A photovoltaic-load joint scene generation method and system based on source-load correlation

Abstract

A photovoltaic-load combined scene generation method and system based on source-load correlation, through the scene generation of the illumination intensity probability distribution model, the node load probability distribution model and the improved Latin hypercube sampling method, and the timing verification, the timing verified photovoltaic independent scene set and the load independent scene set are obtained, the scene sets are combined two by two, the photovoltaic-load scene combination pairs are generated, the scene reduction is carried out through the source-load correlation coefficient deviation, the similar scenes are aggregated and the appearance probability of each typical scene is calculated, and finally the quantity-precise and strong-representative source-load typical combined scenes are efficiently screened out. The design can not only realize the deep fusion of correlation analysis and scene reduction, solve the technical problems of scene redundancy and insufficient representation of traditional reduction methods, effectively improve the scene representation, but also can ensure the statistical accuracy of sampling, have timing rationality and engineering practicality, and effectively improve the accuracy of scene generation.

Description

Technical Field

[0001] This invention relates to a distribution network generation method, and more particularly to a photovoltaic-load joint scenario generation method and system based on source-load correlation. Background Technology

[0002] With the rapid development of distributed photovoltaic (PV) power, its penetration rate in distribution networks continues to increase. However, PV power output is affected by natural factors such as sunlight intensity and ambient temperature, exhibiting significant randomness and volatility. At the same time, the load of the distribution network is constrained by factors such as user electricity consumption behavior and industrial production patterns, resulting in irregular and uncertain fluctuations. These uncertainties on both the power source and load sides pose significant challenges to the safe and stable operation, economic dispatch, and planning and design of the distribution network.

[0003] Traditional scene generation methods mostly deal with the uncertainty of photovoltaic output and load separately, ignoring the potential spatiotemporal correlation between the two (such as the coupling of peak photovoltaic output during the day with peak industrial load, and the matching of peak residential load with low photovoltaic output, etc.), resulting in the generated scene being out of touch with the actual operating state of the distribution network.

[0004] Even if some scene generation methods take into account correlation, they often use single-dimensional correlation analysis, and the scene reduction process lacks targeted optimization, resulting in problems such as insufficient scene representativeness, high computational complexity, and poor adaptability to actual engineering needs.

[0005] While this method of generating distribution networks is feasible, it still has the following drawbacks:

[0006] 1. Existing methods often use fixed thresholds or simple clustering algorithms when reducing scenarios, which makes it difficult to effectively identify and aggregate source-load joint scenarios with high similarity. At the same time, the reduction process is not deeply integrated with correlation analysis, resulting in insufficient representativeness and redundancy of the reduced scenario set.

[0007] 2. Most existing methods deal with the uncertainty of photovoltaic output and load separately, or analyze the correlation between the two through a single dimension, ignoring the two-dimensional correlation between photovoltaic and load in time and space. This results in the generated scenario being out of touch with the actual operating state of the distribution network, failing to accurately reflect the true law of source-load interaction, and having poor accuracy.

[0008] The information disclosed in this background section is intended only to enhance understanding of the overall background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of insufficient scene representativeness and poor accuracy in the existing technology, and to provide a photovoltaic-load joint scene generation method based on source-load correlation that has strong scene representativeness and high accuracy.

[0010] To achieve the above objectives, the technical solution of the present invention is:

[0011] A method for generating photovoltaic-load joint scenarios based on source-load correlation, wherein the grid optimization design method includes the following steps:

[0012] S1. Based on the differentiated fluctuation characteristics of light intensity and distribution network node load, construct a probability distribution model of light intensity and a probability distribution model of node load, respectively.

[0013] S2. Collect historical data of the area to be optimized. Based on the improved Latin hypercube sampling method with weighted key time periods and time series constraints, combine the light intensity probability distribution model and the node load probability distribution model to generate scenarios. Convert the light intensity independent scenario set into a photovoltaic independent scenario set and perform time series verification to obtain the time series verified photovoltaic independent scenario set and load independent scenario set.

[0014] S3. Combine the photovoltaic independent scenario set and the load independent scenario set that meet the time-series constraints in pairs to generate photovoltaic-load scenario combination pairs, and calculate the source-load correlation coefficient deviation. Based on the source-load correlation coefficient deviation The photovoltaic-load scenario combination is reduced to obtain a set of photovoltaic-load joint scenarios with occurrence probabilities.

[0015] In S1, the light intensity probability distribution model includes:

[0016] ;

[0017] In the above formula, The probability density number of light intensity; This represents the actual light intensity. The maximum light intensity during the study period; It is a gamma function; , All are shape parameters of the Beta distribution, including:

[0018] The specific calculation methods for the shape parameters include:

[0019] ;

[0020] In the above formula, Mathematical expectation of historical illumination monitoring data , The variance of historical illumination monitoring data;

[0021] The node load probability distribution model includes:

[0022] ;

[0023] In the above formula, Let be the probability density number of the node load. The average of historical load data. The standard deviation of historical loads can be obtained statistically from historical load data.

[0024] In step S2, the scene generation includes:

[0025] The historical data includes historical illumination data and historical load data. The historical data is aggregated into n time-series dimensions according to the time period of each day. Each time-series dimension includes the data of the corresponding time period of each natural day in the historical data.

[0026] Set the number of scenes, k;

[0027] By fitting the distribution parameters corresponding to each time period t (t=1,2,...,n) using historical illumination data and historical load data, the illumination shape parameters corresponding to each time period are obtained. , and the average of historical load data Standard deviation of historical load ;

[0028] The probability distribution intervals of light intensity and load are uniformly divided into m equally probable sub-intervals, based on sampling weights. Increase the number of sampling points in the sampling interval for critical periods, and maintain equal probability division for non-critical periods to obtain a weighted sampling interval;

[0029] The sampling weight include:

[0030] ;

[0031] ;

[0032] ;

[0033] In the above formula, Let be the sampling weight for time period t; This represents the time-period impact of the normalized time period t. Let t be the time-period influence of the time period. The weighted adjustment coefficient is the weighted adjustment coefficient. Calculated based on the fluctuation characteristics of source load data in the study area; The historical standard deviation of the source / load data for time period t reflects the intensity of force / load fluctuations. The measured values ​​of source / load on day i in time period t; For historical sample days;

[0034] Based on the light intensity distribution model, the load probability distribution model, and the weighted sampling interval, an improved Latin hypercube sampling method is used to generate light intensity independent scene sets and load independent scene sets, respectively.

[0035] The photovoltaic power conversion formula transforms the independent scene set of light intensity into an independent photovoltaic scene set, including:

[0036] The probability density number of light intensity The actual light intensity during time period t is obtained by sampling. After determining the light intensity, the actual photovoltaic output can be calculated by combining the photovoltaic power generation efficiency and the rated power of the photovoltaic array. The specific calculation formula for photovoltaic output includes:

[0037] ;

[0038] In the above formula, The photovoltaic output for time period t; The actual light intensity during time period t. For photovoltaic power generation efficiency; This refers to the rated power of the photovoltaic array;

[0039] Both the photovoltaic independent scenario set and the load independent scenario set have n×k dimensions;

[0040] The initial photovoltaic independent scenario set and load independent scenario set are subjected to time series continuity verification. The change rate of sampled values ​​in adjacent time periods under all scenarios is calculated. If it exceeds the historical maximum time series change rate, the scenario is deleted and a new scenario is generated by resampling, and the verification is performed again. If it does not exceed the historical maximum time series change rate, the scenario is retained, and the photovoltaic independent scenario set and load independent scenario set after time series verification are obtained.

[0041] In S3, the scene reduction includes:

[0042] Calculate the source-load correlation coefficient deviation for each photovoltaic-load scenario combination pair. Eliminating source-load correlation coefficient bias The optimal set of photovoltaic-load related scenarios is obtained by identifying photovoltaic-load scenario combinations that exceed a preset threshold.

[0043] The source-load correlation coefficient deviation include:

[0044] ;

[0045] In the above formula, This is the sample correlation coefficient matrix. The target comprehensive coefficient matrix;

[0046] The sample correlation coefficient matrix By packaging various photovoltaic-load scenario combinations, the comprehensive correlation coefficients between medium load and corresponding photovoltaic time periods are obtained. The sample correlation coefficient matrix is ​​obtained. The dimension is n×n;

[0047] The target comprehensive coefficient matrix By combining the correlation coefficients of different time periods in historical illumination data and historical load data. The target comprehensive coefficient matrix is ​​obtained. The dimension is n×n.

[0048] The comprehensive correlation coefficient include:

[0049] ;

[0050] In the above formula, Let be the comprehensive correlation coefficient for time period t; Let be the temporal correlation coefficient for time period t. Let be the spatial correlation coefficient for time period t. The time-series correlation weighting coefficient. Let be the spatial correlation weight coefficient, and satisfy . ;

[0051] The temporal correlation coefficient of the t-th time period include:

[0052] ;

[0053] In the above formula, For the photovoltaic power output in time period t, Let be the load power in time period t; The time-series average of photovoltaic power output. This represents the time-series average of the load power. The weighting factor for peak load periods is reinforced by an exponential function to strengthen the correlation weighting of peak load periods. The time decay coefficient indicates that the temporal correlation between photovoltaic power output and load power exhibits an objective law of gradual decay over time. The peak load baseline period, The seasonal correction coefficient was obtained by quantitative fitting of the historical operating data of the source and load in the four seasons of the study area over the past 3-5 years.

[0054] The spatial correlation coefficient at time t include:

[0055] ;

[0056] In the above formula, the subscript 0 represents the study area, and the superscript 1 represents the neighboring area.

[0057] A photovoltaic-load joint scenario generation system based on source-load correlation specifically includes: a model building module, a scenario generation module, and a scenario reduction module;

[0058] The model building module is used to: construct a probability distribution model of light intensity and a probability distribution model of node load, respectively, based on the differentiated fluctuation characteristics of light intensity and distribution network node load;

[0059] The scene generation module is used to: collect historical data of the area to be optimized, generate scenes based on the improved Latin hypercube sampling method with weighted key time periods and time series constraints, combined with the light intensity probability distribution model and the node load probability distribution model, convert the light intensity independent scene set into the photovoltaic independent scene set, and perform time series verification to obtain the time series verified photovoltaic independent scene set and the load independent scene set.

[0060] The scenario reduction module is used to: combine the photovoltaic independent scenario set and the load independent scenario set that meet the time constraints in pairs to generate photovoltaic-load scenario combination pairs, and calculate the source-load correlation coefficient deviation. Based on the source-load correlation coefficient deviation The photovoltaic-load scenario combination is reduced to obtain a set of photovoltaic-load joint scenarios with occurrence probabilities.

[0061] In the model building module, the light intensity probability distribution model includes:

[0062] ;

[0063] In the above formula, The probability density number of light intensity; This represents the actual light intensity. The maximum light intensity during the study period; It is a gamma function; , All are shape parameters of the Beta distribution, including:

[0064] The specific calculation methods for the shape parameters include:

[0065] ;

[0066] In the above formula, Mathematical expectation of historical illumination monitoring data , The variance of historical illumination monitoring data;

[0067] The node load probability distribution model includes:

[0068] ;

[0069] In the above formula, Let be the probability density number of the node load. The average of historical load data. The standard deviation of historical loads can be obtained statistically from historical load data.

[0070] In the model building module, the scene generation includes:

[0071] The historical data includes historical illumination data and historical load data. The historical data is aggregated into n time-series dimensions according to the time period of each day. Each time-series dimension includes the data of the corresponding time period of each natural day in the historical data.

[0072] Set the number of scenes, k;

[0073] By fitting the distribution parameters corresponding to each time period t (t=1,2,...,n) using historical illumination data and historical load data, the illumination shape parameters corresponding to each time period are obtained. , and the average of historical load data Standard deviation of historical load ;

[0074] The probability distribution intervals of light intensity and load are uniformly divided into k equally probable sub-intervals, based on sampling weights. Increase the number of sampling points in the sampling interval for critical periods, and maintain equal probability division for non-critical periods to obtain a weighted sampling interval;

[0075] The sampling weight include:

[0076] ;

[0077] ;

[0078] ;

[0079] In the above formula, Let be the sampling weight for time period t; This represents the time-period impact of the normalized time period t. Let t be the time-period influence of the time period. The weighted adjustment coefficient is the weighted adjustment coefficient. Calculated based on the fluctuation characteristics of source load data in the study area; The historical standard deviation of the source / load data for time period t reflects the intensity of force / load fluctuations. The measured values ​​of source / load on day i in time period t; For historical sample days;

[0080] Based on the light intensity distribution model, the load probability distribution model, and the weighted sampling interval, an improved Latin hypercube sampling method is used to generate light intensity independent scene sets and load independent scene sets, respectively.

[0081] The photovoltaic power conversion formula transforms the independent scene set of light intensity into an independent photovoltaic scene set, including:

[0082] The probability density number of light intensity The actual light intensity during time period t is obtained by sampling. After determining the light intensity, the actual photovoltaic output can be calculated by combining the photovoltaic power generation efficiency and the rated power of the photovoltaic array. The specific calculation formula for photovoltaic output includes:

[0083] ;

[0084] In the above formula, The photovoltaic output for time period t; The actual light intensity during time period t. For photovoltaic power generation efficiency; This refers to the rated power of the photovoltaic array;

[0085] Both the photovoltaic independent scenario set and the load independent scenario set have n×k dimensions;

[0086] The initial photovoltaic independent scenario set and load independent scenario set are subjected to time series continuity verification. The change rate of sampled values ​​in adjacent time periods under all scenarios is calculated. If it exceeds the historical maximum time series change rate, the scenario is deleted and a new scenario is generated by resampling, and the verification is performed again. If it does not exceed the historical maximum time series change rate, the scenario is retained, and the photovoltaic independent scenario set and load independent scenario set after time series verification are obtained.

[0087] In the scene reduction module, the scene reduction includes:

[0088] Calculate the source-load correlation coefficient deviation for each photovoltaic-load scenario combination pair. Eliminating source-load correlation coefficient bias The optimal set of photovoltaic-load related scenarios is obtained by identifying photovoltaic-load scenario combinations that exceed a preset threshold.

[0089] The source-load correlation coefficient deviation include:

[0090] ;

[0091] In the above formula, This is the sample correlation coefficient matrix. The target comprehensive coefficient matrix;

[0092] The sample correlation coefficient matrix By packaging various photovoltaic-load scenario combinations, the comprehensive correlation coefficients between medium load and corresponding photovoltaic time periods are obtained. The sample correlation coefficient matrix is ​​obtained. The dimension is n×n;

[0093] The target comprehensive coefficient matrix By combining the correlation coefficients of different time periods in historical illumination data and historical load data. The target comprehensive coefficient matrix is ​​obtained. The dimension is n×n.

[0094] In the scene reduction module, the comprehensive correlation coefficient include:

[0095] ;

[0096] In the above formula, Let be the comprehensive correlation coefficient for time period t; Let be the temporal correlation coefficient for time period t. Let be the spatial correlation coefficient for time period t. The time-series correlation weighting coefficient. Let be the spatial correlation weight coefficient, and satisfy . ;

[0097] The temporal correlation coefficient of the t-th time period include:

[0098] ;

[0099] In the above formula, For the photovoltaic power output in time period t, Let be the load power in time period t; The time-series average of photovoltaic power output. This represents the time-series average of the load power. The weighting factor for peak load periods is reinforced by an exponential function to strengthen the correlation weighting of peak load periods. The time decay coefficient indicates that the temporal correlation between photovoltaic power output and load power exhibits an objective law of gradual decay over time. The peak load baseline period, The seasonal correction coefficient was obtained by quantitative fitting of the historical operating data of the source and load in the four seasons of the study area over the past 3-5 years.

[0100] The spatial correlation coefficient at time t include:

[0101] ;

[0102] In the above formula, the subscript 0 represents the study area, and the superscript 1 represents the neighboring area.

[0103] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0104] 1. In the photovoltaic-load joint scene generation method based on source-load correlation of the present invention, scene generation is performed through a light intensity probability distribution model, a nodal load probability distribution model, and an improved Latin hypercube sampling method, and time-series verification is performed to obtain a time-series verified set of independent photovoltaic scenes and a set of independent load scenes. The scene sets are then combined in pairs to generate photovoltaic-load scene combination pairs, and the source-load correlation coefficient deviation is used to determine the optimal method. By reducing the number of scenarios, aggregating similar scenarios, and calculating the probability of occurrence of each typical scenario, this approach efficiently selects a streamlined number of representative source-load joint scenarios. Therefore, this design achieves a deep integration of correlation analysis and scenario reduction, solving the technical problems of scenario redundancy and insufficient representativeness in traditional reduction methods, and effectively improving scenario representativeness.

[0105] 2. This invention, a photovoltaic-load joint scenario generation method based on source-load correlation, innovatively improves the traditional Latin hypercube sampling method by employing both key-period weighting and temporal constraints. It assigns specific sampling weights to the key spatiotemporal characteristics of photovoltaic output and load power, while simultaneously integrating the temporal synchronicity and spatial complementarity of photovoltaic and load data in different regions. This strengthens the representativeness of sampling data related to key periods and regions, avoiding the loss of key features. Furthermore, by setting constraints on the rate of change of samples in adjacent periods, the rate of change limit is controlled to the maximum value in historical operating data, technically avoiding the problem of sample time-series jumps. Therefore, this design can ensure the accuracy of sampling statistics while taking into account the correlation characteristics of source and load in both time and space, making the generated independent scenarios possess temporal rationality, regional coordination, and engineering practicality, effectively improving the accuracy of scenario generation and the fitting accuracy to the actual distribution network operating characteristics. Thus, this design can ensure the statistical accuracy of sampling while possessing temporal rationality and engineering practicality, effectively improving the accuracy of scenario generation. Attached Figure Description

[0106] Figure 1 This is a flowchart of the method described in this invention.

[0107] Figure 2 This is a structural diagram of the system described in this invention.

[0108] Figure 3 This is a structural diagram of the device described in this invention.

[0109] Figure 4 This is a schematic diagram of a typical source-load scenario and its probability results in Example 2. Detailed Implementation

[0110] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0111] Example 1:

[0112] See Figure 1 A method for generating photovoltaic-load joint scenarios based on source-load correlation, wherein the grid optimization design method includes the following steps:

[0113] S1. Based on the differentiated fluctuation characteristics of light intensity and distribution network node load, construct a probability distribution model of light intensity and a probability distribution model of node load, respectively.

[0114] S2. Collect historical data of the area to be optimized. Based on the improved Latin hypercube sampling method with weighted key time periods and time series constraints, combine the light intensity probability distribution model and the node load probability distribution model to generate scenarios. Convert the light intensity independent scenario set into a photovoltaic independent scenario set and perform time series verification to obtain the time series verified photovoltaic independent scenario set and load independent scenario set.

[0115] S3. Combine the photovoltaic independent scenario set and the load independent scenario set that meet the time-series constraints in pairs to generate photovoltaic-load scenario combination pairs, and calculate the source-load correlation coefficient deviation. Based on the source-load correlation coefficient deviation The photovoltaic-load scenario combination is reduced to obtain a set of photovoltaic-load joint scenarios with occurrence probabilities.

[0116] In S1, the probability distribution model of the light intensity includes:

[0117] ;

[0118] In the above formula, The probability density number of light intensity; This represents the actual light intensity. The maximum light intensity during the study period; It is a gamma function; , All are shape parameters of the Beta distribution, including:

[0119] The specific calculation methods for the shape parameters include:

[0120] ;

[0121] In the above formula, Mathematical expectation of historical illumination monitoring data , The variance of historical illumination monitoring data;

[0122] The node load probability distribution model includes:

[0123] ;

[0124] In the above formula, Let be the probability density number of the node load. The average of historical load data. The standard deviation of historical loads can be obtained statistically from historical load data.

[0125] In step S2, the scene generation includes:

[0126] The historical data includes historical illumination data and historical load data. The historical data is aggregated into n time-series dimensions according to the time period of each day. Each time-series dimension includes the data of the corresponding time period of each natural day in the historical data.

[0127] Set the number of scenes, k;

[0128] Commonly used values ​​for k include 100, 500, or 1000, balancing calculation accuracy and computational cost.

[0129] By fitting the distribution parameters corresponding to each time period t (t=1,2,...,n) using historical illumination data and historical load data, the illumination shape parameters corresponding to each time period are obtained. , and the average of historical load data Standard deviation of historical load ;

[0130] The fitting method is a conventional existing technique such as the least squares method;

[0131] The probability distribution intervals of light intensity and load are uniformly divided into m equally probable sub-intervals, based on sampling weights. Increase the number of sampling points in the sampling interval for critical periods, and maintain equal probability division for non-critical periods to obtain a weighted sampling interval;

[0132] Based on the actual operation patterns of the distribution network, peak photovoltaic output periods (e.g., 10:00-14:00) and peak load periods (e.g., 8:00-10:00 and 18:00-22:00) are identified as the critical periods that have the most significant impact on the operation status of the distribution network. Higher sampling weights are assigned to the sampling intervals of the critical periods (the weight coefficients can be determined comprehensively based on the fluctuation characteristics, peak occurrence probability, and degree of impact on the operation of the distribution network of a large amount of historical source and load data). By assigning higher sampling weights to the critical periods, the number of effective samples and sampling density of the critical periods can be significantly increased, making the generated independent photovoltaic and load scenarios more representative and coverage in the peak output and peak load intervals. This improves the accuracy and credibility of subsequent distribution network operation simulation, scheduling optimization, and planning and design results, and solves the problem of sparse samples and loss of operating characteristics in the critical periods caused by traditional LHS uniform sampling.

[0133] The principle behind this invention is as follows: Traditional Latin hypercube sampling (LHS) uses uniform sampling, resulting in a uniform distribution of samples throughout the entire time period. This can easily lead to sparse samples and feature loss during peak and critical periods. In contrast, this invention expands the sampling interval through weighted sampling, artificially increasing the sampling priority and sample density during critical periods without altering the overall probability distribution characteristics. This concentrates the sampling results in the intervals that have the most significant impact on the safe and economical operation of the distribution network. This method ensures that high-impact intervals are fully sampled from the sampling mechanism, fundamentally avoiding the weakening of key features caused by traditional uniform sampling. This makes the generated scenarios more closely resemble actual operating patterns and is more suitable for distribution network analysis and decision-making.

[0134] The sampling weight include:

[0135] ;

[0136] ;

[0137] ;

[0138] In the above formula, Let be the sampling weight for time period t; This represents the time-period impact of the normalized time period t. Let t be the time-period influence of the time period. The weighted adjustment coefficient is the weighted adjustment coefficient. Calculated based on the fluctuation characteristics of source load data in the study area; The historical standard deviation of the source / load data for time period t reflects the intensity of force / load fluctuations. The measured values ​​of source / load on day i in time period t; For historical sample days;

[0139] By calculating sampling weights, the sampling density is automatically tilted towards key time periods, solving the problem of sparse samples and loss of operational features caused by traditional LHS uniform sampling.

[0140] Based on the light intensity distribution model, the load probability distribution model, and the weighted sampling interval, an improved Latin hypercube sampling method is used to generate light intensity independent scene sets and load independent scene sets, respectively.

[0141] The photovoltaic power conversion formula transforms the independent scene set of light intensity into an independent photovoltaic scene set, including:

[0142] The probability density number of light intensity The actual light intensity during time period t is obtained by sampling. After determining the light intensity, the actual photovoltaic output can be calculated by combining the photovoltaic power generation efficiency and the rated power of the photovoltaic array. The specific calculation formula for photovoltaic output includes:

[0143] ;

[0144] In the above formula, The photovoltaic output for time period t; The actual light intensity during time period t. For photovoltaic power generation efficiency; This refers to the rated power of the photovoltaic array;

[0145] Both the photovoltaic independent scenario set and the load independent scenario set have n×k dimensions;

[0146] The initial photovoltaic independent scenario set and load independent scenario set are subjected to time series continuity verification. The change rate of sampled values ​​in adjacent time periods under all scenarios is calculated. If it exceeds the historical maximum time series change rate, the scenario is deleted and a new scenario is generated by resampling, and the verification is performed again. If it does not exceed the historical maximum time series change rate, the scenario is retained, and the photovoltaic independent scenario set and load independent scenario set after time series verification are obtained.

[0147] In S3, the scene reduction includes:

[0148] Calculate the source-load correlation coefficient deviation for each photovoltaic-load scenario combination pair. Eliminating source-load correlation coefficient bias The optimal set of photovoltaic-load related scenarios is obtained by identifying photovoltaic-load scenario combinations that exceed a preset threshold.

[0149] The probability of each independent scenario in the photovoltaic independent scenario set and the load independent scenario set is 1 / k, and the probability of each pairwise combination is 1 / k², and the overall scenario is of equal probability.

[0150] When the source-load correlation coefficient deviation is removed After the number of photovoltaic-load scenario combinations exceeds the preset threshold, the remaining photovoltaic-load scenario combinations still maintain equal probability. This serves as the premise for subsequent clustering probability calculation. After clustering, the probability of a certain class is equal to the number of scenarios contained in that class / the total number of retained scenarios.

[0151] The source-load correlation coefficient deviation include:

[0152] ;

[0153] In the above formula, This is the sample correlation coefficient matrix. The target comprehensive coefficient matrix;

[0154] The sample correlation coefficient matrix By packaging various photovoltaic-load scenario combinations, the comprehensive correlation coefficients between medium load and corresponding photovoltaic time periods are obtained. The sample correlation coefficient matrix is ​​obtained. The dimension is n×n;

[0155] The target comprehensive coefficient matrix By combining the correlation coefficients of different time periods in historical illumination data and historical load data. The target comprehensive coefficient matrix is ​​obtained. The dimension is n×n.

[0156] The comprehensive correlation coefficient include:

[0157] ;

[0158] In the above formula, Let be the comprehensive correlation coefficient for time period t; Let be the temporal correlation coefficient for time period t. Let be the spatial correlation coefficient for time period t. The time-series correlation weighting coefficient. Let be the spatial correlation weight coefficient, and satisfy . ;

[0159] The temporal correlation coefficient of the t-th time period include:

[0160] ;

[0161] In the above formula, For the photovoltaic power output in time period t, Let be the load power in time period t; The time-series average of photovoltaic power output. This represents the time-series average of the load power. As a weighting factor for peak load periods, an exponential function is used to strengthen the correlation weighting of peak periods, addressing the problem of insufficient emphasis on critical periods in the traditional Pearson coefficient. The time decay coefficient indicates that the temporal correlation between photovoltaic output and load power exhibits an objective law of gradual decay over time. That is, the source-load synergy in adjacent time periods is significantly higher than that in interval time periods. The time decay coefficient can be determined by fitting the autocorrelation analysis statistical results of the source-load time series data of the study area over the past 3-5 years. The peak load baseline period (e.g., 7 PM) is the core period for distribution network operation and source-load coordinated control, and also the key period for reducing load in joint source-load scenarios. The traditional Pearson coefficient, which does not differentiate the weights of peak and off-peak periods, fails to reflect the core needs of scenario reduction. This parameter represents approximately... The larger the weighting factor is during peak load periods; The seasonal correction coefficient is obtained by quantitative fitting of the historical operating data of the source load in the four seasons of the past 3-5 years in the study area (e.g., 1.0 for spring, 1.1 for summer, 1.0 for autumn, and 1.2 for winter), which is adapted to the differences in source load characteristics in different seasons in the region. The window length corresponding to the calculation of the Pearson correlation coefficient using the sliding window method represents the granularity of the time period (if a time period is sampled at the minute level, it contains 60 data points, i.e., W=60). Its essence is a localized Pearson correlation coefficient.

[0162] The spatial correlation coefficient at time t include:

[0163] ;

[0164] In the above formula, the subscript 0 represents the study area, and the superscript 1 represents the neighboring area.

[0165] Example 2:

[0166] To further verify the effectiveness, stability, and superiority of this invention in actual engineering scenarios, based on the historical operation data of distributed photovoltaic power generation in a certain industrial park and its local historical solar irradiance and load monitoring data, the feasibility and engineering application value of the technical solution were verified through calculation, comparison, and analysis.

[0167] The data includes hourly solar irradiance monitoring data and hourly load measurement data for the park over the past three years (2023-2025); typical daily solar power output curves and typical daily load curves for this photovoltaic node; and typical daily solar power output curves and typical daily load curves for neighboring photovoltaic nodes. The rated photovoltaic power of this node is 500kW, with a power generation efficiency of 0.85; the rated load power is 1000kW. The initial number of independent scenarios is set at k=100; the number of typical scenarios after scenario reduction is N=8.

[0168] The photovoltaic-load combined scenario is generated using the model and solution algorithm proposed in this invention. The specific steps are as follows:

[0169] Step 1: Modeling source load uncertainties;

[0170] Based on hourly light intensity monitoring data and hourly load measurement data over the past three years, the probability distribution model parameters for different time periods were fitted, as shown in the table below:

[0171] ;

[0172] Step 2: Improve Latin hypercube sampling to generate source-load independent scenarios;

[0173] (1) Based on historical solar irradiance and load monitoring data, the sampling weights for the corresponding time periods were calculated according to the proposed model. The key time periods include the photovoltaic peak hours of 10:00-14:00, the morning load peak hours of 8:00-10:00, and the evening load peak hours of 18:00-22:00. >1.5.

[0174] (2) For each time period, the photovoltaic and load distribution intervals are divided into 100 equally probable sub-intervals. The non-critical time periods adopt the traditional uniform stratification, and the sub-intervals of the critical time periods expand the high probability area (i.e., increase the number of sampling points).

[0175] (3) Using the improved LHS for sampling, 24x100 dimensional photovoltaic and load initial independent scenarios were generated respectively.

[0176] (4) Perform time-series constraint verification and correction, calculate the maximum change rate of source load in adjacent historical time periods, remove scenarios with change rates exceeding the threshold, and resample until all scenarios meet the time-series constraints to obtain the corresponding independent scenario set.

[0177] Step 3: Combine relevant scenarios for reduction;

[0178] (1) Based on the typical photovoltaic daily power generation curve and typical load daily load curve of the photovoltaic node, the target correlation matrix Robj is calculated using the source-load spatiotemporal two-dimensional correlation quantification method proposed in this invention, with a dimension of 24x24.

[0179] (2) Pair 100 independent photovoltaic scenarios with 100 independent load scenarios to obtain 104 scenario pairs. Let the scenario pair number i=1 and the correlation deviation threshold be 0.1.

[0180] (3) For the i-th scene pair, the source-load spatiotemporal two-dimensional correlation quantization method proposed in this invention is used to calculate the sample correlation matrix Rsam of the i-th scene pair, with a dimension of 24x24.

[0181] (4) According to the correlation deviation ΔR calculation formula defined in this invention, calculate the correlation deviation of the i-th scene pair. If ΔR < 0.1, retain the scene pair; otherwise, discard it. Return to process (3) to perform traversal calculation until all scene pairs have been calculated.

[0182] (5) After calculating the correlation deviation ΔR for each scene and removing scenes that do not meet the requirements, there are approximately 8,000 valid scene pairs remaining.

[0183] Step 4: Determine typical scenarios;

[0184] K-Means clustering was used with 8 clusters. Similar scenes were aggregated using the comprehensive correlation coefficient as the distance indicator, and the probability of occurrence of each typical scene was calculated (probability of a typical scene = number of scenes in that category / total number of scenes). Typical source-load scenes and their probability results are available in [link to relevant documentation]. Figure 4 As shown;

[0185] Comparative analysis with traditional methods:

[0186] To quantitatively verify the superiority of the method of this invention, "traditional Latin hypercube sampling (LHS) + single Pearson correlation coefficient + conventional K-Means clustering" was selected as the traditional control method. The comparison was carried out under the same test case parameters, initial number of scenes, and number of typical scenes. The evaluation was carried out from the dimensions of scene generation quality, correlation representation accuracy, temporal rationality, and scene representativeness. The comparative analysis results are shown in the table below.

[0187] ;

[0188] Compared with traditional scene generation methods, the method of this invention has significantly improved the retention of features in key time periods, the rationality of time sequence, the accuracy of correlation characterization, and the representativeness of the scene. The generated typical photovoltaic-load joint scene is more in line with the actual operating characteristics of the distribution network, and can provide more reliable data support for distribution network operation simulation, scheduling optimization and planning design.

[0189] Example 3:

[0190] See Figure 2A photovoltaic-load joint scenario generation system based on source-load correlation specifically includes: a model building module, a scenario generation module, and a scenario reduction module;

[0191] The model building module is used to: construct a probability distribution model of light intensity and a probability distribution model of node load, respectively, based on the differentiated fluctuation characteristics of light intensity and distribution network node load;

[0192] The scene generation module is used to: collect historical data of the area to be optimized, generate scenes based on the improved Latin hypercube sampling method with weighted key time periods and time series constraints, combined with the light intensity probability distribution model and the node load probability distribution model, convert the light intensity independent scene set into the photovoltaic independent scene set, and perform time series verification to obtain the time series verified photovoltaic independent scene set and the load independent scene set.

[0193] The scenario reduction module is used to: combine the photovoltaic independent scenario set and the load independent scenario set that meet the time constraints in pairs to generate photovoltaic-load scenario combination pairs, and calculate the source-load correlation coefficient deviation. Based on the source-load correlation coefficient deviation The photovoltaic-load scenario combination is reduced to obtain a set of photovoltaic-load joint scenarios with occurrence probabilities.

[0194] In the model building module, the light intensity probability distribution model includes:

[0195] ;

[0196] In the above formula, The probability density number of light intensity; This represents the actual light intensity. The maximum light intensity during the study period; It is a gamma function; , All are shape parameters of the Beta distribution, including:

[0197] The specific calculation methods for the shape parameters include:

[0198] ;

[0199] In the above formula, Mathematical expectation of historical illumination monitoring data , The variance of historical illumination monitoring data;

[0200] The node load probability distribution model includes:

[0201] ;

[0202] In the above formula, Let be the probability density number of the node load. The average of historical load data. The standard deviation of historical loads can be obtained statistically from historical load data.

[0203] In the model building module, the scene generation includes:

[0204] The historical data includes historical illumination data and historical load data. The historical data is aggregated into n time-series dimensions according to the time period of each day. Each time-series dimension includes the data of the corresponding time period of each natural day in the historical data.

[0205] Set the number of scenes, k;

[0206] By fitting the distribution parameters corresponding to each time period t (t=1,2,...,n) using historical illumination data and historical load data, the illumination shape parameters corresponding to each time period are obtained. , and the average of historical load data Standard deviation of historical load ;

[0207] The probability distribution intervals of light intensity and load are uniformly divided into k equally probable sub-intervals, based on sampling weights. Increase the number of sampling points in the sampling interval for critical periods, and maintain equal probability division for non-critical periods to obtain a weighted sampling interval;

[0208] The sampling weight include:

[0209] ;

[0210] ;

[0211] ;

[0212] In the above formula, Let be the sampling weight for time period t; This represents the time-period impact of the normalized time period t. Let t be the time-period influence of the time period. The weighted adjustment coefficient is the weighted adjustment coefficient. Calculated based on the fluctuation characteristics of source load data in the study area; The historical standard deviation of the source / load data for time period t reflects the intensity of force / load fluctuations. The measured values ​​of source / load on day i in time period t; For historical sample days;

[0213] Based on the light intensity distribution model, the load probability distribution model, and the weighted sampling interval, an improved Latin hypercube sampling method is used to generate light intensity independent scene sets and load independent scene sets, respectively.

[0214] The photovoltaic power conversion formula transforms the independent scene set of light intensity into an independent photovoltaic scene set, including:

[0215] The probability density number of light intensity The actual light intensity during time period t is obtained by sampling. After determining the light intensity, the actual photovoltaic output can be calculated by combining the photovoltaic power generation efficiency and the rated power of the photovoltaic array. The specific calculation formula for photovoltaic output includes:

[0216] ;

[0217] In the above formula, The photovoltaic output for time period t; The actual light intensity during time period t. For photovoltaic power generation efficiency; This refers to the rated power of the photovoltaic array;

[0218] Both the photovoltaic independent scenario set and the load independent scenario set have n×k dimensions;

[0219] The initial photovoltaic independent scenario set and load independent scenario set are subjected to time series continuity verification. The change rate of sampled values ​​in adjacent time periods under all scenarios is calculated. If it exceeds the historical maximum time series change rate, the scenario is deleted and a new scenario is generated by resampling, and the verification is performed again. If it does not exceed the historical maximum time series change rate, the scenario is retained, and the photovoltaic independent scenario set and load independent scenario set after time series verification are obtained.

[0220] In the scene reduction module, the scene reduction includes:

[0221] Calculate the source-load correlation coefficient deviation for each photovoltaic-load scenario combination pair. Eliminating source-load correlation coefficient bias The optimal set of photovoltaic-load related scenarios is obtained by identifying photovoltaic-load scenario combinations that exceed a preset threshold.

[0222] The source-load correlation coefficient deviation include:

[0223] ;

[0224] In the above formula, This is the sample correlation coefficient matrix. The target comprehensive coefficient matrix;

[0225] The sample correlation coefficient matrix By packaging various photovoltaic-load scenario combinations, the comprehensive correlation coefficients between medium load and corresponding photovoltaic time periods are obtained. The sample correlation coefficient matrix is ​​obtained. The dimension is n×n;

[0226] The target comprehensive coefficient matrix By combining the correlation coefficients of different time periods in historical illumination data and historical load data. The target comprehensive coefficient matrix is ​​obtained. The dimension is n×n.

[0227] In the scene reduction module, the comprehensive correlation coefficient include:

[0228] ;

[0229] In the above formula, Let be the comprehensive correlation coefficient for time period t; Let be the temporal correlation coefficient for time period t. Let be the spatial correlation coefficient for time period t. The time-series correlation weighting coefficient. Let be the spatial correlation weight coefficient, and satisfy . ;

[0230] The temporal correlation coefficient of the t-th time period include:

[0231] ;

[0232] In the above formula, For the photovoltaic power output in time period t, Let be the load power in time period t; The time-series average of photovoltaic power output. This represents the time-series average of the load power. The weighting factor for peak load periods is reinforced by an exponential function to strengthen the correlation weighting of peak load periods. The time decay coefficient indicates that the temporal correlation between photovoltaic power output and load power exhibits an objective law of gradual decay over time. The peak load baseline period, The seasonal correction coefficient was obtained by quantitative fitting of the historical operating data of the source and load in the four seasons of the study area over the past 3-5 years.

[0233] The spatial correlation coefficient at time t include:

[0234] ;

[0235] In the above formula, the subscript 0 represents the study area, and the superscript 1 represents the neighboring area.

[0236] Example 4:

[0237] See Figure 3 A photovoltaic-load joint scene generation device with improved sampling and two-dimensional correlation quantization, characterized in that it includes a memory and a processor, wherein the memory is used to store computer program code and transmit the computer program code to the processor;

[0238] The processor is configured to execute, according to instructions in the computer program code, the improved sampling and two-dimensional correlation quantization photovoltaic-load joint scenario generation method as described in Example 1.

[0239] Example 5:

[0240] A computer program product includes a computer program, characterized in that the computer program is executed by a processor using the improved sampling and two-dimensional correlation quantization photovoltaic-load joint scenario generation method as described in Example 1.

[0241] The above description is only a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. Any equivalent modifications or changes made by those skilled in the art based on the content disclosed in the present invention should be included within the scope of protection set forth in the claims.

Claims

1. A method for generating photovoltaic-load joint scenarios based on source-load correlation, characterized in that, The space frame optimization design method includes the following steps: S1. Based on the differentiated fluctuation characteristics of light intensity and distribution network node load, construct a probability distribution model of light intensity and a probability distribution model of node load, respectively. S2. Collect historical data of the area to be optimized. Based on the improved Latin hypercube sampling method with weighted key time periods and time series constraints, combine the light intensity probability distribution model and the node load probability distribution model to generate scenarios. Convert the light intensity independent scenario set into a photovoltaic independent scenario set and perform time series verification to obtain the time series verified photovoltaic independent scenario set and load independent scenario set. S3. Combine the photovoltaic independent scenario set and the load independent scenario set that meet the time-series constraints in pairs to generate photovoltaic-load scenario combination pairs, and calculate the source-load correlation coefficient deviation. Based on the source-load correlation coefficient deviation The photovoltaic-load scenario combination is reduced to obtain a set of photovoltaic-load joint scenarios with occurrence probabilities.

2. The photovoltaic-load joint scenario generation method based on source-load correlation according to claim 1, characterized in that, In S1, the light intensity probability distribution model includes: ； In the above formula, The probability density number of light intensity; This represents the actual light intensity. The maximum light intensity during the study period; It is a gamma function; , All are shape parameters of the Beta distribution, including: The specific calculation methods for the shape parameters include: ； In the above formula, Mathematical expectation of historical illumination monitoring data , The variance of historical illumination monitoring data; The node load probability distribution model includes: ； In the above formula, Let be the probability density number of the node load. The average of historical load data. The standard deviation of historical loads can be obtained statistically from historical load data.

3. The photovoltaic-load joint scenario generation method based on source-load correlation according to claim 2, characterized in that, In step S2, the scene generation includes: The historical data includes historical illumination data and historical load data. The historical data is aggregated into n time-series dimensions according to the time period of each day. Each time-series dimension includes the data of the corresponding time period of each natural day in the historical data. Set the number of scenes, k; By fitting the distribution parameters corresponding to each time period t (t=1,2,...,n) using historical illumination data and historical load data, the illumination shape parameters corresponding to each time period are obtained. , and the average of historical load data Standard deviation of historical load ; The probability distribution intervals of light intensity and load are uniformly divided into m equally probable sub-intervals, based on sampling weights. Increase the number of sampling points in the sampling interval for critical periods, and maintain equal probability division for non-critical periods to obtain a weighted sampling interval; The sampling weight include: ； ； ； In the above formula, Let be the sampling weight for time period t; This represents the time-period impact of the normalized time period t. Let t be the time-period influence of the time period. The weighted adjustment coefficient is the weighted adjustment coefficient. Calculated based on the fluctuation characteristics of source load data in the study area; The historical standard deviation of the source / load data for time period t reflects the intensity of force / load fluctuations. The measured values ​​of source / load on day i in time period t; For historical sample days; Based on the light intensity distribution model, the load probability distribution model, and the weighted sampling interval, an improved Latin hypercube sampling method is used to generate light intensity independent scene sets and load independent scene sets, respectively. The photovoltaic power conversion formula transforms the independent scene set of light intensity into an independent photovoltaic scene set, including: The probability density number of light intensity The actual light intensity during time period t is obtained by sampling. After determining the light intensity, the actual photovoltaic output can be calculated by combining the photovoltaic power generation efficiency and the rated power of the photovoltaic array. The specific calculation formula for photovoltaic output includes: ； In the above formula, The photovoltaic output for time period t; The actual light intensity during time period t. For photovoltaic power generation efficiency; This refers to the rated power of the photovoltaic array; Both the photovoltaic independent scenario set and the load independent scenario set have n×k dimensions; The initial photovoltaic independent scenario set and load independent scenario set are subjected to time series continuity verification. The change rate of sampled values ​​in adjacent time periods under all scenarios is calculated. If it exceeds the historical maximum time series change rate, the scenario is deleted and a new scenario is generated by resampling, and the verification is performed again. If it does not exceed the historical maximum time series change rate, the scenario is retained, and the photovoltaic independent scenario set and load independent scenario set after time series verification are obtained.

4. The photovoltaic-load joint scenario generation method based on source-load correlation according to claim 3, characterized in that, In S3, the scene reduction includes: Calculate the source-load correlation coefficient deviation for each photovoltaic-load scenario combination pair. Eliminating source-load correlation coefficient bias The optimal set of photovoltaic-load related scenarios is obtained by identifying photovoltaic-load scenario combinations that exceed a preset threshold. The source-load correlation coefficient deviation include: ； In the above formula, This is the sample correlation coefficient matrix. The target comprehensive coefficient matrix; The sample correlation coefficient matrix By packaging various photovoltaic-load scenario combinations, the comprehensive correlation coefficients between medium load and corresponding photovoltaic time periods are obtained. The sample correlation coefficient matrix is ​​obtained. The dimension is n×n; The target comprehensive coefficient matrix By combining the correlation coefficients of different time periods in historical illumination data and historical load data. The target comprehensive coefficient matrix is ​​obtained. The dimension is n×n.

5. The photovoltaic-load joint scenario generation method based on source-load correlation according to claim 4, characterized in that, The comprehensive correlation coefficient include: ； In the above formula, Let be the comprehensive correlation coefficient for time period t; Let be the temporal correlation coefficient for time period t. Let be the spatial correlation coefficient for time period t. The time-series correlation weighting coefficient. Let be the spatial correlation weight coefficient, and satisfy . ; The temporal correlation coefficient of the t-th time period include: ； In the above formula, For the photovoltaic power output in time period t, Let be the load power in time period t; The time-series average of photovoltaic power output. This represents the time-series average of the load power. The weighting factor for peak load periods is reinforced by an exponential function to strengthen the correlation weighting of peak load periods. The time decay coefficient indicates that the temporal correlation between photovoltaic power output and load power exhibits an objective law of gradual decay over time. The peak load baseline period, The seasonal correction coefficient was obtained by quantitative fitting of historical source water operation data for the past 3-5 years in the study area. The spatial correlation coefficient at time t include: ； In the above formula, the subscript 0 represents the study area, and the superscript 1 represents the neighboring area.

6. A photovoltaic-load joint scenario generation system based on source-load correlation, characterized in that, Specifically, it includes: Model building module, scene generation module, scene reduction module; The model building module is used to: construct a probability distribution model of light intensity and a probability distribution model of node load, respectively, based on the differentiated fluctuation characteristics of light intensity and distribution network node load; The scene generation module is used to: collect historical data of the area to be optimized, generate scenes based on the improved Latin hypercube sampling method with weighted key time periods and time series constraints, combined with the light intensity probability distribution model and the node load probability distribution model, convert the light intensity independent scene set into the photovoltaic independent scene set, and perform time series verification to obtain the time series verified photovoltaic independent scene set and the load independent scene set. The scenario reduction module is used to: combine the photovoltaic independent scenario set and the load independent scenario set that meet the time-series constraints in pairs to generate photovoltaic-load scenario combination pairs, and calculate the source-load correlation coefficient deviation. Based on the source-load correlation coefficient deviation The photovoltaic-load scenario combination is reduced to obtain a set of photovoltaic-load joint scenarios with occurrence probabilities.

7. A photovoltaic-load joint scenario generation system based on source-load correlation according to claim 6, characterized in that, In the model building module, the light intensity probability distribution model includes: ； In the above formula, The probability density number of light intensity; This represents the actual light intensity. The maximum light intensity during the study period; It is a gamma function; , All are shape parameters of the Beta distribution, including: The specific calculation methods for the shape parameters include: ； In the above formula, Mathematical expectation of historical illumination monitoring data , The variance of historical illumination monitoring data; The node load probability distribution model includes: ； In the above formula, Let be the probability density number of the node load. The average of historical load data. The standard deviation of historical loads can be obtained statistically from historical load data.

8. The photovoltaic-load joint scenario generation system based on source-load correlation according to claim 7, characterized in that, In the model building module, the scene generation includes: The historical data includes historical illumination data and historical load data. The historical data is aggregated into n time-series dimensions according to the time period of each day. Each time-series dimension includes the data of the corresponding time period of each natural day in the historical data. Set the number of scenes, k; By fitting the distribution parameters corresponding to each time period t (t=1,2,...,n) using historical illumination data and historical load data, the illumination shape parameters corresponding to each time period are obtained. , and the average of historical load data Standard deviation of historical load ; The probability distribution intervals of light intensity and load are uniformly divided into k equally probable sub-intervals, based on sampling weights. Increase the number of sampling points in the sampling interval for critical periods, and maintain equal probability division for non-critical periods to obtain a weighted sampling interval; The sampling weight include: ； ； ； In the above formula, Let be the sampling weight for time period t; This represents the time-period impact of the normalized time period t. Let t be the time-period influence of the time period. The weighted adjustment coefficient is the weighted adjustment coefficient. Calculated based on the fluctuation characteristics of source load data in the study area; The historical standard deviation of the source / load data for time period t reflects the intensity of force / load fluctuations. The measured values ​​of source / load on day i in time period t; For historical sample days; Based on the light intensity distribution model, the load probability distribution model, and the weighted sampling interval, an improved Latin hypercube sampling method is used to generate light intensity independent scene sets and load independent scene sets, respectively. The photovoltaic power conversion formula transforms the independent scene set of light intensity into an independent photovoltaic scene set, including: The probability density number of light intensity The actual light intensity during time period t is obtained by sampling. After determining the light intensity, the actual photovoltaic output can be calculated by combining the photovoltaic power generation efficiency and the rated power of the photovoltaic array. The specific calculation formula for photovoltaic output includes: ； In the above formula, The photovoltaic output for time period t; The actual light intensity during time period t. For photovoltaic power generation efficiency; This refers to the rated power of the photovoltaic array; Both the photovoltaic independent scenario set and the load independent scenario set have n×k dimensions; The initial photovoltaic independent scenario set and load independent scenario set are subjected to time series continuity verification. The change rate of sampled values ​​in adjacent time periods under all scenarios is calculated. If it exceeds the historical maximum time series change rate, the scenario is deleted and a new scenario is generated by resampling, and the verification is performed again. If it does not exceed the historical maximum time series change rate, the scenario is retained, and the photovoltaic independent scenario set and load independent scenario set after time series verification are obtained.

9. A photovoltaic-load joint scenario generation system based on source-load correlation according to claim 8, characterized in that, In the scene reduction module, the scene reduction includes: Calculate the source-load correlation coefficient deviation for each photovoltaic-load scenario combination pair. Eliminating source-load correlation coefficient bias The optimal set of photovoltaic-load related scenarios is obtained by identifying photovoltaic-load scenario combinations that exceed a preset threshold. The source-load correlation coefficient deviation include: ； In the above formula, This is the sample correlation coefficient matrix. The target comprehensive coefficient matrix; The sample correlation coefficient matrix By packaging various photovoltaic-load scenario combinations, the comprehensive correlation coefficients between medium load and corresponding photovoltaic time periods are obtained. The sample correlation coefficient matrix is ​​obtained. The dimension is n×n; The target comprehensive coefficient matrix By combining the correlation coefficients of different time periods in historical illumination data and historical load data. The target comprehensive coefficient matrix is ​​obtained. The dimension is n×n.

10. A photovoltaic-load joint scenario generation system based on source-load correlation according to claim 9, characterized in that, In the scene reduction module, the comprehensive correlation coefficient include: ； In the above formula, Let be the comprehensive correlation coefficient for time period t; Let be the temporal correlation coefficient for time period t. Let be the spatial correlation coefficient for time period t. The time-series correlation weighting coefficient. Let be the spatial correlation weight coefficient, and satisfy . ; The temporal correlation coefficient of the t-th time period include: ； In the above formula, For the photovoltaic power output in time period t, Let be the load power in time period t; The time-series average of photovoltaic power output. This represents the time-series average of the load power. The weighting factor for peak load periods is reinforced by an exponential function to strengthen the correlation weighting of peak load periods. The time decay coefficient indicates that the temporal correlation between photovoltaic power output and load power exhibits an objective law of gradual decay over time. The peak load baseline period, The seasonal correction coefficient was obtained by quantitative fitting of historical source water operation data for the past 3-5 years in the study area. The spatial correlation coefficient at time t include: ； In the above formula, the subscript 0 represents the study area, and the superscript 1 represents the neighboring area.

Images