Power grid intelligent inflection point identification method based on power load increase

Abstract

The invention discloses a power grid intelligent inflection point identification method based on power load increase, and relates to the technical field of power grid loads, and the method comprises the steps: collecting power data, and obtaining a historical load data sequence; based on the historical load data sequence, obtaining a dual-feature vector containing a load growth value and a load growth rate; defining inflection point indexes and obtaining an inflection point index sequence; performing standardization processing on the double feature vectors, and performing clustering analysis to obtain clustering tags; and based on the inflection point index sequence and the clustering label, uniformly identifying the key form of load increase, and obtaining a differentiated power grid planning strategy. According to the method, the internal mode of load increase can be objectively identified in a data-driven manner by constructing the double feature vectors and carrying out standardized processing and clustering analysis, different load increase key forms can be accurately identified, and a scientific basis is provided for obtaining differentiated power grid planning strategies.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid load, more specifically, the present application relates to a power grid intelligent inflection point identification method based on power load growth. BACKGROUND

[0002] In recent years, with the large-scale grid connection of new energy and the rapid popularization of new types of loads such as electric vehicles, the load structure of the power system is becoming increasingly complex, and its growth process presents significant nonlinearity and high uncertainty. Under this background, the identification of the "inflection point" where the load curve changes from smooth growth to rapid climb has become an important basis for key decisions such as the timing of distribution network expansion, the configuration window of energy storage systems, and the optimization of regional DC transformation sequences.

[0003] Currently, the load forecasting and planning field mainly relies on traditional methods, including trend extrapolation models and rule-based judgment methods based on artificial threshold setting. Trend extrapolation methods such as linear regression and grey prediction models are based on the continuity assumption of historical data and are suitable for scenarios where the load pattern is relatively stable. Threshold methods, on the other hand, identify abnormal growth by setting a fixed growth rate threshold. However, under the background of rapid evolution of load characteristics and intensifying external factors, the above traditional methods gradually reveal the following limitations: (1) Dynamic response lag: Trend extrapolation methods lack sensitivity to step changes in load caused by external events and are difficult to capture fundamental changes in growth patterns, leading to lagging power grid planning behind actual load development and even causing local overload risks.

[0004] (2) Strong subjective dependence and insufficient adaptability: Artificial threshold methods rely on planners' experience to set thresholds and lack adaptive learning ability for historical data distribution characteristics. When load fluctuations intensify or growth patterns change from linear to exponential, the fixed threshold set is likely to cause false positives and missed detections, affecting identification accuracy.

[0005] (3) Single identification mode: Existing methods focus on single-point surge detection and are difficult to effectively distinguish between long-term "rapid growth segments" and short-term "fluctuation noise", lacking the ability to identify complex multi-modal load growth patterns and being difficult to support refined planning needs.

[0006] In summary, there is an urgent need to develop an intelligent identification method that can automatically quantify load growth stage characteristics and accurately identify multi-modal inflection points, to overcome the shortcomings of traditional methods in dynamic response, adaptive learning, and complex pattern recognition, and to achieve adaptive and forward-looking capture of load growth trend mutations and acceleration intervals, providing reliable basis for accurate planning and scientific decision-making of power grid infrastructure. SUMMARY

[0007] In order to overcome the above-mentioned defects of the prior art, embodiments of the present application provide a power load growth-based smart grid inflection point identification method, which solves the problem that traditional methods cannot quickly and accurately identify load growth turning inflection points under the background of increasingly complex power system load growth.

[0008] To achieve the above-mentioned purpose, the embodiments of the present application provide a power load growth-based smart grid inflection point identification method, comprising: collecting power data to obtain a historical load data sequence; obtaining a double-feature vector containing a load growth value and a load growth rate based on the historical load data sequence; defining an inflection point index and obtaining an inflection point index sequence according to the load growth value and the load growth rate; standardizing the double-feature vector and performing clustering analysis to obtain a clustering label; unifying the identification of key forms of load growth based on the inflection point index sequence and the clustering label to obtain differentiated power grid planning strategies.

[0009] In a preferred embodiment, the collecting power data comprises: collecting power time series load data of a target region in a continuous historical period to obtain a historical load data sequence; introducing external source-load-storage development trend data as background information.

[0010] In a preferred embodiment, the obtaining of the double-feature vector containing the load growth value and the load growth rate based on the historical load data sequence comprises: calculating the load growth value and the load growth rate between adjacent historical periods based on the historical load data sequence; combining the load growth value and the load growth rate into a two-dimensional double-feature vector, and combining each double-feature vector to form a double-feature vector set.

[0011] In a preferred embodiment, the defining of the inflection point index and the obtaining of the inflection point index sequence comprise: obtaining the maximum load growth value in the historical period according to the load growth value and the load growth rate, and performing normalization processing on the load growth value and standardization processing on the load growth rate; defining the inflection point index based on the normalized load growth value and the standardized load growth rate, and combining the inflection point index to obtain an inflection point index sequence.

[0012] In a preferred embodiment, the standardization processing of the double-feature vector and the clustering analysis to obtain the clustering label comprise: The load growth value is standardized, and a standardized double feature vector and a double feature vector set are obtained by combining the standardized load growth rate; A density-based spatial clustering algorithm is used to perform clustering analysis on the standardized double feature vector set to obtain a clustering label corresponding to a historical period. The core parameters of the clustering analysis are adaptively adjusted according to the collected power data when the density-based spatial clustering algorithm is used to perform clustering analysis on the standardized double feature vector set. The core parameters include the minimum number of neighborhood samples and the neighborhood radius.

[0013] In a preferred embodiment, the unified identification of the key form of load growth includes identifying a load surge inflection point, identifying a load rapid growth section, and identifying a load slow adjustment stage and a stable growth stage.

[0014] In a preferred embodiment, the identification of the load surge inflection point includes: The mean and standard deviation of the inflection point index are calculated according to the inflection point index sequence, and a first inflection point threshold is adaptively set. The inflection point index is compared with the first inflection point threshold, and a load surge inflection point is screened out according to the clustering label. The load surge inflection point should satisfy: the inflection point index is not less than the first inflection point threshold and / or the clustering analysis is a load surge inflection point.

[0015] In a preferred embodiment, the identification of the load rapid growth section includes: A load growth rate threshold and a second inflection point threshold are adaptively set to obtain a historical period section with a load growth rate not less than the growth rate threshold and an inflection point index not less than the second inflection point threshold. Based on the clustering label, a historical period section with the same clustering label is screened out as a load rapid growth section.

[0016] In a preferred embodiment, the identification of the load slow adjustment stage and the stable growth stage includes: The surge inflection point and the load rapid growth section are excluded, and the development stages are divided and the slow adjustment stage and the stable growth stage are screened out according to the load growth rate, the inflection point index, and the clustering position. The development stages include the explosive development stage, the slow adjustment stage, and the stable growth stage.

[0017] In a preferred embodiment, the differentiated power grid planning strategy is specifically: In the explosive development stage, the main grid expansion is laid out, new key substations are built, and the energy storage and peak regulation capacity construction is strengthened. In the slow adjustment stage, the local optimization framework is optimized to improve the channel capacity and flexibility, and the distributed new energy access position and scale are optimized. In the stable growth stage, the capacity expansion is turned to efficiency improvement, and the fine operation and demand side management are implemented.

[0018] The beneficial effects of the present application are: (1) By combining the absolute growth rate with the relative growth intensity to construct a double feature vector, and performing standardization processing and clustering analysis, the internal mode of load growth can be objectively and data-drivenly identified according to the collected power data, and the subjectivity and hysteresis caused by relying on expert experience in the traditional identification method can be effectively avoided; (2) Based on the inflection point index defined by the load growth value and the load growth rate, the abnormal points in the load growth sequence can be sensitively captured, and early warning can be provided when the load undergoes a revolutionary change; (3) By performing clustering analysis on the double feature vector after standardization processing, the clustering label is obtained, and according to the inflection point index sequence and the clustering label, different load growth key patterns can be automatically and accurately identified, which provides a scientific basis for obtaining differentiated power grid planning strategies, and makes the power grid planning strategy more in line with the actual needs of power grid planning. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 It is a flowchart of the power grid intelligent inflection point identification method based on power load growth; Figure 2 It is a curve graph of the source-load-storage development trend; Figure 3 It is an inflection point index Change trend curve graph; Figure 4 It is a flowchart of density clustering analysis; Figure 5 It is a schematic diagram of clustering result distribution. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0021] Please refer to Figure 1The application discloses a power grid intelligent inflection point identification method based on power load growth, and comprises the following steps: collecting power data to obtain a historical load data sequence; obtaining a double-feature vector containing a load growth value and a load growth rate based on the historical load data sequence; defining an inflection point index and obtaining an inflection point index sequence according to the load growth value and the load growth rate; standardizing the double-feature vector and performing clustering analysis to obtain a clustering label; and uniformly identifying a key form of load growth based on the inflection point index sequence and the clustering label to obtain a differentiated power grid planning strategy.

[0022] By collecting power data to obtain a historical load data sequence, and obtaining a load growth value and a load growth rate based on the historical load data sequence, the load growth value represents an absolute growth amount, and the load growth rate represents a relative growth intensity. The double-feature vector fusing the absolute growth amount and the relative growth intensity is constructed, and the load change characteristics are comprehensively described from the absolute and relative dimensions. According to the constructed double-feature vector, an inflection point index is defined, and an inflection point index sequence is obtained. The inflection point index quantifies the abnormal degree of load growth in a historical period. The double-feature vector is standardized, and clustering analysis is performed on the standardized double-feature vector to obtain a clustering label. The multiple forms of load growth are preliminarily and automatically identified. Finally, different key forms of load growth, such as a sudden increase inflection point, a rapid growth section, a slow adjustment stage and a stable growth stage, are automatically and uniformly identified according to the inflection point index sequence and the clustering label. The identification is accurate. According to the identified key forms of load growth, a differentiated power grid planning strategy corresponding to different key forms can be provided.

[0023] By combining the absolute growth rate and the relative growth intensity to construct the double-feature vector, and by standardizing and performing clustering analysis on the double-feature vector, the internal mode of load growth can be objectively and data-drivenly identified according to the collected power data. The subjectivity and hysteresis caused by the dependence on expert experience in the traditional identification method are effectively avoided. The inflection point index is defined based on the load growth value and the load growth rate. The abnormal points in the load growth sequence can be sensitively captured. Early warning can be provided when the load undergoes a revolutionary change. The clustering label is obtained by performing clustering analysis on the standardized double-feature vector. Different key forms of load growth are automatically and accurately identified according to the inflection point index sequence and the clustering label. A scientific basis is provided for obtaining a differentiated power grid planning strategy. The power grid planning strategy is more in line with the actual needs of power grid planning.

[0024] To further improve the timeliness and objectivity of inflection point identification, the application further limits the collection of power data to obtain a historical load data sequence. Specifically, S1, collecting power data: The power data collection includes: collecting continuous historical periodic power load data of the target area to obtain historical load data sequences; and introducing external source-load-storage development trend data as background information.

[0025] S11. Collect continuous historical periodic power load data of the target area to obtain historical load data sequences; Historical load data: This involves collecting time-series power load data for the target area over several consecutive historical periods; these periods are measured in years, quarters, or months. The collected typical load data for the target area over several consecutive years, i.e., the historical load data sequence, is denoted as:

[0026] In the formula, For the first Annual load statistics can be the maximum load, annual average load, or representative load value confirmed by business operations for that year. The total number of historical years is T. In this embodiment, T=11, corresponding to 2025-2035.

[0027] S12. Introduce external source-load-storage development trend data as background information; Please see Figure 2 External source-load-storage development trend data are introduced, including forecasts of new energy installed capacity, electric vehicle ownership, and large-scale industrial project planning, as background information for auxiliary analysis or subsequent verification, thereby enhancing the system's foresight.

[0028] Based on the collected power data, in order to more objectively identify the inherent patterns of load growth, this embodiment further elaborates on obtaining a dual feature vector containing both load growth value and load growth rate based on historical load data sequences. Specifically: S2. Based on historical load data sequences, obtain a dual feature vector containing load growth value and load growth rate: Based on historical load data sequences, the load growth value and load growth rate between adjacent historical periods are calculated; the load growth value and the load growth rate are combined into a two-dimensional dual feature vector, and the dual feature vectors are combined to form a dual feature vector set.

[0029] S21, Based on historical load data sequences Calculate the load growth value and load growth rate between adjacent historical periods, such as the calculation of the load growth value and load growth rate between adjacent years as follows:

[0030]

[0031] In the formula, the load growth value of the year 2018 relative to the year 2017, i.e., the absolute increment of the load; the load growth rate of the year 2018 relative to the year 2017, i.e., the relative growth rate of the load. the load growth value of the year 2018 relative to the year 2017, i.e., the absolute increment of the load; the load growth rate of the year 2018 relative to the year 2017, i.e., the relative growth rate of the load.

[0032] S22, constructing a double feature vector: In order to simultaneously depict the absolute increment and the relative growth intensity of the load change, i.e., the absolute increment and the relative growth rate, the load growth value and the load growth rate of adjacent historical periods are combined into a two-dimensional double feature vector, for example, the load growth value and the load growth rate of each year are combined into a two-dimensional double feature vector as follows:

[0033] In the formula, the double feature vector of the year 2018; the first dimension describes the absolute increment of the load change; and the second dimension describes the relative growth rate of the load change. S23, composing a double feature vector set:

[0034] The double feature vector set is: In order to more sensitively capture abnormal points in the load growth sequence, the embodiment further describes the definition of the inflection point index and the acquisition of the inflection point index sequence, as follows:

[0035] S3, defining an inflection point index and acquiring an inflection point index sequence according to the load growth value and the load growth rate: The definition of the inflection point index and the acquisition of the inflection point index sequence include: acquiring the maximum load growth value in the historical period according to the load growth value and the load growth rate, and performing normalization processing on the load growth value and standardization processing on the load growth rate; defining an inflection point index based on the normalized load growth value and the standardized load growth rate, and combining the inflection point index to obtain an inflection point index sequence. S31, performing normalization processing on the load growth value and standardization processing on the load growth rate:

[0036] Based on the acquired power data samples in all historical periods, the maximum load growth value is obtained, and in the embodiment, the maximum load growth value is calculated based on all annual samples, i.e., the load growth value is normalized as follows:

[0037] ​​​The mean and standard deviation of the load growth rate, i.e. the standardization of the load growth rate, is calculated as follows:

[0038]

[0039] In the formula, is the mean of the load growth rate, representing the average level of the load growth rate of all years; is the standard deviation of the load growth rate, representing the fluctuation degree of the load growth rate.

[0040] S32, define the inflection point index

[0041] In order to quantitatively depict the "abnormally fast point" appearing in the load growth process, the inflection point index is constructed based on the load growth value and the load growth rate in the embodiment . That is: For each year , the inflection point index is defined as:

[0042] In the formula, is the inflection point index or the growth anomaly index of the th year; is the load growth value of the th year relative to the th year, i.e. the load growth value of the th year; is the maximum value of the load growth value in all annual samples, used for normalization of ; and is the load growth rate of the th year relative to the th year, i.e. the load growth rate of the th year; , are the mean and standard deviation of the load growth rate, respectively; is the relative growth amplitude after normalization; is the Z-score of the load growth rate, reflecting the deviation degree of the growth rate relative to the average level; , are non-negative weight coefficients, satisfying , used to adjust the contribution proportion of "absolute increment" and "relative growth rate" in the K value.

[0043] S33, obtain the inflection point index sequence Please refer to Figure 3In an embodiment, when the load growth value of a certain year is close to the maximum value in the historical period, and the load growth rate is significantly higher than the average level, the corresponding inflection point index will take a higher value, indicating that the year is an abnormally rapid growth or potential inflection point year; when the load growth value and the load growth rate of a certain year are close to the overall average level, the value is lower, indicating that it is in a stable or slow growth stage.

[0044] Based on the inflection point index, the inflection point index sequence is obtained as:

[0045] It can be seen that the inflection point index sequence is actually a comprehensive measurement of "load growth value size + abnormality of growth rate", which can be used for: quickly screening candidate inflection points; as an auxiliary reference for subsequent parameter self-adaptation and threshold setting of density-based spatial clustering algorithm; support for segmented analysis of load growth development stages.

[0046] By constructing the inflection point index which comprehensively measures the load growth value and the load growth rate, it can sensitively capture abnormal points in the historical load data sequence, and combined with the introduced external source-load-storage development trend data, it can provide early warning in the early stage of load revolution, thereby providing valuable preparation time for power grid planning, and realizing the transformation from passive response to active guidance of the power grid.

[0047] To comprehensively analyze the actual situation of load growth, the embodiment standardizes the double feature vectors and performs clustering analysis to obtain clustering labels, which are further described. Specifically as follows: S4, standardizing the double feature vectors and performing clustering analysis to obtain clustering labels The standardization of the double feature vectors and the clustering analysis to obtain the clustering labels include: standardizing the load growth value, combining the standardized load growth rate to obtain the standardized double feature vectors and the double feature vector set; using a density-based spatial clustering algorithm to perform clustering analysis on the standardized double feature vector set to obtain the clustering labels corresponding to the historical period; wherein, when the density-based spatial clustering algorithm performs clustering analysis on the standardized double feature vector set, the core parameters of the clustering analysis are adaptively adjusted according to the collected power data; the core parameters include the minimum number of neighborhood samples and the neighborhood radius.

[0048] S41, standardizing the double feature vectors: Although the load growth value has been normalized before defining the inflection point index , and the load growth rate Standardization was performed, but in order to ensure the weight balance of the two dimensions of load growth value and load growth rate in specific metrics during density clustering, this embodiment still performs standardization on the dual feature vectors.

[0049] The mean and standard deviation of the load growth values ​​are calculated as follows:

[0050]

[0051] In the formula, The mean of the load growth value represents the average level of the load growth value across all years. The standard deviation of the load growth value represents the degree of fluctuation in the load growth value.

[0052] For each year The standardized double feature vectors are as follows:

[0053]

[0054]

[0055] The standardized set of two feature vectors is obtained as follows:

[0056] S42. Using a density-based spatial clustering algorithm, cluster analysis is performed on the standardized dual feature vector set to obtain the cluster labels corresponding to historical periods. Please see Figure 4 A density-based spatial clustering algorithm, such as the DBSCAN algorithm, is used to cluster the standardized dual feature vectors to obtain the clustering label corresponding to each period.

[0057] S421 Adaptive Adjustment of Core Parameters for Cluster Analysis A density-based spatial clustering algorithm is used to cluster the standardized set of dual feature vectors. For density clustering, the core parameter is the minimum number of neighborhood samples. and neighborhood radius .

[0058] The core parameters are adaptively adjusted as follows: Based on the feature dimensions of the two feature vectors and sample size Select Initial value, such as ; Calculate up to the nth for each sample point The distances to the nearest neighbors are plotted as follows: Figure 3inflection point index of load growth trend curve, and automatically find the "bending point" as the neighborhood radius candidate value.

[0059] The density-based spatial clustering algorithm is used to perform density clustering on the standardized double feature vector set, and the core parameters of density clustering, such as the minimum neighborhood sample number and the neighborhood radius, are adaptively determined through grid search or empirical rules based on historical data collected, thereby improving the robustness and adaptability of the method.

[0060] S422, obtain the clustering label corresponding to the historical period, that is, the clustering result: Through DBSCAN algorithm clustering, the annual samples can be divided into several clusters and noise points, and the clustering label is obtained:

[0061] , wherein, is the clustering label of the year ; is a noise point, representing a rare and abnormal inflection point scattered in a low-density area; is a cluster, representing the normal growth of load that forms a high-density area in the feature space.

[0062] During density clustering analysis, the number and type of load growth do not need to be preset. The DBSCAN algorithm can automatically find clusters of any shape based on the double feature vector set, and effectively identify noise points, that is, the accurate identification of point anomalies and segment trends is realized at the same time. The fast growth segment is a transitional high-density path connecting different normal clusters, which solves the problems of response lag and poor adaptability of traditional methods.

[0063] To better identify the inflection point of load growth and obtain a power grid planning strategy suitable for the power grid, the embodiment further describes the unified identification of the key form of load growth based on the inflection point index sequence and the clustering label, and the acquisition of a differentiated power grid planning strategy, specifically: S5, based on the inflection point index sequence and the clustering label, uniformly identify the key form of load growth, and obtain a differentiated power grid planning strategy: S51, uniformly identify the key form of load growth After obtaining the inflection point index sequence and the clustering label After, the embodiment unifies the identification of various key forms in the load growth process, including: load sudden increase inflection point identification, load rapid growth section identification, and load slow adjustment and stable stage identification, that is, the unified identification of the key forms of load growth includes: identifying the load sudden increase inflection point, identifying the load rapid growth section, and identifying the load slow adjustment stage and the stable growth stage.

[0064] S511, load sudden increase inflection point identification: When identifying the load sudden increase inflection point, the mean and standard deviation of the inflection point index are calculated according to the inflection point index sequence, and the first inflection point threshold is adaptively set; the inflection point index is compared with the first inflection point threshold, and the load sudden increase inflection point is screened out according to the clustering label; wherein the load sudden increase inflection point should satisfy: the inflection point index is not less than the first inflection point threshold and / or the clustering analysis is a load sudden increase inflection point.

[0065] S5111, adaptive setting of the first inflection point threshold: According to the inflection point index sequence , the mean and standard deviation are calculated, that is, each inflection point index in the inflection point index sequence is standardized, specifically as follows:

[0066]

[0067] The first inflection point threshold is adaptively set as:

[0068] In the formula, is an adjustment coefficient, preferably .

[0069] S5112, candidate sudden increase point screening If a certain year satisfies: ; and / or the clustering label , that is, it is identified as a low-density outlier point based on the density-based spatial clustering algorithm, then the year is marked as a load sudden increase inflection point candidate.

[0070] S512, load rapid growth section identification: When identifying the load rapid growth section, the load growth rate threshold and the second inflection point threshold are adaptively set, the historical period section with a load growth rate not lower than the growth rate threshold and an inflection point index not less than the second inflection point threshold is obtained; based on the clustering label, the historical period section with the same clustering label is screened out as the load rapid growth section.

[0071] Find a continuous year interval in the historical period, so that: The clustering labels of all years in the interval are the same, i.e.,

[0072] The average load growth rate and average inflection point indicator in the interval are higher, i.e., satisfy the following:

[0073]

[0074] Adaptive setting of the load growth rate threshold value:

[0075] If the following is satisfied: and

[0076] In the formula, is a lower-level threshold value set according to the distribution of the inflection point indicator sequence, i.e., the second inflection point threshold value, which is set to or then the interval is marked as a load rapid growth section.

[0077] Please refer to Figure 3 and Figure 5 In the present embodiment, according to the above rules: The load growth rate of 2027-2029 is significantly higher than the global average level, and the inflection point indicator value is also significantly higher than , so it is identified as an explosive development stage; if there is one or two years in the interval that are marked as noise points by the DBSCAN algorithm, it further emphasizes its sudden increase characteristics.

[0078] S513, slow adjustment stage and stable growth stage identification When identifying the load slow adjustment and stable stage, the sudden increase inflection point and the load rapid growth section are excluded, and the development stage is divided and the slow adjustment stage and the stable growth stage are selected according to the load growth rate, the inflection point indicator and the clustering position; wherein the development stage includes the explosive development stage, the slow adjustment stage and the stable growth stage.

[0079] After excluding the sudden increase inflection point and the load rapid growth section, the remaining years are further divided according to their load growth rate, inflection point indicator and clustering position: If the clustering label of a certain continuous section is stable, the load growth rate is slightly higher than zero but overall slow, and the inflection point indicator value decreases year by year, it is determined to be a slow adjustment stage; If the growth rate of a certain continuous segment fluctuates within a small range, the inflection point indicator If the value is close to 0 and the clusters are mainly of the same low-growth type, then it is determined to be in a stable growth stage.

[0080] Please see Figure 3 and Figure 5 In this embodiment: The load growth rate still has some room for improvement from 2030 to 2033, but its growth rate has slowed significantly compared to the previous period, indicating a turning point. The value decreased year by year, and the clustering results showed that the load growth gradually shifted to a medium-growth cluster, thus it was identified as a slow adjustment phase; the load growth rate and inflection point indicators for 2033-2035 The values ​​remained at a low level, and the clustering results steadily fell into the low-growth cluster, which was identified as a stable growth stage.

[0081] Ultimately, combining Figure 3 inflection point indicator Trend curve and Figure 5 The clustering results distribution diagram shows the conclusion that the entire planning period can be divided into multiple stages: 2027-2029: A period of explosive growth, with inflection point indicators. The value is at a high level, and the load growth rate cluster is located in the high growth rate region.

[0082] 2030-2033: A period of slow adjustment, with inflection point indicators. The value and growth rate gradually declined; the power grid entered a stage of structural optimization, local absorption and peak-shaving capacity enhancement.

[0083] 2033–2035: Stable Growth Phase, Inflection Point Indicators The value is close to zero, and the load growth rate fluctuates little; the system has basically entered a stable operation and refined maintenance phase.

[0084] S52. Obtain differentiated power grid planning strategies: S521. Based on the clustering labels corresponding to the inflection point index sequence and the historical period, output as follows: Figure 3 inflection point indicator Trend curve and such Figure 5 A schematic diagram of the clustering results distribution.

[0085] Please see Figure 3 In this embodiment, the inflection point indicator sequence from 2026 to 2035 is plotted as a graph to represent the "Distributed New Energy Scenarios - K-value Trend Analysis". It can be seen that: Inflection point indicators for 2027 The value will reach its peak and will remain at a high level in 2028; After 2028, inflection point index The value falls year by year and maintains at a medium-low level near 2030-2033; After 2033, inflection point index The value is close to 0 and fluctuates in a small range.

[0086] Please refer to Figure 5 In this embodiment, for the distributed new energy scene, the clustering results of "load growth rate and year" are plotted as follows, and the key year 2029 is marked with a dashed line in the figure. It can be observed that: The points corresponding to 2027-2029 are concentrated in the clustering area with high growth rate; The points corresponding to 2030-2033 gradually transition to the area with medium-low growth rate; The points after 2033 are concentrated in the area with low growth rate and small fluctuations.

[0087] S522, based on the key morphology of unified identification, obtain differentiated planning strategy: Please refer to Table 1 and Table 2, according to the above development stage division and inflection point identification results, differentiated power grid planning strategies can be developed for different stages.

[0088] Table 1: Statistics table of inflection point identification results

[0089] Table 2: Start and end year of rapid growth section and stage division table

[0090] In this embodiment, the differentiated power grid planning strategy obtained is: In the explosive development stage of 2027-2029: layout of main grid expansion and key substation construction in advance; strengthen the construction of energy storage and peak regulation capacity to cope with large fluctuations.

[0091] In the slow adjustment stage of 2030-2033: focus on local grid optimization, channel capacity expansion and flexibility improvement projects; optimize the location and scale of distributed new energy access.

[0092] In the stable growth stage of 2033-2035: from "capacity expansion" to "efficiency improvement", implement fine operation and demand side management strategies.

[0093] Through the complete link of "double eigenvector + inflection point index Value + density clustering + adaptive threshold", the power grid intelligent inflection point identification method based on load growth of the present application realizes the objective identification of load growth inflection point and various development stages, and provides intuitive, quantitative and interpretable decision basis for long-term power grid planning and phased investment.

[0094] The load data of a historical power system is collected, and a double feature vector is constructed by extracting the load growth value and the load growth rate of each historical period. After the double feature vector is standardized, a density-based spatial clustering algorithm is introduced for unsupervised clustering analysis, so as to objectively divide the load growth mode into different categories. Based on the clustering results, the outliers identified by the algorithm are determined as the load sudden increase inflection points, and the dense cluster of high load growth characteristics is determined as the rapid growth section. By introducing density clustering to identify the load growth characteristics, the problems of response lag in the traditional trend extrapolation method, and the dependence on subjective experience and interference from fluctuations in the artificial threshold method are solved, and the mutation points and acceleration period of the load growth can be accurately captured adaptively, thereby providing a forward-looking quantitative basis for power grid expansion, energy storage configuration and other planning decisions.

[0095] The above formulas are dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation. The preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0096] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product.

[0097] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solutions. A person skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0098] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module.

[0099] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0100] Finally: the above only for the preferred embodiments of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the scope of protection of the present application.

Claims

1. A method for identifying intelligent inflection points in power grids based on electricity load growth, characterized in that, include: Collect power data to obtain historical load data sequences; Based on historical load data sequences, a dual feature vector containing load growth value and load growth rate is obtained; Based on the load growth value and the load growth rate, define the inflection point index and obtain the inflection point index sequence; The dual feature vectors are standardized and then subjected to cluster analysis to obtain cluster labels; Based on the inflection point index sequence and the clustering label, key patterns of load growth are identified in a unified manner, and differentiated power grid planning strategies are obtained.

2. The method for intelligent inflection point identification of power grid based on power load growth according to claim 1, characterized in that, The collected power data includes: Collect continuous historical periodic power load data of the target area to obtain historical load data sequences; External source-load-storage development trend data are introduced as background information.

3. The method for intelligent inflection point identification of power grid based on power load growth according to claim 2, characterized in that, The process of obtaining a dual feature vector containing both load growth value and load growth rate based on historical load data sequences includes: Based on historical load data sequences, calculate the load growth value and load growth rate between adjacent historical periods; The load growth value and the load growth rate are combined into a two-dimensional dual feature vector, and the dual feature vectors are combined to form a dual feature vector set.

4. The method for intelligent inflection point identification of power grid based on power load growth according to claim 3, characterized in that, The process of defining inflection point indicators and obtaining an inflection point indicator sequence includes: Based on the load growth value and the load growth rate, the maximum load growth value within the historical period is obtained, and the load growth value is normalized and the load growth rate is standardized. Based on the normalized load growth value and the standardized load growth rate, inflection point indicators are defined, and the inflection point indicators are combined to obtain an inflection point indicator sequence.

5. The method for intelligent inflection point identification of power grid based on power load growth according to claim 4, characterized in that, The standardization process involves processing the dual feature vectors and performing cluster analysis to obtain cluster labels, including: The load growth value is standardized, and the standardized load growth rate is combined to obtain the standardized double feature vector and double feature vector set. A density-based spatial clustering algorithm is used to perform cluster analysis on the standardized dual feature vector set to obtain the cluster labels corresponding to historical periods; Among them, when performing cluster analysis on the standardized dual feature vector set using the density-based spatial clustering algorithm, the core parameters of the cluster analysis are adaptively adjusted according to the collected power data; the core parameters include the minimum number of neighborhood samples and the neighborhood radius.

6. The method for intelligent inflection point identification of power grid based on power load growth according to claim 5, characterized in that, The unified identification of key patterns of load growth includes: identifying load surge inflection points, identifying rapid load growth segments, and identifying slow load adjustment phases and stable growth phases.

7. The method for intelligent inflection point identification of power grid based on power load growth according to claim 6, characterized in that, The identification of the inflection point of sudden load increase includes: The mean and standard deviation of the inflection point indicators are calculated based on the inflection point indicator sequence, and the first inflection point threshold is adaptively set. Compare the inflection point index with the first inflection point threshold, and filter out the load surge inflection points based on the clustering labels; The load surge inflection point should satisfy the following conditions: the inflection point index is not less than the first inflection point threshold and / or cluster analysis indicates a load surge inflection point.

8. The method for intelligent inflection point identification of power grid based on power load growth according to claim 6 or 7, characterized in that, The identification of rapidly increasing load segments includes: Adaptively set a load growth rate threshold and a second inflection point threshold to obtain historical period segments where the load growth rate is not lower than the growth rate threshold and the inflection point index is not lower than the second inflection point threshold; Based on the clustering labels, historical period segments with the same clustering labels are selected as segments with rapid load growth.

9. The method for intelligent inflection point identification of power grid based on power load growth according to claim 8, characterized in that, The identification of the slow adjustment phase and the steady growth phase of the load includes: After eliminating sudden inflection points and rapid load growth periods, the development stages are divided based on load growth rate, inflection point indicators, and cluster positions, and slow adjustment and stable growth stages are selected. The development stages include an explosive growth stage, a slow adjustment stage, and a stable growth stage.

10. The method for intelligent inflection point identification of power grid based on power load growth according to claim 9, characterized in that, The differentiated power grid planning strategy is specifically as follows: During the explosive growth phase, the main grid will be expanded, key substations will be built, and energy storage and peak-shaving capabilities will be strengthened. During the slow adjustment phase, the network structure will be optimized locally to improve channel capacity and flexibility, and the location and scale of distributed new energy access will be optimized. During the stable growth phase, the focus shifts from capacity expansion to efficiency improvement, implementing refined operation and maintenance and demand-side management.

Images