Power grid dispatching optimization method and system based on timing correlation flexible resource identification

Abstract

A power grid scheduling optimization method and system based on timing correlation flexible resource identification, comprising: collecting historical building energy consumption, power grid operation and meteorological data; constructing a power grid friendly building evaluation system to calculate the index value of each evaluation index; the evaluation index includes a grid load inertia coefficient for quantifying the shape similarity between the building load curve and the total grid load curve; calculate the correlation influence weight and statistical characteristic weight of each evaluation index, and fuse them into a comprehensive weight coefficient, the statistical characteristic weight is calculated by the improved CRITIC method of the contrast intensity modified by the grid load sensitivity coefficient; an evaluation model introducing cloud drop overlap similarity is constructed, and the evaluation index value corresponding to each building and the corresponding comprehensive weight are input into the model for sorting; based on the sorting results of each building, a flexible building cluster is screened and a scheduling strategy of each building in the cluster is generated. The present application provides technical support for accurate identification of power grid friendly buildings and optimization of power grid demand side response.

Description

Technical Field

[0001] This invention belongs to the field of power system operation and control technology, and in particular relates to a power grid dispatch optimization method and system based on time-series correlation flexible resource identification. Background Technology

[0002] Building flexible loads have become an important peak-shaving resource for power grid demand-side response. Accurate identification and coordinated scheduling of building flexible resources are core technologies for improving power grid stability and reducing peak-valley differences. Building flexible loads possess significant controllability potential and wide distribution, making them a crucial component of adjustable resources on the power grid demand side. However, blindly utilizing building loads during scheduling can easily lead to power grid disturbances, increased losses from frequent start-ups and shutdowns of distribution equipment, and unsatisfactory peak-shaving and valley-filling effects, hindering the safe and economical operation of the power grid and achieving low-carbon scheduling goals. Furthermore, the time scales of monitoring data from various equipment within buildings often differ, and the identification and evaluation of high-quality buildings typically rely on multi-source data fusion analysis and multi-criteria decision-making methods. Existing technologies commonly employ data fusion methods including interpolation, Dynamic Time Warping (DTW) algorithms, and time-series alignment methods based on machine learning (such as LSTM). In multi-criteria decision-making, methods such as Analytic Hierarchy Process (ANP), Inter-Standard Correlation Analysis (CRITIC), and entropy weighting are frequently used for index weighting, combined with the Top-Order Solution Approximation (TOPSIS) method for comprehensive evaluation. Dynamic Time Warping (DTW) is a method for measuring the similarity between two time series. By flexibly aligning the time axis, it can handle problems such as inconsistent series lengths and phase shifts. Precise selection and coordinated scheduling of flexible building resources are key technologies for tapping demand-side response potential and reducing peak-valley differences in the power grid.

[0003] In existing technologies, the paper "A Method for Identifying Residential Users with Demand Response Potential Based on Association Rule Analysis of Multi-Source Heterogeneous Data," published in the 5th issue of *Power System Technology* in 2023, proposes a method for analyzing the correlation between user load and external factors based on Spearman correlation coefficient. This method includes: collecting multi-source data such as user load, temperature, humidity, and electricity price; calculating the dynamic correlation between load and various external factors using Spearman correlation coefficient; constructing an evaluation index system based on the correlation and determining weights using a weighting method based on correlation and redundancy analysis; and ranking users' demand response potential using the TOPSIS method. The shortcomings of this method are: the Spearman correlation coefficient cannot effectively handle phase shifts and lag effects between time series, and it ignores load response delays caused by factors such as building thermal inertia; the weighting method is singular, relying solely on objective data and failing to reflect causal relationships between indicators and decision-maker preferences, potentially leading to a disconnect between weight allocation and actual management needs; and the lack of fuzziness and randomness handling mechanisms during the evaluation process results in insufficient robustness of the evaluation results when data fluctuations are significant.

[0004] In summary, there is an urgent need to propose a grid-friendly low-carbon building identification and grid dispatch optimization method that takes into account the accuracy of time alignment, the integration of multiple weighting methods, and high robustness. This method would address the shortcomings of existing technologies in areas such as multi-source heterogeneous data processing, indicator weight allocation, and evaluation result stability. It would provide comprehensive technical support for the accurate mining of flexible resources on the building side and the efficient implementation of grid demand-side response, thereby helping to improve the stability of grid operation and the level of low-carbon dispatch. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a power grid dispatch optimization method and system based on time-series correlation flexible resource identification.

[0006] The present invention adopts the following technical solution.

[0007] In a first aspect, the present invention discloses a power grid dispatch optimization method based on time-series correlation flexible resource identification, the method comprising the following steps: Step 1: Collect historical building energy consumption, power grid operation, and meteorological data. After preprocessing and clustering, the data is divided into multiple typical days. The building energy consumption data includes building load, and the power grid operation data includes total power grid load. Step 2: Construct a grid-friendly building evaluation system that includes multiple evaluation indicators, and calculate the index values ​​of each evaluation indicator based on the collected data from the multiple typical days; the evaluation indicators include a grid load inertia coefficient used to quantify the morphological similarity between the building load curve and the total grid load curve; Step 3: Calculate the correlation influence weight and statistical characteristic weight of each evaluation index, and integrate them into the comprehensive weight coefficient of each evaluation index. The statistical characteristic weight is calculated using the improved CRITIC method with the comparison intensity corrected by the network load sensitivity coefficient. The network load sensitivity coefficient is used to characterize the correlation degree of each evaluation index with respect to the network load inertia coefficient. Step 4: Construct an evaluation model that incorporates cloud droplet overlap similarity. Input the evaluation index values ​​corresponding to each building and the comprehensive weight coefficients corresponding to each evaluation index into the model for sorting. The evaluation model adaptively selects the total number of matching cloud droplet pairs participating in the similarity calculation based on the uncertainty of the cloud model. Step 5: Based on the ranking results of each building, select the flexible building clusters that participate in demand response, and generate the scheduling strategy for each building in the flexible building cluster by combining the power grid load regulation requirements of each typical day.

[0008] More preferably, The evaluation indicators also include temperature inertia coefficient, electricity price inertia coefficient, and carbon emission factor inertia coefficient; each inertia coefficient is calculated based on a dynamic time warping algorithm with path bending length constraint, and the path bending length constraint is adaptively determined based on the maximum influence time of the external factor sequence corresponding to the inertia coefficient on the building load sequence; the external factor sequence corresponding to the temperature inertia coefficient is the temperature sequence, the external factor sequence corresponding to the electricity price inertia coefficient is the electricity price sequence, and the external factor sequence corresponding to the carbon emission factor inertia coefficient is the carbon emission factor sequence.

[0009] More preferably, In step 3, the contrast intensity corrected for the net load sensitivity coefficient in the improved CRITIC method is determined as follows:

[0010]

[0011] in, The contrast intensity is adjusted for the net load sensitivity coefficient; For the first j The original contrast strength of each indicator For the first i The building in j Standardized positive scores on each indicator; For the first j The average of all building samples for each indicator; n The total number of buildings to be evaluated; For the first j The network load sensitivity coefficient of each indicator; The correction coefficient is , and satisfies .

[0012] More preferably, The first j The network load sensitivity coefficients for each indicator are determined as follows:

[0013] in, The function for calculating the Pearson correlation coefficient; For the first i The inertia coefficient of the load on a building.

[0014] More preferably, In step 3, the comprehensive weight coefficient of each evaluation indicator is determined as follows:

[0015] in, For the first The comprehensive weighting coefficient of each evaluation indicator is the decision variable; For the first i Importance coefficient of each weighting method; Indicates the first The first weighting method is used to calculate the first weighting method. The weight of each indicator; and The first The maximum and minimum values ​​of each indicator in each weighting method; m This represents the total number of weighting methods. n This represents the total number of indicators.

[0016] More preferably, In step 4, the cloud droplet overlap similarity is used to measure the similarity of clouds based on the degree of overlap of cloud droplet membership, and is specifically determined as follows:

[0017] in, for , Cloud droplet overlap similarity between two cloud models to be compared; For the first k Sample the index values ​​corresponding to the matching cloud droplets; , They are respectively , Membership function; K This represents the total number of matching cloud droplet pairs participating in the similarity calculation.

[0018] More preferably, The membership function is determined as follows:

[0019] in, For indicator value x Membership degree in the cloud model; , These represent the expectation and entropy of the cloud model, respectively.

[0020] More preferably, In step 4, the total number of matching cloud droplet pairs participating in the similarity calculation is determined as follows:

[0021] in, This is the preset minimum entropy value in the cloud model; This is the proportionality coefficient; To match the maximum number of cloud droplets; , These are the entropy and hyperentropy of the cloud model, respectively. This indicates rounding up to the nearest integer.

[0022] More preferably, The step of inputting the evaluation index values ​​corresponding to each building and the comprehensive weight coefficient corresponding to each evaluation index into the model for sorting specifically includes: The cloud model is used to quantify the standardized evaluation index values, construct the standard evaluation cloud corresponding to each evaluation index of each building, and calculate the weighted evaluation cloud corresponding to each evaluation index of each building by combining the comprehensive weight coefficients corresponding to each evaluation index. Based on the cloud model feature parameters of the weighted evaluation cloud, the positive ideal solution cloud and negative ideal solution cloud corresponding to each evaluation index are calculated. For each building's weighted evaluation cloud under each evaluation index, calculate its cloud droplet overlap similarity with the positive and negative ideal solution clouds respectively. After multiple calculations, take the average value as the final cloud droplet overlap similarity, and calculate the confidence interval of the final cloud droplet overlap similarity. For each building, the final cloud droplet overlap similarity of each evaluation index is summarized, and the overall similarity, closeness and confidence interval of each building with the positive and negative ideal solution clouds are calculated. The buildings are ranked based on their proximity and their confidence intervals.

[0023] More preferably, In step 5, the scheduling strategy for each building in the flexible building cluster is determined as follows: Calculate the load allocation weight of each building in the flexible building cluster, and allocate the power grid load regulation demand of each typical day to each building according to the load allocation weight of each building; wherein, the load allocation weight of each building is the ratio of the product of the building's proximity and adjustable load capacity to the sum of the products of the comprehensive proximity and adjustable load capacity of all buildings in the flexible building cluster.

[0024] Secondly, the present invention discloses a power grid dispatch optimization system based on time-series correlation flexible resource identification based on the aforementioned method, including a data acquisition module, an evaluation system construction module, an evaluation index weight calculation module, a flexible building sorting module, and a flexible building cluster dispatch strategy generation module; The data acquisition module collects historical building energy consumption, power grid operation, and meteorological data. After preprocessing and clustering, the data is divided into multiple typical days. The building energy consumption data includes building load, and the power grid operation data includes total power grid load. The evaluation system construction module constructs a grid-friendly building evaluation system that includes multiple evaluation indicators, and calculates the index values ​​of each evaluation indicator based on the collected data of the multiple typical days; the evaluation indicators include the grid load inertia coefficient, which is used to quantify the morphological similarity between the building load curve and the total grid load curve. The evaluation index weight calculation module calculates the correlation influence weight and statistical characteristic weight of each evaluation index, and integrates them into the comprehensive weight coefficient of each evaluation index. The statistical characteristic weight is calculated using the improved CRITIC method with the comparison intensity corrected by the network load sensitivity coefficient. The network load sensitivity coefficient is used to characterize the degree of correlation between each evaluation index and the network load inertia coefficient. The flexible building sorting module constructs an evaluation model that incorporates cloud droplet overlap similarity. The evaluation index values ​​corresponding to each building and the comprehensive weight coefficients corresponding to each evaluation index are input into the model for sorting. The evaluation model adaptively selects the total number of matching cloud droplet pairs participating in the similarity calculation based on the uncertainty of the cloud model. The flexible building cluster scheduling strategy generation module selects flexible building clusters participating in demand response based on the ranking results of each building, and generates the scheduling strategy for each building in the flexible building cluster by combining the power grid load control requirements of each typical day.

[0025] 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 perform the steps of the method according to any one of the first aspects of the present invention.

[0026] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the first aspects of the present invention.

[0027] The beneficial effects of this invention are compared with those of the prior art: 1. This invention improves the temporal alignment accuracy and physical interpretability of correlation analysis between building load and external factors. Traditional methods (such as those based on Spearman or Pearson correlation coefficients) cannot handle phase shifts between time series and are sensitive to inconsistencies in data time scales, ignoring load response lags caused by building thermal inertia, resulting in distorted correlation analysis results. This invention employs a dynamic time warping algorithm to directly calculate the DTW distance between the load series and external series such as temperature and electricity prices, and proposes "inertia coefficient" and "impact timeliness" indicators. It achieves flexible temporal alignment of multi-source heterogeneous data, accurately quantifying the morphological similarity and lag time of the impact of external factors on building load.

[0028] 2. This invention achieves a scientific integration of indicator weights. Existing methods mostly employ traditional weighting methods such as CRITIC, where weights rely entirely on statistical data characteristics. This invention overcomes the shortcomings of the traditional CRITIC method, which calculates weights based solely on the indicator's own fluctuations and correlations, neglecting the relationship between indicators and grid load characteristics. By introducing a grid-load sensitivity coefficient to correct the comparison intensity, the weight allocation proactively tilts towards grid-friendly indicators, better aligning with the actual needs of grid dispatching and low-carbon management. The improved CRITIC method also uses a multiple correlation coefficient as a measure of indicator conflict, enhancing the robustness of statistical feature weights. This invention employs a weight calculation framework that integrates the DANP method and the improved CRITIC method. This framework, for the first time in the field of building identification, systematically combines the correlation influence weights reflecting the causal relationships of indicator networks with the statistical feature weights that enhance the statistical characteristics of data.

[0029] 3. It enhances the robustness, discrimination and anti-interference ability of the comprehensive evaluation model under uncertain environments. In existing technologies, the traditional TOPSIS method is mostly used for ranking. It uses deterministic values ​​and Euclidean distance, which is sensitive to data fluctuations and extreme values. It cannot handle the inherent fuzziness and randomness in the evaluation, and the ranking results are unstable. This invention proposes an evaluation model incorporating cloud droplet overlap similarity. First, a cloud model is used to characterize the fuzzy random scores of the indicators. Then, cloud droplet overlap similarity is used to calculate the similarity between cloud models. The cloud model absorbs the uncertainty of the data through entropy (En) and hyperentropy (He), making the evaluation results less susceptible to interference from individual noise points. Furthermore, the cloud droplet overlap similarity coefficient can more sensitively capture the overlap differences between distributions. When calculating cloud droplet overlap similarity, this invention proposes to adaptively determine the total number of matching cloud droplet pairs participating in the similarity calculation based on the entropy (En) and hyperentropy (He) of the cloud model. This avoids the insufficient accuracy or computational waste caused by fixing the number of cloud droplets. While ensuring the accuracy of complex indicators, it reduces the computational overhead of simple indicators, improving the overall algorithm's running efficiency. During the ranking calculation, this invention introduces confidence intervals as a comparison criterion, incorporating the uncertainty in the cloud model calculation process into the ranking logic. This provides statistical significance support for the ranking results, enhancing the robustness and interpretability of the proposed method.

[0030] 4. It achieves deep synergy between building flexible resources and power grid dispatch, improving the technical and economic efficiency of power grid operation. By accurately identifying building clusters with high response potential through flexible potential identification results, and combining this with power grid load characteristics to generate targeted peak-shaving and valley-filling coordinated dispatch strategies, the accuracy of demand-side dispatch can be effectively improved, reducing the peak-valley difference in the power grid. At the same time, it avoids power grid power disturbances caused by blindly calling upon building loads, reduces the frequency of start-ups and shutdowns of distribution equipment, reduces equipment operating losses, and extends equipment lifespan. In addition, dispatch commands based on proximity matching can improve the synchronization of building load response, further enhancing the peak-shaving and valley-filling effect, and contributing to the low-carbon, stable, and economical operation of the power grid. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the power grid dispatch optimization method based on time-series correlation flexible resource identification in this invention; Figure 2 This is the multi-criteria evaluation system in Embodiment 1 of the present invention; Figure 3 This is the ANP hierarchical structure diagram in Embodiment 1 of the present invention; Figure 4 This is the distribution of index weights under different weighting methods in Embodiment 2 of the present invention; Figure 5 This is the ranking result of the low-carbon interaction potential of various types of buildings in Embodiment 2 of the present invention. Detailed Implementation

[0032] 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.

[0033] like Figure 1 As shown, this invention discloses a power grid dispatch optimization method based on time-series correlation flexible resource identification, comprising the following steps: Step 1: Collect historical building energy consumption, power grid operation, and meteorological data. After preprocessing and clustering, the data is divided into multiple typical days. The building energy consumption data includes building load, and the power grid operation data includes total power grid load. Step 2: Construct a grid-friendly building evaluation system that includes multiple evaluation indicators, and calculate the index values ​​of each evaluation indicator based on the collected data from the multiple typical days; the evaluation indicators include a grid load inertia coefficient used to quantify the morphological similarity between the building load curve and the total grid load curve; The evaluation indicators also include temperature inertia coefficient, electricity price inertia coefficient, and carbon emission factor inertia coefficient; each inertia coefficient is calculated based on a dynamic time warping algorithm with path bending length constraint, and the path bending length constraint is adaptively determined based on the maximum influence time of the external factor sequence corresponding to the inertia coefficient on the building load sequence; the external factor sequence corresponding to the temperature inertia coefficient is the temperature sequence, the external factor sequence corresponding to the electricity price inertia coefficient is the electricity price sequence, and the external factor sequence corresponding to the carbon emission factor inertia coefficient is the carbon emission factor sequence.

[0034] Step 3: Calculate the correlation influence weight and statistical characteristic weight of each evaluation index, and integrate them into the comprehensive weight coefficient of each evaluation index. The statistical characteristic weight is calculated using the improved CRITIC method with the comparison intensity corrected by the network load sensitivity coefficient. The network load sensitivity coefficient is used to characterize the correlation degree of each evaluation index with respect to the network load inertia coefficient. The contrast intensity corrected for the net load sensitivity coefficient in the improved CRITIC method is determined as follows:

[0035]

[0036] in, The contrast intensity is adjusted for the net load sensitivity coefficient; For the first j The original contrast strength of each indicator For the firsti The building in j Standardized positive scores on each indicator; For the first j The average of all building samples for each indicator; n The total number of buildings to be evaluated; For the first j The network load sensitivity coefficient of each indicator; The correction coefficient is , and satisfies .

[0037] The load sensitivity coefficient is determined as follows:

[0038] in, The function for calculating the Pearson correlation coefficient; For the first i The inertia coefficient of the load on a building.

[0039] The comprehensive weighting coefficients of each evaluation indicator are determined as follows:

[0040] in, For the first The comprehensive weighting coefficient of each evaluation indicator is the decision variable; For the first i Importance coefficient of each weighting method; Indicates the first The first weighting method is used to calculate the first weighting method. The weight of each indicator; and The first The maximum and minimum values ​​of each indicator in each weighting method; m This represents the total number of weighting methods. n This represents the total number of indicators.

[0041] Step 4: Construct an evaluation model that incorporates cloud droplet overlap similarity. Input the evaluation index values ​​corresponding to each building and the comprehensive weight coefficients corresponding to each evaluation index into the model for sorting and optimization. The evaluation model adaptively selects the total number of matching cloud droplet pairs participating in the similarity calculation based on the uncertainty of the cloud model.

[0042] The cloud droplet overlap similarity is used to measure the similarity of clouds based on the degree of overlap of cloud droplet membership, and is determined in the following way:

[0043] in, for , Cloud droplet overlap similarity between two cloud models to be compared; For the first k Sample the index values ​​corresponding to the matching cloud droplets; , They are respectively , Membership function; K This represents the total number of matching cloud droplet pairs participating in the similarity calculation.

[0044] The membership function is determined as follows:

[0045] in, For indicator value x Membership degree in the cloud model; , These represent the expectation and entropy of the cloud model, respectively.

[0046] The total number of matching cloud droplet pairs participating in the similarity calculation is determined in the following manner:

[0047] in, This is the preset minimum entropy value in the cloud model; This is the proportionality coefficient; To match the maximum number of cloud droplets; , These are the entropy and hyperentropy of the cloud model, respectively. This indicates rounding up to the nearest integer.

[0048] The step of inputting the evaluation index values ​​corresponding to each building and the comprehensive weight coefficient corresponding to each evaluation index into the model for sorting specifically includes: The cloud model is used to quantify the standardized evaluation index values, construct the standard evaluation cloud corresponding to each evaluation index of each building, and calculate the weighted evaluation cloud corresponding to each evaluation index of each building by combining the comprehensive weight coefficients corresponding to each evaluation index. Based on the cloud model feature parameters of the weighted evaluation cloud, the positive ideal solution cloud and negative ideal solution cloud corresponding to each evaluation index are calculated. For each building's weighted evaluation cloud under each evaluation index, calculate its cloud droplet overlap similarity with the positive and negative ideal solution clouds respectively. After multiple calculations, take the average value as the final cloud droplet overlap similarity, and calculate the confidence interval of the final cloud droplet overlap similarity. For each building, the final cloud droplet overlap similarity of each evaluation index is summarized, and the overall similarity, closeness and confidence interval of each building with the positive and negative ideal solution clouds are calculated. The buildings are ranked based on their proximity and their confidence intervals.

[0049] Step 5: Based on the ranking results of each building, select the flexible building clusters that participate in demand response, and generate the scheduling strategy for each building in the flexible building cluster by combining the power grid load regulation requirements of each typical day.

[0050] The scheduling strategy for each building in the flexible building cluster is determined as follows: Calculate the load allocation weight of each building in the flexible building cluster, and allocate the power grid load regulation demand of each typical day to each building according to the load allocation weight of each building; wherein, the load allocation weight of each building is the ratio of the product of the building's proximity and adjustable load capacity to the sum of the products of the comprehensive proximity and adjustable load capacity of all buildings in the flexible building cluster.

[0051] Example 1: like Figure 1 As shown, this invention discloses a power grid dispatch optimization method based on time-series correlation flexible resource identification, comprising the following steps: Step 1: Collect historical building energy consumption, power grid operation, and meteorological data. After preprocessing and clustering, the data is divided into multiple typical days. The building energy consumption data includes building load, and the power grid operation data includes total power grid load. Specifically, in this embodiment, summer and winter data on building energy consumption, power grid operation, and meteorology are collected and clustered to obtain four types of typical daytime time series data, providing a standardized analytical basis for subsequent evaluation.

[0052] In step 1, the following three types of time-series data are collected for buildings in the target area during summer and winter: building energy consumption, power grid operation, and meteorology. Then, based on calendar information, the dates of the year are divided into four predefined categories: Summer Workdays (SWD), Summer Holidays (SHD), Winter Workdays (WWD), and Winter Holidays (WHD).

[0053] Building energy consumption data, including the total active power sequence of a building collected through a building energy management system or smart meters, i.e., building load data, is denoted as... The recommended time resolution is 15 minutes or 1 hour. This is a typical day category.

[0054] Power grid operation data, including the total active power load sequence of the regional power grid obtained from the power dispatch center. That is, the total load of the power grid and the time-of-use electricity price sequence. and carbon emission intensity sequence The time resolution is typically 30 minutes.

[0055] Meteorological data, including outdoor dry-bulb temperature series obtained from local weather stations. Relative humidity sequence and total solar radiation intensity sequence The time resolution is typically 15 minutes or 1 hour.

[0056] The summer and winter data were cleaned and aligned, including imputation of missing values ​​and handling of outliers. To preserve complete information on building energy consumption patterns, this invention directly uses the building daily load curve as the feature vector for cluster analysis. The summer and winter building load data were segmented by day to obtain... Daily load curves, each curve contains Each sampling point (e.g., 96 points correspond to a 15-minute resolution). Daily building load vector Represented as: (1) in, The sampling time points within a day , as well as This represents the total active power of the building at each sampling time point.

[0057] For each date category subset (in The load curves for all days within this subset are then averaged arithmetically to obtain the typical daily load curve for that category. : (2) in Representing a subset The number of days included.

[0058] For each category, a representative daily load curve is selected from that subset. The selection criterion is that this curve resembles the average curve. The Euclidean distance is the smallest: (3) in This represents the Euclidean distance. From this, four representative days are determined: For each selected representative day Other data from that day were extracted to form a multi-source heterogeneous dataset. These four datasets constituted the baseline scenario dataset for subsequent evaluation of all buildings.

[0059] Step 2: Construct a grid-friendly building evaluation system that includes multiple evaluation indicators, and calculate the index values ​​of each evaluation indicator based on the collected data from the multiple typical days; the evaluation indicators include a grid load inertia coefficient used to quantify the morphological similarity between the building load curve and the total grid load curve; like Figure 2As shown, the evaluation system for grid-friendly buildings includes five primary indicators: energy consumption characteristics, interaction potential, environmental protection and low carbon emissions, inertia coefficient, and geographic space. The energy consumption characteristics include the following secondary indicators: daily load factor, daily peak-to-valley difference, and building energy efficiency ratio; The interactive potential includes the following secondary indicators: flexible load ratio, typical daily difference coefficient, and level of intelligence; The environmental protection and low carbon standards include the following secondary indicators: renewable energy utilization rate and carbon emission intensity per unit area; The inertia coefficient includes the following secondary indicators: temperature inertia coefficient, electricity price inertia coefficient, grid load inertia coefficient, and carbon emission factor inertia coefficient; The geospatial data includes the following secondary indicators: building cluster distribution density, regional power grid capacity margin, and utilization rate of vacant space.

[0060] Preferably, the inertia coefficients of each building load in relation to temperature, electricity price, grid load, and carbon emission factors are calculated based on a dynamic regularization algorithm, and the evaluation index scores of each building are calculated; the specific method for calculating the inertia coefficients is as follows: (4) in, For the first s A typical daily normalized building load sequence For the first s The normalized sequence of external factors for a typical day includes external environmental signals such as outdoor temperature sequence, electricity price sequence, total grid load sequence, and carbon emission factor sequence. For the first The time weight of a typical day represents the total duration of the original data sample represented by that typical day. This is the DTW distance calculation function; and These are the lengths of the building load sequence and the external drive time series, respectively.

[0061] Taking temperature as an example, the building load data and outdoor temperature data are normalized to obtain two sets of time series. , The DTW distance between the two sets of data is calculated as the inertia coefficient. The larger the inertia coefficient, the greater the difference in the changing trends between the building load curve and the outdoor temperature curve, and the less the building load is affected by the outdoor temperature. The specific calculation method is as follows: (5) In the formula: For the first The time weight of a typical day represents the total duration of the original data sample represented by that typical day. This is the DTW distance calculation function; and These are the normalized load data and outdoor temperature time series for the s-th typical day, respectively. and These represent the lengths of the load data and the outdoor temperature time series, respectively.

[0062] The DTW distance calculation process in this invention mainly consists of two parts: constructing the cumulative distance matrix and finding the shortest path. For building load time series... X =( x 1, x 2,…, x i ,…, x n ), and external factor data time series Y =( y 1, y 2,…, y i ,…, y m ), the DTW distance between the two The calculation method is as follows: (6) (7) (8) (9) (10) (11) (12) In the formula: d ( x i , y j ) represents time series x i and y j The distance between them; C This is the cumulative distance matrix. C ( i , j )express arrive The minimum cumulative distance is obtained by recursion as shown in formula (6); formulas (8) and (9) are the cumulative distance matrices respectively. C Boundary conditions for the first row and first column; formula ((10) is from Back to The backtracking rules are because the DTW algorithm is based on...C ( n , m ) Back to C (1,1) is used to find the optimal alignment path, and the final DTW distance is obtained. It is the bottom right element of the cumulative distance matrix. C ( n , m As shown in formula (11).

[0063] It should be noted that formula (12) represents the path bending length constraint that needs to be satisfied when calculating each inertia coefficient. This constraint can effectively limit the maximum number of consecutive corresponding points between time series data. The maximum path curvature length is adaptively determined by the maximum influence time of the external factor sequence corresponding to this inertia coefficient on the building load sequence. Preferably, for the temperature inertia coefficient, its corresponding... A 2-hour timeframe is used in summer and a 4-hour timeframe in winter because the air conditioning thermal response cycle is shorter in summer, while the building envelope has a higher thermal inertia in winter. Regarding the electricity price inertia coefficient and the carbon emission factor inertia coefficient... Take 1 hour as the settlement cycle for the grid time-of-use pricing and daily carbon emission scheduling; for the grid load inertia coefficient, Take 0.

[0064] Formula (12) introduces a maximum path bending length constraint mechanism based on the interaction cycle between building thermophysical characteristics and power grid, which ensures that the alignment path between the load time series and the external factor time series will not have a cross-cycle forced matching that is completely contrary to the building thermal inertia and the power grid operation cycle. At the same time, it avoids a large number of cross-cycle calculations and adapts to the rapid identification needs of batch buildings in the power grid dispatch scenario.

[0065] The specific calculation methods for each secondary indicator are as follows: 1. Energy consumption characteristics (1) Daily load factor B 1 The daily load factor is the ratio of the building's average daily load to its maximum daily load, used to characterize the stability of the building's load. The calculation formula is as follows: (13) In the formula: For the first The representative duration of a typical day, that is, the total duration of the raw data represented by that typical day within the statistical period; and For the first The average and maximum values ​​of a typical daily load.

[0066] (2) Daily peak-valley difference rate B 2 The daily peak-to-valley difference rate is the ratio of the daily peak-to-valley difference of the building load to the maximum load. The calculation formula is as follows: (14) In the formula: , These represent the maximum and minimum daily loads of the building on the s-th typical day, respectively.

[0067] (3) Building energy efficiency ratio B 3 The building energy efficiency ratio is the ratio of a building's energy consumption to its total floor area. The calculation formula is as follows: (15) In the formula: This represents the total area of ​​the building, expressed in square meters.

[0068] 2. Interactive Potential Indicators (1) Proportion of flexible load B 4 The proportion of flexible loads is the average proportion of flexible loads such as air conditioning, lighting, and energy storage in the building's peak load. The calculation formula is as follows: (16) In the formula: Indicates the peak period specified by the power grid; , , and They are respectively t The power consumption of air conditioning, lighting, energy storage, and total building power at any given time.

[0069] (2) Typical daily variation coefficient B 5 The greater the difference in curve shape between typical days, the stronger the user's ability to adjust electricity consumption patterns at different times. This difference can be described using the Spearman correlation coefficient between the load data of each typical day. (17) In the formula: The number of typical days; , They represent the first n , m A typical daily load data vector; This represents the Spearman correlation coefficient for load curve data on different typical days.

[0070] (3) Level of intelligence B 6 The level of building intelligence is primarily assessed and scored based on three aspects: the deployment level of intelligent equipment systems, data and communication capabilities, and user-friendliness. The deployment level of intelligent equipment systems can be comprehensively evaluated by the number, integration level, and coverage of building automation system modules and intelligent sensors. Data and communication capabilities can be comprehensively evaluated by real-time data acquisition capabilities, communication protocol compatibility, and data storage and analysis capabilities. User-friendliness mainly depends on whether the user interface design for operating the intelligent system is simple and intuitive, and whether user feedback channels are readily available.

[0071] 3. Environmental protection and low-carbon indicators (1) Renewable energy utilization rate B 7 The renewable energy utilization rate is the ratio of the amount of renewable energy used in a building to the building's total energy consumption, calculated using the following formula: (18) In the formula: For the first s A typical day t Renewable energy consumption during the period For the first s A typical day t Total building energy consumption during a given period.

[0072] (2) Carbon emission intensity per unit area B 8 Building carbon intensity is defined as the ratio of a building's total carbon emissions over its entire life cycle to its total load. The formula for calculating building carbon intensity is as follows: (19) (20) (twenty one) In the formula: The total carbon emissions throughout the building's entire life cycle; These are the carbon emissions during the building operation, construction, transportation, demolition, and recycling phases; Total building load; for t The building's carbon emission factor for electricity consumption at any given time; For the first z Carbon emissions per unit consumption of a certain fuel; For building consumption z The consumption of this type of fuel.

[0073] 4. Environmental inertia index (1) Temperature inertia coefficient B 9 The temperature inertia coefficient is calculated using the following formula: (twenty two) The maximum bending length r of DTW is 2 hours in summer and 4 hours in winter. (2) Electricity price inertia coefficient B 10 (twenty three) In the formula: This is the normalized electricity price time series.

[0074] (3) Net load inertia coefficient B 11 The grid load inertia coefficient is used to quantify the morphological similarity between building load curves and the total grid load curve. When building loads and grid loads highly overlap in time series, it indicates that the building's electricity consumption behavior is one of the main factors contributing to grid peak loads. Implementing demand response measures for such buildings (e.g., reducing their flexible loads during peak hours) can most directly reduce the overall peak load of the grid, thus achieving effective peak shaving. Simultaneously, load filling for such buildings during grid off-peak periods (e.g., energy storage charging) can also significantly increase off-peak loads, achieving a valley-filling effect. This paper uses the inertia coefficient to calculate grid load similarity, while setting the maximum path curvature length to 0 to ensure synchronization. The formula for calculating the grid load overlap coefficient is as follows: (twenty four) In the formula: This is the time series of the normalized overall load of the power grid.

[0075] (4) Carbon emission factor inertia coefficient B 12 (25) In the formula: This is the normalized carbon emission factor sequence.

[0076] 5. Spatial location indicators (1) Building cluster distribution density B 13 Building density in an area refers to the number of buildings or the area of ​​a unit of area. This is because areas with higher density can achieve greater carbon reduction potential through clustered demand response. The calculation formula is as follows: (26) In the formula: Represents architecture Located in the region ; The building area; This refers to the total area occupied by the building.

[0077] (2) Regional distribution network capacity margin B 14 The regional distribution network capacity margin is defined as the ratio of the remaining capacity of the distribution network where the building is located to the peak load. The calculation formula is as follows: (27) In the formula: The capacity of the tie line for connecting the building to the power distribution network.

[0078] (3) Utilization rate of free space B 15 The utilization rate of vacant space is defined as the ratio of vacant space area to the total building area, and the calculation formula is as follows: (28) In the formula: This refers to the vacant area of ​​the building.

[0079] Step 3: Calculate the correlation influence weight and statistical characteristic weight of each evaluation index, and integrate them into the comprehensive weight coefficient of each evaluation index. The statistical characteristic weight is calculated using the improved CRITIC method with the comparison intensity corrected by the network load sensitivity coefficient. The network load sensitivity coefficient is used to characterize the correlation degree of each evaluation index with respect to the network load inertia coefficient. Preferably, in step 3, calculating the correlation influence weights based on the DANP method specifically includes: using expert discussions, questionnaires, and other methods to score the influence relationships between each indicator using a 0-9 scale to form an initial direct relationship matrix. O : (29) In the formula: The number of evaluation indicators; Indicators For indicators The degree of influence, where the diagonal elements are 0.

[0080] For direct relation matrix O The standardized influence matrix is ​​obtained by standardization. E And calculate the comprehensive impact matrix. T : (30) (31) in, for m An identity matrix of order 1.

[0081] Based on the calculated comprehensive impact matrix Calculate the measure of the degree of influence of each indicator in the evaluation system: influence degree Degree of influence Centrality causal degree Furthermore, a causal relationship diagram was drawn based on the centrality and causality constructs, and the correlations between the indicators were analyzed: (32) (33) (34) (35) In the formula, It is a comprehensive influence matrix T The element in represents the first element. The first indicator for the first The combined impact of each indicator.

[0082] The above describes the DEMATEL method, which constructs a direct influence matrix based on the logical relationships between factors in the system. This matrix is ​​used to calculate the influence and affected degree of each factor on other factors, thereby deriving the causality and centrality of each factor. The ANP structure is as follows: Figure 3 The algorithm can be divided into a control layer and a network layer. The control layer can be further divided into a target layer and a criterion layer. In the criterion layer, each criterion is independent and governed only by the target layer. Each criterion governs not simply independent elements, but a network structure where elements are interdependent and mutually influential. The network layer consists of multiple independent elements or sets of elements, each of which can influence or be influenced by other elements. Specialized software can be used to calculate the limiting hypermatrix in the ANP method, thus obtaining the limiting relative ranking of each dimension and indicator. The final correlation influence weight of each indicator is any column of the limiting hypermatrix.

[0083] The steps for calculating the statistical feature weights based on the improved CRITIC method are as follows: (1) To n One building to be evaluated, used m The evaluation score is calculated for each evaluation indicator. The evaluation indicator data is then forward standardized to obtain the standardized score for the [number]th indicator. i The first building j Evaluation scores for each indicator .

[0084] (2) Calculate the ratio of the mean difference to the mean. As an indicator j The original contrast intensity is obtained, and a sensitivity coefficient is introduced to correct the original contrast intensity to obtain the corrected contrast intensity; The traditional CRITIC method calculates comparative strength solely based on the statistical dispersion of indicator data, assuming that the numerical dispersion differences of all indicators contribute equally to the grid-friendly assessment. However, in real-world scenarios, fluctuations in some indicator values ​​can accurately reflect a building's grid-load matching and peak-shaving / valley-filling potential; while the numerical dispersion of other indicators is merely random fluctuation unrelated to grid-load synergy. The traditional CRITIC method fails to distinguish these differences, treating both types of fluctuations equally. This leads to the erroneous amplification of noise differences unrelated to grid-friendly performance, weakening the physical relevance of the comparative strength.

[0085] This invention introduces the aforementioned load inertia coefficient to characterize the load matching capability of a single building, and then calculates the correlation between each evaluation index and the load inertia coefficient, defining it as the load sensitivity coefficient. This sensitivity coefficient is used to adjust the contrast intensity of each index; the higher the correlation and the stronger the load differentiation capability, the greater the amplification of the contrast intensity, ultimately resulting in a higher weight in the weighting process. The contrast intensity after correction by the load sensitivity coefficient is calculated as follows: (36) (37) (38) in, For the first j The original contrast strength of each indicator For the first i The building in j Standardized positive scores on each indicator; For the first j The average of all building samples for each indicator; For the first j The load sensitivity coefficient of each indicator is used to measure the degree of correlation between the indicator and the building's load matching ability. The larger the value, the better the indicator can reflect the building's load matching potential. The Pearson correlation coefficient calculation function is used to measure the correlation between two indicators; The correction coefficient is , and satisfies This is used to control the strength of the correction and prevent the weights from being over-amplified; This is the corrected contrast intensity.

[0086] Existing CRITIC methods typically use the Pearson correlation coefficient to measure the correlation between indicators when calculating conflict, which leads to the neglect of multicollinearity among indicators and may underestimate information redundancy. To address this issue, this invention introduces a double logarithmic multiple regression multiple correlation coefficient to calculate conflict. By transforming each indicator into a dimensionless elasticity value through logarithmic transformation, a multiple regression model can simultaneously capture the indicators. jThe method's conflict measure, which considers the joint dependencies of all other indicators, exhibits dimensional consistency, nonlinear adaptability, and multivariate structural integrity, making it more suitable for the actual situation in the evaluation of grid-friendly buildings, where the indicators have complex dimensions, nonlinear relationships, and significant multicollinearity.

[0087] Calculate the double log-regression multiple correlation coefficient Conflicts as indicators.

[0088] (39) (40) In the formula: This is the constant term in the regression model. Represents the regression coefficients of each explanatory variable, reflecting the influence of other indicators on the first variable. j The degree of influence of each indicator; Refers to the first The average value of each indicator; and Each refers to The average and predicted values.

[0089] (3) Multiplication is combined to form information carrying capacity The information carrying capacity is normalized to obtain the statistical feature weights for each scenario. Based on the weight ratio of various typical daily scenarios in the annual power grid dispatch, the statistical feature weights of the single scenarios are weighted and fused to finally obtain statistical feature weights that adapt to the dispatch requirements of all scenarios. .

[0090] (41) (42) After obtaining the correlation influence weight and statistical feature weight, this invention uses the following method to calculate the comprehensive weight coefficient, and its mathematical model is as follows: (43) In the formula: It is the importance coefficient of the method, reflecting the preference for the weighting method; Indicates the first The first weighting method obtained The weight of each indicator; Let be the desired overall weight, and be the decision variable; and The first The maximum and minimum values ​​of each indicator in each weighting method.

[0091] Step 4: Construct an evaluation model that incorporates cloud droplet overlap similarity. Input the evaluation index values ​​corresponding to each building and the comprehensive weight coefficients corresponding to each evaluation index into the model for sorting. The evaluation model adaptively selects the total number of matching cloud droplet pairs participating in the similarity calculation based on the uncertainty of the cloud model. In step 4, the specific steps of the building ranking method based on the evaluation model that incorporates cloud droplet overlap similarity are as follows: The cloud model is used to quantify the standardized evaluation results to obtain the cloud decision-making standard matrix. ,in, m The number of buildings being evaluated. n The number of evaluation indicators is determined; and the weighting matrix is ​​calculated using equation (45). .

[0092] (44) (45) In the formula: Represents architecture Medium evaluation indicators The standard evaluation cloud; among them, , , Buildings Medium evaluation indicators The expected value, entropy, and hyperentropy of the standardized score are three indicators used to characterize the uncertainty and ambiguity of the score. For architecture Medium evaluation indicators Weighted evaluation cloud; Evaluation indicators The overall weight.

[0093] Determining the positive and negative ideal solution clouds: The ideal decision cloud for each indicator is determined by comparing the expected value, entropy, and hyperentropy in order of priority. First, the expected value is compared, with the one with the larger expected value being better. If the expected values ​​are equal, then the entropy is compared, with the one with the smaller entropy being better. If the entropy is also equal, then the hyperentropy is compared, with the one with the smaller hyperentropy being better.

[0094] (46) (47) in, For a positive ideal decision cloud matrix, For the first j The positive ideal solution of each evaluation index; For negative ideal decision cloud matrix, For the first j The negative ideal solution cloud for each evaluation indicator.

[0095] This invention proposes a cloud droplet overlap similarity method to calculate the similarity between the weighted evaluation clouds of each building and the clouds of the positive and negative ideal solutions. This method does not use Euclidean distance, but rather measures cloud similarity based on the degree of overlap of cloud droplet membership, thus more sensitively capturing overlap differences between distributions.

[0096] For each pair of matched cloud droplets ,in, , Define a set of cloud droplets; define two clouds. , Cloud droplet overlap similarity It can be represented as: (48) (49) In the formula: , Two cloud models are to be compared; For the first k Sample the index values ​​corresponding to the matching cloud droplets; K The total number of matching cloud droplet pairs participating in the similarity calculation; , They are respectively , Membership function; Sampled value The membership degree corresponding to the cloud model, , These are the expectation and entropy of the cloud model, respectively. In formula (48), the numerator represents the smaller membership degree of the two cloud droplets at the same location, i.e., the degree of overlap; the denominator represents the larger membership degree of the two cloud droplets at the same location, i.e., the coverage area.

[0097] To avoid insufficient accuracy or computational waste caused by a fixed number of cloud droplets, this invention proposes to adaptively determine the total number of matching cloud droplet pairs participating in similarity calculation based on the entropy (En) and hyperentropy (He) of the cloud model. K The higher the entropy and hyperentropy, the greater the uncertainty of the data, and the more cloud droplets are needed to stably represent its distribution.

[0098] The total number of matching cloud droplet pairs participating in similarity calculation K The calculation formula is as follows: (50) In the formula: The preset minimum entropy value is preferably 0.01; This is a proportionality coefficient, preferably 500 in this embodiment; To match the maximum number of cloud droplets, in this embodiment, it is preferably 2000; This indicates rounding up to the nearest integer.

[0099] Experiments have verified that for clouds with low uncertainty (i.e. En + He <0.1), adaptive K Between 200 and 500; for high uncertainty clouds (i.e. En + He >0.5), K It can reach 1500~2000, thus maintaining the stability of similarity estimation while ensuring computational efficiency.

[0100] For architecture i In terms of indicators j Weighted evaluation cloud Interpreting the ideal cloud First, determine the adaptive number of cloud droplets in the unified sampling point set, and then select two clouds. K The larger value. Generated using a two-dimensional cloud normal generator. K Each cloud droplet yields a sampling point. And calculate the membership degree. Define the weighted evaluation cloud. Interpreting the ideal cloud The cloud droplet overlap similarity is: (51) Similarly, Negative ideals and cloud interpretation Cloud droplet overlap similarity The calculation method is the same, specifically: (52) in, To weighted evaluate cloud at sampling points Membership degree of the location; , Positive and negative ideal solution clouds at sampling points Membership degree of the location; , These represent the cloud droplet overlap similarity between the weighted evaluation cloud and the positive and negative ideal solution clouds, respectively.

[0101] Because cloud droplet generation is random, the similarity obtained from a single calculation fluctuates. To obtain robust similarity estimates and quantify their uncertainty, this invention repeats cloud droplet generation and similarity calculation for each cloud pair for a total of [number missing]. In this embodiment, 30 times is preferred, and the average of the samples is taken as the final similarity of cloud droplet overlap. The overlap similarity between the final cloud droplets and the ideal solution cloud is: (53) Simultaneously calculate the sample standard deviation. And give a 90% confidence interval: (54) The same processing is applied to the negative ideal solution cloud to obtain the final cloud droplet overlap similarity with the negative ideal solution cloud. and its confidence interval, building i Mean of overall similarity with positive and negative ideal solution clouds , They are respectively: (55) architecture i Proximity for: (56) The buildings are ranked based on their proximity and their confidence intervals, specifically including: First, based on the proximity scores of each building, the buildings are initially sorted in descending order from highest to lowest. After the initial sorting, the proximity scores of adjacent buildings in the sorted sequence are iterated sequentially. If the difference in proximity between two adjacent buildings is less than a preset threshold (preferably 0.05), relying solely on a single value cannot reliably distinguish the scheduling quality of the two buildings. In this case, a robust comparison verification is performed using confidence intervals. The specific judgment rule is: if the difference in proximity between two adjacent buildings is less than a preset threshold (preferably 0.05), the difference in proximity between two adjacent buildings cannot reliably distinguish the scheduling quality of the two buildings. In this case, a robust comparison verification is performed using confidence intervals. The specific judgment rule is: if the building ranked higher... i The lower bound of the confidence interval for proximity is greater than that of buildings ranked lower. k If the upper bound of the confidence interval for proximity is reached, then the original ranking is maintained, and the building can be determined. i Superior to architecture k If the confidence intervals of the two buildings overlap, the ranking result is not statistically significant, and the two buildings are marked as "ranking uncertain" and are placed in the same priority echelon during scheduling. If the difference in proximity between two adjacent buildings is greater than or equal to the preset threshold, the initial single-point ranking result is directly retained without the need for secondary verification of the confidence interval.

[0102] Step 5: Based on the ranking results of each building, select the flexible building clusters that participate in demand response, and generate the scheduling strategy for each building in the flexible building cluster by combining the power grid load regulation requirements of each typical day.

[0103] The specific steps are as follows: Step 5.1: Screening of flexible building clusters; Based on the ranking results of each building calculated in step 4, select the top-ranked buildings with a preset proportion (preferably 20%) or a proximity greater than the selection threshold. The buildings constitute a cluster of flexible buildings that respond to regional demand. The selection threshold is calculated using the following formula:

[0104] in, This represents the maximum approximation of flexibility potential among all buildings; The selection factor is set to a value within the range of 0.6. 0.9, which can be dynamically adjusted according to the power grid's peak-shaving needs and the total number of buildings in the region.

[0105] The adjustable peak-shaving capacity and adjustable valley-filling capacity of each building in the cluster are statistically analyzed, and the total adjustable peak-shaving capacity of the cluster is calculated using the following formula. Total adjustable valley filling capacity :

[0106]

[0107] in, Represents architecture i A collection belonging to flexible building clusters; For the first s Typical of the sun i The maximum adjustable peak capacity of the building; For the first s Typical of the sun i The maximum adjustable valley capacity of the building.

[0108] Step 5.2: Calculate the load allocation weight of each building in the flexible building cluster, and allocate the power grid load regulation demand of each typical day to each building according to the load allocation weight of each building; The load allocation weight of each building is the ratio of the product of the building's proximity and adjustable load capacity to the sum of the products of the comprehensive proximity and adjustable load capacity of all buildings in the flexible building cluster.

[0109] Specifically, the following steps are included: Step 5.2.1: Generation of peak shaving scheduling strategy; For the peak power grid period of a typical day of type s Based on the total peak shaving demand issued by the power grid dispatching station, the peak shaving load instructions are allocated according to the principle of "potential priority and capacity constraint" by using the product of building proximity and adjustable peak shaving capacity as the allocation weight.

[0110] The physical meaning of this allocation logic is that the higher the overall proximity, the better the building's overall performance in terms of grid load matching, load adjustability, response stability, and low-carbon attributes. The higher the scheduling benefits under the same capacity, the higher the proportion of control load is allocated, so the ranking result directly determines the allocation priority of scheduling resources.

[0111] The peak load allocation weight for each building is calculated using the following formula:

[0112]

[0113] in, For the first s Typical of the sun i The peak-shaving allocation weight for each building.

[0114] Based on the total peak shaving demand of the power grid and the allocation weight, calculate the peak-hour peak shaving command value for each building:

[0115]

[0116] in, For the first s Typical of the sun i The peak shaving command value for the building; the above formula is a capacity limit constraint, which means that if the calculation result exceeds the maximum adjustable capacity of the building, the upper limit value is taken, and the remaining peak shaving demand is redistributed to the remaining buildings that have not exceeded the limit according to the same weight rule, until the total peak shaving demand is completely distributed.

[0117] Step 5.2.2: Generation of valley-filling scheduling strategy; Regarding the first s During the off-peak hours of a typical day, based on the total grid demand for valley filling. Similarly, the product of comprehensive proximity and adjustable valley filling capacity is used as the allocation weight to generate a valley filling charging scheduling strategy during off-peak hours.

[0118]

[0119]

[0120] In the formula, For the first s Typical of the sun i The weighting of valley filling for each building.

[0121] Based on the total grid demand for valley filling and the allocation weight, calculate the valley filling charging instruction value for each building during off-peak hours:

[0122]

[0123] in, For the first s Typical of the sun i The peak reduction command value for the building; the excess part adopts the same secondary allocation rule as the peak reduction.

[0124] Step 5.3: Strategy Issuance and Execution: The peak-shaving and valley-filling scheduling strategies for typical days are integrated into a regional power grid collaborative scheduling strategy, which is then sent to the power grid dispatching master station through a dedicated power communication network. At the same time, the time-sharing control instructions for the corresponding buildings are sent to the building energy management systems of each building. The system automatically controls the operating power and start / stop status of flexible loads such as air conditioning, energy storage, and lighting. Under the premise of ensuring the energy comfort of the building, the system completes the power grid demand response control instructions, ultimately achieving the technical goals of peak shaving and valley filling, smoothing power grid power fluctuations, and reducing losses from frequent equipment start-stop.

[0125] Example 2: This embodiment uses 169 public buildings in a large city in my country as examples to implement the identification method described in this invention.

[0126] Complete data for the target buildings during the summer (June-August) and winter (December-February) of 2023 were collected, including: hourly building load, half-hourly grid load and electricity price, hourly temperature, humidity, and solar radiation data from meteorological stations. After cleaning and alignment, the daily load curves for each day of the summer and winter seasons for each building were used as features, and K-means clustering was performed. Based on the main date attributes of the samples within each cluster, four typical day categories were automatically identified. Subsequently, the actual date closest to the cluster center was selected from each cluster as the typical day of that category, and the corresponding aligned complete multi-source time-series dataset was extracted as the benchmark scenario for subsequent unified evaluation.

[0127] Figure 4 The distribution of indicator weights under different weighting methods in this embodiment shows that the present invention, by improving the multiple correlation coefficient in the CRITIC method, can more comprehensively measure the information content of the indicators; at the same time, the introduction of the mean difference to consider the influence of the indicator average value can amplify the weight difference between indicators from the two dimensions of conflict and contrast intensity. Regarding the calculation of correlation influence weights, the DANP method is an improved analysis method based on Dematel influence relationships, which can allocate weights based on the mutual influence relationships between indicators, resulting in more reasonable results. For example, for the classic demand response indicator, the daily peak-to-valley difference B2, the DANP method can assign it a higher and more reasonable weight. There are significant differences in the evaluation perspectives of correlation influence weights and statistical characteristic weights. The improved minimum range method combined weighting used in this invention differs from the standard range method mainly in the setting of preference coefficients and the constraint of weight range. The weight distribution obtained by the method of this invention is closer to the correlation influence weights, preserving the rationality of the correlation influence weights while fully considering the consistency with objective data.

[0128] After implementing this invention, the ranking results of the low-carbon interaction potential of various types of buildings are as follows: Figure 5As shown, sports buildings and government office buildings ranked highly, with similarity scores of 0.78 and 0.72 respectively, while medical and health buildings ranked low, with a similarity score of 0.41. Furthermore, sensitivity analysis of the evaluation method shows that the sensitivity of the method in this invention is on average 14.71% higher than that of the traditional TOPSIS method.

[0129] 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.

[0130] 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.

[0131] 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.

[0132] 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.

[0133] 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 power grid dispatch optimization method based on time-series correlation flexible resource identification, characterized in that, The method includes the following steps: Step 1: Collect historical building energy consumption, power grid operation, and meteorological data. After preprocessing and clustering, the data is divided into multiple typical days. The building energy consumption data includes building load, and the power grid operation data includes total power grid load. Step 2: Construct a grid-friendly building evaluation system that includes multiple evaluation indicators, and calculate the index values ​​of each evaluation indicator based on the collected data from the multiple typical days; the evaluation indicators include a grid load inertia coefficient used to quantify the morphological similarity between the building load curve and the total grid load curve; Step 3: Calculate the correlation influence weight and statistical characteristic weight of each evaluation index, and integrate them into the comprehensive weight coefficient of each evaluation index. The statistical characteristic weight is calculated using the improved CRITIC method with the comparison intensity corrected by the network load sensitivity coefficient. The network load sensitivity coefficient is used to characterize the correlation degree of each evaluation index with respect to the network load inertia coefficient. Step 4: Construct an evaluation model that incorporates cloud droplet overlap similarity. Input the evaluation index values ​​corresponding to each building and the comprehensive weight coefficients corresponding to each evaluation index into the model for sorting. The evaluation model adaptively selects the total number of matching cloud droplet pairs participating in the similarity calculation based on the uncertainty of the cloud model. Step 5: Based on the ranking results of each building, select the flexible building clusters that participate in demand response, and generate the scheduling strategy for each building in the flexible building cluster by combining the power grid load regulation requirements of each typical day.

2. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 1, characterized in that: The evaluation indicators also include temperature inertia coefficient, electricity price inertia coefficient, and carbon emission factor inertia coefficient; each inertia coefficient is calculated based on a dynamic time warping algorithm with path bending length constraint, and the path bending length constraint is adaptively determined based on the maximum influence time of the external factor sequence corresponding to the inertia coefficient on the building load sequence; the external factor sequence corresponding to the temperature inertia coefficient is the temperature sequence, the external factor sequence corresponding to the electricity price inertia coefficient is the electricity price sequence, and the external factor sequence corresponding to the carbon emission factor inertia coefficient is the carbon emission factor sequence.

3. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 1, characterized in that: In step 3, the contrast intensity corrected for the net load sensitivity coefficient in the improved CRITIC method is determined as follows: in, The contrast intensity is adjusted for the net load sensitivity coefficient; For the first j The original contrast strength of each indicator For the first i The building in j Standardized positive scores on each indicator; For the first j The average of all building samples for each indicator; n The total number of buildings to be evaluated; For the first j The network load sensitivity coefficient of each indicator; The correction coefficient is , and satisfies .

4. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 3, characterized in that: The first j The network load sensitivity coefficients for each indicator are determined as follows: in, The function for calculating the Pearson correlation coefficient; For the first i The inertia coefficient of the load on a building.

5. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 1, characterized in that: In step 3, the comprehensive weight coefficient of each evaluation indicator is determined as follows: in, For the first The comprehensive weighting coefficient of each evaluation indicator is the decision variable; For the first i Importance coefficient of each weighting method; Indicates the first The first weighting method is used to calculate the first weighting method. The weight of each indicator; and The first The maximum and minimum values ​​of each indicator in each weighting method; m This represents the total number of weighting methods. n This represents the total number of indicators.

6. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 1, characterized in that: In step 4, the cloud droplet overlap similarity is used to measure the similarity of clouds based on the degree of overlap of cloud droplet membership, and is specifically determined as follows: in, for , Cloud droplet overlap similarity between two cloud models to be compared; For the first k Sample the index values ​​corresponding to the matching cloud droplets; , They are respectively , Membership function; K This represents the total number of matching cloud droplet pairs participating in the similarity calculation.

7. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 6, characterized in that: The membership function is determined as follows: in, For indicator value x Membership degree in the cloud model; , These represent the expectation and entropy of the cloud model, respectively.

8. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 6, characterized in that: In step 4, the total number of matching cloud droplet pairs participating in the similarity calculation is determined as follows: in, This is the preset minimum entropy value in the cloud model; This is the proportionality coefficient; To match the maximum number of cloud droplets; , These are the entropy and hyperentropy of the cloud model, respectively. This indicates rounding up to the nearest integer.

9. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 6, characterized in that: The step of inputting the evaluation index values ​​corresponding to each building and the comprehensive weight coefficient corresponding to each evaluation index into the model for sorting specifically includes: The cloud model is used to quantify the standardized evaluation index values, construct the standard evaluation cloud corresponding to each evaluation index of each building, and calculate the weighted evaluation cloud corresponding to each evaluation index of each building by combining the comprehensive weight coefficients corresponding to each evaluation index. Based on the cloud model feature parameters of the weighted evaluation cloud, the positive ideal solution cloud and negative ideal solution cloud corresponding to each evaluation index are calculated. For each building's weighted evaluation cloud under each evaluation index, calculate its cloud droplet overlap similarity with the positive and negative ideal solution clouds respectively. After multiple calculations, take the average value as the final cloud droplet overlap similarity, and calculate the confidence interval of the final cloud droplet overlap similarity. For each building, the final cloud droplet overlap similarity of each evaluation index is summarized, and the overall similarity, closeness and confidence interval of each building with the positive and negative ideal solution clouds are calculated. The buildings are ranked based on their proximity and their confidence intervals.

10. The power grid dispatch optimization method based on time-series correlation flexible resource identification according to claim 9, characterized in that: In step 5, the scheduling strategy for each building in the flexible building cluster is determined as follows: Calculate the load allocation weight of each building in the flexible building cluster, and allocate the power grid load regulation demand of each typical day to each building according to the load allocation weight of each building; wherein, the load allocation weight of each building is the ratio of the product of the building's proximity and adjustable load capacity to the sum of the products of the comprehensive proximity and adjustable load capacity of all buildings in the flexible building cluster.

11. A power grid dispatch optimization system based on time-series correlation flexible resource identification according to the method of any one of claims 1-10, comprising a data acquisition module, an evaluation system construction module, an evaluation index weight calculation module, a flexible building ranking module, and a flexible building cluster dispatch strategy generation module, characterized in that: The data acquisition module collects historical building energy consumption, power grid operation, and meteorological data. After preprocessing and clustering, the data is divided into multiple typical days. The building energy consumption data includes building load, and the power grid operation data includes total power grid load. The evaluation system construction module constructs a grid-friendly building evaluation system that includes multiple evaluation indicators, and calculates the index values ​​of each evaluation indicator based on the collected data of the multiple typical days; the evaluation indicators include the grid load inertia coefficient, which is used to quantify the morphological similarity between the building load curve and the total grid load curve. The evaluation index weight calculation module calculates the correlation influence weight and statistical characteristic weight of each evaluation index, and integrates them into the comprehensive weight coefficient of each evaluation index. The statistical characteristic weight is calculated using the improved CRITIC method with the comparison intensity corrected by the network load sensitivity coefficient. The network load sensitivity coefficient is used to characterize the degree of correlation between each evaluation index and the network load inertia coefficient. The flexible building sorting module constructs an evaluation model that incorporates cloud droplet overlap similarity. The evaluation index values ​​corresponding to each building and the comprehensive weight coefficients corresponding to each evaluation index are input into the model for sorting. The evaluation model adaptively selects the total number of matching cloud droplet pairs participating in the similarity calculation based on the uncertainty of the cloud model. The flexible building cluster scheduling strategy generation module selects flexible building clusters participating in demand response based on the ranking results of each building, and generates the scheduling strategy for each building in the flexible building cluster by combining the power grid load control requirements of each typical day.

12. 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 1-10.

13. 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 1-10.

Images