative days from ordinary day clusters, the weight is set to the total number of natural days contained in the ordinary day cluster to which it belongs. The two adjacent scene days in the scene chain sequence are grouped into ordered pairs, and the set of sequential relationships between adjacent scene days is obtained by summarizing them. The hourly scene time period set, scene weight set, and adjacent relationship set are collectively encapsulated into the structured interface for output.

[0084] In this embodiment, the annual hourly index correspondence refers to the fixed position of each natural day in the 8760-hour sequence throughout the year. Scene weight is a numerical value representing the original number of natural days represented by each scene day in the scene chain, used in the downstream planning process to reconstruct the total annual statistics through weighted averaging. An ordered pair is a pair of two elements arranged in a specific order. The adjacency set is a set of ordered pairs consisting of all adjacent scene days in the scene chain sequence, explicitly defining the temporal sequence connection logic.

[0085] In a specific implementation process, the final scenario chain natural day sequence can be set as Equation (19). (19); in, For the final scenario chain daily sequence, This indicates that the data is ordered according to the original time sequence throughout the year. The mathematical meaning of equation (17) is to uniformly map the multi-source sample set into an ordered natural day sequence; its engineering meaning is to ensure that the scene chain read by the subsequent planning program still maintains the true time sequence. Let... Dc ={ d (1), d (2), …, d ( M )}, where M is the total number of natural days in the scene chain, and d (1) to d ( M The original time sequence has been used.

[0086] For each day in the sequence, the start and end positions of that day in the annual hour index can be calculated. Specifically, if the annual hour count starts from 1, then the hour index corresponding to the d-th day is from the 24(d-1)+1 hour to the 24d-th hour. All 24 hour indices corresponding to d can be extracted. Then, in the order of d(1), d(2), ..., d(M), all the extracted hour indices are concatenated and merged to form a total hour index list, as shown in the following formula (20): (20); in, The final output is a set of hourly scene time periods, with symbols... The union represents the combination of all the hours corresponding to the retained natural days. The mathematical meaning of Equation (20) is to expand the day-level compression result into a set of hour-level indices that can be directly calculated; its engineering significance is to facilitate seamless integration with existing new energy base planning programs and production simulation programs. Explicitly tell the downstream program: it only needs to simulate these hours, and does not need to simulate the 8760 hours of the whole year.

[0087] Furthermore, for each scenario day in the scenario chain, its weight needs to be determined, i.e., how many natural days in the original data that scenario day represents. The weight assignment rule is related to its source: if the scenario day comes from the anchored sample set (i.e., it is an extreme natural day or its adjacent day), then its weight is set to 1, indicating that it only represents itself and must be accurately simulated. If the scenario day is a representative day extracted from a normal day cluster, then its weight is set to the total number of natural days contained in the normal day cluster to which the representative day belongs. For example, if a representative day comes from a cluster containing 15 ordinary days, then its weight is 15, meaning that in subsequent planning calculations, the time-series data of this day will be used to represent those 15 days. This forms the scenario day weight set. When constructing the objective function (such as minimizing total cost), downstream programs can multiply the cost of each scenario day by its weight, thereby statistically reconstructing the overall situation for the whole year.

[0088] Furthermore, to support simulations requiring time sequence information, such as cross-day status transmission of energy storage and the connection of continuous low-output events, it is necessary to explicitly inform the downstream program of the connection relationships between each day in the scenario chain. The specific construction method is as follows: starting from the first day of the scenario chain sequence... Begin, and make it the second day Form an ordered pair Then and Forming ordered pairs And so on, until... and Forming ordered pairs All these ordered pairs constitute a set. When the downstream program finishes simulating... In the last hour, it is necessary to transition to the next scenario day by querying... You will then know which way to jump. The first hour, thus achieving a seamless connection in terms of timing.

[0089] Furthermore, the constructed hourly scene time period set, scene weight set, and adjacency relationship set can be encapsulated into a structured output object, namely the scene chain output interface. As shown in equation (21): (twenty one); In this context, the mathematical meaning of equation (21) is to encapsulate the time index, scene representativeness, and temporal sequence connection relationship into a unified structured output object; the engineering meaning is to enable downstream planning programs to read three types of information at once: "which time periods are there, how many original natural days these time periods represent, and how these natural days are connected." Downstream programs only need to call this one interface to obtain all the temporal sequence information, statistical weight information, and temporal sequence connection information required to perform planning calculations.

[0090] This embodiment solves the problem of a single output format and difficulty in being used by downstream programs by constructing a structured interface that includes hourly time periods, scene weights, and adjacency relationships. This enables the method to be seamlessly integrated into downstream programs such as capacity planning and energy storage configuration.

[0091] The following is in conjunction with the appendix Figure 2 This document provides a specific application example of the new energy scenario chain construction method based on extreme daily anchoring and order-preserving clustering, as described in this specification, to further illustrate the method. Figure 2 This is a flowchart of another method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering, provided by an embodiment of the present invention. The process includes the following steps.

[0092] Step 201: Read the hourly wind power resource sequence, photovoltaic resource sequence, power transmission demand sequence, and threshold parameters and weight parameters required for scenario construction throughout the year, and calculate the renewable power output and net load according to equations (1) to (3).

[0093] Step 202: Calculate the climbing intensity, power abandonment sensitivity, continuous low output intensity and joint extreme score according to equations (4) to (8) respectively, and obtain the hourly risk characteristic sequence for the whole year.

[0094] Step 203: Map the hourly extreme scores to natural days according to equations (11) and (12) to filter high-risk extreme natural days; then add adjacent days according to equations (13) and (14) to obtain the set of remaining ordinary days.

[0095] Step 204: Perform order-preserving clustering on the remaining ordinary days according to equations (15) to (18) to extract the true representative day of each ordinary day cluster.

[0096] Step 205: According to Equations (19) to (21), merge the extreme day anchor set and the ordinary day representative set, and output the final scene chain natural day sequence, hour-level scene time period set and structured interface in the original time order.

[0097] Step 206: Calculate the capture rate of key extreme features of the scene chain according to Equations (9) and (10), and compare and verify it with the typical daily baseline of the month.

[0098] To further illustrate the beneficial effects of the present invention, this embodiment uses a set of specific wind power, photovoltaic and load data for verification, and compares the present invention (hereinafter referred to as "time-series-preserving extreme scenario chain") with the traditional "monthly typical daily baseline" method.

[0099] Figure 3 A daily distribution map for the extreme scenario chain is provided. This map shows the distribution of key scenario days anchored or completed by the method of this invention throughout the 365 calendar days of the year. The horizontal axis represents the calendar day number (from day 1 to day 365), and the vertical axis indicates whether each calendar day was selected into the final scenario chain sequence. Figure 2 As can be seen, the selected natural days are not evenly distributed throughout the year, but are concentrated in certain high-risk periods, such as the hot and calm periods in summer or the continuous rainy days in winter. This figure visually illustrates that the present invention can effectively identify and anchor those extremely high-risk dates that determine the planning boundaries.

[0100] Figure 4 This is a comparison chart of extreme feature capture capabilities. The chart, presented as a bar chart, compares the capture rates of the "time-series-maintaining extreme scenario chain" and the "monthly typical daily baseline" of this invention across four key extreme indicators. These four indicators are: peak net load capture rate, maximum ramp capture rate, high-risk power outage capture rate, and continuous low output coverage rate. Figure 3 As can be clearly seen, the method of this invention significantly outperforms the typical monthly daily baseline in terms of maximum uphill capture rate, high-risk power shortage capture rate, and continuous low-output coverage. In particular, the result of this invention exceeds 1 in the high-risk power shortage capture rate, indicating that it is more sensitive to the exposure to this type of extreme risk than the original annual sequence, which is of great value for conservative planning.

[0101] Table 1 compares the extreme feature capture rates of the time-preserving extreme scenario chain with the typical monthly daily baseline.

[0102] Table 1

[0103] As shown in Table 1, although the peak net load capture rate of this invention is close to that of the baseline method, it is significantly superior in the dimensions of extreme ramping, high-risk power shortage, and continuous low-output coverage, which truly determine the planning boundaries of new energy bases. In particular, the high-risk power shortage capture rate reaches 1.2863, indicating that this invention not only retains the high-risk power shortage window throughout the year but also more sensitively exposes boundary situations that traditional typical day compression methods easily miss. Calculation results show that the average extreme feature capture rate of the scenario chain constructed by the method of this invention is 0.8528, while the monthly typical day baseline is only 0.6701. Therefore, this invention is more suitable as a scenario input module in new energy base capacity planning, energy storage planning, and robustness verification procedures.

[0104] Corresponding to the above method embodiments, this specification also provides an embodiment of a new energy scenario chain construction device based on extreme daily anchoring and order-preserving clustering. Figure 5 This diagram illustrates a structural schematic of a new energy scenario chain construction device based on extreme daily anchoring and order-preserving clustering, according to one embodiment of this specification. Figure 5 As shown, the device includes: Module 51 is used to acquire hourly data on new energy resources and power transmission demand throughout the year; Calculation module 52 is used to calculate the hourly joint extreme score based on the new energy resource data and power transmission demand data; The filtering module 53 is used to filter out extreme natural days from the natural days of the whole year according to the hourly joint extreme score, and selectively incorporate each extreme natural day and its adjacent natural days before and after into the anchored sample set, and treat the remaining natural days as ordinary days. Clustering module 54 is used to perform order-preserving clustering on the ordinary days to obtain multiple ordinary day clusters, and extract the real natural day as the representative day from each ordinary day cluster; The construction module 55 is used to merge the natural days in the anchored sample set with the extracted representative days, sort them according to the original time order, and construct the scene chain sequence and the structured interface corresponding to the scene chain sequence.

[0105] The above is a schematic scheme of a new energy scenario chain construction device based on extreme daily anchoring and order-preserving clustering according to this embodiment. It should be noted that the technical solution of this new energy scenario chain construction device based on extreme daily anchoring and order-preserving clustering belongs to the same concept as the technical solution of the aforementioned new energy scenario chain construction method based on extreme daily anchoring and order-preserving clustering. Details not described in detail in the technical solution of the new energy scenario chain construction device based on extreme daily anchoring and order-preserving clustering can be found in the description of the technical solution of the aforementioned new energy scenario chain construction method based on extreme daily anchoring and order-preserving clustering.

[0106] Figure 6A structural block diagram of a computing device 600 according to one embodiment of this specification is shown. The components of the computing device 600 include, but are not limited to, a memory 610 and a processor 620. The processor 620 is connected to the memory 610 via a bus 630, and a database 650 is used to store data.

[0107] The computing device 600 also includes an access device 640, which enables the computing device 600 to communicate via one or more networks 660. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 640 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0108] In one embodiment of this specification, the above-described components of the computing device 600 and Figure 4 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 4 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0109] The computing device 600 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 600 can also be a mobile or stationary server.

[0110] The processor 620 executes the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering. The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the above-described method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering belong to the same concept. Details not described in detail in the technical solution of the computing device can be found in the description of the technical solution of the above-described method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering.

[0111] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described method for constructing a new energy scenario chain based on extreme day anchoring and order-preserving clustering.

[0112] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the above-described method for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the above-described method for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering.

[0113] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, the computer is instructed to perform the steps of the above-described method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering.

[0114] The above is an illustrative scheme of a computer program according to this embodiment. It should be noted that the technical solution of this computer program belongs to the same concept as the technical solution of the above-mentioned method for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering. For details not described in detail in the technical solution of the computer program, please refer to the description of the technical solution of the above-mentioned method for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering.

[0115] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0116] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0117] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0118] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0119] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method for constructing a new energy scene chain by extreme day anchoring and order-preserving clustering, characterized in that, include: Obtain hourly data on new energy resources and power transmission demand throughout the year; Based on the aforementioned new energy resource data and power transmission demand data, calculate the hourly joint extreme score; Based on the hourly joint extreme scores, extreme days are selected from the natural days of the whole year, and each extreme day and its adjacent days are selectively incorporated into the anchored sample set, and the remaining natural days are treated as ordinary days. Perform order-preserving clustering on the ordinary days to obtain multiple ordinary day clusters, and extract the real natural day as the representative day from each ordinary day cluster; The natural days in the anchored sample set are merged with the extracted representative days and sorted according to the original time order to construct a scene chain sequence and the structured interface corresponding to the scene chain sequence.

2. The method of claim 1, wherein, The new energy resource data includes wind power resource sequences and photovoltaic resource sequences, and the power transmission demand data includes power transmission demand sequences; based on the new energy resource data and power transmission demand data, an hourly joint extreme score is calculated, including: Calculate hourly renewable power output based on the wind power resource sequence, photovoltaic resource sequence, and preset wind power and photovoltaic installed capacity; Based on the hourly renewable output and the power supply demand sequence, calculate multiple hourly risk characteristics; The risk characteristics are normalized, and the time-by-time joint extreme score is calculated based on each normalized risk characteristic and its corresponding feature weight.

3. The method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering according to claim 2, characterized in that, The steps for obtaining the feature weights include: Based on the key focus of the plan, determine the types of extreme risks that the new energy base should focus on; Based on the extreme risk types and normalized risk characteristics, multiple sets of candidate feature weights are set; each set of candidate feature weights corresponds to a different risk preference. Based on the weights of each set of candidate features, the joint extreme scores are calculated sequentially over time to obtain the candidate scene chain sequence corresponding to each set of candidate feature weights. Calculate the average extreme feature capture rate for each group of candidate scene chain sequences, and count the total number of natural days contained in each group of candidate scene chain sequences; Each candidate scenario chain sequence is input into the downstream capacity planning, energy storage configuration, and robustness verification program, and the fluctuation range of the output results corresponding to each candidate scenario chain sequence is recorded. The set of candidate feature weights with the highest average extreme feature capture rate, the total number of natural days in the scene chain not exceeding the preset scale limit, and the smallest fluctuation in downstream output results are determined as the feature weights to be used in the official application.

4. The method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering according to claim 3, characterized in that, Calculate the average extreme feature capture rate for each candidate scene chain sequence, including: Select at least one target risk feature from multiple risk features to calculate the average extreme feature capture rate; For any target risk feature, extract the first feature statistical value of the target risk feature in the original sequence of the whole year, and extract the second feature statistical value of the target risk feature in the candidate scenario chain sequence; wherein, the original sequence of the whole year includes the hourly wind power resource sequence, the hourly photovoltaic resource sequence and the hourly power transmission demand sequence of the whole year; Divide the second feature statistical value by the first feature statistical value to obtain the single feature capture rate of the target risk feature; The average extreme feature capture rate of each candidate scenario chain sequence is obtained by summing the single feature capture rates of all target risk features and then dividing by the total number of target risk features.

5. The method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering according to claim 1, characterized in that, Based on the hourly joint extreme scores, a preset number of extreme natural days are selected from the natural days of the year, including: For each natural day, the maximum value of the hourly joint extreme scores across all time periods within that natural day is taken as the comprehensive extreme score for that natural day; Based on the overall extreme scores from highest to lowest, the top K natural days are selected as extreme natural days, where K is an integer greater than or equal to 1; when the overall extreme scores of multiple natural days are tied, the natural day with the earlier time is retained, or the tie is broken by using the single risk feature with preset priority and then the extreme natural day is added.

6. The method for constructing a new energy scenario chain based on extreme daily anchoring and order-preserving clustering according to claim 1, characterized in that, Each extreme natural day and its preceding and following natural days are selectively incorporated into the anchored sample set, including: If the extreme natural day and its adjacent natural days do not exceed the annual boundary and are not covered by the neighborhood of other extreme natural days, then the extreme natural day and its adjacent natural days are included in the anchored sample set. If the adjacent natural days before and after an extreme natural day are covered by the neighborhood of other extreme natural days or exceed the annual boundary, then only that extreme natural day itself is included in the anchored sample set.

7. The method for constructing new energy scenario chains based on extreme daily anchoring and order-preserving clustering according to claim 1, characterized in that, Construct the structured interface corresponding to the scene chain sequence, including: Each scene day in the scene chain sequence is expanded according to the annual hourly index correspondence, and then merged to obtain an hourly scene time period set; The scene weights are assigned based on the day of the scene: for natural days from the anchored sample set, the weight is set to 1; for representative days from ordinary day clusters, the weight is set to the total number of natural days contained in the ordinary day cluster to which it belongs. The two adjacent scene days in the scene chain sequence are grouped into ordered pairs, and the set of sequential relationships between adjacent scene days is obtained by summarizing them. The hourly scene time period set, scene weight set, and adjacent relationship set are collectively encapsulated into the structured interface for output.

8. A new energy scenario chain construction device based on extreme daily anchoring and order-preserving clustering, characterized in that, include: The acquisition module is used to acquire hourly data on new energy resources and power transmission demand throughout the year; The calculation module is used to calculate the hourly joint extreme score based on the new energy resource data and power transmission demand data; The filtering module is used to filter out extreme natural days from the natural days of the whole year based on the hourly joint extreme score, and selectively incorporate each extreme natural day and its adjacent natural days into the anchored sample set, and treat the remaining natural days as ordinary days. The clustering module is used to perform order-preserving clustering on the ordinary days to obtain multiple ordinary day clusters, and extract the real natural day as the representative day from each ordinary day cluster; The module is used to merge the natural days in the anchored sample set with the extracted representative days, sort them according to the original time order, and construct the scene chain sequence and the structured interface corresponding to the scene chain sequence.

9. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the new energy scenario chain construction method of extreme day anchoring and order-preserving clustering as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed by the processor, they implement the steps of the new energy scenario chain construction method of extreme day anchoring and order-preserving clustering as described in any one of claims 1 to 7.