A method and system for quantifying and optimally scheduling standby demand considering the space-time correlation characteristics of source load

Abstract

The application discloses a kind of considering source load space-time correlation characteristics backup demand quantification and optimization scheduling method and system, belong to power system optimization scheduling technical field.First, construct the high-dimensional uncertainty analysis framework that fuses timing and contemporaneous static correlation, through Markov chain, predicted power space adaptive partition and joint modeling, generate output scene set by sampling;Second, based on scenario set quantification multi-time interval net load climbing demand, and introduce tail sample adequacy detection and high-risk interval reverse directional expansion closed-loop mechanism;Finally, construct the scheduling model considering unit and energy storage physical constraint and cross-time section tidal flow transmission limit safety constraint, by generating timing polyhedral backup deliverable safety domain envelope, establish main energy and climbing backup joint optimization model.The application realizes the accurate quantification of backup demand and full-cycle physical send-out guarantee, effectively improves the safety and economy of high-proportion new energy power grid operation.

Description

Technical Field

[0001] This invention belongs to the field of power system optimization scheduling, and specifically relates to a method and system for quantifying and optimizing reserve demand considering the spatiotemporal correlation characteristics of source and load. Background Technology

[0002] The installed capacity of existing new energy sources such as wind power and photovoltaics continues to grow rapidly, gradually transforming from supplementary energy to primary energy. A high proportion of new energy grid connection has become a typical feature of the new power system. The large-scale, fluctuating, and intermittent access of new energy sources has significantly increased the amplitude and frequency of active power fluctuations in the power grid, leading to a surge in the demand for system net load ramping and continuously increasing regulation pressure. This places higher demands on the precise quantification, optimized allocation, and reliable dispatch of flexible ramping reserves.

[0003] Current traditional power system optimization scheduling models have three major flaws: First, the modeling accuracy of source-load uncertainty is insufficient. The temporal correlation between wind, solar and load and the static coupling characteristics of multiple sources are not fully integrated, resulting in an inaccurate characterization of the joint probability distribution of source-load prediction errors. This makes it difficult to generate output scenarios that fit the actual operating conditions, directly leading to a large deviation between the quantitative results of system ramp-up demand and the actual operating conditions.

[0004] Secondly, the modeling of ramp-up reserve suffers from physical adaptability defects. Existing models mostly adopt a static instantaneous power reservation approach, constraining reserve capacity only at a single moment, without considering the differences in ramp-up rates under different start-up and shutdown states of thermal power units, or the energy constraints of energy storage state of charge (SOC) on continuous regulation capabilities. Furthermore, they fail to construct full-cycle physical constraints for the multi-period continuous call characteristics of ramp-up reserve, easily leading to the problem of "sufficient reserve capacity but actual unusable capacity." The scheduling scheme becomes disconnected from the actual physical operating characteristics of the units and energy storage, failing to guarantee the safe and stable operation of the system.

[0005] Third, there is a lack of cross-period safety verification of the power flow transmission capacity of the power grid. Traditional dispatching often assumes that reserved reserves can be transmitted to the load side without loss, ignoring the dynamic congestion of transmission sections caused by large-scale power flow shifts during long-term reserve calls. This easily leads to the phenomenon of invalid pseudo-reserves where the generating units have the physical capacity to ramp up, but cannot transmit due to network topology limitations.

[0006] Therefore, accurately characterizing the spatiotemporal uncertainties of power source and load, scientifically quantifying ramping demand across multiple time scales, and constructing a flexible ramping optimization scheduling model that adapts to the physical characteristics of regulating resources have become key technical issues that urgently need to be addressed for the safe and economical dispatch of power systems. Summary of the Invention

[0007] To address the shortcomings of traditional scheduling models, such as neglecting the spatiotemporal correlation between source and load, the sparse long-tail samples during extreme ramping, and the limitations of continuous reserve deployment and network power flow transmission, this invention provides a method and system for quantifying and optimizing reserve demand based on the spatiotemporal correlation characteristics of source and load. First, this invention constructs a high-dimensional uncertainty analysis framework that integrates time-series and synchronous static correlations. Through Markov chains, adaptive partitioning of the predicted power space, and D-Vine Copula joint modeling, a power output scenario set is generated by sampling. Second, based on the scenario set, the net load ramping demand across multiple time intervals is quantified, and a tail sample sufficiency detection and a high-risk interval reverse directional expansion closed-loop mechanism are introduced. Finally, a scheduling model considering the physical constraints of generating units and energy storage, as well as the cross-time-period cross-sectional power flow transmission limit safety constraints, is constructed. By generating a time-series polyhedral reserve transportable safety domain envelope, a joint optimization model for main energy and ramping reserve is established. This invention achieves accurate quantification of reserve demand and full-cycle physical transportability assurance, effectively improving the safety and economy of high-proportion renewable energy grid operation.

[0008] The present invention adopts the following technical solution.

[0009] In a first aspect, the present invention provides a method for quantifying and optimizing the scheduling of reserve requirements, taking into account the spatiotemporal correlation characteristics of source and load, including: Acquire historical actual power output data, historical predicted power output data, and planned daily predicted power output data of wind, solar, and load. Determine the set of physical mutation critical power based on the physical characteristics of wind and solar power output, and generate physical mutation weights based on the set of physical mutation critical power. Combine the historical actual power output data of wind, solar, and load with the historical predicted power output data to adaptively partition the continuous predicted power space, and use the physical mutation weights to correct the adaptive partitioning process of the predicted power space to construct a conditional probability density model that integrates the planned daily predicted power output data. Establish a joint probability distribution model based on the historical actual power output data of wind, solar, and load. Combine the conditional probability density model with the joint probability distribution model to generate a multi-time period joint power output scenario set. Based on the multi-time period joint output scenario set, the net load ramp distribution at different time scales is calculated, and the tail sample sufficiency detection is performed. If the sample size in the extreme ramp interval on one side is lower than the preset sufficiency threshold, the scenario that causes the extreme ramp is reverse-mapped to a high-risk net output power interval, and the high-risk net output power interval is fed back to the predictive power space adaptive partitioning process for directional expansion until the sample sufficiency requirement is met, and the up and down ramp backup requirements are output. Based on the aforementioned ramp-up and ramp-up backup requirements, a flexible ramp-up optimization scheduling model is constructed that takes into account the physical constraints of the generating units and energy storage, as well as the safety constraints of cross-time section power flow transmission limits. The final scheduling scheme is obtained by solving the flexible ramp-up optimization scheduling model.

[0010] Preferably, a conditional probability density model integrating the predicted data for the planned days is constructed, and a joint probability distribution model is established based on the historical actual power output data of the wind, solar, and load. The conditional probability density model and the joint probability distribution model are combined to generate a multi-time period joint power output scenario set, specifically including: By introducing the Markov chain hypothesis, the temporal correlation between adjacent time periods in the historical actual power output data of wind and solar load is extracted, and the temporal joint distribution of power output in multiple time periods is modeled and reduced to a two-dimensional dependency relationship between adjacent time periods. Using the predicted data for the planned day as a priori conditions, combined with the two-dimensional dependency relationship, and based on the adaptive partitioning process of the predicted power space after the physical mutation weight correction, kernel density estimation is performed to obtain the conditional probability density of the power output at the current moment, thus forming the conditional probability density model. A multi-source synchronous static correlation model of wind and solar loads is constructed using a D-Vine Copula structure. During the calculation process, the actual output of each source load in the historical actual output data of wind and solar loads is transformed into a uniform distribution variable by fitting their respective marginal cumulative distribution functions. The uniform distribution variable is then mapped and transformed by the t-distribution inverse cumulative distribution function, and the mapping result is substituted into the t-Copula density function corresponding to the D-Vine Copula structure for joint calculation to obtain the joint probability distribution model. The multi-time period joint output scenario set is generated by combining the conditional probability density model and the joint probability distribution model.

[0011] Preferably, the process of determining a set of physical abrupt change critical power values ​​based on the physical characteristics of wind and solar power output, generating physical abrupt change weights based on the set of physical abrupt change critical power values, and using the physical abrupt change weights to correct the adaptive partitioning process of the predicted power space specifically includes: The set of critical power for physical abrupt change is determined by the wind turbine's power generation capacity when it is switched in, the wind turbine's rated full power, the wind turbine's power output when it is switched out and the photovoltaic power. A matching tolerance threshold is set. When recursively dividing the continuous predicted power space, if the absolute value of the difference between the candidate segmentation point and any element in the physical mutation critical power set is less than or equal to the matching tolerance threshold, the physical mutation weight takes a preset amplification factor; otherwise, the physical mutation weight takes a value of 1; the amplification factor is greater than 1. Calculate the decrease in integral mean square error before and after the predicted power space partitioning, and multiply the decrease in integral mean square error by the physical mutation weight to obtain the weighted error decrease. The candidate segmentation point that maximizes the reduction in the weighted error is selected as the optimal segmentation point, and the continuous predicted power space is divided into two sub-intervals based on the optimal segmentation point.

[0012] Preferably, tail sample sufficiency testing is performed, specifically including: Calculate the net load ramp-up of the multi-time period joint output scenario set at multiple time intervals, and statistically analyze the ramp-up quantiles and ramp-down quantiles under a preset confidence level; The expected number of theoretical one-sided tail samples is determined based on the preset confidence level and the total number of scenarios in the multi-time period joint output scenario set. The expected number of theoretical one-sided tail samples is then multiplied by a preset sufficiency verification coefficient to obtain the sufficiency threshold. The number of extreme uphill samples that are greater than or equal to the uphill quantile and the number of extreme downhill samples that are less than or equal to the downhill quantile are counted respectively.

[0013] Preferably, if the sample size within a single extreme ramp interval is lower than a preset sufficiency threshold, the scenario triggering the extreme ramp is reverse-mapped into a high-risk net power output interval, and the high-risk net power output interval is fed back into the adaptive partitioning process of the predicted power space for targeted expansion, specifically including: If the number of extreme uphill samples is less than the adequacy threshold, then the minimum net load at the corresponding moment that caused the extreme uphill phenomenon in the generated scenario, together with the historical maximum net load at the current moment, constitutes the high-risk net output power range for uphill. If the number of extreme downhill samples is less than the adequacy threshold, then the maximum net load value at the corresponding moment that caused the extreme downhill phenomenon in the generated scenario, together with the historical minimum net load value at the current moment, constitutes the downhill high-risk net output power range. The ratio of the adequacy threshold to the actual number of extreme climbing samples is calculated as the local oversampling coefficient, and the corresponding high-risk net output power interval is fed back to the predictive power space adaptive partitioning process. The local oversampling coefficient is used to increase the sampling weight within the interval.

[0014] Preferably, a flexible ramp-up optimization scheduling model is constructed that takes into account the physical constraints of generating units and energy storage, as well as the safety constraints of cross-time section power flow transmission limits, specifically including: The objective function is to minimize the sum of the no-load cost, start-up cost, shutdown cost, operating fuel cost of thermal power units, and the backup cost of thermal power units and energy storage modules providing flexible ramp-up backup. The physical constraints of the generating units and energy storage include: the output and ramping constraints of the thermal power units, the charging and discharging power constraints of the energy storage, the state of charge constraints of the energy storage, the energy storage power balance constraints, and the energy storage ramping constraints determined by the state of charge; when balancing the flexible ramping reserve requirements of the entire system, the ramping reserve capacity reserved by the thermal power units and the ramping reserve capacity provided by the energy storage modules are taken into account. Establish the static topological mapping relationship between the unit's basic power generation output, energy storage charging and discharging power, and power flow between source and load nodes and transmission sections, and calculate the baseline power flow. The cross-time section power flow transmission limit safety constraint includes: constraining the power flow increase or decrease mapped to the transmission section after the benchmark power flow is superimposed on the full utilization of the ramp-up reserve capacity of the thermal power unit and the energy storage module, so as not to exceed the forward transmission limit and reverse transmission limit of the transmission section within any continuous reserve call duration, so as to generate a time-series polyhedron reserve transportable safety domain envelope.

[0015] Preferably, the power output and ramping constraints of the thermal power unit specifically include cross-time period power output boundary constraints and continuous ramping rate constraints that take into account the differences in start-up and shutdown states. The cross-time period output boundary constraint is used to: synchronously verify the upper and lower limits of the output of the thermal power unit at the current time and the time when the standby call ends, and constrain the basic power generation output of the thermal power unit at the current time and the time when the standby call ends, so that after superimposing the reserved upward ramp standby capacity or deducting the reserved downward ramp standby capacity, it does not exceed the maximum and minimum output limits of the thermal power unit. The continuous ramp rate constraint that takes into account the difference between start-up and shutdown states is used to: combine the ramp rate capability per unit time of the thermal power unit in the operating state, start-up state and shutdown state, respectively calculate the maximum allowable upward ramp rate and the maximum allowable downward ramp rate within the standby continuous call duration, so as to limit the upward ramp standby capacity and the downward ramp standby capacity.

[0016] Secondly, the present invention provides a backup demand quantification and optimized scheduling system that considers the spatiotemporal correlation characteristics of source and load, comprising: The probability scenario generation module is used to acquire historical actual power output data, historical predicted power output data, and planned day predicted power output data of wind, solar and load. It determines the set of physical mutation critical power based on the physical characteristics of wind and solar power output, and generates physical mutation weights based on the set of physical mutation critical power. Combining the historical actual power output data and the historical predicted power output data of wind, solar and load, it adaptively partitions the continuous predicted power space, and uses the physical mutation weights to correct the adaptive partitioning process of the predicted power space, thereby constructing a conditional probability density model that integrates the planned day predicted power output data. It also establishes a joint probability distribution model based on the historical actual power output data of wind, solar and load, and combines the conditional probability density model with the joint probability distribution model to generate a multi-time period joint power output scenario set. The backup demand quantification module is used to calculate the net load ramp distribution at different time scales based on the multi-time period joint output scenario set, and to perform tail sample sufficiency detection. If the sample size in a single extreme ramp interval is lower than the preset sufficiency threshold, the scenario that triggers the extreme ramp will be reverse-mapped to a high-risk net output power interval, and the high-risk net output power interval will be fed back to the predictive power space adaptive partitioning process for directional expansion until the sample sufficiency requirement is met, and the backup demand for ramping up and down will be output. The optimization scheduling solution module is used to combine the uphill and downhill backup requirements to construct a flexible ramp optimization scheduling model that takes into account the physical constraints of the generating units and energy storage, as well as the safety constraints of cross-time section power flow transmission limits, and solve the flexible ramp optimization scheduling model to obtain the final scheduling scheme.

[0017] Preferably, the probability scenario generation module is specifically used to construct a conditional probability density model that integrates the predicted data for the planned days, and to establish a joint probability distribution model based on the historical actual output data of wind, solar and load. The conditional probability density model and the joint probability distribution model are then combined to generate a multi-time period joint output scenario set, specifically including: By introducing the Markov chain hypothesis, the temporal correlation between adjacent time periods in the historical actual power output data of wind and solar load is extracted, and the temporal joint distribution of power output in multiple time periods is modeled and reduced to a two-dimensional dependency relationship between adjacent time periods. Using the predicted data for the planned day as a priori conditions, combined with the two-dimensional dependency relationship, and based on the adaptive partitioning process of the predicted power space after the physical mutation weight correction, kernel density estimation is performed to obtain the conditional probability density of the power output at the current moment, thus forming the conditional probability density model. A multi-source synchronous static correlation model of wind and solar loads is constructed using a D-Vine Copula structure. During the calculation process, the actual output of each source load in the historical actual output data of wind and solar loads is transformed into a uniform distribution variable by fitting their respective marginal cumulative distribution functions. The uniform distribution variable is then mapped and transformed by the t-distribution inverse cumulative distribution function, and the mapping result is substituted into the t-Copula density function corresponding to the D-Vine Copula structure for joint calculation to obtain the joint probability distribution model. The multi-time period joint output scenario set is generated by combining the conditional probability density model and the joint probability distribution model.

[0018] Preferably, the probabilistic scenario generation module is specifically used to determine the set of physical abrupt change critical power based on the physical characteristics of wind and solar power output, generate physical abrupt change weights based on the set of physical abrupt change critical power, and use the physical abrupt change weights to correct the adaptive partitioning process of the predicted power space, specifically including: The set of critical power for physical abrupt change is determined by the wind turbine's power generation capacity when it is switched in, the wind turbine's rated full power, the wind turbine's power output when it is switched out and the photovoltaic power. A matching tolerance threshold is set. When recursively dividing the continuous predicted power space, if the absolute value of the difference between the candidate segmentation point and any element in the physical mutation critical power set is less than or equal to the matching tolerance threshold, the physical mutation weight takes a preset amplification factor; otherwise, the physical mutation weight takes a value of 1; the amplification factor is greater than 1. Calculate the decrease in integral mean square error before and after the predicted power space partitioning, and multiply the decrease in integral mean square error by the physical mutation weight to obtain the weighted error decrease. The candidate segmentation point that maximizes the reduction in the weighted error is selected as the optimal segmentation point, and the continuous predicted power space is divided into two sub-intervals based on the optimal segmentation point.

[0019] Preferably, the backup demand quantification module is specifically used to perform tail sample sufficiency detection, and specifically includes: Calculate the net load ramp-up of the multi-time period joint output scenario set at multiple time intervals, and statistically analyze the ramp-up quantiles and ramp-down quantiles under a preset confidence level; The expected number of theoretical one-sided tail samples is determined based on the preset confidence level and the total number of scenarios in the multi-time period joint output scenario set. The expected number of theoretical one-sided tail samples is then multiplied by a preset sufficiency verification coefficient to obtain the sufficiency threshold. The number of extreme uphill samples that are greater than or equal to the uphill quantile and the number of extreme downhill samples that are less than or equal to the downhill quantile are counted respectively.

[0020] Preferably, the backup demand quantification module is specifically used to, if the sample size within a single-sided extreme ramp interval is lower than a preset sufficiency threshold, reverse-map the scenario that triggers the extreme ramp into a high-risk net output power interval, and feed the high-risk net output power interval back to the predicted power space adaptive partitioning process for targeted expansion, specifically including: If the number of extreme uphill samples is less than the adequacy threshold, then the minimum net load at the corresponding moment that caused the extreme uphill phenomenon in the generated scenario, together with the historical maximum net load at the current moment, constitutes the high-risk net output power range for uphill. If the number of extreme downhill samples is less than the adequacy threshold, then the maximum net load value at the corresponding moment that caused the extreme downhill phenomenon in the generated scenario, together with the historical minimum net load value at the current moment, constitutes the downhill high-risk net output power range. The ratio of the adequacy threshold to the actual number of extreme climbing samples is calculated as the local oversampling coefficient, and the corresponding high-risk net output power interval is fed back to the predictive power space adaptive partitioning process. The local oversampling coefficient is used to increase the sampling weight within the interval.

[0021] Preferably, the optimized scheduling solution module is specifically used to construct a flexible ramp-up optimized scheduling model that takes into account the physical constraints of generating units and energy storage, as well as the safety constraints of cross-time section power flow transmission limits. Specifically, it includes: The objective function is to minimize the sum of the no-load cost, start-up cost, shutdown cost, operating fuel cost of thermal power units, and the backup cost of thermal power units and energy storage modules providing flexible ramp-up backup. The physical constraints of the generating units and energy storage include: the output and ramping constraints of the thermal power units, the charging and discharging power constraints of the energy storage, the state of charge constraints of the energy storage, the energy storage power balance constraints, and the energy storage ramping constraints determined by the state of charge; when balancing the flexible ramping reserve requirements of the entire system, the ramping reserve capacity reserved by the thermal power units and the ramping reserve capacity provided by the energy storage modules are taken into account. Establish the static topological mapping relationship between the unit's basic power generation output, energy storage charging and discharging power, and power flow between source and load nodes and transmission sections, and calculate the baseline power flow. The cross-time section power flow transmission limit safety constraint includes: constraining the power flow increase or decrease mapped to the transmission section after the benchmark power flow is superimposed on the full utilization of the ramp-up reserve capacity of the thermal power unit and the energy storage module, so as not to exceed the forward transmission limit and reverse transmission limit of the transmission section within any continuous reserve call duration, so as to generate a time-series polyhedron reserve transportable safety domain envelope.

[0022] Preferably, the power output and ramping constraints of the thermal power unit specifically include cross-time period power output boundary constraints and continuous ramping rate constraints that take into account the differences in start-up and shutdown states. The cross-time period output boundary constraint is used to: synchronously verify the upper and lower limits of the output of the thermal power unit at the current time and the time when the standby call ends, and constrain the basic power generation output of the thermal power unit at the current time and the time when the standby call ends, so that after superimposing the reserved upward ramp standby capacity or deducting the reserved downward ramp standby capacity, it does not exceed the maximum and minimum output limits of the thermal power unit. The continuous ramp rate constraint that takes into account the difference between start-up and shutdown states is used to: combine the ramp rate capability per unit time of the thermal power unit in the operating state, start-up state and shutdown state, respectively calculate the maximum allowable upward ramp rate and the maximum allowable downward ramp rate within the standby continuous call duration, so as to limit the upward ramp standby capacity and the downward ramp standby capacity.

[0023] Thirdly, the present invention provides a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method.

[0024] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method.

[0025] The beneficial effects of this invention are compared with those of the prior art: (1) This invention proposes a physical mutation weight correction method for adaptive partitioning of predicted power space. Physical mutation weighting coefficients are introduced into the recursive binary partitioning objective function based on the optimal criterion of integral mean square error (IMSE). Physical mutation partitioning points such as wind turbine cut-in power, rated full power, cut-out shutdown power, and photovoltaic inverter saturation power are assigned weights greater than 1 to amplify the error reduction and form an adaptive binning mechanism driven by both statistical error and power generation physical mechanism. It realizes the differentiated predicted power space partitioning with automatic refinement of sample dense area, automatic coarsening of sample sparse area, and adaptive densification of physical mutation area. It overcomes the defects of traditional pure data-driven binning that ignores power generation mechanism and insufficient accuracy of probability density fitting in mutation interval. It provides a basic guarantee for accurate modeling of probability density under predicted information conditions and the quality of subsequent scene generation.

[0026] (2) This invention constructs a closed-loop mechanism for tail sample sufficiency verification and reverse resampling. Based on the target confidence level, a single-sided tail sample sufficiency threshold is set to quantitatively determine the number of samples in the extreme ramp interval. When the sample is insufficient, the high-risk net output power interval that induces extreme ramp is reverse mapped and sent back to the predictive power space adaptive partitioning module. The extreme scenario is expanded in a directional manner according to the local oversampling coefficient, and the scenario set is iteratively expanded until the verification is passed. This closed-loop mechanism makes up for the defect of missing tail samples in traditional sampling, ensures the physical credibility of the reserve capacity quantification in the extreme ramp scenario of high-energy power grid, and avoids the risk of insufficient reserve to scheduling operation.

[0027] (3) This invention establishes a joint optimization scheduling model for main energy and flexible ramping reserve that takes into account energy storage and cross-time section power flow safety domain. It constructs cross-time output boundary constraints, continuous ramping rate constraints that take into account start-stop state differences, and energy storage capacity verification constraints that are oriented towards continuous call duration. Based on the DC power flow node power transfer factor matrix, it adds a backup transportability coupling constraint that dynamically binds the unit's ramping output increment across all time periods with the transmission limit of each transmission section across time periods, forming a time-series polyhedral safety domain envelope that expands dynamically with the backup duration. It realizes the ramping reserve feasibility verification of the four-dimensional whole chain from "unit power - duration - energy storage energy - power flow", eliminates pseudo-reserves that "the unit has ramping capacity but the network cannot transmit", solves the problem that traditional single-moment verification cannot cover the entire process of backup continuous call, and the problem of dynamic congestion of transmission sections caused by large-scale power flow transfer during backup call. It realizes accurate quantification of multi-time scale ramping reserve and full-cycle, full-dimensional safety optimization configuration.

[0028] (4) This invention proposes a joint modeling method for high-dimensional uncertainty of wind-solar-load by integrating Markov chain dimensionality reduction, conditional probability modeling of prediction information and D-Vine Copula. It reduces the temporal joint distribution of power output in multiple time periods to a two-dimensional dependency relationship between adjacent time periods, and incorporates the predicted power output at the current time as a condition into the scene sampling. At the same time, it portrays the temporal correlation of source and load and the static correlation of wind, solar and load in the same period. It solves the problems of traditional Markov chain scene generation not utilizing prediction priors, insufficient scene accuracy and high complexity of high-dimensional temporal joint distribution modeling. The generated scene has both temporal smoothness and multi-source coupling characteristics, providing high-quality scene input for hill-climbing demand assessment.

[0029] (5) Based on the generated multi-scenario net load curve, this invention statistically analyzes the upper and lower quantiles of the net load ramp distribution at a set confidence level at multiple time intervals such as 15min, 60min, and 120min. It realizes the probabilistic quantification of ramp demand at different times and multiple time scales, which changes the shortcomings of traditional deterministic ramp demand estimation that cannot reflect the level of uncertainty. It provides a direct quantitative basis for the differentiated and refined configuration of ramp reserve capacity at different confidence levels. Attached Figure Description

[0030] Figure 1 This is an overall flowchart of a backup demand quantification and optimized scheduling method that considers the spatiotemporal correlation characteristics of source and load provided by the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0032] Example 1: like Figure 1 As shown, a method for quantifying and optimizing the scheduling of reserve demand considering the spatiotemporal correlation characteristics of source loads is presented. The method includes the following steps: S1: Obtain historical actual power output data, historical predicted power output data, and planned daily predicted power output data of wind, solar, and load; determine the set of physical mutation critical power based on the physical characteristics of wind and solar power output, and generate physical mutation weights based on the set of physical mutation critical power; combine the historical actual power output data of wind, solar, and load with the historical predicted power output data to adaptively partition the continuous predicted power space, and use the physical mutation weights to correct the adaptive partitioning process of the predicted power space to construct a conditional probability density model that integrates the planned daily predicted power output data; and combine the conditional probability density model with the established joint probability distribution model based on the historical actual power output data of wind, solar, and load to generate a multi-time period joint power output scenario set.

[0033] Specifically, the historical actual operating power data of wind power, photovoltaic power and loads are obtained through the power system's data acquisition and monitoring control (SCADA) system or new energy power prediction system as historical actual output data of wind, solar and loads, and the corresponding historical day-ahead predicted power data are obtained as historical predicted output data. At the same time, the day-ahead predicted power data of wind power, photovoltaic power and loads for each time period of the planned day issued by the dispatch center are obtained as the planned day prediction data.

[0034] S1.1: First, a wind / solar / load time correlation model that integrates temporal correlation and forecast information is proposed; S1.1.1: Introducing the Markov chain hypothesis, by extracting the temporal correlation of power output at adjacent times in the historical actual power output data of wind and solar load, the temporal joint distribution of power output in multiple time periods is modeled and reduced to a two-dimensional dependency relationship between adjacent times.

[0035] In a preferred but non-limiting embodiment of the present invention, S1.1.1 specifically includes: To address the rapid decay of the short-term temporal correlation between renewable energy and load output, a Markov chain hypothesis is introduced. The temporal resolution of historical actual output data for wind and solar loads is 15 minutes (or 1 hour, depending on scheduling requirements). When processing historical actual output data, continuous time-series data sequences are reconstructed into two-dimensional data pairs consisting of the current moment and the previous moment, retaining only the temporal correlation between outputs at adjacent moments. Through this Markov hypothesis, the modeling problem of the original high-dimensional joint distribution of temporal correlations is reduced to modeling the two-dimensional dependency relationships between adjacent moments, significantly reducing the modeling complexity and computational cost of the joint distribution of outputs across multiple time periods.

[0036] S1.1.2: Using the predicted data for the planned day as a priori conditions, and combining it with two-dimensional dependencies, the conditional probability density of the output at the current moment is obtained through conditional probability density estimation, thus forming the conditional probability density model.

[0037] In a preferred but non-limiting embodiment of the present invention, S1.1.2 specifically includes: The conditional probability density model for the power output at the current moment is constructed within a Markov chain modeling framework, integrating prediction information. This model uses the power output prediction information extracted from the planned day prediction data as an additional condition to build the conditional probability density model for the power output at the current moment. .

[0038] in, This refers to the actual output of wind power, solar power, or load at the current time t. The actual output of wind power, solar power, or load at the previous moment. This is the predicted output for the current moment extracted from the predicted data for the planned day. The model utilizes... and While constructing a two-dimensional dependency relationship, it effectively integrates known prediction priors, solving the problems of unused prediction information and insufficient scene accuracy in traditional Markov chain scenario generation.

[0039] According to the chain rule of conditional probability density, both ternary joint density and binary joint density can be decomposed into the form of "conditional density × marginal density". The conditional probability density model of the output at the current moment can be calculated as follows: (1) According to Sklar's theorem for conditional copulas, the joint conditional probability density can be expressed as: (2) Therefore, the conditional probability density model of the force output at the current moment can be transformed into: (3) in: This represents the actual output of wind power, solar power, or load at time t; Given the prediction information, the probability density function is calculated using the power space adaptive partitioning and kernel density estimation method described in S1.1.3 below; Let X be the corresponding conditional cumulative distribution function, where X is a random variable, specifically referring to the output of wind power, photovoltaic power, or load; Given the conditional Copula density function with known prediction information, the t-Copula is used to describe the non-Gaussian skewness characteristics of wind power, photovoltaic power, or load output. The specific calculation is as follows: General expression for binary standard t-Copula density: (4) In the formula That is, the cumulative distribution function and correlation coefficient under the aforementioned predictive information conditions. Degrees of freedom .against The calculation, taking into account the smooth evolution of the physical state between adjacent time points, approximates the cumulative distribution of the output at the previous time point using an engineering approach, employing the cumulative distribution function under the condition of its own predictive information at the previous time point instead. .

[0040] In particular, , ,in For degrees of freedom The inverse cumulative distribution function of the standard univariate t-distribution.

[0041] By constructing the conditional t-Copula log-likelihood function, and then substituting it into the bivariate t-Copula density, and splitting the logarithmic terms, we can obtain the following formula: (5) in, , , , represent the values ​​of the k-th sample after inverse mapping through the t-distribution; N is the total number of samples.

[0042] Solving the maximum likelihood function using a numerical iterative algorithm Initial value of correlation coefficient By calculating all The Kendall rank correlation coefficient (or Spearman rank correlation coefficient) is used, and the correlation transformation formula is applied as follows: Obtain, among which, Kendall's rank correlation coefficient; initial values ​​of degrees of freedom. In the candidate set Grid traversal, substituting into the likelihood function, taking the values ​​that make the grid traversal... The largest value is used as the initial degree of freedom.

[0043] Then, the optimal parameters are found iteratively, respectively for... Find the first-order partial derivative This forms the gradient vector, which is then used to calculate the current gradient. Lower log-likelihood The upward direction; moving slightly along the upward direction. Inspection scope after relocation: If the value exceeds the limit, pull it back to the boundary; repeatedly move the parameters and recalculate the likelihood value; when two consecutive iterations occur, If the likelihood value hardly changes, or stops increasing, then... That is the optimal estimate.

[0044] S1.1.3: The distribution of wind and solar power output / load exhibits significant heterogeneity with varying power levels. A single global probability density model cannot characterize the distribution differences across different prediction intervals, leading to biases in subsequent conditional probability modeling. To address this, this invention proposes a method of "adaptive partitioning of the predicted power space + interval-based kernel density estimation." This method divides the continuous predicted power space into several state intervals, where the actual power output distribution exhibits high homogeneity. Then, nonparametric kernel density estimation is used to fit the interval PDF, achieving accurate modeling of the conditional probability density.

[0045] The core improvements to the "adaptive partitioning of predicted power space + interval kernel density estimation" are as follows: The aforementioned adaptive partitioning of the predicted power space is a recursive binning method based on the optimal integral mean square error (IMSE) criterion. With the goal of minimizing the global statistical error of probability density estimation, it adaptively generates partitioning intervals through a recursive binary search strategy, thereby automatically narrowing intervals in dense sample areas and automatically widening intervals in sparse sample areas.

[0046] Let the sample set of historical predicted power and actual power output be . The range of values ​​is The actual output value range is The goal is to adaptively divide the data into K non-overlapping sub-intervals. This minimizes the integral mean square error (IMSE) of the global probability density estimation after kernel density estimation is performed separately in each sub-interval.

[0047] First, define the core evaluation indicators: (6) (1) Adopt a recursive binary search strategy: starting from the entire prediction power space, insert dividing points one by one to divide the original interval into two; starting from the initial global interval Begin by dividing the current interval into segments. Insert split point Divide it into two sub-intervals: (7) The split point s is taken from the adjacent points of the sorted samples in the current interval, ensuring that the candidate split point satisfies the condition that both sub-intervals after splitting contain at least 1000 points. A historical sample.

[0048] (2) The selection of each split is based on the criterion of maximizing the decrease in local IMSE; Local IMSE definition: for any interval Its local error is: (8) in This represents the range of values ​​for the actual output force. To predict power belonging to the interval The actual conditional probability density function corresponding to the time is denoted as [the conditional symbol is omitted in subsequent formulas for simplicity]. ; The probability density function is estimated using interval kernel density estimation: (9) in, For Gaussian kernel function, The number of data points within the interval. For core bandwidth, , Belongs to subinterval Actual output sample set The standard deviation.

[0049] Due to the true probability density function f (x) is unknown. A computable approximate estimator of the integral mean square error is constructed using leave-one-out cross-validation. : (10) In the formula, To remove the i-th sample pair Then, based on sub-intervals The actual output of the remaining samples, and the kernel density estimate calculated at that point.

[0050] Existing recursive binary search methods based on IMSE decrease rely solely on pure statistical error to select segmentation points, representing a general data binning approach that fails to consider the inherent physical nonlinear boundaries of wind and solar power output. To integrate the power physics mechanism, this embodiment introduces a physical mutation weighting coefficient into the error decrease objective function. The method assigns a weight greater than 1 to the segmentation points of the physical mechanism abrupt change intervals to amplify the error reduction. The modified optimal segmentation criterion takes into account both the probability density fitting accuracy and the physical characteristics of the output. Under the same statistical error, it prioritizes the completion of interval cutting at the physical abrupt change location, realizing adaptive densification of the abrupt change region. This overcomes the shortcomings of traditional pure data-driven binning that ignores the generator mechanism and has insufficient fitting accuracy of the abrupt change interval, forming a statistical-physical dual-driven adaptive power space partitioning method.

[0051] In a preferred but non-limiting embodiment of the present invention, the specific steps include: determining the wind turbine's power generation capacity, rated full-capacity power generation, power generation capacity, and rated output power of the photovoltaic system to form the physical mutation critical power set; setting a matching tolerance threshold, wherein when recursively dividing the continuous predicted power space, if the absolute value of the difference between the candidate segmentation point and any element in the physical mutation critical power set is less than or equal to the matching tolerance threshold, the physical mutation weight is set to a preset amplification factor; otherwise, the physical mutation weight is set to 1; the amplification factor is greater than 1; calculating the decrease in integral mean square error before and after the division of the predicted power space, multiplying the decrease in integral mean square error by the physical mutation weight to obtain the weighted error decrease; selecting the candidate segmentation point that maximizes the weighted error decrease as the optimal segmentation point, and dividing the continuous predicted power space into two sub-intervals based on the optimal segmentation point. The implementation details are as follows: Define the set of critical power sets for physical abrupt changes in wind and solar power output. : (11) In the formula: To cut off the power generation capacity of the wind turbine, The two are the cut-out shutdown power of the wind turbine, specifically the corresponding critical power values ​​obtained by mapping the wind turbine's cut-in and cut-out wind velocities through the wind turbine power curve, respectively. This refers to the rated full-power output of the wind turbine. This refers to the saturated rated output power of the photovoltaic inverter.

[0052] In this embodiment, when the photovoltaic output approaches the inverter's rated output power, the actual photovoltaic output characteristics and prediction error distribution will experience boundary effects due to the inverter's capacity saturation limitation. Therefore, the rated output power is used as the physical abrupt change critical point of the prediction power space.

[0053] Constructing the physical abrupt change weighting coefficients at the split points : (12) in The amplification factor has an engineering value range of 2 to 5; the larger the value, the higher the priority of dividing the physical mutation boundary. p represents the set of critical power levels for physical mutations. Elements in; This is the set matching tolerance threshold. Since the actual sample split point is unlikely to be exactly equal to the theoretical mutation value, this tolerance threshold (or when s is the effective sample gap closest to the physical mutation point) is introduced to ensure the effective activation of physical prior information.

[0054] Let the local error of the parent interval I be... Dividing the data into I1 and I2, the error reduction after introducing physical weighting correction is calculated as follows: (13) The optimal split point is calculated as follows, with the reduction in the weighted physical-statistical joint error as the optimization objective: (14) in This is the set of all feasible partition points within the current interval.

[0055] (3) Greedy recursion termination and adaptive result (formulaic termination condition) Repeat steps (1) and (2) until any of the following termination conditions are met: The total number of intervals has reached the preset limit: ; The error decrease is less than the convergence threshold: ; The interval has reached the constraint boundary: ( Current interval width; (Number of samples in the interval).

[0056] The final partitioning result satisfies: It achieves adaptive partitioning by automatically refining dense sample regions and automatically coarsening sparse sample regions.

[0057] (4) Adaptive matching mapping of conditional probability density: After completing the spatial adaptive partitioning and kernel density estimation of each sub-interval, predict the output for any given current time. Find the specific sub-interval to which it belongs based on its numerical value. (i.e., satisfy) Then, the probability density function required in step S1.1.2 under the given prediction information conditions. The predicted power range can be used. Corresponding kernel density estimation model To approximate it, that is:

[0058] Correspondingly, its conditional cumulative distribution function The result is obtained by integrating the partition kernel density estimation model:

[0059] in, This is the physical lower bound corresponding to the actual output of wind power, photovoltaic power, or load.

[0060] S1.2: A joint probability distribution model considering the synchronous static correlation of multiple source random variables of wind and solar load is proposed; The D-Vine Copula is used to characterize the static correlation of wind and solar loads across multiple variables during the same period. In the calculation process, the actual output of each source load in the historical actual output data of wind and solar loads is converted into a uniform distribution variable through its respective marginal cumulative distribution function. The uniform distribution variable is then mapped and transformed through the t-distribution inverse cumulative distribution function. The mapping result is then substituted into the t-Copula density function corresponding to the D-Vine Copula structure for joint calculation to obtain the joint probability distribution model.

[0061] In a preferred but non-limiting embodiment of the present invention, S1.2 specifically includes: Suppose that at a certain time t, the wind power output is Photovoltaic power output The load is Then its joint probability density function is: (15) in: These are the marginal probability density functions for wind power, photovoltaic power, and load output, respectively. This is the cumulative distribution function of the corresponding variable; A binary Copula function, used to characterize the direct correlation between variables; It is a conditional bivariate Copula function that characterizes the correlation between wind power and load under the condition that the photovoltaic output is known; S1.3: A joint probability distribution model of wind-solar-load output is proposed, which considers the integration of multivariate coupling characteristics and time-series correlation characteristics.

[0062] The multi-time-period joint output scenario set is generated by combining the conditional probability density model and the joint probability distribution model. In a preferred but non-limiting embodiment of the present invention, S1.3 specifically includes: S1.3.1: Initialize the first time-stamp sample (t=1), directly from the static joint density. Monte Carlo sampling is performed to obtain the starting point that considers the synchronous static correlation characteristics of wind-solar-load.

[0063] (16) S1.3.2: Combining the existing time-series correlation model of wind / solar / load with the predicted information at time t, i.e. Generate the time-series prediction value for each variable at the current time: (17) Obtain the time series candidate vector: (18) S1.3.3: Due to It satisfies the time series characteristics but not the joint correlation characteristics of wind-solar-load, so it needs to be projected onto the joint distribution. This is done in advance using a joint probability density model. Generate a static candidate sample set of size N (e.g., N=1000) by sampling a large number of samples. Let the samples in this set be... Then, find the closest one in the candidate set. The sample is used as the actual scene value at the current moment, which satisfies the joint correlation characteristics of wind-solar-load while being as close as possible to the time series transfer result.

[0064] A standardized Euclidean distance metric is used to eliminate the influence of differences in the absolute dimensions of different energy outputs, and to evaluate static candidate samples. With time series candidates How close to the subject: (19) in, , , These are the standard deviations of historical output for wind power, solar power, and load, respectively. The above-mentioned scale N follows a joint density. Find the closest sample in the sample set. The vector is used as the actual scene value at the current moment. : (20) S1.3.4: Let t=t+1, repeat steps S1.3.2 to S1.3.3 until t=T, to form a complete time series scenario of wind-solar-load output that integrates multiple correlations.

[0065] S2: Calculate the net load ramp distribution at different time scales based on the multi-time period joint output scenario set, and perform tail sample sufficiency detection; if the sample size in the extreme ramp interval on one side is lower than the preset sufficiency threshold, the scenario that causes the extreme ramp will be reverse-mapped to a high-risk net output power interval, and the high-risk net output power interval will be fed back to the predictive power space adaptive partitioning process for directional expansion until the sample sufficiency requirement is met, and the up and down ramp backup requirements will be output.

[0066] S2.1: First, based on the generated series of scenarios, calculate the net load time series curve for each scenario i; (twenty one) in, This represents the net load at time t in scenario i. The load at time t under scenario i; Let be the wind power output at time t in scenario i; Provide photovoltaic power output at time t under scenario i; S2.2: Calculate multiple time intervals at time t for each scenario. ( The net load ramp-up (e.g., 15min, 60min, 120min, etc.) was calculated, and the ramp-up quantiles and ramp-down quantiles under the preset confidence level were statistically analyzed to obtain the ramp-up demand at different time intervals t.

[0067] (twenty two) (twenty three) The above formula, For scenario i, time t to t+ Sample values ​​of net load ramp distribution; Let time t be the time interval from t to t+, which consists of the set of all scenes. The random variable representing the net load ramp-up distribution; Confidence level; , From time t to t+ Net load ramp-up at confidence level The upper and lower quantiles are given. Therefore, the uphill climb requirement at different time intervals at each time t is... Downhill climbing requirements are .

[0068] Traditional one-way Monte Carlo sampling has sparse samples in the long tail of extreme net load ramping. Directly using Equation (23) to calculate quantiles will cause the ramping reserve demand to be distorted, introducing operational risks to the subsequent scheduling model.

[0069] To address this, a tail sample sufficiency check and scenario resampling closed-loop process have been added: The expected number of theoretically unilateral tail samples is determined based on the preset confidence level and the total number of scenarios in the multi-time-period joint output scenario set. This expected number of theoretically unilateral tail samples is then multiplied by a preset sufficiency verification coefficient to obtain the sufficiency threshold. Specifically, this includes: (twenty four) in, This is the sufficiency threshold. This represents the total number of scenes generated in step S1.3; denoted as , where b is the expected number of samples that should theoretically fall into the single-tailed region; b is the tail sample sufficiency verification coefficient (often taken as 1.2-2.0 in engineering) used to ensure that the number of samples in extreme scenarios has a certain risk-resistance margin.

[0070] First, calculate the confidence level using the following formula. Corresponding sufficiency threshold Count the number of extreme uphill samples that are greater than or equal to the uphill quantile. and the number of extreme downhill samples less than or equal to the downhill quantile. Determine if there is sufficient tail sample: (25) If the sample size within a single extreme ramp interval is lower than a preset sufficiency threshold, the scenario that triggers the extreme ramp will be mapped inversely to a high-risk net power output interval. Specifically, it includes: If the number of extreme uphill samples Less than the sufficiency threshold This will be the minimum net load at the corresponding moment that triggers the extreme uphill phenomenon in the generated scenario. Compared with the historical maximum net load at the current moment This constitutes a high-risk net power output range for uphill climbing. If the number of extreme downhill samples Less than the sufficiency threshold Then the maximum net load value at the corresponding moment that caused the extreme downhill climbing phenomenon in the generated scenario will be... Compared with the historical minimum net load at the current moment This constitutes a high-risk net power output range during downhill climbing.

[0071] This mapping identifies the power output range that is highly susceptible to inducing extreme uphill climbs. The calculation formula is as follows: (26) For a single-sided interval that triggers resampling, the ratio of the sufficiency threshold to the actual number of extreme uphill samples is calculated as the local oversampling coefficient. : (27) In the formula, This represents the actual extreme climbing sample size, which is used when uphill resampling is triggered. When resampling is triggered on the downhill slope, take .

[0072] and the corresponding high-risk net output power range Feedback is fed back to the adaptive partitioning process of the predicted power space for targeted augmentation, utilizing the local oversampling coefficients. Increase the sampling weight within the interval to generate a new set of scenarios. : (28) Based on equation (28), the original scene and the newly grown tail scene are merged to obtain the augmented scene set. , among which, in execution When generating new scenarios, the net output power range at the system level needs to be considered. Based on the established sensitivity of wind and solar power output reduction under extreme climbing scenarios (or the rules for combinations of extremely poor operating conditions), the reverse decoupling mapping is applied to the predicted power space of each wind and solar power unit, utilizing local oversampling coefficients. Increase the sampling probability of the corresponding extreme wind and light conditions sub-interval; re-enter the S2.2 iterative calculation formulas (22) and (23) until the tail sample meets the sufficiency requirement, and output reliable uphill and downhill backup requirements. This closed-loop mechanism compensates for the lack of tail samples in traditional sampling, ensuring the physical reliability of reserve capacity quantification under extreme ramping scenarios in high-energy power grids.

[0073] S3: Combining the aforementioned ramp-up and ramp-up backup requirements, a flexible ramp-up optimization scheduling model is constructed that takes into account the physical constraints of the generating units and energy storage, as well as the safety constraints of cross-time section power flow transmission limits. The final scheduling scheme is obtained by solving the flexible ramp-up optimization scheduling model, thereby achieving accurate quantification and optimized configuration of ramp-up backup.

[0074] S3.1: Based on the classic multi-period deterministic unit combination model, the system's flexible ramp-up reserve requirements and capacity constraints are incorporated into the model to establish a scheduling model that jointly optimizes main energy and flexible ramp-up reserve.

[0075] The objective function is to minimize the sum of the no-load cost, start-up cost, shutdown cost, operating fuel cost of the thermal power unit, and the backup cost of the thermal power unit and energy storage module providing flexible ramp-up backup, as follows: (29) In the formula, G represents the set of thermal power units; IS represents the set of energy storage modules. Let g be a binary variable representing the on / off state of unit g at time t (1 for on, 0 for off). Let g be a binary variable indicating whether unit g is started at time t (1 for start, 0 for no start). The variable is a binary variable characterizing whether unit g is shut down at time t (1 for shutdown, 0 for no shutdown). , , These are the no-load cost, start-up cost, and shutdown cost of unit g, respectively. The active power output of unit g at time t; Let g be the fuel cost function for unit g. , The cost of providing uphill and downhill flexible ramp-up standby capacity for either unit g or energy storage module s; and These represent the time intervals reserved by unit g or energy storage module s at time t, respectively. Uphill reserve capacity and downhill reserve capacity.

[0076] It should be noted that the different time intervals in this embodiment (e.g., 15min, 60min, 120min) represent different time scales for flexible ramp-up backup demand and continuous call-up capability, used to characterize short-term, medium-term, and long-term backup demand and their continuous call-up constraints respectively; different time scales do not repeatedly price the same physical backup capacity, but rather form backup demand constraints and physical feasibility verification for the corresponding duration.

[0077] S3.2: The physical constraints of the generating unit and energy storage include: the output and ramping constraints of the thermal power unit, the charging and discharging power constraints of the energy storage, the state of charge constraints of the energy storage, the energy storage power balance constraints, and the energy storage ramping constraints determined by the state of charge; when balancing the flexible ramping reserve requirements of the entire system, the ramping reserve capacity reserved by the thermal power unit and the ramping reserve capacity provided by the energy storage module are taken into account.

[0078] To address the shortcomings of traditional ramp-up standby modeling, which often focuses only on instantaneous power reservation at a single moment and does not constrain the continuous feasibility of standby deployment, making it difficult to adapt to the needs of standby products across multiple time scales, a ramp-up standby unit output reservation mechanism that takes into account the duration constraints of multiple time periods was constructed, realizing the full-cycle physical constraints of standby capacity: The power output and ramping constraints of the thermal power units specifically include cross-time period power output boundary constraints and continuous ramping rate constraints that take into account the differences in start-up and shutdown states. Equations (30)-(33) construct cross-time period output boundary constraints, which are used to: synchronously verify the thermal power unit's output at the current time and the standby call end time t+. The upper and lower limits of the output constrain the basic power generation output of the thermal power unit at the current time and the time when the standby call ends. After adding the reserved upward ramp standby capacity or deducting the reserved downward ramp standby capacity, the output does not exceed the maximum and minimum output limits of the thermal power unit, thus avoiding the defect that traditional single-time constraints cannot guarantee the compliance of output over continuous periods. Equations (34)-(35) propose a continuous ramp rate constraint that takes into account the difference between start-up and shutdown states. The continuous ramp rate constraint that takes into account the difference between start-up and shutdown states is used to: combine the ramp rate capability per unit time of the thermal power unit in the running state, start-up state and shutdown state, respectively calculate the maximum allowable ramp rate upward and the maximum allowable ramp rate downward within the standby continuous call duration, so as to limit the standby capacity upward and the standby capacity downward, and ensure that the ramp rate action of the unit in the entire standby call window conforms to the physical limits.

[0079] Equations (36)-(37) constrain the balance of flexible ramp-up backup requirements for the entire system.

[0080] Thermal power unit output and gradient constraints: , (30) , (31) , (32) , (33) , (34) , (35) , (36) , (37) In the above formula, , These represent the unit's upward climbing ability per unit time at time t, in both the running and startup states. , These represent the unit's downward ramping capability per unit time at time t, in both the operating and shutdown states. , For the maximum and minimum output of the unit g. , These represent the uphill and downhill climbing requirements at different time intervals at various times t; , These provide flexible ramp-up and ramp-down backup capacity for the energy storage module s.

[0081] Equation (38) and subsequent formulas establish energy storage charging and discharging and energy capacity verification constraints for continuous call duration, ensuring from an energy perspective that energy storage will not be overcharged / over-discharged during the entire standby call period, thus solving the problem of traditional methods that only verify power and do not consider continuous call energy support.

[0082] Energy storage operation constraints: (1) Energy storage charging and discharging power constraints: (38) (39) (40) In the formula, , The actual charging and discharging power of energy storage module s at time t; , These are the maximum charging and discharging power of the energy storage, respectively. , These are the charging and discharging state variables of energy storage, respectively, during charging. The value is 1, during discharge. The value is 1.

[0083] (2) Energy storage state of charge constraints: (41) (42) (3) Energy storage power balance constraints (43) In the formula Let be the state of charge of the energy storage s at time t; , These represent the maximum and minimum states of charge of the energy storage s, respectively. Let be the amount of energy stored in s at time t; The rated capacity (maximum electrical energy that can be stored) of energy storage s. , The charging and discharging efficiency of energy storage s.

[0084] (4) Energy storage ramping constraints determined by state of charge (44) (45) It should be noted that when converting between electricity and power in equations (43)-(45), the duration of continuous standby call is... Standardized base unit conversion is required, for example when When the time is set to 15 minutes, it is converted to 0.25 hours in the formula to ensure the consistency of physical dimensions in the calculation of system energy and power.

[0085] S3.3: Existing multi-period continuous ramp-up standby scheduling models only verify standby adjustability from the perspectives of unit power, duration, and energy storage capacity. They assume that the unit's reserved ramp-up standby can be fully transmitted to the load side without loss, completely ignoring the dynamic congestion problem of transmission sections caused by large-scale power flow shifts during long-sequence standby calls. This easily leads to the pseudo-standby defect of "the unit physically possesses ramp-up capacity, but the network cannot transmit it." To address this, this patent adds a multi-period standby transmittance safety domain envelope constraint with dynamic power flow congestion delay. Based on the DC power flow node power transfer factor matrix, a time-series topology coupling inequality is constructed, dynamically binding the unit's ramp-up output increment across all time periods to the upper limit of transmission capacity of each transmission section, generating a variable that varies with the standby duration. The dynamically expanding temporal polyhedral security domain envelope completes the ramp-up standby feasibility verification from the four-dimensional full chain of "unit power - duration - energy storage energy - grid power flow", filling the network topology blind spot of traditional standby modeling.

[0086] (1) Definition of basic power flow topology mapping: Establish the static topological mapping relationship between the unit's basic power generation output, energy storage charging and discharging power, and the power flow between source-load nodes and transmission sections, and calculate the baseline power flow: Let F be the set of transmission sections, and let the DC power flow node transfer factor matrix be... Matrix elements This represents the linear contribution coefficient of the unit output increment of unit g to the power flow at section f; similarly, The transfer factor of the node where energy storage s is located to section f; For the basic power output of unit g at time t, , These represent the charging and discharging power of the energy storage, respectively. Let f be the reference power flow at section f at time t. Establish the static topology mapping relationship between unit output and power flow at section f: (46) In the formula, The reference tidal force component generated by wind and solar power output and system load at section f.

[0087] (2) Limit safety constraints on cross-time section power flow transmission The cross-time section power flow transmission limit safety constraint includes: constraining the power flow increase or decrease mapped to the transmission section after the benchmark power flow is superimposed on the full utilization of the ramp-up reserve capacity of the thermal power unit and the energy storage module, so as not to exceed the forward transmission limit and reverse transmission limit of the transmission section within any continuous reserve call duration, so as to generate a time-series polyhedron reserve transportable safety domain envelope.

[0088] set up , These represent the maximum permissible transmission power limits for the forward and reverse directions of the transmission section f, respectively. Reserved duration for continuous invocation at time t Upward and downward ramp reserve capacity.

[0089] To ensure any duration Throughout the entire process of full and continuous ramp-up reserve power supply, the transmission section does not experience power flow exceeding limits or dynamic congestion. After superimposing the benchmark power flow with the ramp-up reserve power flow increments of all thermal power units and energy storage devices, the power flow always falls within the safe transmission range of the section. (47) in, This set of inequalities achieves mathematical locking between the time-series ramp trajectory and the grid cross-sectional safety boundary: Because the power transfer factor of the grid topology is directional, any increase or decrease in node output can cause the cross-sectional power flow to shift in the forward or reverse direction. Therefore, whether the ramp reserve is fully activated for both upward and downward ramps, the cross-sectional power flow distribution after superimposing the baseline power flow must be synchronously verified to ensure it does not exceed the forward transmission limit of the cross-section and does not fall below the reverse transmission limit; this must be done synchronously across all time intervals. It simulates the cumulative offset effect of power flow caused by long-term continuous ramping, avoiding the shortcomings of traditional single instantaneous cross-section verification that cannot cover continuous call congestion delay.

[0090] (3) Temporal polyhedron standby transportable secure domain envelope By combining the aforementioned unit output boundaries, continuous ramp rate, energy storage constraints, and new power flow topology coupling constraints, a joint feasible safety domain envelope is defined, which includes unit time-series output, multi-duration ramp reserve, and grid power flow safety boundaries. Satisfying equations (30)-(47): (48) Only scheduling schemes that fall within the envelope of the time-series polyhedron can simultaneously meet the three conditions of unit physical adjustability, continuous energy support from energy storage, and full transmission through the grid channel. This eliminates invalid pseudo-backup that cannot be transmitted through the grid, and achieves full-cycle, multi-dimensional safety verification of ramp-up backup across multiple time scales.

[0091] In a preferred but non-limiting embodiment of the present invention, the flexible ramp-up optimization scheduling model is input into a mixed-integer optimization solver for solving. The optimization objective is to minimize the objective function. Within the feasible region that satisfies the requirements of generating units, energy storage, reserve capacity, and transmission section safety constraints, the start-up and shutdown status of generating units, unit output, ramp-up and ramp-up reserve capacity, energy storage charging and discharging power, and energy storage state of charge are determined for each time period, outputting the optimal scheduling scheme. This optimal scheduling scheme is then converted into physical execution commands, issuing start-up and shutdown control and active power setpoints for each time period to the distributed control system of the corresponding thermal power unit, and issuing charging and discharging control signals to the energy storage management system to drive the physical equipment to perform flexible ramp-up adjustments, ensuring the physical operational safety of the high-proportion renewable energy power grid.

[0092] Example 2: This invention provides a backup demand quantification and optimized scheduling system that considers the spatiotemporal correlation characteristics of source and load, comprising: The probability scenario generation module is used to acquire historical actual power output data, historical predicted power output data, and planned day predicted power output data of wind, solar and load. It determines the set of physical mutation critical power based on the physical characteristics of wind and solar power output, and generates physical mutation weights based on the set of physical mutation critical power. Combining the historical actual power output data and the historical predicted power output data of wind, solar and load, it adaptively partitions the continuous predicted power space, and uses the physical mutation weights to correct the adaptive partitioning process of the predicted power space, thereby constructing a conditional probability density model that integrates the planned day predicted power output data. It also establishes a joint probability distribution model based on the historical actual power output data of wind, solar and load, and combines the conditional probability density model with the joint probability distribution model to generate a multi-time period joint power output scenario set. The backup demand quantification module is used to calculate the net load ramp distribution at different time scales based on the multi-time period joint output scenario set, and to perform tail sample sufficiency detection. If the sample size in a single extreme ramp interval is lower than the preset sufficiency threshold, the scenario that triggers the extreme ramp will be reverse-mapped to a high-risk net output power interval, and the high-risk net output power interval will be fed back to the predictive power space adaptive partitioning process for directional expansion until the sample sufficiency requirement is met, and the backup demand for ramping up and down will be output. The optimization scheduling solution module is used to combine the uphill and downhill backup requirements to construct a flexible ramp optimization scheduling model that takes into account the physical constraints of the generating units and energy storage, as well as the safety constraints of cross-time section power flow transmission limits, and solve the flexible ramp optimization scheduling model to obtain the final scheduling scheme.

[0093] In one possible implementation, the probability scenario generation module is specifically used to construct a conditional probability density model that integrates the predicted data for the planned days, and to establish a joint probability distribution model based on the historical actual output data of wind, solar, and load. The conditional probability density model is then combined with the joint probability distribution model to generate a multi-time period joint output scenario set, specifically including: By introducing the Markov chain hypothesis, the temporal correlation between adjacent time periods in the historical actual power output data of wind and solar load is extracted, and the temporal joint distribution of power output in multiple time periods is modeled and reduced to a two-dimensional dependency relationship between adjacent time periods. Using the predicted data for the planned day as a priori conditions, combined with the two-dimensional dependency relationship, and based on the adaptive partitioning process of the predicted power space after the physical mutation weight correction, kernel density estimation is performed to obtain the conditional probability density of the power output at the current moment, thus forming the conditional probability density model. A multi-source synchronous static correlation model of wind and solar loads is constructed using a D-Vine Copula structure. During the calculation process, the actual output of each source load in the historical actual output data of wind and solar loads is transformed into a uniform distribution variable by fitting their respective marginal cumulative distribution functions. The uniform distribution variable is then mapped and transformed by the t-distribution inverse cumulative distribution function, and the mapping result is substituted into the t-Copula density function corresponding to the D-Vine Copula structure for joint calculation to obtain the joint probability distribution model. The multi-time period joint output scenario set is generated by combining the conditional probability density model and the joint probability distribution model.

[0094] In one possible implementation, the probabilistic scenario generation module is specifically used to determine a set of physical abrupt change critical power values ​​based on the physical characteristics of wind and solar power output, generate physical abrupt change weights based on the set of physical abrupt change critical power values, and use the physical abrupt change weights to correct the adaptive partitioning process of the predicted power space, specifically including: The set of critical power for physical abrupt change is determined by the wind turbine's power generation capacity when it is switched in, the wind turbine's rated full power, the wind turbine's power output when it is switched out and the photovoltaic power. A matching tolerance threshold is set. When recursively dividing the continuous predicted power space, if the absolute value of the difference between the candidate segmentation point and any element in the physical mutation critical power set is less than or equal to the matching tolerance threshold, the physical mutation weight takes a preset amplification factor; otherwise, the physical mutation weight takes a value of 1; the amplification factor is greater than 1. Calculate the decrease in integral mean square error before and after the predicted power space partitioning, and multiply the decrease in integral mean square error by the physical mutation weight to obtain the weighted error decrease. The candidate segmentation point that maximizes the reduction in the weighted error is selected as the optimal segmentation point, and the continuous predicted power space is divided into two sub-intervals based on the optimal segmentation point.

[0095] In one possible implementation, the backup demand quantification module, specifically used to perform tail sample sufficiency detection, includes: Calculate the net load ramp-up of the multi-time period joint output scenario set at multiple time intervals, and statistically analyze the ramp-up quantiles and ramp-down quantiles under a preset confidence level; The expected number of theoretical one-sided tail samples is determined based on the preset confidence level and the total number of scenarios in the multi-time period joint output scenario set. The expected number of theoretical one-sided tail samples is then multiplied by a preset sufficiency verification coefficient to obtain the sufficiency threshold. The number of extreme uphill samples that are greater than or equal to the uphill quantile and the number of extreme downhill samples that are less than or equal to the downhill quantile are counted respectively.

[0096] In one possible implementation, the backup demand quantification module is specifically used to, if the sample size within a single-sided extreme ramp interval is lower than a preset sufficiency threshold, reverse-map the scenario that triggers the extreme ramp to a high-risk net output power interval, and feed the high-risk net output power interval back to the predicted power space adaptive partitioning process for targeted expansion, specifically including: If the number of extreme uphill samples is less than the adequacy threshold, then the minimum net load at the corresponding moment that caused the extreme uphill phenomenon in the generated scenario, together with the historical maximum net load at the current moment, constitutes the high-risk net output power range for uphill. If the number of extreme downhill samples is less than the adequacy threshold, then the maximum net load value at the corresponding moment that caused the extreme downhill phenomenon in the generated scenario, together with the historical minimum net load value at the current moment, constitutes the downhill high-risk net output power range. The ratio of the adequacy threshold to the actual number of extreme climbing samples is calculated as the local oversampling coefficient, and the corresponding high-risk net output power interval is fed back to the predictive power space adaptive partitioning process. The local oversampling coefficient is used to increase the sampling weight within the interval.

[0097] In one possible implementation, the optimization scheduling solution module is specifically used to construct a flexible ramp optimization scheduling model that takes into account the physical constraints of generating units and energy storage, as well as the cross-time section power flow transmission limit safety constraints. Specifically, it includes: The objective function is to minimize the sum of the no-load cost, start-up cost, shutdown cost, operating fuel cost of thermal power units, and the backup cost of thermal power units and energy storage modules providing flexible ramp-up backup. The physical constraints of the generating units and energy storage include: the output and ramping constraints of the thermal power units, the charging and discharging power constraints of the energy storage, the state of charge constraints of the energy storage, the energy storage power balance constraints, and the energy storage ramping constraints determined by the state of charge; when balancing the flexible ramping reserve requirements of the entire system, the ramping reserve capacity reserved by the thermal power units and the ramping reserve capacity provided by the energy storage modules are taken into account. Establish the static topological mapping relationship between the unit's basic power generation output, energy storage charging and discharging power, and power flow between source and load nodes and transmission sections, and calculate the baseline power flow. The cross-time section power flow transmission limit safety constraint includes: constraining the power flow increase or decrease mapped to the transmission section after the benchmark power flow is superimposed on the full utilization of the ramp-up reserve capacity of the thermal power unit and the energy storage module, so as not to exceed the forward transmission limit and reverse transmission limit of the transmission section within any continuous reserve call duration, so as to generate a time-series polyhedron reserve transportable safety domain envelope.

[0098] In one possible implementation, the power output and ramping constraints of the thermal power unit specifically include cross-time period power output boundary constraints and continuous ramping rate constraints that take into account the differences in start-up and shutdown states. The cross-time period output boundary constraint is used to: synchronously verify the upper and lower limits of the output of the thermal power unit at the current time and the time when the standby call ends, and constrain the basic power generation output of the thermal power unit at the current time and the time when the standby call ends, so that after superimposing the reserved upward ramp standby capacity or deducting the reserved downward ramp standby capacity, it does not exceed the maximum and minimum output limits of the thermal power unit. The continuous ramp rate constraint that takes into account the difference between start-up and shutdown states is used to: combine the ramp rate capability per unit time of the thermal power unit in the operating state, start-up state and shutdown state, respectively calculate the maximum allowable upward ramp rate and the maximum allowable downward ramp rate within the standby continuous call duration, so as to limit the upward ramp standby capacity and the downward ramp standby capacity.

[0099] Example 3: An electronic device includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the processor executes the program, it implements the steps of a method for quantifying and optimizing the scheduling of standby demand that takes into account the spatiotemporal correlation characteristics of source and load.

[0100] Example 4: A computer-readable storage medium storing computer program instructions, which, when read and executed by a processor, implement the steps of a method for quantifying and optimizing the scheduling of standby demand considering the spatiotemporal correlation characteristics of source and load.

[0101] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0102] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0103] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0104] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for quantifying and optimizing the scheduling of reserve demand considering the spatiotemporal correlation characteristics of source and load, characterized in that, include: Acquire historical actual power output data, historical predicted power output data, and planned daily predicted power output data of wind, solar, and load. Determine the set of physical mutation critical power based on the physical characteristics of wind and solar power output, and generate physical mutation weights based on the set of physical mutation critical power. Combine the historical actual power output data of wind, solar, and load with the historical predicted power output data to adaptively partition the continuous predicted power space, and use the physical mutation weights to correct the adaptive partitioning process of the predicted power space to construct a conditional probability density model that integrates the planned daily predicted power output data. Establish a joint probability distribution model based on the historical actual power output data of wind, solar, and load. Combine the conditional probability density model with the joint probability distribution model to generate a multi-time period joint power output scenario set. Based on the multi-time period joint output scenario set, the net load ramp distribution at different time scales is calculated, and the tail sample sufficiency detection is performed. If the sample size in the extreme ramp interval on one side is lower than the preset sufficiency threshold, the scenario that causes the extreme ramp is reverse-mapped to a high-risk net output power interval, and the high-risk net output power interval is fed back to the predictive power space adaptive partitioning process for directional expansion until the sample sufficiency requirement is met, and the up and down ramp backup requirements are output. Based on the aforementioned ramp-up and ramp-up backup requirements, a flexible ramp-up optimization scheduling model is constructed that takes into account the physical constraints of the generating units and energy storage, as well as the safety constraints of cross-time section power flow transmission limits. The final scheduling scheme is obtained by solving the flexible ramp-up optimization scheduling model.

2. The method for quantifying and optimizing the scheduling of reserve demand considering the spatiotemporal correlation characteristics of source and load as described in claim 1, characterized in that, A conditional probability density model integrating the predicted data for the planned days is constructed, and a joint probability distribution model is established based on the historical actual output data of wind and solar loads. The conditional probability density model and the joint probability distribution model are combined to generate a multi-time period joint output scenario set, specifically including: By introducing the Markov chain hypothesis, the temporal correlation between adjacent time periods in the historical actual power output data of wind and solar load is extracted, and the temporal joint distribution of power output in multiple time periods is modeled and reduced to a two-dimensional dependency relationship between adjacent time periods. Using the predicted data for the planned day as a priori conditions, combined with the two-dimensional dependency relationship, and based on the adaptive partitioning process of the predicted power space after the physical mutation weight correction, kernel density estimation is performed to obtain the conditional probability density of the power output at the current moment, thus forming the conditional probability density model. A multi-source synchronous static correlation model of wind and solar loads is constructed using a D-Vine Copula structure. During the calculation process, the actual output of each source load in the historical actual output data of wind and solar loads is transformed into a uniform distribution variable by fitting their respective marginal cumulative distribution functions. The uniform distribution variable is then mapped and transformed by the t-distribution inverse cumulative distribution function, and the mapping result is substituted into the t-Copula density function corresponding to the D-Vine Copula structure for joint calculation to obtain the joint probability distribution model. The multi-time period joint output scenario set is generated by combining the conditional probability density model and the joint probability distribution model.

3. The method for quantifying and optimizing the scheduling of reserve demand considering the spatiotemporal correlation characteristics of source and load as described in claim 2, characterized in that, The process involves determining a set of physical abrupt change critical power values ​​based on the physical characteristics of wind and solar power output, generating physical abrupt change weights based on these weights, and using these weights to correct the adaptive partitioning process of the predicted power space. Specifically, this includes: The set of critical power for physical abrupt change is determined by the wind turbine's power generation capacity when it is switched in, the wind turbine's rated full power, the wind turbine's power output when it is switched out and the photovoltaic power. A matching tolerance threshold is set. When recursively dividing the continuous predicted power space, if the absolute value of the difference between the candidate segmentation point and any element in the physical mutation critical power set is less than or equal to the matching tolerance threshold, the physical mutation weight takes a preset amplification factor; otherwise, the physical mutation weight takes a value of 1; the amplification factor is greater than 1. Calculate the decrease in integral mean square error before and after the predicted power space partitioning, and multiply the decrease in integral mean square error by the physical mutation weight to obtain the weighted error decrease. The candidate segmentation point that maximizes the reduction in the weighted error is selected as the optimal segmentation point, and the continuous predicted power space is divided into two sub-intervals based on the optimal segmentation point.

4. The method for quantifying and optimizing the scheduling of reserve demand considering the spatiotemporal correlation characteristics of source and load as described in claim 1, characterized in that, Perform tail sample sufficiency testing, specifically including: Calculate the net load ramp-up of the multi-time period joint output scenario set at multiple time intervals, and statistically analyze the ramp-up quantiles and ramp-down quantiles under a preset confidence level; The expected number of theoretical one-sided tail samples is determined based on the preset confidence level and the total number of scenarios in the multi-time period joint output scenario set. The expected number of theoretical one-sided tail samples is then multiplied by a preset sufficiency verification coefficient to obtain the sufficiency threshold. The number of extreme uphill samples that are greater than or equal to the uphill quantile and the number of extreme downhill samples that are less than or equal to the downhill quantile are counted respectively.

5. The method for quantifying and optimizing the scheduling of reserve demand considering the spatiotemporal correlation characteristics of source and load according to claim 4, characterized in that, If the sample size within a single extreme climbing interval is lower than a preset sufficiency threshold, the scenario triggering the extreme climbing will be reverse-mapped into a high-risk net power output interval, and the high-risk net power output interval will be fed back into the adaptive partitioning process of the predicted power space for targeted expansion, specifically including: If the number of extreme uphill samples is less than the adequacy threshold, then the minimum net load at the corresponding moment that caused the extreme uphill phenomenon in the generated scenario, together with the historical maximum net load at the current moment, constitutes the high-risk net output power range for uphill. If the number of extreme downhill samples is less than the adequacy threshold, then the maximum net load value at the corresponding moment that caused the extreme downhill phenomenon in the generated scenario, together with the historical minimum net load value at the current moment, constitutes the downhill high-risk net output power range. The ratio of the adequacy threshold to the actual number of extreme climbing samples is calculated as the local oversampling coefficient, and the corresponding high-risk net output power interval is fed back to the predictive power space adaptive partitioning process. The local oversampling coefficient is used to increase the sampling weight within the interval.

6. The method for quantifying and optimizing the scheduling of reserve demand considering the spatiotemporal correlation characteristics of source and load according to claim 1, characterized in that, A flexible ramp-up optimization scheduling model is constructed, taking into account both the physical constraints of generating units and energy storage, as well as the ultimate safety constraints of cross-time section power flow transmission. Specifically, this includes: The objective function is to minimize the sum of the no-load cost, start-up cost, shutdown cost, operating fuel cost of thermal power units, and the backup cost of thermal power units and energy storage modules providing flexible ramp-up backup. The physical constraints of the generating units and energy storage include: the output and ramping constraints of the thermal power units, the charging and discharging power constraints of the energy storage, the state of charge constraints of the energy storage, the energy storage power balance constraints, and the energy storage ramping constraints determined by the state of charge; when balancing the flexible ramping reserve requirements of the entire system, the ramping reserve capacity reserved by the thermal power units and the ramping reserve capacity provided by the energy storage modules are taken into account. Establish the static topological mapping relationship between the unit's basic power generation output, energy storage charging and discharging power, and power flow between source and load nodes and transmission sections, and calculate the baseline power flow. The cross-time section power flow transmission limit safety constraint includes: constraining the power flow increase or decrease mapped to the transmission section after the benchmark power flow is superimposed on the full utilization of the ramp-up reserve capacity of the thermal power unit and the energy storage module, so as not to exceed the forward transmission limit and reverse transmission limit of the transmission section within any continuous reserve call duration, so as to generate a time-series polyhedron reserve transportable safety domain envelope.

7. A method for quantifying and optimizing the scheduling of reserve demand considering the spatiotemporal correlation characteristics of source and load, as described in claim 6, is characterized in that... The power output and ramping constraints of the thermal power units specifically include cross-time period power output boundary constraints and continuous ramping rate constraints that take into account the differences in start-up and shutdown states. The cross-time period output boundary constraint is used to: synchronously verify the upper and lower limits of the output of the thermal power unit at the current time and the time when the standby call ends, and constrain the basic power generation output of the thermal power unit at the current time and the time when the standby call ends, so that after superimposing the reserved upward ramp standby capacity or deducting the reserved downward ramp standby capacity, it does not exceed the maximum and minimum output limits of the thermal power unit. The continuous ramp rate constraint that takes into account the difference between start-up and shutdown states is used to: combine the ramp rate capability per unit time of the thermal power unit in the operating state, start-up state and shutdown state, respectively calculate the maximum allowable upward ramp rate and the maximum allowable downward ramp rate within the standby continuous call duration, so as to limit the upward ramp standby capacity and the downward ramp standby capacity.

8. A backup demand quantification and optimization scheduling system considering the spatiotemporal correlation characteristics of source and load, characterized in that, include: The probability scenario generation module is used to acquire historical actual power output data, historical predicted power output data, and planned day predicted power output data of wind, solar and load. It determines the set of physical mutation critical power based on the physical characteristics of wind and solar power output, and generates physical mutation weights based on the set of physical mutation critical power. Combining the historical actual power output data and the historical predicted power output data of wind, solar and load, it adaptively partitions the continuous predicted power space, and uses the physical mutation weights to correct the adaptive partitioning process of the predicted power space, thereby constructing a conditional probability density model that integrates the planned day predicted power output data. It also establishes a joint probability distribution model based on the historical actual power output data of wind, solar and load, and combines the conditional probability density model with the joint probability distribution model to generate a multi-time period joint power output scenario set. The backup demand quantification module is used to calculate the net load ramp distribution at different time scales based on the multi-time period joint output scenario set, and to perform tail sample sufficiency detection. If the sample size in a single extreme ramp interval is lower than the preset sufficiency threshold, the scenario that triggers the extreme ramp will be reverse-mapped to a high-risk net output power interval, and the high-risk net output power interval will be fed back to the predictive power space adaptive partitioning process for directional expansion until the sample sufficiency requirement is met, and the backup demand for ramping up and down will be output. The optimization scheduling solution module is used to combine the uphill and downhill backup requirements to construct a flexible ramp optimization scheduling model that takes into account the physical constraints of the generating units and energy storage, as well as the safety constraints of cross-time section power flow transmission limits, and solve the flexible ramp optimization scheduling model to obtain the final scheduling scheme.

9. A backup demand quantification and optimized scheduling system considering the spatiotemporal correlation characteristics of source and load according to claim 8, characterized in that, A conditional probability density model integrating the predicted data for the planned days is constructed, and a joint probability distribution model is established based on the historical actual output data of wind and solar loads. The conditional probability density model and the joint probability distribution model are combined to generate a multi-time period joint output scenario set, specifically including: By introducing the Markov chain hypothesis, the temporal correlation between adjacent time periods in the historical actual power output data of wind and solar load is extracted, and the temporal joint distribution of power output in multiple time periods is modeled and reduced to a two-dimensional dependency relationship between adjacent time periods. Using the predicted data for the planned day as a priori conditions, combined with the two-dimensional dependency relationship, and based on the adaptive partitioning process of the predicted power space after the physical mutation weight correction, kernel density estimation is performed to obtain the conditional probability density of the power output at the current moment, thus forming the conditional probability density model. A multi-source synchronous static correlation model of wind and solar loads is constructed using a D-Vine Copula structure. During the calculation process, the actual output of each source load in the historical actual output data of wind and solar loads is transformed into a uniform distribution variable by fitting their respective marginal cumulative distribution functions. The uniform distribution variable is then mapped and transformed by the t-distribution inverse cumulative distribution function, and the mapping result is substituted into the t-Copula density function corresponding to the D-Vine Copula structure for joint calculation to obtain the joint probability distribution model. The multi-time period joint output scenario set is generated by combining the conditional probability density model and the joint probability distribution model.

10. A backup demand quantification and optimized scheduling system considering the spatiotemporal correlation characteristics of source and load according to claim 8, characterized in that, The process involves determining a set of physical abrupt change critical power values ​​based on the physical characteristics of wind and solar power output, generating physical abrupt change weights based on these weights, and using these weights to correct the adaptive partitioning process of the predicted power space. Specifically, this includes: The set of critical power for physical abrupt change is determined by the wind turbine's power generation capacity when it is switched in, the wind turbine's rated full power, the wind turbine's power output when it is switched out and the photovoltaic power. A matching tolerance threshold is set. When recursively dividing the continuous predicted power space, if the absolute value of the difference between the candidate segmentation point and any element in the physical mutation critical power set is less than or equal to the matching tolerance threshold, the physical mutation weight takes a preset amplification factor; otherwise, the physical mutation weight takes a value of 1; the amplification factor is greater than 1. Calculate the decrease in integral mean square error before and after the predicted power space partitioning, and multiply the decrease in integral mean square error by the physical mutation weight to obtain the weighted error decrease. The candidate segmentation point that maximizes the reduction in the weighted error is selected as the optimal segmentation point, and the continuous predicted power space is divided into two sub-intervals based on the optimal segmentation point.

11. A backup demand quantification and optimized scheduling system considering the spatiotemporal correlation characteristics of source and load as described in claim 8, characterized in that, Perform tail sample sufficiency testing, specifically including: Calculate the net load ramp-up of the multi-time period joint output scenario set at multiple time intervals, and statistically analyze the ramp-up quantiles and ramp-down quantiles under a preset confidence level; The expected number of theoretical one-sided tail samples is determined based on the preset confidence level and the total number of scenarios in the multi-time period joint output scenario set. The expected number of theoretical one-sided tail samples is then multiplied by a preset sufficiency verification coefficient to obtain the sufficiency threshold. The number of extreme uphill samples that are greater than or equal to the uphill quantile and the number of extreme downhill samples that are less than or equal to the downhill quantile are counted respectively.

12. A backup demand quantification and optimized scheduling system considering the spatiotemporal correlation characteristics of source and load as described in claim 11, characterized in that, If the sample size within a single extreme climbing interval is lower than a preset sufficiency threshold, the scenario triggering the extreme climbing will be reverse-mapped into a high-risk net power output interval, and the high-risk net power output interval will be fed back into the adaptive partitioning process of the predicted power space for targeted expansion, specifically including: If the number of extreme uphill samples is less than the adequacy threshold, then the minimum net load at the corresponding moment that caused the extreme uphill phenomenon in the generated scenario, together with the historical maximum net load at the current moment, constitutes the high-risk net output power range for uphill. If the number of extreme downhill samples is less than the adequacy threshold, then the maximum net load value at the corresponding moment that caused the extreme downhill phenomenon in the generated scenario, together with the historical minimum net load value at the current moment, constitutes the downhill high-risk net output power range. The ratio of the adequacy threshold to the actual number of extreme climbing samples is calculated as the local oversampling coefficient, and the corresponding high-risk net output power interval is fed back to the predictive power space adaptive partitioning process. The local oversampling coefficient is used to increase the sampling weight within the interval.

13. A backup demand quantification and optimized scheduling system considering the spatiotemporal correlation characteristics of source and load as described in claim 8, characterized in that, A flexible ramp-up optimization scheduling model is constructed, taking into account both the physical constraints of generating units and energy storage, as well as the ultimate safety constraints of cross-time section power flow transmission. Specifically, this includes: The objective function is to minimize the sum of the no-load cost, start-up cost, shutdown cost, operating fuel cost of thermal power units, and the backup cost of thermal power units and energy storage modules providing flexible ramp-up backup. The physical constraints of the generating units and energy storage include: the output and ramping constraints of the thermal power units, the charging and discharging power constraints of the energy storage, the state of charge constraints of the energy storage, the energy storage power balance constraints, and the energy storage ramping constraints determined by the state of charge; when balancing the flexible ramping reserve requirements of the entire system, the ramping reserve capacity reserved by the thermal power units and the ramping reserve capacity provided by the energy storage modules are taken into account. Establish the static topological mapping relationship between the unit's basic power generation output, energy storage charging and discharging power, and power flow between source and load nodes and transmission sections, and calculate the baseline power flow. The cross-time section power flow transmission limit safety constraint includes: constraining the power flow increase or decrease mapped to the transmission section after the benchmark power flow is superimposed on the full utilization of the ramp-up reserve capacity of the thermal power unit and the energy storage module, so as not to exceed the forward transmission limit and reverse transmission limit of the transmission section within any continuous reserve call duration, so as to generate a time-series polyhedron reserve transportable safety domain envelope.

14. A backup demand quantification and optimized scheduling system considering the spatiotemporal correlation characteristics of source and load as described in claim 13, characterized in that, The power output and ramping constraints of the thermal power units specifically include cross-time period power output boundary constraints and continuous ramping rate constraints that take into account the differences in start-up and shutdown states. The cross-time period output boundary constraint is used to: synchronously verify the upper and lower limits of the output of the thermal power unit at the current time and the time when the standby call ends, and constrain the basic power generation output of the thermal power unit at the current time and the time when the standby call ends, so that after superimposing the reserved upward ramp standby capacity or deducting the reserved downward ramp standby capacity, it does not exceed the maximum and minimum output limits of the thermal power unit. The continuous ramp rate constraint that takes into account the difference between start-up and shutdown states is used to: combine the ramp rate capability per unit time of the thermal power unit in the operating state, start-up state and shutdown state, respectively calculate the maximum allowable upward ramp rate and the maximum allowable downward ramp rate within the standby continuous call duration, so as to limit the upward ramp standby capacity and the downward ramp standby capacity.

15. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-7.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-7.

Images