A method and system for approving a multi-element user industry power supply scheme

Abstract

A method and system for approving a multi-user industrial expansion power supply scheme. The method comprises collecting voltage, current and power time sequence sampling sequences of each user metering point in the park, and constructing a user industrial expansion load characteristic observation set; determining the load rate fluctuation interval, power factor drift direction and harmonic current phase angle offset of the same user in the continuous observation window, and constructing a modal drift characteristic vector; similarity measurement is performed on the modal drift characteristic vectors of different users in the park, and a diffusion graph is established; it is judged whether the migration time lag is less than a preset first threshold value and the drift acceleration exceeds a preset second threshold value at the same time; based on the modal characteristics of the previous user migration, the installation application capacity of the to-be-expanded user is corrected, and an industrial expansion power supply scheme approval result is generated.

Description

Technical Field

[0001] This invention belongs to the field of power load analysis, and specifically relates to a method and system for verifying power supply schemes for multiple users. Background Technology

[0002] During the expansion of services for users in industrial parks, load characteristics are typically established based on the user's industry type, equipment configuration, and historical electricity consumption curves, and the user's new electricity demand is assessed accordingly. For example, for software development users, load analysis is usually conducted based on office computing scenarios, and their electricity capacity requirements are determined according to the scale of computing equipment and the distribution of working hours. However, in actual operation, some users in the park may migrate their equipment usage due to changes in business direction. For example, computing resources originally used for program development and office computing may gradually be used for continuous computing tasks such as data processing and model training. This type of business migration usually occurs first with a small number of users, and the load change of individual users is limited, making it difficult to directly affect the overall power supply status of the park. However, because the business model is replicable, when previous users benefit from this business migration or improve resource utilization efficiency, it may induce other similar users to make similar adjustments, thus causing continuous changes in the regional load structure.

[0003] Existing load analysis methods for business expansion typically rely on user industry attributes and historical load data, focusing on existing load changes and assuming stable user business models. However, in the early stages of business migration, individual user load changes may not yet be significant, making it difficult for traditional methods to identify underlying business migration trends or determine whether such changes could spread to other users. This results in an inability to accurately anticipate changes in overall electricity demand within the industrial park.

[0004] Therefore, there is an urgent need for a system and method for verifying power supply expansion schemes for diverse users, which can improve the accuracy of power supply expansion scheme verification. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for verifying power supply schemes for multi-user businesses, thereby solving the technical problem of improving the accuracy of power supply scheme verification.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution.

[0007] This invention first discloses a method for verifying power supply schemes for multi-user business expansion, the method comprising the following steps: Collect voltage, current and power time-series sampling sequences of each user metering point in the park, and simultaneously obtain the list of power receiving equipment, operating shifts and historical capacity expansion records in the business expansion application files. Combine instantaneous active power, reactive power and harmonic distortion rate, equipment rated capacity and starting characteristics to construct a user business expansion load characteristic observation set. Based on the user business load characteristic observation set, determine the load rate fluctuation range, power factor drift direction and harmonic current phase angle offset of the same user within the continuous observation window, and construct a mode drift feature vector based on the coupling deviation and timing misalignment among the three. The similarity of the modal drift feature vectors of different users in the park is measured. Based on the transfer path of the phase angle offset of the harmonic current, the correlation strength between the previous users and the users waiting to expand their business is extracted in terms of the narrowing rate of the load rate fluctuation range and the rising slope of the harmonic distortion rate. A migration and diffusion map of electricity consumption behavior is established. Based on the diffusion map, the migration start point is the moment when the previous user deviates from the industry benchmark mode, and the diffusion response point is the moment when the user to be expanded enters the feature neighborhood after the previous user's migration. The migration delay and mode drift acceleration are calculated, and it is determined whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold at the same time. If both conditions are met, the application capacity of the user to be expanded is modified based on the modal characteristics after the previous user migration. The transformer capacity is verified and the feeder access margin is calculated based on the modification results, and the power supply scheme verification result for the expanded business is generated.

[0008] The present invention further includes the following preferred embodiments: The construction of the user load characteristic observation set includes: Collect three-phase voltage, current, active power and reactive power data of each user metering point in the park within a continuous sampling period, and synchronize the sampling sequences of different metering devices in time. The instantaneous power state at each sampling moment is calculated based on the phase relationship between voltage and current, and the load rate change trajectory, power factor change trend and harmonic distortion rate distribution characteristics are extracted. Map the rated capacity, starting current multiple, and operating shift information of the power receiving equipment in the business expansion application file to the corresponding load sampling sequence to establish the correlation between equipment operating status and changes in power parameters; Based on the power change characteristics corresponding to the equipment startup, stable operation and shutdown stages, the continuous load sequence is divided into state segments, and the load response characteristics under different operating states are extracted. The instantaneous active power, reactive power, harmonic distortion rate, capacity utilization, and equipment response characteristics are correlated according to the user number to form a user business load characteristic observation set.

[0009] The determination of the load rate fluctuation range, power factor drift direction, and harmonic current phase angle offset for the same user within a continuous observation window includes: Based on the user business load characteristic observation set, the load status within the continuous observation window of the same user is divided, and the maximum load rate, minimum load rate and load rate change range corresponding to each observation window are obtained. Based on the relationship between active and reactive power changes, determine the drift direction of the power factor relative to the stable operating state, and extract the corresponding change amplitude; The harmonic currents within each observation window are analyzed by component analysis, and the phase angle change trajectory corresponding to the main harmonic components is extracted to obtain the phase angle offset of the harmonic currents.

[0010] The construction of the modal drift feature vector based on the coupling deviation and temporal misalignment among the three factors includes: Based on the changes in modal discrimination parameters within the continuous observation window, the degree of coupling deviation between the change in load rate fluctuation range and the change in power factor drift direction is calculated. Based on the time correspondence between the phase angle shift process of harmonic current and the load rate change process, the time misalignment of harmonic change relative to load change is extracted. The synchronization deviation between the power factor change process and the harmonic phase angle shift process is calculated to obtain the multi-parameter coupling change characteristics; The load rate fluctuation range variation, power factor drift, harmonic phase angle offset, coupling deviation, and timing misalignment are combined to form the user mode drift feature vector.

[0011] The similarity measurement of modal drift feature vectors of different users within the park includes: Obtain the modal drift feature vectors corresponding to each user in the park, and classify them according to industry type, equipment composition and historical load scale; Feature alignment is performed on the load rate change direction, power factor drift direction, and harmonic phase angle change trajectory in the feature vectors corresponding to different users; Calculate the directional consistency and magnitude difference of drift features in each dimension among different users to obtain the modal drift similarity among users; Users with similar behavioral change trends are filtered based on modal drift similarity to establish a set of related users.

[0012] The establishment of the electricity consumption behavior migration and diffusion map includes: Based on the associated user set, users are sorted according to the occurrence time of different user modal drift to determine previous users and users awaiting business expansion; Using the phase angle offset of harmonic current as a transmission feature of user behavior changes, we can extract the corresponding relationship between the direction and timing of harmonic phase angle changes among different users. Calculate the difference in the rate of change of the load rate fluctuation range between the previous users and the users awaiting expansion, and extract the matching coefficient between the rising slope of the harmonic distortion rate of the two. Based on the time correlation between the harmonic phase angle change path, the load rate change rate, and the harmonic distortion rate change trend, a migration and diffusion diagram of user electricity consumption behavior is established.

[0013] The calculation of migration delay and modal drift acceleration includes: Based on the diffusion graph, time-series data of modal drift characteristics between previous users and users awaiting employment expansion were obtained; The time when the previous user first deviated from the industry benchmark mode was determined as the migration starting point, and the time when the user awaiting business expansion entered the feature neighborhood after the previous user's migration was determined as the diffusion response point. Based on the time change relationship of the modal drift feature vectors of previous users and users awaiting business expansion, the migration starting point corresponding to the deviation of previous users from the industry benchmark mode and the diffusion response point of users awaiting business expansion entering the corresponding migration feature range are determined, and the migration time lag is determined based on the time interval between the two. The modal drift rate increment is calculated based on the rate of change of load rate fluctuation range, the rate of change of harmonic distortion rate, and the rate of change of power factor after the unemployed user enters the characteristic neighborhood. Based on the change process of modal drift characteristics within the continuous observation window of the unemployed users, the changing trends of the load rate fluctuation range, the harmonic distortion rate, and the power factor drift rate are extracted to determine the modal drift acceleration.

[0014] The determination of whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold simultaneously includes: Based on the migration time delay and modal drift acceleration corresponding to the diffusion map, the migration process of different users is classified. Analyze the response time range of similar migration relationships within the historical operating cycle to determine the first threshold; The modal drift threshold is determined based on the range of load changes caused by normal user production adjustments, seasonal variations, and equipment fluctuations. When the current migration delay is less than the first threshold and the modal drift acceleration exceeds the second threshold, the corresponding migration type is marked as a migration type with propagation correction value.

[0015] The approved results of the power supply scheme for business expansion include: Obtain stable modal characteristics of previous users after migration corresponding to migration types with transmission correction value, including stable load rate range, power factor operating range, and harmonic distortion rate level; Based on the deviations between the load rate variation range, power factor variation trend, and harmonic distortion rate variation level in the stable modal characteristics and the predicted load characteristics of the users awaiting business expansion, the load state offset corresponding to the users awaiting business expansion is extracted. Based on the differences in user behavior responses corresponding to migration time lags, the migration weights are adjusted according to the original demand coefficients of users to be expanded, and the simultaneity rate parameter is corrected according to the degree of overlap of modal drift characteristics of multiple associated users. Based on the revised demand factor and simultaneity rate parameters, and combined with the maximum load demand corresponding to the applied capacity, the transformer capacity margin and feeder access margin are checked. Based on the capacity margin verification results, the applied capacity, transformer configuration capacity, and feeder access margin are comprehensively assessed to form the approval result of the business expansion power supply scheme that takes into account the impact of user behavior migration and diffusion.

[0016] This invention also discloses a system for verifying power supply schemes for multi-user business expansion using the aforementioned method, comprising: The feature acquisition module is used to collect the voltage, current and power time-series sampling sequences of each user metering point in the park, and simultaneously obtain the list of power receiving equipment, operating shifts and historical capacity expansion records in the business expansion application files. Combined with instantaneous active power, reactive power and harmonic distortion rate, equipment rated capacity and starting characteristics, a user business expansion load feature observation set is constructed. The modal analysis module is used to determine the load rate fluctuation range, power factor drift direction and harmonic current phase angle offset of the same user within a continuous observation window based on the user business load characteristic observation set, and to construct a modal drift feature vector based on the coupling deviation and timing misalignment among the three. The migration identification module is used to measure the similarity of modal drift feature vectors of different users in the park. Based on the transfer path of harmonic current phase angle offset, it extracts the correlation strength between previous users and users waiting to expand their business in terms of the narrowing rate of load rate fluctuation range and the rising slope of harmonic distortion rate, and establishes a migration and diffusion map of electricity consumption behavior. The value determination module is used to calculate the migration delay and mode drift acceleration based on the diffusion map, taking the moment when the previous user deviates from the industry benchmark mode as the migration starting point and the moment when the user to be expanded enters the feature neighborhood after the previous user's migration as the diffusion response point, and determine whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold at the same time. The scheme verification module is used to modify the application capacity of the user to be expanded based on the modal characteristics after the previous user migration if the conditions are met. Based on the modification results, the transformer capacity is verified and the feeder access margin is calculated to generate the scheme verification result for the expansion power supply.

[0017] Accordingly, this application also discloses a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the aforementioned multi-user business expansion power supply scheme verification method.

[0018] Accordingly, this application also discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned multi-user business expansion power supply scheme verification method.

[0019] The beneficial effects of this invention are as follows: Compared with the prior art, this invention provides a method and system for verifying power supply schemes for multi-user business expansion. By integrating user metering operation data and business expansion archive information, it constructs a business expansion load characteristic observation system for multi-users, breaking through the limitations of traditional business expansion analysis that relies solely on historical load curves and fixed industry experience for capacity assessment, and achieving dynamic perception of changes in users' actual electricity consumption behavior. By extracting the coupling relationship between load rate fluctuations, power factor drift, and harmonic characteristics, it can identify modal drift phenomena caused by changes in user electricity consumption patterns with production methods, equipment structures, and operating strategies, improving the ability to characterize unstable electricity consumption characteristics. Furthermore, through cross-user characteristic correlation analysis, it can explore the relationships between different users within the park. By studying the migration and diffusion patterns of electricity consumption behavior, we can identify the potential impact of previous user load changes on the future electricity consumption status of users undergoing business expansion. This avoids load forecasting errors caused by existing methods neglecting the transmission relationship between user behaviors. By introducing migration time delay and mode drift acceleration judgment mechanisms, we can screen for electricity consumption behavior migration types with practical correction value, improve the targeting of business expansion capacity adjustments, and reduce the computational complexity caused by ineffective corrections. During the power supply scheme verification process, we can dynamically correct the applied capacity, demand factor, and simultaneity rate based on the characteristics of the migrated load. This makes the transformer capacity configuration and feeder access margin calculation more consistent with the future operating status of users, thereby improving the accuracy and adaptability of business expansion power supply scheme verification and enhancing the ability to support the load change trends of users in the park. Attached Figure Description

[0020] Figure 1 This is a flowchart of the multi-user business expansion power supply scheme verification method in this invention.

[0021] Figure 2 This is a schematic diagram of the structure of the multi-user business expansion power supply scheme verification system in this invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0023] The embodiments described in this application are merely some, not all, embodiments of the present invention. Based on the spirit of the present invention, other embodiments obtained by those skilled in the art without inventive effort are all within the protection scope of the present invention.

[0024] See Figure 1 As shown, the method for verifying the power supply scheme for multi-user business expansion disclosed in this invention includes the following steps: Step 1: Collect the voltage, current and power time-series sampling sequences of each user metering point in the park, and simultaneously obtain the list of power receiving equipment, operating shifts and historical capacity expansion records in the business expansion application files. Combine the instantaneous active power, reactive power and harmonic distortion rate, equipment rated capacity and starting characteristics to construct the user business expansion load characteristic observation set.

[0025] In a further embodiment, constructing the user load characteristic observation set includes: Collect three-phase voltage, current, active power and reactive power data of each user metering point in the park within a continuous sampling period, and synchronize the sampling sequences of different metering devices in time; uniformly set the continuous sampling period of all user metering terminals in the park, and collect three-phase voltage, three-phase current, active power and reactive power of each user at high frequency to obtain raw time-series sampling data with timestamps. To address the clock deviation issue among metering devices of different models and sampling frequencies within the park, this invention uses a unified time synchronization clock within the station as a benchmark to perform global time alignment on all sampling sequences. Based on the unified time synchronization clock of the park, the initial time offset and clock drift rate of each metering device are calculated, and the local timestamp of each original sampling point is corrected: the corrected timestamp equals the local timestamp minus the initial offset, and then minus the product of the clock drift rate and the time length between the local timestamp and the most recent time synchronization time. Simultaneously, a correction confidence level is assigned based on the residual of the drift rate fitting; the smaller the residual, the higher the confidence level. Next, a unified time grid is set, and for each grid point, the sampling point with the closest timestamp in the corrected sequences of each device is searched. If multiple sampling points meet the matching tolerance, they are weighted and fused using the product of the reciprocal of the time deviation and the correction confidence level. Grid points exceeding the matching tolerance are marked as interpolation points. Finally, the interpolation points are processed according to their missing type: short-term isolated missing points are interpolated using linear interpolation with adjacent valid points. The missing point value equals the previous valid point value plus the difference between the previous and subsequent valid point values ​​multiplied by the ratio of the time from the missing point to the previous valid point to the time between the two valid points. Short-term continuous missing points are interpolated using cubic spline interpolation. A piecewise cubic curve is constructed using 3 to 5 valid points before and after the missing segment as control points. The value of each point within the missing segment is determined by the value of this curve at the corresponding time point. If historical data from the same period is available, the interpolation result is corrected using the rate of change during that period. Long-term boundary missing points are marked as invalid intervals and are not forced to interpolate; they are skipped during subsequent feature extraction. After interpolation, a physical consistency check is performed to check for anomalies such as a power factor greater than 1 or a sudden negative power change. If an anomaly is found, the interpolation point is rolled back and marked as invalid. The formula for multi-device sampling fusion based on time deviation and correction confidence is: w(i,g)=[C(i,j (i,g)) / (d(i,g)+ε)] / Σ q=1→Ng [C(q,j (q,g)) / (d(q,g)+ε)] Where w(i,g) is the fusion weight of the sampling point corresponding to the i-th metering device at the g-th unified time grid point; C(i,j (i,g)) represents the time correction confidence level of the corresponding sampling point; d(i,g) represents the time deviation between the corrected timestamp of the corresponding sampling point and the time of the unified time grid; ε is a very small positive number to prevent the denominator from being abnormal due to a zero time deviation; Ng is the number of sampling points that meet the preset matching tolerance; j (i,g) is the sampling point number in the i-th metering device that is closest to the time of the g-th time grid point.

[0026] Eliminate equipment sampling time difference and data misalignment, unify the time series benchmark of all user power data, and output time-aligned, dimension-consistent, continuous and complete user power sampling sequences to provide a standardized data source for subsequent unified analysis of load characteristics; Based on the time-synchronized sampling sequences of three-phase voltage, three-phase current, active power, and reactive power, the instantaneous power state is calculated point-by-point according to the sampling time. Instantaneous active power is obtained by summing the products of the instantaneous values ​​of the in-phase voltage and current over three phases; instantaneous reactive power is obtained by summing the products of the instantaneous values ​​of the in-phase voltage and the quadrature components of the current over three phases; and apparent power is obtained by taking the square root of the sum of the squares of instantaneous active power and instantaneous reactive power. Using the sum of the rated capacities of all the user's electrical equipment in the business expansion application file as the denominator and the point-by-point instantaneous active power as the numerator, the instantaneous power state at each sampling time is calculated. Load rate is calculated, and the instantaneous load rates at consecutive sampling times are arranged in ascending order by timestamp to form a load rate change trajectory; the real-time power factor is calculated by the ratio of instantaneous active power to apparent power at each point, and arranged in ascending order by timestamp to form a power factor change trend, with the power factor sign distinguishing between leading and lagging directions and the numerical change distinguishing between drift amplitude; the collected current waveforms are subjected to Fourier harmonic decomposition according to a preset sampling window to separate the fundamental current and each harmonic current component, and the total harmonic distortion rate is calculated by the ratio of the square root of the sum of the squares of the effective values ​​of each harmonic current to the effective value of the fundamental current, arranged in ascending order by timestamp. The sequence is arranged to form the distribution characteristics of harmonic distortion rate, and the maximum, minimum, average, and abrupt change times of the harmonic distortion rate over the entire time period are statistically analyzed. Simultaneously with feature extraction, the equipment list, operating shifts, and historical capacity expansion records are retrieved from the business expansion application files according to user number. Using the equipment identifier in the equipment list as an index, the rated capacity and starting current multiple of each piece of equipment are associated with the load sampling sequence of the metering point connected to that equipment. Furthermore, using the equipment start-up and shutdown times in the operating shifts as anchor points, the shift information is mapped to the time axis of the load sampling sequence, establishing the relationship between equipment operating status and changes in electrical parameters. The correlation between these parameters is as follows: at the moment of equipment startup, there is a short-term rapid increase in active power and load rate, and an instantaneous increase in reactive power and harmonic current in the load sampling sequence, with the increase amplitude being positively correlated with the startup current multiple; during the stable operation period of the equipment, there are small fluctuations in active power, reactive power, load rate, and power factor, which are generally stable; at the moment of equipment shutdown, there is a rapid decrease in power and a continuous drop in load; thus, the static equipment parameters and dynamic power consumption time sequence parameters are matched one-to-one according to the user number, equipment identifier, and timestamp, forming a user business load characteristic observation set; The formula relating equipment startup characteristics to dynamic electrical parameters is as follows: A(i,k)=[ΔP(i,k) / P_r(i)]×[ΔQ(i,k) / Q_r(i)]×[ΔH(i,k) / H_r(i)]×K_s(i) Where A(i,k) is the dynamic load correlation value corresponding to the start-up time of the kth power receiving equipment of user i; ΔP(i,k) is the increase in active power before and after equipment start-up; P_r(i) is the rated active power of equipment k; ΔQ(i,k) is the increase in reactive power before and after equipment start-up; Q_r(i) is the rated reactive power of equipment k; ΔH(i,k) is the increase in harmonic current characteristics before and after equipment start-up; H_r(i) is the reference harmonic current characteristic value under normal operating conditions of equipment; and K_s(i) is the starting current multiple of equipment k.

[0027] Using synchronized three-phase voltage and current time-series data, instantaneous active, reactive, and apparent power are calculated point-by-point based on the real-time phase relationship between voltage and current to obtain the actual power operation status of users at each moment. Combined with the rated capacity of the equipment, the instantaneous load rate is calculated point-by-point, forming a continuous time-series load rate change trajectory, characterizing the user load rise, fluctuation, and fall process. Based on the ratio of active power to apparent power, the real-time power factor is calculated frame-by-frame, continuously tracking the dynamic drift trend of power factor leading and lagging. Simultaneously, harmonic decomposition is performed on the acquired current waveform to extract each harmonic component. The total harmonic distortion rate (THD) is calculated, and the distribution range, fluctuation range, and abrupt change characteristics of harmonic distortion over the entire time period are statistically analyzed. This results in a multi-dimensional load characteristic profile that includes load rate time-series characteristics, power factor evolution characteristics, and power quality distortion characteristics. Instantaneous active power is obtained by summing the products of instantaneous voltage and current over three phases; instantaneous reactive power is obtained by summing the products of instantaneous voltage and the quadrature components of current over three phases; and apparent power is obtained by taking the square root of the sum of the squares of instantaneous active power and instantaneous reactive power. The instantaneous load rate is based on the rated load of all electrical equipment of the user as specified in the business expansion application file. The sum of capacities is used as the denominator, and the instantaneous active power at each point is used as the numerator. These are then arranged in ascending order by timestamp to form a load factor change trajectory with time on the horizontal axis and instantaneous load factor on the vertical axis. The real-time power factor is calculated as the ratio of instantaneous active power to apparent power at each point, and arranged in ascending order by timestamp to form a power factor evolution sequence. The power factor sign distinguishes between leading and lagging directions, and the power factor difference between adjacent sampling points distinguishes between drift amplitude and drift rate. The acquired current waveform is subjected to Fourier harmonic decomposition according to a preset sampling window to separate the fundamental current and each harmonic current component. The total harmonic distortion (THD) rate is calculated by the ratio of the square root of the sum of squares of the effective values ​​of the fundamental current to the effective value of the fundamental current. This THD is then arranged in ascending order by timestamp to form a THD time series. The maximum, minimum, mean, standard deviation, and abrupt change times of the THD are statistically analyzed for the entire time period. The abrupt change time is determined when the difference in THD between adjacent sampling windows exceeds a preset abrupt change threshold. Finally, the load factor change trajectory, power factor evolution sequence, and THD time series are aligned by timestamp to form a multidimensional load characteristic that includes load factor time series characteristics, power factor evolution characteristics, and power quality distortion characteristics. The formula for aligning load factor, power factor, and harmonic distortion rate with timestamps to form multidimensional load characteristics is as follows: F(g)=[L(g),PF(g),THD(g),ΔPF(g),ΔTHD(g)] Wherein, F(g) is the multidimensional load characteristic corresponding to the g-th unified sampling time; L(g) is the instantaneous load rate at the g-th sampling time; PF(g) is the real-time power factor at the g-th sampling time; THD(g) is the total harmonic distortion rate at the g-th sampling time; ΔPF(g) is the power factor change between adjacent sampling times; ΔTHD(g) is the total harmonic distortion rate change between adjacent sampling times; and g is the unified sampling time number arranged in ascending order of timestamps.

[0028] Based on the power change characteristics corresponding to the equipment startup, stable operation, and shutdown phases, the continuous load sequence is divided into state segments. Load response characteristics under different operating states are extracted, and three operating condition discrimination criteria (startup, steady state, and shutdown) are constructed based on equipment operating characteristics and power parameter change patterns. According to these criteria, the continuous long-term load sequence is automatically segmented and labeled with operating conditions, eliminating invalid data segments such as equipment maintenance, power outages, and fault disturbances. Specific load response characteristics for each operating condition are extracted. The discrimination criterion for the startup condition is: within a preset startup discrimination window, the instantaneous active power increase exceeds a certain proportion of the rated capacity, and the increase duration is less than the preset startup time. The system is characterized by a long duration of active power and a simultaneous increase in reactive power and total harmonic distortion exceeding their respective preset incremental thresholds. The steady-state condition is defined as follows: within a preset steady-state judgment window, the fluctuations in instantaneous active power, reactive power, instantaneous load rate, and power factor are all less than their respective preset steady-state fluctuation thresholds, and there are no segments within the window that meet the start-up or shutdown judgment conditions. The shutdown condition is defined as follows: within a preset shutdown judgment window, the instantaneous active power decrease exceeds a certain proportion of the rated capacity, and the duration of the decrease is less than the preset shutdown duration, while the instantaneous load rate continuously falls back below the preset residual load threshold. During automatic segmentation, the system traverses the continuous long-time-series load sequence in ascending order of timestamps, determining the current load point by point. The sampling points meet the operating condition criteria. When multiple consecutive sampling points meet the same operating condition criterion and the duration exceeds the preset minimum segment duration, the continuous interval is marked as the corresponding operating condition segment. If a power surge, current interruption, voltage over-limit, or harmonic distortion rate over-limit occurs simultaneously within a segment, it is determined to be an equipment maintenance, power outage, or fault disturbance segment and is removed. After the operating conditions are marked, the specific load response characteristics under each operating condition are extracted. The power rise rate during the startup phase is determined by the ratio of the instantaneous active power increment to the corresponding time increment within the startup segment. The startup impact load is determined by the difference between the maximum instantaneous active power within the startup segment and the steady-state active power before startup. The instantaneous harmonic increment is also determined. The load factor is determined by the difference between the maximum total harmonic distortion (THD) within the startup section and the steady-state THD before startup. The average load factor during the steady-state phase is determined by the arithmetic mean of the instantaneous load factors at each sampling point within the steady-state section. The power factor stability interval is composed of the maximum and minimum real-time power factor values ​​within the steady-state section. The normal harmonic level is determined by the arithmetic mean of the THD within the steady-state section. The load decline rate during the shutdown phase is determined by the ratio of the instantaneous active power decrease amplitude to the corresponding time increment within the shutdown section. The residual load characteristics are determined by the arithmetic mean of the instantaneous active power within a preset duration after the shutdown phase ends. This enables refined decomposition of load time-series data and hierarchical extraction of features. The formula for the joint judgment of the starting condition is: K_start=I(ΔP≥ηP_r)×I(0<Δt≤T_start)×I(ΔQ≥δQ)×I(ΔTHD≥δH) Where: K_start is the start-up condition judgment result; ΔP is the instantaneous active power increase amplitude within the start-up judgment window; P_r is the rated capacity of the corresponding power receiving equipment; η is the judgment ratio of the active power increase amplitude relative to the rated capacity; Δt is the duration of active power increasing from the pre-start level to the current level; T_start is the preset start-up duration threshold; ΔQ is the reactive power increase amplitude within the start-up judgment window; δQ is the reactive power increment judgment threshold; ΔTHD is the total harmonic distortion (THD) increase amplitude within the start-up judgment window; δH is the THD increment judgment threshold; I(·) is the condition indicator function, taking 1 when the condition in parentheses is met, and 0 otherwise. When K_start=1, the current continuous segment is judged as the start-up condition.

[0029] Instantaneous active power, instantaneous reactive power, harmonic distortion rate, capacity utilization, and equipment response characteristics are correlated according to user ID. Using independent users as the basic unit, the instantaneous active power, reactive power, harmonic distortion rate, real-time load rate, equipment capacity utilization, and load response characteristics corresponding to each operating condition are structurally integrated and uniformly correlated to form a user-specific load characteristic observation set. The correlation establishes a three-level index relationship using user ID as the primary key, timestamp as the time-series index, and equipment identifier as the equipment index. The instantaneous active power, instantaneous reactive power, total harmonic distortion rate, and instantaneous load rate of each metering point under the same user are aligned by timestamp and then pieced together point by point to form a dynamic [database / database]. The power consumption time-series parameter vector; the capacity utilization level is characterized by the arithmetic mean and maximum value of the instantaneous load rate at each point, where the arithmetic mean reflects the overall capacity utilization level of the user, and the maximum value reflects the peak capacity utilization level of the user; the equipment response characteristics are composed of the power rise rate, start-up impact load, and instantaneous harmonic increment during the startup phase, the average load rate, power factor stability range, and normal harmonic level during the steady-state phase, and the load fall-off rate and residual load characteristics during the shutdown phase, and are associated with the corresponding power receiving equipment according to the equipment identifier; during structured integration, the static equipment parameters, dynamic power consumption time-series parameters, and operating condition segmentation characteristics under the same user are aggregated into a unified feature structure according to the user number, where the static The parameters of the equipment include the rated capacity of the receiving equipment, the starting current multiple, and the operating shift. The dynamic power consumption time-series parameters include instantaneous active power, instantaneous reactive power, total harmonic distortion rate, instantaneous load rate, and real-time power factor arranged by timestamp. The operating condition segmentation characteristics include the start and end timestamps of each operating condition segment, the operating condition type, and the corresponding load response characteristics. When normalizing dimensions, the physical dimensions and numerical ranges of each characteristic parameter are unified. Power parameters are unified to the same power unit, and load rate, power factor, and harmonic distortion rate are unified to dimensionless ratios. Missing dimensions are marked by user number and not included in subsequent calculations. When filtering anomalies, instantaneous load rates exceeding the preset upper limit and power factors exceeding negative one to one are excluded. Sampling points with a range and total harmonic distortion exceeding the preset distortion limit are excluded, and sections marked as equipment maintenance, power outage, or fault disturbance are removed. During standardized coding, each type of characteristic parameter is assigned a unique feature identifier according to the preset coding rules, so that the static equipment parameters, dynamic power consumption time series parameters, and operating condition segment characteristics of the same user at any timestamp can be uniquely indexed by the user number, timestamp, and equipment identifier. This ensures that the static equipment parameters, dynamic power consumption time series parameters, and operating condition segment characteristics of the same user correspond one-to-one and are time-matched, forming a user business load characteristic observation set that covers equipment attributes, operating conditions, load fluctuation characteristics, and power quality characteristics and can be directly used for modal analysis and behavior recognition.

[0030] The formula for constructing the dynamic electricity consumption time-series parameter vector of the user business load characteristic observation set is as follows: F_u(t)=[P_u(t),Q_u(t),H_u(t),L_u(t),PF_u(t)] Where: F_u(t) is the dynamic power consumption time sequence parameter vector corresponding to user number u at timestamp t; P_u(t) is the instantaneous active power; Q_u(t) is the instantaneous reactive power; H_u(t) is the total harmonic distortion rate; L_u(t) is the instantaneous load rate; PF_u(t) is the real-time power factor.

[0031] Step 2: Based on the user business load characteristic observation set, determine the load rate fluctuation range, power factor drift direction and harmonic current phase angle offset of the same user within the continuous observation window, and construct the mode drift feature vector based on the coupling deviation and timing misalignment among the three.

[0032] In a further embodiment, determining the load rate fluctuation range, power factor drift direction, and harmonic current phase angle offset of the same user within a continuous observation window includes: Based on the user business expansion load characteristic observation set, the load status within the same user's continuous observation window is divided. A fixed-duration sliding observation window is used to continuously and comprehensively segment the user's daily time-series load data. The observation window duration is uniformly set to a fixed hourly or daily window and a slightly overlapping sliding method is used to ensure the continuity and complete transition of load status evolution. For each independent observation window, using the start and end timestamps of the window as boundaries, all instantaneous load rate sampling points falling within the boundaries are extracted from the user business expansion load characteristic observation set in ascending order of timestamp to form an in-window load rate sampling sequence. The maximum and minimum load rates within the window are initialized as the instantaneous load rates of the first sampling point in this sequence. Subsequently, each subsequent sampling point is read sequentially in ascending order of timestamp, and the instantaneous load rate of that point is compared with the current maximum and minimum load rates within the window. If it is greater than the current maximum load rate within the window, the maximum load rate within the window is updated; otherwise, it is updated. If the minimum load rate is within a new window, it remains unchanged otherwise. After traversing to the last sampling point, the maximum and minimum load rates within the window are output, and the timestamps of the corresponding sampling points are recorded simultaneously as the temporal positions of the extreme points of load fluctuation. If there are multiple sampling points with the same instantaneous load rate within the window, the value and timestamp of the first sampling point are taken. If there are sampling points with discontinuous time within the window due to abnormal filtering or invalid segment removal, only the actual valid sampling points are traversed, and the number of valid sampling points and time coverage are recorded as the confidence level of the statistical results of the window. After traversal, the load rate fluctuation range of the current observation window is obtained by the difference between the maximum and minimum load rates within the window. By continuously calculating the fluctuation amplitude and stability of the user's load at different times through window-by-window statistics, different load states such as stable load operation, intermittent load fluctuation, and large load anomalies are distinguished, providing quantitative load boundary indicators for subsequent determination of modal changes in electricity consumption behavior.

[0033] Based on the relationship between active and reactive power changes, the drift direction of the power factor relative to the stable operating state is determined, and the corresponding change amplitude is extracted. The average power factor during the long-term stable production operation of user equipment is selected as the benchmark steady-state power factor to construct a user-specific reactive power operation benchmark. Within each continuous observation window, active and reactive power sampling data are synchronously matched at each time step. The real-time power factor sequence is calculated through the dynamic ratio of active and reactive power. The overall distribution characteristics of the real-time power factor within the window are compared with the steady-state benchmark. When the reactive power ratio increases and the active power utilization efficiency decreases, the power factor is determined to drift in a lagging negative direction; when reactive power consumption decreases and the active power utilization rate increases, the power factor is determined to drift in a leading positive direction. The drift is determined to clarify the specific drift direction of the power factor. Simultaneously, the maximum absolute value of the power factor deviation relative to the steady-state benchmark within the observation window is calculated to obtain the power factor drift amplitude corresponding to this window, thus achieving a quantitative characterization of the user's reactive power operation state deviation trend and degree. The benchmark steady-state power factor is selected as follows: from the user's business load characteristic observation set, segments marked as steady-state with a duration exceeding a preset benchmark duration and instantaneous active power fluctuation amplitude less than a preset benchmark fluctuation threshold are selected. The arithmetic mean of the real-time power factor at each sampling point within this segment is taken as the user's benchmark steady-state power factor, and the corresponding time range and number of sampling points are recorded as the benchmark. Reliability; The real-time power factor sequence is obtained as follows: Within each continuous observation window, the instantaneous active power and instantaneous reactive power are read point by point in ascending order of timestamp. The real-time power factor of the sampling point is calculated by the ratio of instantaneous active power to apparent power, where apparent power is obtained by taking the square root of the sum of the squares of instantaneous active power and instantaneous reactive power. The real-time power factors of each sampling point within the window are arranged in ascending order of timestamp to form a real-time power factor sequence. The drift direction is determined as follows: The arithmetic mean of the real-time power factor sequence within the window is calculated. This mean is compared with the reference steady-state power factor. If the mean is less than the reference steady-state power factor and the mean ratio of instantaneous reactive power to instantaneous active power within the window is greater than the reference range, then the drift direction is determined. If the average ratio is greater than the reference steady-state power factor and the average ratio of instantaneous reactive power to instantaneous active power within the window is less than the average ratio of the corresponding reference segment, then the power factor is determined to be drifting towards the leading direction. If the absolute value of the difference between the average value and the reference steady-state power factor is less than the preset drift determination threshold, then the power factor is determined not to have drifted. The drift amplitude is extracted as follows: in the real-time power factor sequence within the window, the difference between the real-time power factor and the reference steady-state power factor at each sampling point is calculated point by point, and the maximum value of the absolute values ​​of all differences is taken as the power factor drift amplitude of the observation window. At the same time, the timestamp of the sampling point corresponding to the maximum value is recorded as the position of the drift extreme point.If there are discontinuous sampling points within the window due to abnormal filtering or invalid segment removal, the above calculation is only performed on the actual valid sampling points, and the number of valid sampling points and time coverage are recorded as the confidence level of the window drift determination result.

[0034] The formula for extracting the power factor drift amplitude is: D_PF(k)=max_t|PF(k,t)-PF0| Where: D_PF(k) is the power factor drift amplitude of the k-th continuous observation window; PF(k,t) is the real-time power factor corresponding to timestamp t in the k-th observation window; PF0 is the user reference steady-state power factor; max_t is the maximum value among all valid sampling points in the observation window.

[0035] Harmonic currents within each observation window are analyzed to extract the phase angle change trajectories corresponding to the main harmonic components, thus obtaining the harmonic current phase angle offset. Fourier harmonic decomposition is performed on the user-side sampled current waveform within each load observation window to separate the fundamental current from each harmonic current component. The lowest-order characteristic harmonics, which are most sensitive to equipment operating conditions and have the highest distortion weight, are selected as the analysis objects. The phase angle changes of each main characteristic harmonic within the observation window are continuously tracked, and phase time-series evolution data are continuously recorded to form harmonic phase angle change trajectories. The harmonic phase angle under the user's standard steady-state operating conditions is selected as the reference phase angle and compared with the observation windows. The angle difference between the real-time harmonic phase angle and the reference phase angle is defined as the harmonic current phase angle offset of the current window. This parameter reflects the shift in power consumption distortion characteristics caused by changes in user equipment switching, load structure adjustments, and changes in the proportion of nonlinear loads, capturing changes in users' implicit power consumption behavior. Fourier harmonic decomposition uses the current sampling sequence within the observation window as input, performs integer-cycle truncation and windowing on the sampling sequence according to the fundamental frequency, and separates the fundamental current component and the second to the preset highest-order harmonic current components through discrete Fourier transform, obtaining the amplitude and phase angle of each harmonic current. Low-order characteristic harmonics are then filtered. The method is as follows: Among the second to the highest preset harmonics, the harmonics are sorted from largest to smallest according to the proportion of their effective current values ​​to the sum of all effective harmonic current values. Several harmonics with the lowest order and a proportion exceeding a preset weight threshold are selected as the main characteristic harmonics. Simultaneously, considering the type of nonlinear equipment in the user's business expansion application file, harmonics matching the typical characteristic harmonic order of that type of equipment are also included in the analysis. The method for obtaining the harmonic phase angle change trajectory is as follows: Within the observation window, the phase angles of each main characteristic harmonic are extracted point by point in ascending order of timestamp. The phase angles of the same characteristic harmonic at each sampling point are arranged in ascending order of timestamp to form the phase angle of that harmonic. The phase angle time sequence is obtained, and the jump points where the phase angle difference between adjacent sampling points exceeds ±180 degrees are corrected by periodic extension to eliminate the pseudo-jump caused by phase scrambling. After correction, the continuously changing harmonic phase angle change trajectory is obtained. The reference phase angle is selected by screening the section where the operating condition is marked as steady state and the duration exceeds the preset reference duration, and the total harmonic distortion rate fluctuation amplitude within the time period is less than the preset reference fluctuation threshold. The arithmetic mean of the phase angle of each main characteristic harmonic at each sampling point in the section is taken as the reference phase angle of the corresponding characteristic harmonic, and the time range and the number of sampling points corresponding to the section are recorded as the reference confidence level.The harmonic current phase angle offset is obtained as follows: Within the observation window, for each major characteristic harmonic, the difference between the phase angle of the sampling point and the corresponding reference phase angle is calculated point by point. This difference is normalized to the range of -180 degrees to 180 degrees using a periodic extension method. The maximum absolute value of the normalized difference is taken as the phase angle offset of the characteristic harmonic within the observation window. The phase angle offsets of each major characteristic harmonic are then weighted and summed according to the harmonic order to obtain the harmonic current phase angle offset of the observation window. The weights are determined by the proportion of the effective value of each characteristic harmonic current to the sum of the effective values ​​of all characteristic harmonic currents. If there are discontinuous sampling points within the observation window due to abnormal filtering or invalid segment removal, the above decomposition, tracking, and offset calculation are only performed on the actual valid sampling points. The number of valid sampling points and the time coverage are recorded as the confidence level of the offset determination result for the window.

[0036] The formula for weighting the phase angle offsets of each major characteristic harmonic to form the phase angle offset of the window-level harmonic current is: D_H(k)=Σ[m=1→M]wm×max_t|Δφm(k,t)| Where: D_H(k) is the phase angle offset of the harmonic current in the k-th observation window; M is the number of main characteristic harmonics; wm is the weight of the m-th characteristic harmonic, determined according to the proportion of the effective value of the harmonic current to the sum of the effective values ​​of all characteristic harmonic currents; Δφm(k,t) is the difference between the phase angle of the m-th characteristic harmonic at timestamp t and the corresponding reference phase angle after periodic extension normalization in the k-th observation window; max_t is the maximum phase angle offset among the effective sampling points in the observation window.

[0037] The load rate fluctuation range, power factor drift direction and corresponding amplitude, and harmonic current phase angle offset obtained from the above-mentioned observation window are unified in dimension and aligned in time sequence, and integrated into a user electricity consumption behavior mode discrimination parameter system. Among them, the load rate fluctuation range represents the stability change of the user's load capacity utilization, the power factor drift characteristic represents the shift trend of the user's reactive power operation mode, and the harmonic phase angle offset represents the abnormal characteristics of the user's equipment operation structure and power consumption distortion state. It can be used as a quantitative basis for identifying the user's steady-state power consumption mode, judging mode shift, and distinguishing different power consumption behavior modes.

[0038] The construction of the modal drift feature vector based on the coupling deviation and temporal misalignment among the three factors includes: Based on the changes in modal discrimination parameters within the continuous observation window, the degree of coupling deviation between the load rate fluctuation range change and the power factor drift direction change is calculated. Within adjacent continuous sliding observation windows, the load rate fluctuation range parameters and power factor drift parameters for each window are read. The time-series changes in the load rate fluctuation range and the power factor drift are obtained sequentially through subtraction. Using the synchronous change patterns of these two parameters under normal steady-state power consumption modes as the benchmark coupling relationship, the consistency between the load rate change trend and the power factor drift trend within the current window is calculated. When the load rate narrowing or expanding and the power factor positive or negative drift are asynchronous, reversed, or mismatched in magnitude, a parameter coupling deviation effect occurs. The real-time coupling deviation of two core parameters is calculated using a normalized coupling deviation calculation model to characterize the degree of coordinated variation between the user's load capacity utilization state and reactive power operation state, effectively identifying the implicit power consumption mode shift phenomenon where a single parameter is normal and multiple parameters are linked abnormally. Specifically, the time-series change in load rate fluctuation range is obtained by subtracting the load rate fluctuation range of the previous adjacent observation window from the current observation window's load rate fluctuation range; the time-series change in power factor drift is obtained by subtracting the power factor drift amplitude of the previous adjacent observation window from the power factor drift amplitude of the current observation window. The power factor drift amplitude is quantized with a sign, taking negative values ​​for lagging negative drift and positive values ​​for leading positive drift. The reference coupling relationship is constructed as follows: in the user... The load characteristic observations of business expansion identify segments marked as steady-state with a duration exceeding a preset benchmark duration. Within these segments, the time-series changes in load rate fluctuations and power factor drift are calculated pairwise for adjacent observation windows. The sign agreement ratio and the average amplitude ratio of these two types of changes are statistically analyzed. The sign agreement ratio is used as the benchmark unidirectional coefficient, and the average amplitude ratio is used as the benchmark amplitude ratio coefficient, together forming the benchmark coupling relationship for the user. The matching consistency is calculated as follows: for the current adjacent window pair, the signs of the time-series changes in load rate fluctuations and power factor drift are taken respectively. If the signs are the same, it is determined to be a unidirectional change; if the signs are opposite, it is determined to be a unidirectional change; if the absolute value of either change is less than... The preset change threshold is used to determine that there is no significant change. The solution method of the normalized coupling deviation calculation model is as follows: the time series change of the load rate fluctuation range is divided by the user's reference steady-state load rate fluctuation range to obtain the first normalized change. The time series change of the power factor drift is divided by the user's reference steady-state power factor drift amplitude to obtain the second normalized change. The absolute value of the difference between the first normalized change and the reference amplitude proportional coefficient multiplied by the second normalized change is used as the amplitude deviation component. The absolute value of the difference between the reference homing coefficient and the current window's actual homing determination result is used as the direction deviation component. The direction deviation component is multiplied by the preset direction weight, and the amplitude deviation component is multiplied by the preset amplitude weight and then summed to obtain the coupling deviation degree of the current adjacent window pair.When the change is determined to be in the same direction, the directional deviation component is zero, and the coupling deviation is composed only of the amplitude deviation component. When the change is determined to be in the opposite direction, the directional deviation component reaches its maximum value, and the coupling deviation increases significantly. When the change is determined to be insignificant, the coupling deviation takes the preset minimum deviation amount. If there are time-discontinuous sampling points in the current observation window or the previous adjacent observation window due to abnormal filtering or invalid segment removal, the above change amount calculation is performed only on the valid sampling points that exist in both windows, and the number of valid sampling points and time coverage in both windows are recorded as the confidence level of the coupling deviation calculation result.

[0039] The formula for calculating the coupling deviation degree by combining the reference coupling relationship with the magnitude and direction deviation of the current window is as follows: D(k)=α×|ΔL(k) / L0-R0×ΔPF(k) / PF0|+β×|S0-S(k)| Where D(k) is the coupling deviation of the current k-th adjacent observation window pair; ΔL(k) is the time-series change of the load rate fluctuation range between the current window and the previous adjacent window; ΔPF(k) is the time-series change of the signed power factor drift between the current window and the previous adjacent window; L0 is the reference load rate fluctuation range under the user's steady-state power consumption mode; PF0 is the reference power factor drift amplitude under the user's steady-state power consumption mode; R0 is the reference amplitude proportional coefficient; S0 is the reference homing coefficient; S(k) is the actual homing determination result of the current window pair; α is the amplitude deviation weight; β is the direction deviation weight.

[0040] Based on the time correspondence between the phase angle shift process of harmonic current and the load rate change process, the time misalignment of harmonic change relative to load change is extracted. The synchronization deviation between the power factor change process and the harmonic phase angle shift process is calculated to obtain the multi-parameter coupling change characteristics; The load rate fluctuation range variation, power factor drift, harmonic phase angle offset, coupling deviation, and timing misalignment are combined to form a user mode drift feature vector. The feature parameters are normalized to eliminate the weight bias caused by the difference in the numerical range of each parameter. The normalized load rate fluctuation range variation, power factor drift, harmonic current phase angle offset, coupling deviation, and timing misalignment are then sequentially spliced ​​in a fixed dimension order to construct a multi-dimensional user mode drift feature vector with fixed dimensions, comparable timing, and horizontal comparability between users.

[0041] Step 3: Measure the similarity of the modal drift feature vectors of different users in the park. Based on the transfer path of the phase angle offset of the harmonic current, extract the correlation strength between the previous users and the users waiting to expand their business in terms of the narrowing rate of the load rate fluctuation range and the rising slope of the harmonic distortion rate, and establish a migration and diffusion map of electricity consumption behavior.

[0042] In a further embodiment, the similarity measurement of modal drift feature vectors of different users within the park includes: This invention acquires modal drift feature vectors for each user within the park and classifies them according to industry type, equipment composition, and historical load scale. It batch-reads multi-dimensional modal drift feature vectors generated by all resident users within each continuous observation window, forming a user modal feature dataset. To avoid interference from differences in electricity consumption characteristics among users of different industries, equipment configurations, and load volumes in similarity judgment, this invention employs a multi-dimensional clustering classification mechanism for pre-grouping users: First, users are classified into manufacturing, scientific and technological innovation, commercial office, and supporting service industries according to industrial and commercial classification standards. Second, user equipment composition types are distinguished according to the equipment list in the business expansion archives, classifying users into purely resistive load users, inductive dynamic load users, nonlinear rectifier load users, and mixed load users based on equipment structure. Finally, users are classified into large, medium, and small load scale levels based on their historical maximum load, average annual electricity consumption, and applied capacity range. Through joint screening based on industry, equipment structure, and load scale, users with similar characteristic attributes are grouped into the same cluster, achieving homogeneous user pre-classification.

[0043] Feature alignment is performed on the load rate change direction, power factor drift direction, and harmonic phase angle change trajectory in the feature vectors corresponding to different users. Within the same user cluster, the modal drift feature vectors of each user are aligned in terms of time series and feature rules. The observation window duration, sampling frequency, and feature calculation caliber of all users are unified to eliminate system bias caused by inconsistent data statistical scales. For the load rate dimension, the criteria for the change direction of load expansion and load contraction are uniformly defined. For the power factor dimension, the criteria for judging positive drift and negative drift are unified. For the harmonic feature dimension, the phase reference correction is performed on the harmonic phase angle change trajectory of each user. The industry steady-state reference phase angle is uniformly used as the reference origin to eliminate the influence of the initial phase deviation of the equipment. At the same time, the time axis registration of the feature change time series curves of different users is performed to achieve complete alignment of the power consumption modal drift trajectories of different users in terms of time series scale, change direction, and reference state.

[0044] The similarity of modal drift between users is obtained by calculating the directional consistency and amplitude differences of drift features across different dimensions. Based on feature alignment, the similarity of modal drift between each pair of users is quantified dimension by dimension. The consistency of the change directions of three features—load rate fluctuation, power factor drift, and harmonic phase shift—is determined. The number of dimensions changing in the same direction and the number of dimensions changing in opposite directions are counted to obtain the feature directional consistency coefficient. The relative deviation of the amplitude of each feature change is calculated, and the deviations of load rate fluctuation range, power factor drift amplitude, and harmonic phase shift amplitude are calculated separately. The multi-dimensional deviations are weighted, fused, and normalized to obtain the modal drift amplitude difference between users. This difference is then combined with the directional consistency coefficient and the amplitude difference for analysis. Calculate and generate user modal drift similarity values ​​in the interval [0, 1]. The method for determining the direction of change in each dimension is as follows: For each user, take the time series of three features within the continuous observation window: load rate fluctuation range change, power factor drift, and harmonic current phase angle offset. Perform linear fitting on the time series of each dimension, using the sign of the fitting slope as the sign of the user's direction of change in that dimension. If the absolute value of the slope is less than a preset direction determination threshold, it is determined to be without significant change and is set to zero. The method for calculating the direction consistency coefficient is as follows: Pair users together and compare the signs of their direction of change in each dimension. If the signs are the same and neither is zero, it is recorded as a same-direction dimension; if the signs are opposite, it is recorded as a opposite-direction dimension; if either is zero or both are zero, it is recorded as a non-zero dimension. For dimensions without significant similarity, the directional consistency coefficient is obtained by dividing the number of dimensions in the same direction by the sum of the number of dimensions in the same direction and the number of dimensions in opposite directions. If the number of dimensions without significant similarity exceeds a preset threshold, a preset low similarity value is used as the directional consistency coefficient. The calculation method for the amplitude deviation of each dimension is as follows: For each dimension, the absolute value of the slope fitted by two users on the time series of that dimension is taken as their respective change amplitude. The relative deviation of that dimension is obtained by dividing the absolute value of the difference between the two change amplitudes by the larger of the two change amplitudes. If both change amplitudes are less than a preset amplitude judgment threshold, a preset minimum deviation value is used. The calculation method for the modal drift amplitude difference between users is as follows: The deviation of load rate fluctuation range, power factor drift amplitude, and harmonic phase angle offset amplitude are calculated separately. The results are multiplied by their respective preset dimension weights and then summed. The dimension weights are determined by the variance contribution rate of each dimension in the user's business load characteristic observation set. The weighted summation result is normalized according to a preset normalization interval to obtain the difference in modal drift amplitude between users in the interval [0, 1]. The closer the value is to zero, the closer the change amplitude of each dimension of the two users is. The joint calculation method of modal drift similarity between users is as follows: multiply the directional consistency coefficient by the preset directional weight, subtract the difference in modal drift amplitude between users from one and then multiply by the preset amplitude weight, and sum the two to obtain the similarity value in the interval [0, 1]. The closer the value is to one, the more consistent the power consumption modal drift pattern and the power consumption behavior evolution trend of the two users are. The closer it is to zero, the more significant the difference in power consumption behavior changes.If either user has discontinuous sampling points in their time series due to anomaly filtering or invalid segment removal, then the above fitting and bias calculations are performed only on the time periods where both users have valid sampling points. The number of valid sampling points and the time coverage are recorded as the confidence level of the similarity calculation result.

[0045] The formula for calculating the modal drift similarity between users by combining the directional consistency coefficient and the amplitude difference is as follows: S(i,j)=λd×Cd(i,j)+λa×[1-Da(i,j)] Where: S(i,j) is the modal drift similarity between user i and user j; Cd(i,j) is the directional consistency coefficient of the three types of drift features between the two users; Da(i,j) is the weighted difference of the drift amplitude of each dimension between the two users; λd is the directional consistency weight; λa is the amplitude difference weight, and λd+λa=1.

[0046] Users with similar behavioral trends are selected based on modal drift similarity to establish a set of associated users. A fixed similarity screening threshold is pre-set, and the similarity calculation results of all users in the park are compared with the threshold. User pairs with similarity higher than the threshold are selected, and these users are identified as having highly consistent electricity consumption modal drift patterns and behavioral evolution trends. The associated users that meet the similarity conditions are collected, deduplicated, and clustered. Users with the same source of electricity consumption change trends and the same type of modal drift characteristics are integrated into a whole to form a set of associated users.

[0047] In a further embodiment, establishing the electricity consumption behavior migration and diffusion map includes: Based on the associated user set, users are sorted according to the occurrence time of different modal drifts to identify previous users and users awaiting business expansion. On the basis of the constructed associated user set, the timestamp of the first significant modal drift of each user in the set is extracted. The timestamp is determined by the moment when the user's modal drift feature vector breaks through the industry steady-state benchmark range. All associated users are sorted in global ascending order according to the chronological order of modal anomalies. Users who first generate stable power consumption modal shifts and first complete the adjustment of power consumption behavior structure are marked as previous users. Users of the same type who gradually show similar drift characteristics in subsequent time series and are currently in the state of business expansion application, capacity application, or power consumption structure adjustment are uniformly classified as users awaiting business expansion.

[0048] Using harmonic current phase angle offset as a transmission feature of user behavior changes, the direction and timing of harmonic phase angle changes among different users are extracted. Harmonic current phase angle offset is selected as the core transmission identification clue for cross-user electricity behavior diffusion. Compared with load rate and power factor parameters, harmonic phase angle changes are more sensitive to user equipment switching, nonlinear load ratio adjustment, and production condition migration, and have stronger predictive characteristics. The direction, trend, and inflection point timing of harmonic phase angle offsets of previous users and users awaiting business expansion are extracted one by one within the continuous observation window. The change paths of harmonic phase angle rise, fall, and offset saturation of the two types of users are compared. The timing transmission chain of harmonic characteristic anomalies of previous users first and anomalies of the same source of users awaiting business expansion are sorted out. By matching the same direction, continuity, and lag of the harmonic evolution trajectory of multiple users, the characteristic transmission path of electricity behavior diffusion from previous users to surrounding similar users in the park is determined.

[0049] The difference in the rate of change of the load rate fluctuation range narrowing process between existing users and users awaiting expansion is calculated, and the matching coefficient between the rising slopes of harmonic distortion rates of the two is extracted. Differential calculations are performed on the time-series load data and power quality data of both existing users and users awaiting expansion to determine the rate of change of the load rate fluctuation range narrowing over time. Simultaneously, the time-series variation curve of harmonic distortion rate within a continuous observation window is fitted, and the rising slope of harmonic distortion rate is solved using the least squares method to calculate the strengthening rate of nonlinear electricity consumption characteristics of users. The time-series difference and amplitude difference in the load narrowing rate between existing users and users awaiting expansion are compared, and the matching coefficient of the rising slopes of harmonic distortion rates of the two is calculated to measure the correlation strength between subsequent users and previous users in the evolution of electricity consumption structure distortion. The method for calculating the narrowing rate of the load rate fluctuation range is as follows: For each user, take the time series of their load rate fluctuation range within a continuous observation window. Using the timestamp of the observation window as the independent variable and the load rate fluctuation range within the window as the dependent variable, perform a first-order differential on this time series. That is, calculate the difference between the load rate fluctuation range of the next window and the load rate fluctuation range of the previous window for each pair of adjacent observation windows, and then divide by the difference between the timestamps of the two windows to obtain the narrowing rate of each pair of adjacent windows. When this rate is negative, it indicates that the load rate fluctuation range is narrowing. Take the arithmetic mean of all negative rates as the narrowing rate of the user's load rate fluctuation range, and record the starting point when the user first experiences multiple consecutive windows with negative narrowing rates. The time stamp is used as the starting time of narrowing; the slope of the harmonic distortion rate rise is obtained as follows: for each user, the total harmonic distortion rate time series within the continuous observation window is taken, and a linear regression model is constructed with the timestamp of the observation window as the independent variable and the total harmonic distortion rate within the window as the dependent variable. The slope of the regression line is solved by the least squares method, and this slope is taken as the slope of the harmonic distortion rate rise for that user. The larger the slope, the faster the nonlinear electricity consumption characteristics are strengthened. The starting time stamp of the first continuous rise in harmonic distortion rate for multiple consecutive windows for that user is recorded as the distortion start time; the difference in the rate of change is calculated as follows: the rate of change of the load rate fluctuation range of the user awaiting expansion is subtracted from the rate of change of the load rate fluctuation range of the previous user to obtain the amplitude difference. The value is obtained by subtracting the starting time of the narrowing of the previous user from the starting time of the narrowing of the current user. The amplitude difference and the timing difference together represent the degree of difference between the two in the load narrowing process. The matching coefficient is calculated as follows: take the absolute values ​​of the rising slope of the harmonic distortion rate of the previous user and the rising slope of the harmonic distortion rate of the current user, and divide the absolute value of the slope of the current user by the absolute value of the slope of the previous user to obtain the slope ratio. If the ratio falls within the preset matching interval, it is determined that the harmonic rising trends of the two are matched. The matching coefficient is the reciprocal of the absolute value of the difference between the ratio and one and normalized to the interval [0, 1]. The closer the value is to one, the closer the nonlinear power consumption characteristics of the two are to each other and the higher the correlation strength of the subsequent user to the previous user.If any user's time series sequence contains discontinuous sampling points due to abnormal filtering or invalid segment removal, the above differentiation and fitting calculations are only performed on the time periods where both users have valid sampling points. The number of valid sampling points and the time coverage are recorded as the confidence level of the narrowing rate of change, rising slope, and matching coefficient calculation results.

[0050] The formula for the difference in the rate of change of the load factor fluctuation range is: ΔRclose=Vclose,new Vclose,old Where: ΔRclose is the difference in the rate of narrowing of the load rate fluctuation range between the user awaiting expansion and the previous user; Vclose,new is the rate of narrowing of the load rate fluctuation range between the user awaiting expansion; and Vclose,old is the rate of narrowing of the load rate fluctuation range between the previous user.

[0051] Based on the temporal correlation between harmonic phase angle change path, load rate change rate, and harmonic distortion rate change trend, a user electricity consumption behavior migration and diffusion map is established. Using the harmonic phase angle offset transmission path as an index, the linkage temporal relationship between three types of characteristics—load rate fluctuation narrowing rate and temporal evolution law, and harmonic distortion rate growth trend—is coupled to construct a multi-dimensional temporal correlation constraint mechanism. Using the entire modal drift evolution process of previous users as a standard template, the following rhythm, change pattern, and offset trend of subsequent characteristic evolution of users awaiting expansion are matched. The temporal causality and trend correlation between the two in multi-parameter linkage evolution are calculated. False correlation samples caused by random load fluctuations, seasonal disturbances, and short-term equipment fluctuations are eliminated. Finally, a system is established with previous users as the source and... This is a migration and diffusion diagram of electricity consumption behavior, targeting similar user groups and exhibiting temporal sequence, shared characteristics, and transmissible trends. The harmonic phase angle offset propagation path begins at the moment when the harmonic current phase angle offset of a previous user first consistently exceeds a preset offset threshold, and ends at the moment when the user awaiting business expansion first exhibits a unidirectional harmonic current phase angle offset that consistently exceeds a preset threshold. The direction, amplitude, and occurrence sequence of the harmonic phase angle offsets of each user between the starting and ending points are arranged in ascending order by timestamp. A multi-dimensional temporal correlation constraint mechanism pairs each group of previous and awaiting business expansion users, extracting the load rate fluctuation narrowing rate sequence, harmonic distortion rate rise slope sequence, and harmonic distortion rate before and after the migration starting point for each previous user. The wave phase offset sequence forms a standard template. Three corresponding feature sequences of the user awaiting business expansion before and after the diffusion response point are extracted to form a following sequence. The two sequences are aligned by timestamps, and the unidirectionality and lag of the three features are calculated point-by-point. Unidirectionality is determined by the consistency of the sign of the change direction at corresponding points in the two sequences, while lag is determined by the time difference between the user's feature change point and the previous user's feature change point. Temporal causality and trend correlation are obtained by dividing the number of feature points of the three types that are unidirectional and whose lag time difference falls within a preset propagation window by the total number of feature points participating in the comparison. The mean of the lag time differences of the three types of features is used as the average propagation lag. The temporal causality is obtained by weighted summing of the trend correlation coefficient and the reciprocal of the average propagation lag. The higher the coefficient, the clearer the migration and transmission relationship between the two users. The method for eliminating pseudo-association samples is as follows: for each paired sample, if its average transmission delay exceeds the upper limit of the preset transmission time window, or the proportion of the same-direction feature points in the three types of features is lower than the preset same-direction proportion threshold, or the change amplitude of the user characteristics is less than the preset significant change threshold, it is judged as a pseudo-association sample and eliminated. The electricity consumption behavior migration and diffusion diagram uses users as nodes, the temporal causality coefficient of paired samples as edge weights, and the average transmission delay as the edge direction attribute. The retained valid paired samples are connected by edges in the direction from the previous user to the user waiting for business expansion, and in the case where there are multiple previous user sources for the same user waiting for business expansion, the edges are sorted from largest to smallest by edge weight and a preset number of edges are retained.If two users in any paired sample have discontinuous sampling points due to abnormal filtering or invalid segment removal, then the above alignment and calculation are only performed on time periods where both users have valid sampling points, and the number of valid sampling points and time coverage are recorded as the confidence basis for calculating the edge weights of that paired sample.

[0052] The formula for the trend correlation coefficient is: C_trend(i,j)=N_same,lag / N_total Where: C_trend(i,j) is the trend correlation coefficient between the previous user i and the user j waiting for business expansion; N_same,lag is the number of feature points in the three types of features of the two users that have the same direction of change and the time difference between the corresponding change points is within the preset propagation time window; N_total is the total number of valid feature points participating in the comparison.

[0053] Step 4: Based on the diffusion map, take the moment when the previous user deviates from the industry benchmark mode as the migration starting point, and take the moment when the user to be expanded enters the feature neighborhood after the previous user's migration as the diffusion response point. Calculate the migration delay and mode drift acceleration, and determine whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold at the same time.

[0054] In a further embodiment, the calculation of migration delay and modal drift acceleration includes: Based on the diffusion map, time-series data of modal drift characteristics of previous users and users awaiting expansion are obtained. According to the established user electricity consumption behavior migration diffusion map, paired samples of previous users and users awaiting expansion are determined. The time-series datasets of the two types of users in the associated time period are read respectively. The full-dimensional parameters of the user modal drift feature vector at each time node are extracted in batches using a unified sliding observation window, including the change in load rate fluctuation range, power factor drift, harmonic phase angle offset, coupling deviation and time misalignment. The parameters are arranged in ascending order by timestamp to construct the modal drift time-series sequence of previous users and the modal drift time-series sequence of users awaiting expansion.

[0055] The time when a previous user first deviates from the industry benchmark mode is determined as the migration starting point, and the time when the user awaiting business expansion enters the feature neighborhood after the previous user's migration is determined as the diffusion response point. Based on the historical steady-state data of users in the same industry, with the same equipment type and load scale, the feature threshold range of the industry benchmark mode is defined in advance as the criterion for judging the normal and stable power consumption status of users. The time-series feature sequences of previous users are traversed to retrieve the initial moment when their modal drift feature vector first breaks through the industry benchmark modal threshold as a whole and maintains the offset state for multiple consecutive observation windows. This moment is fixed as the migration starting point of this power consumption behavior migration. At the same time, based on the steady-state drift feature space formed after the previous user completes the modal offset, the corresponding feature neighborhood range is constructed. This neighborhood is the feature convergence interval of the new power consumption behavior mode of the previous user. The real-time modal feature changes of the user awaiting business expansion are continuously monitored. When the modal drift feature vector of the user awaiting business expansion falls into the aforementioned feature neighborhood for the first time and stays stably, this moment is recorded as the diffusion response point.

[0056] Based on the temporal change relationship of the modal drift feature vectors of previous users and potential new users, the migration starting point corresponding to the deviation of previous users from the industry benchmark mode and the diffusion response point of potential new users entering the corresponding migration feature range are determined, and the migration lag is determined based on the time interval between the two. The timestamps of the migration starting point and the diffusion response point are aligned under a unified time-series coordinate axis. By calculating the difference between the diffusion response point time and the migration starting point time, the time interval for the migration diffusion of previous user behavior to potential new users is obtained. This time interval is defined as the migration lag. The migration lag quantifies the propagation response speed of the imitation of electricity consumption behavior of similar users in the park, business model migration, and load structure adjustment.

[0057] Based on the rate of change of load rate fluctuation interval, harmonic distortion rate, and power factor drift after the user enters the characteristic neighborhood, the modal drift rate increment is calculated. Taking the user's arrival at the diffusion response point as the starting time, characteristic change data from multiple consecutive observation windows are extracted. Time-series differentiation is performed on the three core parameters of load rate fluctuation interval, harmonic distortion rate, and power factor to solve for the change of each parameter per unit time, obtaining the rate of narrowing of load rate fluctuation, the rate of increase of harmonic distortion, and the rate of drift of power factor. The historical steady-state change rate before the user enters the new mode is used as the reference rate, and the rate difference of each stage of the three parameters is calculated to obtain the rate increment of each dimension.

[0058] Based on the changes in modal drift characteristics within the continuous observation window of users awaiting business expansion, the trends of load rate fluctuation interval change rate, harmonic distortion rate change rate, and power factor drift rate are extracted to determine the modal drift acceleration. Based on the continuously changing characteristic rate sequences of users awaiting business expansion over multiple time periods, the load rate change rate, harmonic distortion rate change rate, and power factor drift rate are again subjected to time-series trend fitting and differential solution to characterize the evolution trend of the change rate of each dimension. The dynamic evolution trends of the three types of rates are weighted and fused to obtain the rate of change of the overall modal characteristics, which is defined as the modal drift acceleration. The modal drift acceleration is used to characterize the transition of user electricity consumption modes from a steady state to a new mode. The dynamic intensity is assessed. The rate of change of the load rate fluctuation interval is obtained as follows: using the timestamp of the continuous observation window as the independent variable and the load rate fluctuation interval within the window as the dependent variable, the difference between the load rate fluctuation interval of the subsequent window and the load rate fluctuation interval of the preceding window is calculated for each pair of adjacent observation windows. This difference is then divided by the difference between the timestamps of the preceding and following windows to obtain the rate of change of the load rate fluctuation interval for each pair of adjacent windows. These are then arranged in ascending order of timestamps to form a sequence of load rate change rates. The rate of change of the harmonic distortion rate is obtained as follows: using the timestamp of the continuous observation window as the independent variable and the total harmonic distortion rate within the window as the dependent variable, the difference between the total harmonic distortion rate of the subsequent window and the total harmonic distortion rate of the preceding window is calculated for each pair of adjacent observation windows. This difference is then divided by the difference between the timestamps of the continuous observation window and the total harmonic distortion rate of the preceding window. The difference between the timestamps of two consecutive windows yields the rate of change of harmonic distortion for each pair of adjacent windows. These are then arranged in ascending order of timestamps to form a sequence of harmonic distortion rate change rates. The power factor drift rate is obtained by using the timestamp of the continuous observation window as the independent variable and the power factor drift amplitude within the window as the dependent variable. For each pair of adjacent observation windows, the difference between the power factor drift amplitude of the subsequent window and the power factor drift amplitude of the preceding window is calculated, and then divided by the difference between the timestamps of the two consecutive windows. This yields the power factor drift rate for each pair of adjacent windows, which is then arranged in ascending order of timestamps to form a sequence of power factor drift rates. The evolution trend of the rate of change in each dimension is characterized by: analyzing the load factor change rate sequence and the harmonic distortion rate change rate, respectively. A linear regression model is constructed using the observation window timestamp as the independent variable for the sequence and power factor drift rate sequence. The slope of the regression line for each sequence is solved using the least squares method. This slope represents the trend of the rate of change of the corresponding dimension over time. The larger the absolute value of the slope, the more drastic the increase or decrease in the rate of change of that dimension. The rate of change of the overall modal characteristic is obtained by multiplying the regression slope of the load rate change rate, the regression slope of the harmonic distortion rate change rate, and the regression slope of the power factor drift rate by their respective preset dimension weights and then summing them. The dimension weights are determined by the variance contribution rate of the three rate sequences in the historical observation data of the users awaiting business expansion. The absolute value of the weighted sum is taken to obtain the rate of change of the overall modal characteristic.The modal drift acceleration is determined as follows: the rate of change of the overall modal characteristics is compared with a preset acceleration benchmark value. If it is greater than the preset acceleration benchmark value, the modal drift acceleration is determined to be positive and equal to the rate of change of velocity, indicating that the user's power consumption mode transition is accelerating. If it is less than the preset acceleration benchmark value, the modal drift acceleration is determined to be zero or negative, indicating that the user's power consumption mode is stabilizing or the transition is decelerating. If the rate sequence of the user awaiting business expansion has discontinuous sampling points due to abnormal filtering or invalid segment removal, the above fitting and differentiation calculations are only performed on the actual valid rate sampling points, and the number of valid sampling points and time coverage are recorded as the confidence level of the modal drift acceleration calculation result.

[0059] The formula for the weighted fusion of the overall modal characteristic velocity change rate is: Am=|w_L×K_L+w_H×K_H+w_PF×K_PF| Where: Am is the overall modal characteristic velocity change rate, i.e., the basic quantity of modal drift acceleration; K_L is the regression slope of the load rate fluctuation range change rate sequence; K_H is the regression slope of the harmonic distortion rate change rate sequence; K_PF is the regression slope of the power factor drift rate sequence; w_L, w_H, and w_PF are the preset weights corresponding to the three dimensions of load rate, harmonic distortion rate, and power factor. The three K values ​​are obtained by least-squares linear fitting of the corresponding rate sequences and observation window timestamps. The larger Am is, the faster the change rate of multiple modal drift characteristics is changing, i.e., the stronger the dynamic degree of the user's electricity consumption behavior transitioning from the original steady state to a new form.

[0060] In a further embodiment, determining whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold simultaneously includes: Based on the migration time delay and modal drift acceleration corresponding to the diffusion map, the migration processes of different users are classified. For all previously established users and users awaiting business expansion in the park with established diffusion maps, the migration time delay parameters and modal drift acceleration parameters corresponding to each pair of associated samples are uniformly extracted to construct a two-dimensional migration feature dataset. The transmission speed of user electricity consumption behavior diffusion is distinguished according to the size of the migration time delay, and the severity of user electricity consumption mode transition is distinguished according to the size of the modal drift acceleration. The migration processes of various electricity consumption behaviors in the park are classified and divided into four types of migration samples: short-term high-speed migration, long-term low-speed migration, short-term low-speed disturbance, and long-term high-speed gradual change. Among them, pseudo-migration samples caused by random load fluctuations, seasonal load fluctuations, and short-term equipment disturbances show the characteristics of disordered time delay and extremely low acceleration, while effective migration samples caused by real business model migration and electricity consumption structure adjustment have stable time delay intervals and high acceleration characteristics.

[0061] The response time range corresponding to similar migration relationships within the historical operating cycle is statistically analyzed to determine the first threshold. A historical database of user electricity consumption mode migration over at least one year is retrieved, and historical behavior diffusion cases of users in the same industry, with the same equipment structure and load scale are collected. Historical migration time delay data corresponding to various effective electricity consumption behavior migrations are statistically analyzed to form a sample set of similar migration response times. Mathematical statistical analysis is performed on the sample set to remove abnormal extreme values. The maximum response time delay, average response time delay, and concentrated distribution interval of effective migration cases are extracted. Using the maximum critical response time of historical effective migration as a benchmark, and appropriately adjusted according to the park's load evolution pattern, a fixed first threshold for migration time delay is finally determined. This first threshold represents the maximum response time delay of stable electricity consumption behavior migration between users in the park.

[0062] Based on the load variation range caused by normal user production adjustments, seasonal changes, and equipment fluctuations, a modal drift threshold is determined. When the current migration delay is less than the first threshold and the modal drift acceleration exceeds the second threshold, the corresponding migration type is marked as a migration type with transmission correction value. A dual-condition joint judgment mechanism is used to complete the screening. Only when both the migration delay constraint and the modal drift acceleration constraint are met can it be identified as a valid migration. The migration delay of the currently identified user is compared with the preset first threshold. If the current migration delay is smaller, it indicates that the electricity consumption behavior has a fast diffusion and transmission speed, strong user behavior replicability, and high trend stability. The current modal drift acceleration is simultaneously compared with the second threshold. If the drift acceleration is significantly higher than the threshold, excluding routine load disturbances, it proves that the current modal change is a structural upgrade and mode migration of electricity consumption behavior. Migration types that simultaneously meet both constraints are marked and archived, and it is determined that the electricity consumption mode migration of this type of previous user has a significant cross-user transmission demonstration effect. The first threshold is determined as follows: The historical database of user electricity consumption mode migration over the past year in the park is read, historical behavior diffusion cases of users in the same industry, with the same equipment structure, and the same load scale are collected, and the migration time lag corresponding to each valid migration case is statistically analyzed to form a sample set of similar migration response times. After removing abnormal extreme values ​​in the sample set that exceed a preset standard deviation multiple, the threshold is taken as... The maximum response time delay of the remaining samples is used as the first threshold, and the first threshold is moderately adjusted upward according to the preset correction coefficient based on the park load evolution law; the second threshold is determined as follows: in the user business load characteristic observation set, the section marked as steady state and lasting for more than the preset benchmark time is screened, and the normal fluctuation range of the load rate fluctuation interval, harmonic distortion rate change rate and power factor drift rate caused by normal production adjustment, seasonal changes and equipment fluctuations are statistically analyzed. The modal drift acceleration in this section is calculated according to the aforementioned modal drift acceleration calculation method, and the maximum value of the modal drift acceleration in this section is taken as the upper limit of normal disturbance. The upper limit of normal disturbance is multiplied by the preset safety factor. The second threshold is obtained by multiplying the full value, making the second threshold higher than the modal drift acceleration level that can be generated by normal production adjustments, seasonal changes, and equipment fluctuations; the execution method of the dual-condition joint judgment mechanism is as follows: for each pair of previous users and pending expansion users, their migration lag and modal drift acceleration are read. First, the migration lag is compared with the first threshold. If the migration lag is greater than or equal to the first threshold, it is directly judged as an invalid migration and the judgment of the pairing sample ends. If the migration lag is less than the first threshold, the modal drift acceleration is compared with the second threshold. If the modal drift acceleration is less than or equal to the second threshold, it is judged as an invalid migration. If the modal drift acceleration is greater than the second threshold, it is judged as an valid migration.For paired samples determined to be valid migrations, their migration type is marked as a migration type with transmission correction value and archived in the migration type database. Simultaneously, the previous user identifier, the user identifier awaiting business expansion, the migration delay, the modal drift acceleration, and the determination timestamp of the paired sample are recorded as the triggering basis for capacity correction in subsequent business expansion power supply scheme approval. If the characteristic sequence of any user in the paired sample has discontinuous sampling points due to abnormal filtering or invalid segment removal, the above migration delay and modal drift acceleration calculations are only performed on the time periods when both users have valid sampling points. The number of valid sampling points and time coverage are recorded as the confidence level of the determination result. Paired samples with a confidence level lower than a preset confidence threshold are not included in the valid migration archive.

[0063] The formula for joint determination based on two conditions is: I_valid(i,j)=I[τ(i,j)<T_1]×I[Am(i,j)> T_2]×I[C(i,j)≥Cmin] Where: I_valid(i,j) is the determination result of whether the previous user i and the user j waiting for business expansion constitute a valid migration; τ(i,j) is the migration lag between the two users; T_1 is the first threshold, that is, the maximum allowable response lag in similar historical valid migration cases; Am(i,j) is the modal drift acceleration of the corresponding paired sample; T_2 is the second threshold, that is, the modal drift acceleration threshold determined by normal production adjustments, seasonal changes and equipment fluctuations; C(i,j) is the calculated confidence level of the paired sample; Cmin is the preset confidence threshold; I[·] is 1 when the condition is met, and 0 otherwise.

[0064] Step 5: If all conditions are met, then based on the modal characteristics after the previous user migration, the application capacity of the user to be expanded is modified. Based on the modification result, the transformer capacity is verified and the feeder access margin is calculated to generate the verification result of the power supply scheme for the expanded business.

[0065] In a further embodiment, the generation of the power supply scheme verification result includes: To obtain the stable mode characteristics of previous users after migration for migration types with conduction correction value, including stable load rate range, power factor operating range, and harmonic distortion rate level, the following steps are taken: For the aforementioned effective migration types determined to have conduction correction value, corresponding previous user samples are identified. During the subsequent observation period after the previous users have completed mode drift and their power consumption status tends to stabilize, continuous power characteristic data under steady-state conditions are continuously collected. By statistically analyzing the load operating parameters within the steady-state observation window, the upper and lower limits of stable load rate, the normal power factor fluctuation operating range, and the average level and distribution range of steady-state harmonic distortion rate under the new mode of the previous users are converged, thus obtaining the previous user... The established mode characteristics of electricity consumption after a user completes the migration of their electricity consumption behavior are used as a standard reference template for the future evolution of the electricity consumption structure of users awaiting expansion. The determination method for users who have completed mode drift and whose electricity consumption status tends to be stable is as follows: within multiple consecutive observation windows after the migration start point, the absolute values ​​of the load rate fluctuation range change rate, harmonic distortion rate change rate, and power factor drift rate are all less than a preset stability determination threshold, and the mode drift acceleration is less than or equal to a preset acceleration benchmark value. The starting time of the first observation window that meets this condition is then determined as the stable start time. The statistical method for stable mode characteristics is as follows: starting from the stable start time, a preset stable... The subsequent observation period after the observation duration is taken as the steady-state observation period. Within this steady-state observation period, each observation window is selected to identify the segment marked as steady-state and whose duration exceeds the preset benchmark duration as the effective steady-state window. The following statistics are performed on all effective steady-state windows: The upper and lower limits of the stable load rate are obtained as follows: the minimum instantaneous load rate of each sampling point within all effective steady-state windows is taken as the lower limit of the stable load rate, and the maximum instantaneous load rate is taken as the upper limit of the stable load rate. The lower and upper limits constitute the stable load rate range. At the same time, the timestamps corresponding to the upper and lower limits and the number of effective steady-state windows are recorded as the basis for statistical confidence. The normal power factor fluctuation range is obtained as follows: all observation windows are selected as the upper and lower limits. The minimum real-time power factor of each sampling point within the effective steady-state window is taken as the lower limit of the operating range, and the maximum value is taken as the upper limit of the operating range. The lower limit and the upper limit constitute the normal power factor fluctuation operating range, and the midpoint of the range is taken as the center value of the normal power factor. At the same time, the proportion of the power factor leading and lagging directions is recorded. The average level and distribution range of steady-state harmonic distortion rate are obtained as follows: the arithmetic mean of the total harmonic distortion rate of each sampling point within all effective steady-state windows is taken as the average level of steady-state harmonic distortion rate. The minimum and maximum values ​​are taken to form the distribution range of steady-state harmonic distortion rate. The upper quartile and lower quartile are taken according to the preset quartile to form the concentrated distribution range of steady-state harmonic distortion rate.The output method for the fixed-mode power consumption characteristics is as follows: The stable load rate range, normal power factor fluctuation operating range, and average level of steady-state harmonic distortion rate are associated and stored with the distribution range according to user identifier, migration type identifier, and stable start time. This forms the fixed-mode power consumption characteristics of the previous user after migration, serving as a standard reference template for the future evolution of the power consumption structure of users awaiting expansion. If there are discontinuous sampling points within the steady-state observation period due to abnormal filtering or invalid segment removal, the above statistics are only performed on the actual valid steady-state windows, and the number of valid steady-state windows and time coverage are recorded as the confidence level of the fixed-mode power consumption characteristics statistical results. Migration types with a confidence level lower than the preset confidence threshold are not included in the standard reference template.

[0066] Based on the deviations between the load rate variation range, power factor variation trend, and harmonic distortion rate variation level in the stable modal characteristics and the predicted load characteristics of the users awaiting business expansion, the corresponding load state offset of the users awaiting business expansion is extracted. Based on the equipment configuration, operating mode, and historical load patterns of the users awaiting business expansion during their current business expansion application stage, initial predicted load characteristics under the traditional assessment mechanism are generated, obtaining the initial predicted load rate range, predicted power factor operating range, and predicted harmonic distortion level. The initial predicted characteristics of the users awaiting business expansion are compared dimension-by-dimensionally with the steady-state modal characteristics of previous users after migration, calculating the load rate range deviation, power factor steady-state offset deviation, and harmonic distortion level deviation. Through multi-dimensional deviation weighted fusion, the unpredictable load state offset of the users awaiting business expansion is obtained. The overall load state offset of the power structure quantifies the magnitude of load characteristic changes after the user seeking expansion follows the previous user's power consumption mode migration, characterizing the degree of difference between traditional static load assessment and future dynamic load evolution. The initial predicted load characteristics are generated from the list of power-receiving equipment, operating shifts, and historical load patterns during the user's expansion application phase. Specifically, the initial predicted maximum load is obtained by multiplying the sum of the rated capacity of the power-receiving equipment by the industry's typical demand factor; the initial predicted load rate range is obtained based on the simultaneous operation ratio of equipment corresponding to the operating shift; the predicted power factor operating range is obtained based on the typical power factor range corresponding to the equipment type; and the predicted harmonic distortion level is obtained based on the typical harmonic distortion level corresponding to the equipment type. The load factor range deviation is obtained by subtracting the lower limit of the initial predicted load factor range for the user awaiting expansion from the lower limit of the previous stable load factor range, and then subtracting the upper limit from the upper limit of the previous user's stable load factor range. The sum of the absolute values ​​of these two differences is divided by the width of the previous user's stable load factor range. The steady-state power factor deviation is obtained by subtracting the midpoint of the predicted power factor operating range for the user awaiting expansion from the midpoint of the previous user's normal power factor fluctuation operating range. The absolute value of this difference is divided by the width of the previous user's normal power factor fluctuation operating range, and the direction of the power factor sign is used as the indicator of directional deviation. The harmonic distortion level deviation is obtained by subtracting the predicted harmonic distortion level for the user awaiting expansion from the average steady-state harmonic distortion rate of the previous user. The absolute value of this difference is divided by the average steady-state harmonic distortion rate of the previous user. The average level of harmonic distortion rate is obtained. If the predicted harmonic distortion level of the user awaiting business expansion falls within the distribution range of the steady-state harmonic distortion rate of the previous user, the preset minimum deviation value is taken. The overall load state offset is obtained by multiplying the load rate interval deviation, the power factor steady-state offset deviation, and the harmonic distortion level deviation by their respective preset dimension weights and then summing them. The dimension weights are determined by the variance contribution rate of each dimension in the migration type database of similar migration samples. When the direction deviation is marked as inconsistent, a preset direction penalty is superimposed on the power factor dimension weighting term. The weighted summation result is normalized to the [0, 1] interval. The larger the value, the greater the change in load characteristics of the user awaiting business expansion after following the previous user's power consumption mode migration and the more significant the difference from the traditional static load assessment.If the historical load patterns of users awaiting expansion contain discontinuous sampling points due to missing data or the removal of invalid segments, then only the actual valid sampling points will be used to generate initial predicted load characteristics. The number of valid sampling points and the time coverage will be recorded as the confidence level for the overall load state offset calculation.

[0067] Based on the differences in user behavior response corresponding to migration lag, the migration weights are adjusted for the original demand coefficients of the user awaiting business expansion. The simultaneity rate parameter is then corrected based on the overlap of modal drift features corresponding to multiple associated users. During migration weight adjustment, the migration lag is divided by a first threshold to obtain the lag ratio. This lag ratio is used as a response speed factor. The original demand coefficient of the user awaiting business expansion is multiplied by the sum of the products of the response speed factor and the overall load state offset to obtain the adjusted demand coefficient. The original demand coefficient is determined by the equipment configuration and typical industry demand coefficients during the user's business expansion application phase. When correcting the simultaneity rate parameter, multiple associated users with valid migration pairings with the user awaiting business expansion are selected from the associated user set. For each associated user, its modal drift feature vector is extracted. The overlap of modal drift feature vectors between pairs of associated users in each dimension is calculated. The overlap is determined by the consistency of the numerical signs in the corresponding dimensions of the two vectors. The overlap coefficient is obtained by dividing the number of dimensions with amplitude deviation less than the preset overlap threshold by the total number of dimensions. The overlap coefficient is obtained by taking the arithmetic mean of the overlap degree corresponding to all associated users. The corrected overlap coefficient parameter is obtained by multiplying the original simultaneity rate of the users to be expanded by one and adding the overlap coefficient. The overall load state offset is obtained by multiplying the load rate interval deviation, power factor steady-state offset deviation and harmonic distortion level deviation by their respective preset dimension weights and then summing them. The dimension weights are determined by the variance contribution rate of each dimension in the same type of migration sample in the migration type database. The weighted summation result is normalized to the [0, 1] interval. If there are discontinuous sampling points in the historical load pattern of the users to be expanded due to missing data or invalid segment removal, the initial predicted load feature generation is performed only on the actual valid sampling points, and the number of valid sampling points and time coverage are recorded as the confidence basis for the overall load state offset calculation result.

[0068] The formula for adjusting the demand factor jointly by migration time delay and overall load state offset is: Kadj=K0×[1+α×(1-τ / T_1)×D_L] Wherein, Kadj is the adjusted demand coefficient; K0 is the original demand coefficient for users awaiting business expansion; τ is the migration lag between users awaiting business expansion and associated users; T_1 is the first threshold; α is the preset migration adjustment coefficient; and D_L is the overall load state offset, ranging from 0 to 1. (1-τ / T_1) is the response speed factor obtained from the migration lag; the smaller the migration lag, the larger this factor. It is then multiplied by the overall load state offset D_L, so that the demand coefficient is simultaneously affected by the behavior transmission speed and the degree of load state offset.

[0069] Based on the corrected demand factor and simultaneity rate parameters, and combined with the maximum load demand corresponding to the applied capacity, the transformer capacity margin and feeder connection margin are checked. The dynamically corrected demand factor and simultaneity rate are substituted into the traditional business expansion load calculation model. Combined with the user's declared equipment installed capacity and maximum labor load demand, the user's future actual maximum operating load is recalculated. Using the corrected predicted maximum load as a benchmark, the rated configuration capacity of the transformer to be connected is compared to calculate the transformer's current remaining capacity margin and long-term operating margin. This verifies whether the transformer has insufficient capacity redundancy or over-configuration issues. Simultaneously, considering the existing load level of the feeder, the line's limited capacity, and the newly added corrected load, the feeder connection margin is calculated to determine whether the line's carrying capacity meets the user's long-term power demand after mode migration. This completes the dual capacity margin check at the transformer level and the feeder level. Specifically, the user's future actual maximum operating load is obtained by multiplying the user's declared equipment installed capacity by the corrected demand factor and then by the corrected simultaneity rate parameter. The transformer's current remaining capacity margin is calculated by subtracting the corrected predicted maximum load from the rated configuration capacity of the transformer to be connected and dividing by the transformer's rated configuration capacity. The rated configuration capacity is determined as follows: if it is less than the preset minimum margin threshold, it is considered insufficient capacity redundancy; if it is greater than the preset maximum margin threshold, it is considered over-configuration. The long-term operating margin of the transformer is obtained by adding the corrected predicted maximum load to the preset load growth factor to obtain the long-term predicted load, and then dividing the difference between the transformer's rated configuration capacity and the long-term predicted load by the transformer's rated configuration capacity. The feeder connection remaining margin is obtained by subtracting the feeder's current load level from the line's quota capacity to obtain the feeder's available capacity, and then subtracting the corrected predicted maximum load. A negative value indicates that the line's carrying capacity does not meet the user's requirements. If the long-term electricity demand after the migration is positive and less than the preset minimum feeder margin threshold, it is judged as critical access requiring early warning. After the dual capacity margin verification is completed, the current remaining capacity margin of the transformer, the long-term operating margin, and the remaining feeder access margin are stored in association according to the user identifier, migration type identifier, and verification timestamp, which serves as the basis for determining the verification result of the business expansion power supply scheme. If the data submitted by the user to be expanded is missing or abnormal, it is replaced by the average value of the corrected demand coefficient and simultaneity rate parameter of the same migration sample, and the data replacement identifier is recorded as the basis for the confidence level of the verification result.

[0070] Based on the capacity margin verification results, the applied capacity, transformer configuration capacity, and feeder access margin are comprehensively assessed to form the approval result of the business expansion power supply scheme that takes into account the impact of user behavior migration and diffusion.

[0071] The beneficial effects of this invention are as follows: Compared with the prior art, this invention provides a method and system for verifying power supply schemes for multi-user business expansion. By integrating user metering operation data and business expansion archive information, it constructs a business expansion load characteristic observation system for multi-users, breaking through the limitations of traditional business expansion analysis that relies solely on historical load curves and fixed industry experience for capacity assessment, and achieving dynamic perception of changes in users' actual electricity consumption behavior. By extracting the coupling relationship between load rate fluctuations, power factor drift, and harmonic characteristics, it can identify modal drift phenomena caused by changes in user electricity consumption patterns with production methods, equipment structures, and operating strategies, improving the ability to characterize unstable electricity consumption characteristics. Furthermore, through cross-user characteristic correlation analysis, it can explore the relationships between different users within the park. By studying the migration and diffusion patterns of electricity consumption behavior, we can identify the potential impact of previous user load changes on the future electricity consumption status of users undergoing business expansion. This avoids load forecasting errors caused by existing methods neglecting the transmission relationship between user behaviors. By introducing migration time delay and mode drift acceleration judgment mechanisms, we can screen for electricity consumption behavior migration types with practical correction value, improve the targeting of business expansion capacity adjustments, and reduce the computational complexity caused by ineffective corrections. During the power supply scheme verification process, we can dynamically correct the applied capacity, demand factor, and simultaneity rate based on the characteristics of the migrated load. This makes the transformer capacity configuration and feeder access margin calculation more consistent with the future operating status of users, thereby improving the accuracy and adaptability of business expansion power supply scheme verification and enhancing the ability to support the load change trends of users in the park.

[0072] This invention can be a system, method, and / or computer program product. This invention also discloses a system for determining the power supply scheme for multi-user business expansion based on the aforementioned method for determining such schemes, comprising: The feature acquisition module is used to collect the voltage, current and power time-series sampling sequences of each user metering point in the park, and simultaneously obtain the list of power receiving equipment, operating shifts and historical capacity expansion records in the business expansion application files. Combined with instantaneous active power, reactive power and harmonic distortion rate, equipment rated capacity and starting characteristics, a user business expansion load feature observation set is constructed. The modal analysis module is used to determine the load rate fluctuation range, power factor drift direction and harmonic current phase angle offset of the same user within a continuous observation window based on the user business load characteristic observation set, and to construct a modal drift feature vector based on the coupling deviation and timing misalignment among the three. The migration identification module is used to measure the similarity of modal drift feature vectors of different users in the park. Based on the transfer path of harmonic current phase angle offset, it extracts the correlation strength between previous users and users waiting to expand their business in terms of the narrowing rate of load rate fluctuation range and the rising slope of harmonic distortion rate, and establishes a migration and diffusion map of electricity consumption behavior. The value determination module is used to calculate the migration delay and mode drift acceleration based on the diffusion map, taking the moment when the previous user deviates from the industry benchmark mode as the migration starting point and the moment when the user to be expanded enters the feature neighborhood after the previous user's migration as the diffusion response point, and determine whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold at the same time. The scheme verification module is used to modify the application capacity of the user to be expanded based on the modal characteristics after the previous user migration if the conditions are met. Based on the modification results, the transformer capacity is verified and the feeder access margin is calculated to generate the scheme verification result for the expansion power supply.

[0073] Based on the spirit of this invention, those skilled in the art will readily conceive of obtaining a computer program product based on the aforementioned multi-user business expansion power supply scheme verification method. The computer program product may include a computer-readable storage medium on which computer-readable program instructions are loaded to cause a processor to implement various aspects of this disclosure. That is, this application also includes a terminal comprising a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps according to the aforementioned multi-user business expansion power supply scheme verification method.

[0074] 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 (step-RAM), 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.

[0075] 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.

[0076] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (I-Step A) 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 Prototype Malltalk, 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 execute the computer-readable program instructions, thereby implementing various aspects of this disclosure.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for verifying power supply schemes for diversified user businesses, characterized in that, Includes the following steps: Collect voltage, current and power time-series sampling sequences of each user metering point in the park, and simultaneously obtain the list of power receiving equipment, operating shifts and historical capacity expansion records in the business expansion application files. Combine instantaneous active power, reactive power and harmonic distortion rate, equipment rated capacity and starting characteristics to construct a user business expansion load characteristic observation set. Based on the user business load characteristic observation set, determine the load rate fluctuation range, power factor drift direction and harmonic current phase angle offset of the same user within the continuous observation window, and construct a mode drift feature vector based on the coupling deviation and timing misalignment among the three. The similarity of the modal drift feature vectors of different users in the park is measured. Based on the transfer path of the phase angle offset of the harmonic current, the correlation strength between the previous users and the users waiting to expand their business is extracted in terms of the narrowing rate of the load rate fluctuation range and the rising slope of the harmonic distortion rate. A migration and diffusion map of electricity consumption behavior is established. Based on the diffusion map, the migration start point is the moment when the previous user deviates from the industry benchmark mode, and the diffusion response point is the moment when the user to be expanded enters the feature neighborhood after the previous user's migration. The migration delay and mode drift acceleration are calculated, and it is determined whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold at the same time. If both conditions are met, the application capacity of the user to be expanded is modified based on the modal characteristics after the previous user migration. The transformer capacity is verified and the feeder access margin is calculated based on the modification results, and the power supply scheme verification result for the expanded business is generated.

2. The system and method for verifying power supply schemes for multiple users as described in claim 1, characterized in that, The construction of the user load characteristic observation set includes: Collect three-phase voltage, current, active power and reactive power data of each user metering point in the park within a continuous sampling period, and synchronize the sampling sequences of different metering devices in time. The instantaneous power state at each sampling moment is calculated based on the phase relationship between voltage and current, and the load rate change trajectory, power factor change trend and harmonic distortion rate distribution characteristics are extracted. Map the rated capacity, starting current multiple, and operating shift information of the power receiving equipment in the business expansion application file to the corresponding load sampling sequence to establish the correlation between equipment operating status and changes in power parameters; Based on the power change characteristics corresponding to the equipment startup, stable operation and shutdown stages, the continuous load sequence is divided into state segments, and the load response characteristics under different operating states are extracted. The instantaneous active power, reactive power, harmonic distortion rate, capacity utilization, and equipment response characteristics are correlated according to the user number to form a user business load characteristic observation set.

3. The system and method for verifying power supply schemes for multi-user business expansion according to claim 1, characterized in that, The determination of the load rate fluctuation range, power factor drift direction, and harmonic current phase angle offset for the same user within a continuous observation window includes: Based on the user business load characteristic observation set, the load status within the continuous observation window of the same user is divided, and the maximum load rate, minimum load rate and load rate change range corresponding to each observation window are obtained. Based on the relationship between active and reactive power changes, determine the drift direction of the power factor relative to the stable operating state, and extract the corresponding change amplitude; The harmonic currents within each observation window are analyzed by component analysis, and the phase angle change trajectory corresponding to the main harmonic components is extracted to obtain the phase angle offset of the harmonic currents.

4. The system and method for verifying power supply schemes for multiple users as described in claim 1, characterized in that, The construction of the modal drift feature vector based on the coupling deviation and temporal misalignment among the three factors includes: Based on the changes in modal discrimination parameters within the continuous observation window, the degree of coupling deviation between the load rate fluctuation range change and the power factor drift direction change is calculated. Based on the time correspondence between the phase angle shift process of harmonic current and the load rate change process, the time misalignment of harmonic change relative to load change is extracted. The synchronization deviation between the power factor change process and the harmonic phase angle shift process is calculated to obtain the multi-parameter coupling change characteristics; The load rate fluctuation range variation, power factor drift, harmonic phase angle offset, coupling deviation, and timing misalignment are combined to form the user mode drift feature vector.

5. The system and method for verifying power supply schemes for multi-user business expansion according to claim 1, characterized in that, The similarity measurement of modal drift feature vectors of different users within the park includes: Obtain the modal drift feature vectors corresponding to each user in the park, and classify them according to industry type, equipment composition and historical load scale; Feature alignment is performed on the load rate change direction, power factor drift direction, and harmonic phase angle change trajectory in the feature vectors corresponding to different users; Calculate the directional consistency and magnitude difference of drift features in each dimension among different users to obtain the modal drift similarity among users; Users with similar behavioral change trends are filtered based on modal drift similarity to establish a set of related users.

6. The system and method for verifying power supply schemes for multi-user business expansion according to claim 1, characterized in that, The establishment of the electricity consumption behavior migration and diffusion map includes: Based on the associated user set, users are sorted according to the occurrence time of different user modal drift to determine previous users and users awaiting business expansion; Using the phase angle offset of harmonic current as a transmission feature of user behavior changes, we can extract the corresponding relationship between the direction and timing of harmonic phase angle changes among different users. Calculate the difference in the rate of change of the load rate fluctuation range between the previous users and the users awaiting expansion, and extract the matching coefficient between the rising slope of the harmonic distortion rate of the two. Based on the time correlation between the harmonic phase angle change path, the load rate change rate, and the harmonic distortion rate change trend, a migration and diffusion diagram of user electricity consumption behavior is established.

7. The system and method for verifying power supply schemes for multiple users as described in claim 1, characterized in that, The calculation of migration delay and modal drift acceleration includes: Based on the diffusion graph, time-series data of modal drift characteristics between previous users and users awaiting employment are obtained; The time when the previous user first deviated from the industry benchmark mode was determined as the migration starting point, and the time when the user awaiting business expansion entered the feature neighborhood after the previous user's migration was determined as the diffusion response point. Based on the time change relationship of the modal drift feature vectors of previous users and users awaiting business expansion, the migration starting point corresponding to the deviation of previous users from the industry benchmark mode and the diffusion response point of users awaiting business expansion entering the corresponding migration feature range are determined, and the migration time lag is determined based on the time interval between the two. The modal drift rate increment is calculated based on the rate of change of load rate fluctuation range, the rate of change of harmonic distortion rate, and the rate of change of power factor after the unemployed user enters the characteristic neighborhood. Based on the change process of modal drift characteristics within the continuous observation window of the unemployed users, the changing trends of the load rate fluctuation range, the harmonic distortion rate, and the power factor drift rate are extracted to determine the modal drift acceleration.

8. The method for verifying a multi-user business expansion power supply scheme according to claim 1, characterized in that, Determining whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold simultaneously includes: Based on the migration time delay and modal drift acceleration corresponding to the diffusion map, the migration process of different users is classified. Analyze the response time range of similar migration relationships within the historical operating cycle to determine the first threshold; The modal drift threshold is determined based on the range of load changes caused by normal user production adjustments, seasonal variations, and equipment fluctuations. When the current migration delay is less than the first threshold and the modal drift acceleration exceeds the second threshold, the corresponding migration type is marked as a migration type with propagation correction value.

9. The method for verifying a multi-user business expansion power supply scheme according to claim 1, characterized in that, The results of the power supply plan approval for business expansion are generated, including: Obtain stable modal characteristics of previous users after migration corresponding to migration types with transmission correction value, including stable load rate range, power factor operating range, and harmonic distortion rate level; Based on the deviations between the load rate variation range, power factor variation trend, and harmonic distortion rate variation level in the stable modal characteristics and the predicted load characteristics of the users awaiting business expansion, the load state offset corresponding to the users awaiting business expansion is extracted. Based on the differences in user behavior responses corresponding to migration time lags, the migration weights are adjusted according to the original demand coefficients of users to be expanded, and the simultaneity rate parameter is corrected according to the degree of overlap of modal drift characteristics of multiple associated users. Based on the revised demand factor and simultaneity rate parameters, and combined with the maximum load demand corresponding to the applied capacity, the transformer capacity margin and feeder access margin are checked. Based on the capacity margin verification results, the applied capacity, transformer configuration capacity, and feeder access margin are comprehensively assessed to form the approval result of the business expansion power supply scheme that takes into account the impact of user behavior migration and diffusion.

10. A system for verifying power supply schemes for multi-user business expansion, characterized in that, include: The feature acquisition module is used to collect the voltage, current and power time-series sampling sequences of each user metering point in the park, and simultaneously obtain the list of power receiving equipment, operating shifts and historical capacity expansion records in the business expansion application files. Combined with instantaneous active power, reactive power and harmonic distortion rate, equipment rated capacity and starting characteristics, a user business expansion load feature observation set is constructed. The modal analysis module is used to determine the load rate fluctuation range, power factor drift direction and harmonic current phase angle offset of the same user within a continuous observation window based on the user business load characteristic observation set, and to construct a modal drift feature vector based on the coupling deviation and timing misalignment among the three. The migration identification module is used to measure the similarity of modal drift feature vectors of different users in the park. Based on the transfer path of harmonic current phase angle offset, it extracts the correlation strength between previous users and users waiting to expand their business in terms of the narrowing rate of load rate fluctuation range and the rising slope of harmonic distortion rate, and establishes a migration and diffusion map of electricity consumption behavior. The value determination module is used to calculate the migration delay and mode drift acceleration based on the diffusion map, taking the moment when the previous user deviates from the industry benchmark mode as the migration starting point and the moment when the user to be expanded enters the feature neighborhood after the previous user's migration as the diffusion response point, and determine whether the migration delay is less than a preset first threshold and the drift acceleration exceeds a preset second threshold at the same time. The scheme verification module is used to modify the application capacity of the user to be expanded based on the modal characteristics after the previous user migration if the conditions are met. Based on the modification results, the transformer capacity is verified and the feeder access margin is calculated to generate the scheme verification result for the expansion power supply.

11. 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 multi-user business expansion power supply scheme verification method according to any one of claims 1-9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the steps of the multi-user business expansion power supply scheme verification method as described in any one of claims 1-9.

Images