Enterprise intelligent management method and system based on big data

Abstract

This invention relates to the field of enterprise data management technology, specifically disclosing an enterprise intelligent management method and system based on big data. It acquires multi-dimensional electrical parameter data, such as the high-frequency active power sequence of the target circuit, and the line parameter matrix. It iteratively extracts multi-scale autocorrelation functions using a preset multi-time-scale sliding window, and generates a power consumption rhythm baseline through fusion and weighting. Simultaneously, it performs forward differential operations on the active power sequence. When a preset dynamic jump threshold is exceeded, it adaptively extracts power envelope data blocks and extracts transient feature vectors. If a match is not found with the equipment feature map library, an unknown equipment access marker is generated. By locking down unauthorized capacity increases and combining this with line thermal field boundary analysis to derive safety boundary load data, it overcomes the passive lag limitation of traditional fixed demand thresholds, achieving an active integrated judgment capability from behavioral rhythm disturbances to physical safety boundaries. This significantly reduces the time delay and false alarm risk of unauthorized capacity increases in the plant-within-a-plant mode.

Description

Technical Field

[0001] This invention belongs to the field of enterprise data management technology, and relates to enterprise intelligent management methods and systems based on big data. Background Technology

[0002] In the factory-within-a-factory model, some tenant companies, in order to meet their production expansion needs, often connect new electrical equipment without completing capacity reporting, leading to problems such as line overload, transformer load imbalance, and unclear sharing boundaries between tenants. For the power supply side, if the new load exceeds the original design limit, it may lead to increased circuit temperature rise, increased voltage drop, and decreased power supply stability. For tenant companies, if the expansion of electricity consumption is not effectively identified and restrained, it can easily cause deviations in operating cost accounting and affect the overall power order of the park.

[0003] Furthermore, unauthorized capacity expansion is not simply a matter of increased electricity consumption; it is often accompanied by synchronous changes in the original electricity consumption rhythm, equipment start-up and shutdown patterns, and circuit voltage characteristics. Normal production activities typically exhibit relatively stable daily, weekly, and monthly rhythms. However, the addition of new equipment easily disrupts these rhythms, resulting in abnormal coupling relationships in power fluctuations, voltage drop characteristics, and time-of-day distribution. Therefore, judging solely based on electricity consumption figures at a single point in time is insufficient to distinguish between production adjustments, short-term fluctuations, and genuine unauthorized capacity expansion, and it is also difficult to establish a continuous and traceable basis for judgment.

[0004] However, existing capacity expansion identification methods still rely primarily on data verification at the end of the settlement cycle, resulting in significant time lag and an inability to promptly reflect real-time electricity consumption changes on the tenant's side. Furthermore, the single-indicator identification logic does not adequately consider circuit rhythm changes, voltage drop linkages, and cross-circuit coordinated changes, easily leading to missed detections and misjudgments. Especially in factory-within-a-factory scenarios, tenant companies' production plans and weekday / non-weekday loads differ significantly. Without comprehensive analysis of historical electricity consumption patterns and circuit correlation characteristics, it is difficult to accurately distinguish between normal capacity expansion, temporary scheduling, and unauthorized unauthorized capacity increases, thus limiting the real-time nature and accuracy of electricity consumption management. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention provides an enterprise intelligent management method and system based on big data to solve the above-mentioned technical problems.

[0006] To achieve the above and other objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides an enterprise intelligent management method based on big data, the method comprising: Obtain the multidimensional electrical parameter data sequence of the target circuit and the corresponding line parameter matrix. The multidimensional electrical parameter data sequence shall include at least the high-frequency active power sequence and the terminal voltage RMS value sequence. The high-frequency active power sequence is subjected to a sliding window iterative calculation with preset multi-time scales to extract the corresponding multi-scale autocorrelation function sequence, and the power consumption rhythm baseline data is generated by fusion and weighting. A forward differential operation is performed on the high-frequency active power sequence to generate a differential power sequence. When the value of the differential power sequence exceeds the preset dynamic jump threshold, the power envelope data block of the corresponding time interval is adaptively extracted. The power envelope data block is parsed to extract a set of transient feature vectors, and the set of transient feature vectors is compared with the device feature map library by similarity vector mapping. If no match is found, unknown device access marker data is generated. Calculate the total power increment data within the target monitoring period based on the high-frequency active power sequence; When an unknown device accesses the marked data, the steady-state power increment data corresponding to all power envelope data blocks within the target monitoring period are accumulated to obtain the cumulative unknown power increment data. Calculate the proportion of cumulative unknown power increment data to total power increment data. When the proportion is greater than a preset proportion threshold and the multi-scale autocorrelation function sequence at the current time node decays and deforms relative to the electricity consumption rhythm baseline data, generate evidence data of unauthorized capacity expansion. Based on the cumulative unknown power increment data and the line parameter matrix, thermal field boundary analysis is performed to generate safe boundary load data. The safe boundary load data is then structurally spliced ​​with the evidence data of unauthorized capacity expansion to output the management judgment result.

[0007] A second aspect of the present invention provides an enterprise intelligent management system based on big data, the system comprising: The data acquisition module is used to acquire the multidimensional electrical parameter data sequence of the target circuit and the corresponding line parameter matrix. The multidimensional electrical parameter data sequence includes at least the high-frequency active power sequence and the terminal voltage RMS value sequence. The rhythm modeling module is used to perform iterative calculations of the high-frequency active power sequence using a sliding window at preset multiple time scales, extract the corresponding multi-scale autocorrelation function sequence, and generate baseline data of electricity consumption rhythm through fusion and weighting. At the same time, it performs forward differential operation on the high-frequency active power sequence to generate a differential power sequence. When the value of the differential power sequence exceeds the preset dynamic jump threshold, it adaptively extracts the power envelope data block of the corresponding time interval. The feature recognition module is used to parse the power envelope data block, extract the transient feature vector set, and compare the transient feature vector set with the device feature map library through similarity vector mapping. If no match is found, unknown device access marker data is generated. The incremental determination module is used to calculate the total power increment data within the target monitoring period based on the high-frequency active power sequence. When the unknown device access marker data is identified, the module accumulates the steady-state power increment data corresponding to all power envelope data blocks within the target monitoring period to obtain the cumulative unknown power increment data. The module also calculates the proportion of the cumulative unknown power increment data to the total power increment data. When the proportion is greater than a preset proportion threshold and the multi-scale autocorrelation function sequence at the current time node decays and deforms relative to the electricity consumption rhythm baseline data, the module generates evidence data of unauthorized capacity expansion. The boundary output module is used to perform thermal field boundary analysis based on the cumulative unknown power increment data and the line parameter matrix, generate safe boundary load data, and structurally splice the safe boundary load data with the evidence data of unauthorized capacity expansion to output the management judgment result.

[0008] As described above, the enterprise intelligent management method and system based on big data provided by this invention has at least the following beneficial effects: This invention acquires multi-dimensional electrical parameter data, such as the high-frequency active power sequence and the effective value sequence of the terminal voltage of the target circuit, and introduces the corresponding line parameter matrix. It then performs iterative autocorrelation calculations on the active power sequence using a sliding window across preset multi-timescales, and generates a baseline data of electricity consumption rhythm reflecting the inherent production rhythm through fusion and weighting. Simultaneously, it performs forward differential operations on the same active power sequence. When the differential value exceeds a preset dynamic jump threshold, it adaptively extracts the power envelope data block of the corresponding time interval, parses out the transient feature vector set, and compares it with a similarity vector mapping of an equipment feature map library. If a match fails, it generates unknown equipment access marker data. This processing path combines the periodicity modeling of electricity consumption behavior with the capture of transient physical characteristics of equipment start-up and shutdown. It can keenly detect production rhythm disturbances caused by newly added unregistered equipment, effectively avoiding the shortcomings of traditional methods that rely solely on fixed demand thresholds or manual inspections, such as long lag times and the inability to distinguish between normal production fluctuations and illegal capacity increases. This makes the anomaly identification process more closely reflect the actual load evolution characteristics of the circuit.

[0009] This invention, after obtaining total power increment data by statistically analyzing the high-frequency active power sequence within the target monitoring period, accumulates the steady-state power increments corresponding to all power envelope data blocks upon identifying an unknown device access marker. This yields the cumulative unknown power increment data, which is then calculated as a percentage of the total power increment. When this percentage exceeds a preset threshold, and the multi-scale autocorrelation function sequence at the same time point shows significant attenuation relative to the baseline data of electricity consumption rhythm, evidence data of unauthorized capacity expansion is generated. Subsequently, thermal field boundary analysis is performed based on the cumulative unknown power increment data and the line parameter matrix to generate safety boundary load data characterizing the circuit's safe carrying capacity. The safety boundary load data and the evidence data of unauthorized capacity expansion are then structurally concatenated to output the management judgment result. This method of correlating the strength of evidence of abnormal electricity consumption behavior with the physical safety boundary of the line significantly improves the accuracy of unauthorized capacity expansion judgment and early warning capabilities. It also provides managers with current-limiting suggestions or capacity expansion upgrades for specific circuits, thereby greatly enhancing the autonomy and precision of enterprise-side electricity management. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram showing the connections between the steps of the method of the present invention.

[0012] Figure 2 This is a schematic diagram showing the connections of the various modules in the system of the present invention. Detailed Implementation

[0013] The following description, in conjunction with the implementation of this invention, is merely an example and illustration of the concept of this invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the inventive concept or exceed the scope defined in these claims, all of which should fall within the protection scope of this invention. Example

[0014] Please see Figure 1 As shown, enterprise intelligent management methods based on big data include: Obtain the multidimensional electrical parameter data sequence of the target circuit and the corresponding line parameter matrix. The multidimensional electrical parameter data sequence shall include at least the high-frequency active power sequence and the terminal voltage RMS value sequence.

[0015] The high-frequency active power sequence is subjected to a sliding window iterative calculation with preset multi-time scales to extract the corresponding multi-scale autocorrelation function sequence, and the electricity consumption rhythm baseline data is generated by fusion and weighting, including: For high-frequency active power sequences, sampling interval sequences matching one natural day, one natural week, and one natural month are used as sliding windows to extract targeted first-scale autocorrelation function sequences, second-scale autocorrelation function sequences, and third-scale autocorrelation function sequences. Calculate the standard deviation of the load rate of the high-frequency active power sequence within the current sliding window; The standard deviation of the load rate is used as an independent variable to input into a preset nonlinear activation function, which inversely maps out the dynamic weight coefficient vectors corresponding to the first-scale autocorrelation function sequence, the second-scale autocorrelation function sequence, and the third-scale autocorrelation function sequence. By using dynamic weight coefficient vectors to perform dot product fitting on autocorrelation function sequences at three scales, baseline data of electricity consumption rhythm expressed by feature matrices are generated.

[0016] In the specific implementation of this invention, a high-frequency intelligent gateway deployed at the end of the transformer substation collects multi-dimensional electrical parameter data sequences and corresponding line parameter matrices of the target circuit in real time according to a preset sampling frequency. Based on this, a continuous high-frequency active power sequence and an effective voltage value sequence at the end are obtained, and the rated design power parameters of the circuit are extracted from the line parameter matrix simultaneously.

[0017] After obtaining the high-frequency active power sequence, three independent benchmark sliding windows were used, each spanning a natural day (24 hours), a natural week (168 hours), and a natural month (approximately 720 hours). Within each sliding window, the periodic similarity features of the current sequence were extracted using a discrete autocorrelation function, generating a first-scale autocorrelation function sequence, a second-scale autocorrelation function sequence, and a third-scale autocorrelation function sequence. Any element in the aforementioned autocorrelation function sequences... The calculation is based on a variation of the classic Pearson correlation coefficient formula, which is defined as follows: ; In the formula, k represents the scale level of the sliding window. Corresponding to day, week, and month respectively. This is the time delay step, measured in seconds. This represents the total number of data points within the corresponding scale window. Let be the sampled value of the high-frequency active power at time t, with the dimension kilowatt; This is the arithmetic mean of the active power within the corresponding window. Its value ranges from [-1, 1] and is used to purely characterize the topological self-similarity of electricity consumption rhythm on the time axis.

[0018] Further calculate the load factor standard deviation of the high-frequency active power sequence within the current sliding window. At this point, the pre-extracted rated design power parameters are invoked. Through formula The real-time load rate sequence is calculated point by point, and then the statistical standard deviation of the load rate within that window is obtained. ; This reflects the discrete distribution of the intensity of electricity consumption by enterprises during the current assessment period.

[0019] The statistical standard deviation of the load rate is then used as the independent variable and input into a preset nonlinear activation function, which inversely maps out the dynamic weight coefficient vector corresponding to the autocorrelation function sequence of the three scales mentioned above. This nonlinear activation function is not a standard function in general neural networks, but a time-decaying softmax function with a penalty factor specifically designed for industrial power consumption scenarios. Its calculation formula is as follows: In the formula, The output is the dynamic weight coefficient corresponding to the k-th scale; The preset scale-based perturbation penalty factor is determined based on a physical empirical model, where short-term data exhibits the weakest perturbation resistance while long-term data demonstrates the strongest. Therefore, the penalty factor is assigned sequentially from smallest to largest scale, with the specific configuration as follows: If the company's electricity consumption has been unstable recently (i.e. (Too large), affected by the exponential negative decay amplification, representing the daily scale (k=1) weight. It will be drastically compressed, representing a lunar scale. weight This is relatively prominent, which means that when the system experiences drastic fluctuations in operating conditions, it will automatically anchor the benchmark for judging what constitutes a normal power consumption rhythm to long-term historical patterns, preventing the baseline from being contaminated by short-term abnormal random fluctuations; conversely, if the power consumption is extremely stable ( (Approaching 0), the three weights will automatically tend towards equal distribution or tilt towards a day-scale, thereby improving sensitivity to short-term, minute changes. Furthermore, the three weight coefficients... The sum of these is always 1.

[0020] Finally, the dynamic weight coefficient vector is obtained. Then, the system uses this vector to perform element-wise dot product fusion fitting on the previously generated first, second, and third scale autocorrelation function sequences, i.e., performs... This process generates baseline data on electricity consumption patterns expressed in a matrix. .

[0021] A forward differential operation is performed on the high-frequency active power sequence to generate a differential power sequence. When the value of the differential power sequence exceeds the preset dynamic jump threshold, the power envelope data block of the corresponding time interval is adaptively extracted.

[0022] Preferably, the process of generating the preset dynamic transition threshold includes: Extract a fixed-length historical sliding sequence segment ahead of the current acquisition point of the high-frequency active power sequence; Calculate the numerical discrete variance of historical sliding sequence segments; The discrete variance of the data is multiplied by a preset sensitivity multiplier coefficient, and the resulting product is used as the preset dynamic jump threshold for the current acquisition period, so that the preset dynamic jump threshold is adaptively scaled according to the background fluctuation of the power grid.

[0023] Specifically, the system performs a forward differential operation on the currently acquired high-frequency active power sequence in the buffer stream. Considering the causality in real-time stream processing, the system introduces a small buffer delay of one sampling period. For data point i corresponding to the current time t, the power difference between it and the next time point i+1 is extracted, and the calculation formula is defined as follows: In this formula, and All are instantaneous active power samples obtained directly from the underlying sensors, and their units are kilowatts; To dynamically determine whether a power surge at any given moment constitutes a genuine device start-up or shutdown event, the system synchronously executes logic to generate a preset dynamic transition threshold. Using the current sampling point i as a reference, the system extracts a fixed-length historical sliding sequence segment ahead of the current timeline (i.e., in the historical direction where events have already occurred). The length N of this segment is preferably N = 100 sampling points.

[0024] Calculate the variance of the historical sliding sequence segment using classical statistical formulas. The calculation formula is: ; in, For all power history points within this fixed-length window The arithmetic mean of the units is expressed in kilowatts.

[0025] Then, the discrete variance of the data is calculated. Multiply by the preset sensitivity coefficient Perform multiplication and use the resulting product as the preset dynamic jump threshold for the current acquisition cycle. The comparison is performed. The great advantage of this squared variation is that it can exponentially amplify large-span abrupt signal at the physical level, while simultaneously suppressing minute background jitter, thereby increasing the signal-to-noise ratio and enabling the characteristics of the device at startup to be captured extremely sharply.

[0026] When the judgment condition When this condition is met, it indicates that the jump energy of the differential power sequence has completely exceeded the dynamic jump threshold, immediately triggering the adaptive truncation mechanism. The system takes the current trigger point i as the starting point, backtracks a preset number of sampling points as a steady-state background margin, and continues to track the squared difference values ​​of subsequent data until the squared difference values ​​of 5 consecutive sampling points fall back to the threshold. The level is below the threshold, thus defining the termination time of this transient event. All high-frequency power raw points within this complete time interval are extracted and packaged into power envelope data blocks.

[0027] The transient feature vector set is extracted by parsing the power envelope data block, and the transient feature vector set is compared with the device feature map library by similarity vector mapping. If no match is found, unknown device access marker data is generated.

[0028] Preferably, the transient feature vector set is extracted by parsing the power envelope data block, including: In the power envelope data block, locate the data at the starting point of the mutation that triggers the threshold to exceed the limit, as well as the data at the new steady-state endpoint after the mutation that enters a range of numerical fluctuations smaller than the preset steady-state residual value. Extract the power value difference between the new steady-state endpoint data and the abrupt change initiation point data as a steady-state increment feature; Within the time span between the mutation initiation point data and the new steady-state endpoint data, the maximum power peak data is extracted, and the ratio of the maximum power peak data to the mutation initiation point data is calculated as the initiation shock multiple feature. Calculate the time difference between the data rising from the mutation initiation point to the numerical node where the corresponding 90% steady-state increment feature is located, and use it as the rising slope feature. By combining the steady-state incremental features, the initiation impact multiple features, and the rising slope features in a predetermined dimension order, a set of transient feature vectors for the current out-of-limit event is constructed.

[0029] Preferably, the transient feature vector set is compared with the device feature map library using similarity vector mapping. If no match is found, unknown device access marker data is generated, including: Retrieve the registered and legitimate equipment feature vector clusters from the equipment feature map library; Calculate the multidimensional Euclidean distance between the transient feature vector set and each legitimate device feature vector cluster; Minimum value search is performed on all multidimensional Euclidean distance values ​​to obtain the minimum distance identifier value; When the minimum distance identifier value is greater than the preset safety boundary radius parameter, it is determined that a non-registered load merging event has occurred, and the current time point and transient feature vector set are encapsulated as unknown device access marker data.

[0030] It should be added that the method for constructing the device feature map library is as follows: Step A: Conduct an isolated power-on start-up test on each of the legally registered electrical devices within the park. During the test, ensure that the target circuit is connected only to the device under test, and all other loads are disconnected.

[0031] Step B: For each legally registered device, perform an independent startup test at least five times. For each test, extract a set of transient feature vectors using the transient feature vector extraction method described in this invention. Calculate the arithmetic mean of the steady-state increment feature, startup impact multiple feature, and rise slope feature extracted from multiple tests. This arithmetic mean constitutes the baseline expected value of the device in the c-th dimension. Simultaneously, its standard deviation is calculated to form the characteristic standard deviation range of the device in the c-th dimension. .

[0032] Step C: Register each legally registered device The equipment identification information (equipment name, model, rated power, etc.) is packaged into a valid equipment feature vector cluster and stored in the equipment feature map library.

[0033] Step D: When a new legal device is added to the park or when the characteristics of an operating device change significantly after a major overhaul, steps A to C are repeated for that device to update the baseline expected value and the characteristic standard deviation range and write them over the corresponding record in the device characteristic map library.

[0034] Specifically, the system first locates two key topological anchor points on the time axis of the power envelope data block: one is the data at the start of a sudden change that triggers a threshold exceedance, and the other is the data at the new steady-state endpoint after the sudden change ends and the data enters a period of numerical flattening. When defining the new steady-state endpoint, the system needs to determine whether the amplitude of subsequent fluctuations converges within a preset steady-state residual value. It is recommended that the preset steady-state residual value be set to an adjustable range of one to three percent of the current loop's rated permissible load, with two percent being a preferred fixed margin.

[0035] After locking these two anchor points, the high-frequency power value corresponding to the new steady-state endpoint data is directly subtracted from the power value of the mutation initiation point data to obtain the steady-state incremental feature, with the physical dimension of kilowatts. Next, within the time axis span between the two anchor points, the maximum power peak data is globally scanned, and the pure numerical ratio of this peak data to the mutation initiation point data is calculated. This ratio serves as the starting impact multiple feature reflecting the severity of the equipment impact. Subsequently, the system calculates the time difference between the data rising from the mutation initiation point along the time axis to the numerical node corresponding to 90% of the steady-state incremental feature, using this as the rising slope feature characterizing the mechanical inertia of the equipment, with the physical dimension of seconds. At this point, the steady-state incremental feature, starting impact multiple feature, and rising slope feature are encapsulated in an array according to a specific dimensional order to construct a set of transient feature vectors for the current out-of-limit event.

[0036] Because the aforementioned transient feature vector set contains three completely different physical dimensions—kilowatt, dimensionless, and second—a dimensionless Euclidean distance variant formula based on the inherent tolerance variance constraint of the equipment is introduced after retrieving the legally registered equipment feature vector clusters from the equipment feature map library. This variant formula is used to calculate the multidimensional Euclidean distance between the transient feature vector set and each legally registered equipment feature vector cluster in the library. The structure of this variant formula is defined as follows: ; This represents the multidimensional Euclidean distance between the currently extracted set of transient feature vectors and the m-th legal device in the database. The cursor index for the feature dimension; The c-th dimension transient feature is the true observed value; The baseline expected value of the m-th legal device in the c-th dimension, pre-entered in the spectral library. and The dimensions correspond to each other; The c-th dimension of the characteristic standard deviation that the device is allowed to undergo due to long-term aging or fluctuations in power grid conditions in the spectral library. The pre-allocated dimensionless dimension weight parameters satisfy the condition that the sum of the three is 1.

[0037] After traversing the entire graph library, the system performs a minimum value optimization and filtering on all output multidimensional Euclidean distance values ​​to extract the minimum distance identifier value. Finally, the system compares this minimum distance identifier value with a preset safety boundary radius parameter. It is recommended that the preset safety boundary radius parameter be set to a dimensionless adjustable range of 0.2 to 0.5. To avoid false alarms caused by slight deviations in startup characteristics due to physical factors such as seasonal thickening of lubricating oil from legitimate devices of the same model, and to ensure the blocking of non-registered high-power energy-consuming devices disguised as low-power appliances from joining the network, a value of 0.35 is recommended for field implementation. When the minimum distance identifier value is greater than this safety boundary radius parameter, the system determines that there is a load joining event that cannot match the historical registration whitelist. It immediately extracts the current time point and encapsulates it with a transient feature vector set, solidifying it as unknown device access marker data.

[0038] The total power increment data within the target monitoring period is calculated based on the high-frequency active power sequence.

[0039] When an unknown device is identified to access the marked data, the steady-state power increment data corresponding to all power envelope data blocks within the target monitoring period are accumulated to obtain the cumulative unknown power increment data.

[0040] Calculate the proportion of cumulative unknown power increment data to total power increment data. When the proportion is greater than a preset proportion threshold and the multi-scale autocorrelation function sequence at the current time point decays and deforms relative to the electricity consumption rhythm baseline data, generate evidence data of unauthorized capacity expansion.

[0041] Preferably, the multidimensional electrical parameter data sequence also includes a sequence of effective values ​​of the first-terminal voltage, and before calculating the proportion of the cumulative unknown power increment data to the total power increment data, it also includes: Simultaneously extract the effective value sequence of the first-end voltage and the effective value sequence of the last-end voltage corresponding to the current time node, and perform quotient calculation on the voltage difference sequence between the two and the load current sequence calculated corresponding to the high-frequency active power sequence to generate the equivalent impedance sequence of the loop. Extract the historical equivalent impedance base value before the corresponding time period of the unknown device access tag data, and compare the deviation with the loop equivalent impedance sequence; When the equivalent impedance sequence of the loop remains in a preset stable state and the voltage difference sequence descent gradient and the cumulative unknown power increment data show positive covariance in the corresponding time period, an impedance backtracking verification pass mark is generated and superimposed on the private capacity expansion anomaly evidence data.

[0042] Preferably, after generating evidence data of unauthorized capacity expansion, a verification step to prevent collaborative misjudgment is also included: Capture the high-frequency active power sequence of other parallel circuits under the same transformer area, and extract the negative power increment event sequence that matches the time period corresponding to the unknown device access marker data. The negative power increment event sequence is the transient power difference sequence in the parallel circuit where the active power drops sharply. Perform time-axis cross-correlation scans and algebraic sum operations on the absolute value sequences of the steady-state power increment sequence and the negative power increment event sequence representing the access of unknown devices; When the cross-correlation peak reaches the preset synchronization threshold, and the algebraic sum of the sum of steady-state power increment data and the sum of negative power increment event data approaches the preset zero-point band range, the physical phenomenon is defined as load transfer scheduling, and a clearing instruction is generated to block and intercept the current private capacity increase abnormal evidence data and return the value to zero.

[0043] In this embodiment of the invention, the total power increment data is calculated within the target monitoring period. The specific calculation logic is as follows: To define the target monitoring period, the system uses the current judgment time as the endpoint and traces back a fixed time span as a statistical window; this fixed time span can be preset according to the production scheduling pattern of the industrial park, typically taking one natural week. The high-frequency active power sequence within this period is marked as the monitoring dataset.

[0044] Based on this, a sliding piecewise linear fitting method is used to extract the baseline trend line within the monitoring period. After subtracting the trend increment corresponding to the baseline trend line from the net increment of the high-frequency active power sequence from the first sampling point to the last sampling point, the total power increment data truly composed of sudden equipment start-ups and shutdowns and unexplained fluctuations is obtained. The calculation formula is as follows: ; and These are the active power values ​​at the first and last sampling points within the monitoring window, respectively, in kilowatts. The total time span of the monitoring window is measured in hours. The trend slope is obtained by least-squares linear fitting of the power sequence within the monitoring window, with the dimension kilowatt-hour.

[0045] The system extracts the steady-state power increment data of all corresponding power envelope data blocks within the monitoring period only when an unknown device access marker is identified, and performs algebraic accumulation to obtain the cumulative unknown power increment data. By calculating the ratio between the two... This is used to quantify the degree to which unauthorized expansion dominates the overall increase in capacity.

[0046] Determine if the ratio value is greater than a preset ratio threshold. This preset ratio threshold is to avoid misjudgments caused by the company's normal, very small amount of temporary maintenance power consumption; a value of 0.20 is recommended for on-site implementation. When When the limit is exceeded, the system further retrieves the multi-scale autocorrelation function sequence of the current time point and calculates the dynamic time warping distance between it and the baseline data of the previous electricity consumption rhythm. If this distance exceeds the preset tolerance, it indicates that the company's original electricity consumption pattern has been severely disrupted by unknown loads (i.e., attenuation and deformation have occurred), and the system at this time initially generates evidence data of unauthorized capacity expansion.

[0047] The determination of the attenuation deformation is achieved by calculating the standardized Euclidean distance between the multi-scale fused autocorrelation function sequence at the current time point and the electricity consumption rhythm baseline. The system synchronously performs the following processing: First, extract the sequence of multi-scale fused autocorrelation functions calculated in real time within the current sliding window. Compared with the previously established baseline data on electricity consumption patterns The composition is compared.

[0048] Secondly, calculate the deformation deviation of the two using the following formula. :

[0049] L is the total number of maximum delay steps in the autocorrelation function sequence; Used to characterize the overall offset of the power consumption rhythm structure.

[0050] Finally, The data is compared with a preset deformation tolerance threshold. If the former is greater than the latter, the multi-scale autocorrelation function sequence is determined to have undergone attenuation deformation relative to the electricity consumption rhythm baseline data. This threshold can be customized according to the historical production fluctuation characteristics of different tenants.

[0051] To prevent the aforementioned power amplification from being a false power fluctuation caused by overall grid voltage dips or aging and overheating of line terminals, a verification based on the longitudinal physical space perspective of the line is introduced to prevent misjudgment. The specific verification logic is as follows: Synchronously extract the effective value sequence of the first-end voltage corresponding to the current time node. and the effective value sequence of terminal voltage The difference between the two is defined as the voltage drop sequence. Simultaneously, the high-frequency active power sequence Divide by the effective value sequence of terminal voltages The equivalent load current sequence characterizing pure power consumption is approximately calculated. ; The equivalent impedance sequence of the loop is obtained by performing a quotient operation on the two. Next, extract the equivalent impedance base value of the circuit under normal conditions from the historical runtime library. The deviation is compared with the current impedance. To integrate the impedance steady state with the voltage drop gradient for determination, the following calculation formula is used:

[0052] This represents the physical confidence level of the output. The Pearson correlation coefficient between the voltage drop sequence and the unknown power increment sequence is used to quantify the positive physical covariance of voltage drop increase caused by large current. This is a preset dimensionless sensitivity constant, and it is recommended that the value range be between 5.0 and 10.0.

[0053] If the power increase in the above formula is purely due to the connection of an unauthorized load, the physical impedance of the line itself will not change abruptly. In this case, the penalty term approaches 1, and the voltage drop increases in tandem with the load. Approaching 1, physical confidence level Maintain a high level; conversely, if the anomaly is caused by a surge in resistance due to oxidation or rust at the circuit terminals, the penalty term will shrink drastically, causing the physical confidence level to drop to an extremely low level instantly. When When the stability exceeds the preset confidence threshold, the system generates an impedance backtracking verification pass flag and adds it to the private capacity expansion anomaly evidence data.

[0054] In addition, considering that there are often compliant workshop-level load scheduling operations in industrial sites (such as moving a large welding machine from the first floor to the second floor for plug-in operation), in order to prevent such internal switching transfers from being mistaken for unauthorized capacity expansion, a horizontal anti-collaboration misjudgment verification was also initiated simultaneously. The specific verification steps are as follows: Capture the high-frequency active power sequence of other parallel circuits under the same transformer substation, and extract the negative power decrease event sequence with matching time periods. Its corresponding final steady-state increment The value is negative, with units of kW. To quantify whether this increase and decrease pertain to the same physical device, the system inputs both into the formula: in: In the target loop, the power increment time series contained in the power envelope data block marked as accessed by an unknown device is the sequence of active power sample values ​​relative to the power value at the start of the change from the abrupt change to the new steady-state endpoint, with the unit being kilowatts. It is a sequence of absolute values ​​of negative power increment events occurring within a corresponding time period in a parallel circuit. The specific construction method is as follows: extract the power drop event that occurs synchronously with the unknown device access event, take the absolute value of the negative difference between the active power value of each sampling point in the event and the power value at the power change starting point of the change, and form a positive power change sequence with the dimension of kilowatt. Let be the total steady-state power increment of the unknown device access event in the target loop, which is a scalar with the dimension kW; This represents the total steady-state power reduction of this negative power increment event in the parallel circuit, which is a negative value with the dimension of kW; This is a preset balance factor, dimensionless, with a preferred value of 5.0, used to adjust the penalty strength of the energy conservation term on the overall correlation. Time sliding offset; If and only if the correlation degree When the value exceeds the preset synchronization threshold, it indicates that not only are the start-stop waveforms strikingly similar, but the power algebra and value also perfectly approach the preset zero-point range. Therefore, it is defined as a legitimate load transfer scheduling, and a clearing instruction is immediately generated to block and intercept the current abnormal evidence data of unauthorized capacity increase and to completely reset the value to zero.

[0055] Based on the cumulative unknown power increment data and the line parameter matrix, thermal field boundary analysis is performed to generate safe boundary load data. The safe boundary load data is then structurally spliced ​​with the evidence data of unauthorized capacity expansion to output the management judgment result.

[0056] Preferably, thermal field boundary analysis is performed based on the cumulative unknown power increment data and the line parameter matrix to generate safe boundary load data, including: The cumulative unknown power increment data and the current high-frequency active power sequence are summed and converted into estimated current data. Extract conductor cross-section data, laying method coefficient, and rated permissible temperature data from the line parameter matrix; The estimated current data, conductor cross-section data, and laying method coefficient are substituted into the set equivalent thermal circuit differential equations for integration and solution, and the estimated temperature rise time series data within the future preset time window is derived. Extract the highest peak value from the estimated temperature rise time series data. When the highest peak value is greater than the extreme value after the rated permissible temperature data is attenuated according to a preset ratio, calculate the corresponding critical area limit value as the safety boundary load data.

[0057] Preferably, the estimated current data, conductor cross-section data, and laying method coefficient are substituted into a set of equivalent thermal circuit differential equations for integration and solution, thereby deriving the estimated temperature rise time series data for a future preset time window, including: Equivalent thermal resistance parameters and equivalent heat capacity parameters are constructed based on conductor cross-section data and laying method coefficients; The transient temperature change rate is represented by the first-order differential term, and an equivalent thermal circuit differential equation system is constructed, which includes the equivalent thermal resistance parameter, the equivalent heat capacity parameter, and the square term of the estimated current data. Using the ambient reference temperature data as the initial boundary condition, Runge-Kutta numerical iteration is performed on the equivalent thermal circuit differential equations to output the estimated temperature rise time series data, which is composed of multiple discrete time points and corresponding temperature prediction values.

[0058] In this embodiment of the invention, the cumulative unknown power increment data and the steady-state value of the current high-frequency active power sequence are summed to obtain the true total active power consumed by the current circuit.

[0059] By combining the effective value of the terminal voltage collected from the front end with the preset power factor reference value, and using the classic three-phase power backward calculation formula, the pure power value is converted into estimated current data that characterizes the intensity of the directional movement of charge. .

[0060] Extract the conductor cross-section data S and laying method coefficient of the corresponding cable from the underlying line parameter matrix. and the rated permissible temperature data of the cable insulation material. Among them, the laying method coefficient The heat conduction penalty parameter is preset based on long-term engineering experience. It is recommended to set it to an adjustable range of 0.6 to 1.2 according to the actual site conditions.

[0061] Based on this, using the formula Calculate the resistivity per unit length of wire ,in The conductivity constant of the copper / aluminum material solidified in the system, and the resistivity per unit length of wire. The dimension is ohms per meter.

[0062] Meanwhile, based on the conductor cross-section data S and the laying method coefficient The equivalent thermal resistance parameters were derived. With equivalent heat capacity parameters The formula is defined as follows: ; in Let be the transient temperature of the target cable at time t; the first term on the right side of the equation The dimensionless Joule heating rate per unit length driven by the estimated current is given by... That is, watts per meter; the second term on the right side of the equation is the Newtonian cooling term for heat dissipation from the cable to the surrounding environment. The ambient reference temperature data is used. The equivalent heat capacity parameter is shown on the left side of the equation. The dimension of is joules per meter.

[0063] After establishing the equations, the system sets the ambient reference temperature data. Using the fourth-order Runge-Kutta numerical iterative algorithm as the initial boundary conditions, the temperature trajectory within a preset time window is integrally solved, thus outputting a predicted temperature rise time series consisting of multiple discrete time points and their corresponding predicted temperature values. The highest peak value is then extracted from this predicted temperature rise time series. and compare it with the rated permissible temperature data. Preset ratio The extreme values ​​after attenuation are compared. The preset ratio here... The recommended value is 0.90, whose preset logic is to forcibly retain a 10% heat margin to cope with sudden high-temperature heat waves in summer. When the predicted peak data... This indicates that continuing to operate the current unknown expansion equipment will inevitably lead to irreversible thermal breakdown of the cable insulation layer. At this point, the system immediately performs steady-state inversion calculations on the above differential equations, setting the heat accumulation rate on the left side... The critical safe current is derived by reversing the terms and taking the square root. Then, multiply back by the corresponding voltage system to calculate the load limit value corresponding to this critical area limit, which is used as the safety boundary load data.

[0064] It should be added that, based on the conductor cross-section data S and the laying method coefficient... Constructing equivalent thermal resistance parameters and equivalent heat capacity parameters The specific construction formula is as follows: Equivalent thermal resistance parameters The calculation formula is: ; Equivalent heat capacity parameters The calculation formula is: ; in: This refers to the outer diameter of the conductor, in meters. It can be calculated by converting the conductor cross-sectional area S to a circular cross-section. This refers to the total outer diameter of the conductor, including the insulation layer, in meters. It can be obtained by referring to the conductor standard specification table. Thermal conductivity of insulating materials, measured in watts per meter per degree Celsius, is determined based on the material of the conductor insulation layer. The specific heat capacity of the conductor material is expressed in joules per kilogram per degree Celsius; for copper conductors it is 385, and for aluminum conductors it is 900. Density of the conductor material, expressed in kilograms per cubic meter; Specific heat capacity of insulating materials, in units of The determination is based on the insulation material; Density of insulating material, unit: The determination is based on the insulation material; This is the laying method coefficient, used to correct for heat dissipation conditions under different laying environments, and its value range is [value range missing]. When the wires are laid in a sealed metal cable tray, Take 0.6; when the wire is laid openly in free air, Take 1.2; for other laying methods, linear interpolate within the interval according to the heat dissipation conditions.

[0065] The equivalent thermal resistance parameter The thermal resistance per unit length of a conductor to its surroundings, measured in degrees Celsius per meter per watt; the equivalent heat capacity parameter It is the equivalent heat capacity per unit length of conductor and its insulation layer, with dimensions of joules per degree Celsius per meter.

[0066] Example 2: like Figure 2 As shown, the enterprise intelligent management system based on big data includes a data acquisition module, a rhythm modeling module, a feature recognition module, an incremental judgment module, and a boundary output module. The various modules are connected via wired and / or wireless connections to enable data transmission between them; The data acquisition module is used to acquire the multidimensional electrical parameter data sequence of the target circuit and the corresponding line parameter matrix. The multidimensional electrical parameter data sequence includes at least the high-frequency active power sequence and the terminal voltage RMS value sequence. The rhythm modeling module is used to perform iterative calculations of the high-frequency active power sequence using a sliding window at preset multiple time scales, extract the corresponding multi-scale autocorrelation function sequence, and generate baseline data of electricity consumption rhythm through fusion and weighting. At the same time, it performs forward differential operation on the high-frequency active power sequence to generate a differential power sequence. When the value of the differential power sequence exceeds the preset dynamic jump threshold, it adaptively extracts the power envelope data block of the corresponding time interval. The feature recognition module is used to parse the power envelope data block, extract the transient feature vector set, and compare the transient feature vector set with the device feature map library through similarity vector mapping. If no match is found, unknown device access marker data is generated. The incremental determination module is used to calculate the total power increment data within the target monitoring period based on the high-frequency active power sequence. When the unknown device access marker data is identified, the module accumulates the steady-state power increment data corresponding to all power envelope data blocks within the target monitoring period to obtain the cumulative unknown power increment data. The module also calculates the proportion of the cumulative unknown power increment data to the total power increment data. When the proportion is greater than a preset proportion threshold and the multi-scale autocorrelation function sequence at the current time node decays and deforms relative to the electricity consumption rhythm baseline data, the module generates evidence data of unauthorized capacity expansion. The boundary output module is used to perform thermal field boundary analysis based on the cumulative unknown power increment data and the line parameter matrix, generate safe boundary load data, and structurally splice the safe boundary load data with the evidence data of unauthorized capacity expansion to output the management judgment result.

[0067] Example 3: As a further improvement to the present invention, when the access transient feature matching is successful, but the system is still suspected of unauthorized capacity expansion, a secondary fine identification mechanism based on steady-state harmonic feature spectrum is initiated.

[0068] In the logic for generating unknown device access markers, if the calculated minimum distance marker value is not greater than the preset safety boundary radius parameter, but the relevant circuit still triggers subsequent evidence data of unauthorized capacity increases (i.e., highly abnormal power consumption rhythm and power increment characteristics), the system determines that there may be an access event of a non-registered device of the same model. At this time, the following secondary verification steps are initiated: Step 1: After the new steady-state endpoint data of the power envelope data block, extract a segment of steady-state high-frequency current waveform data of a preset length (e.g., 10 power frequency cycles). Perform a Fast Fourier Transform on this steady-state current waveform to extract the amplitudes of the fundamental wave and the 3rd, 5th, 7th, 9th, and 11th odd harmonics. Construct a high-dimensional steady-state harmonic feature vector. ,in The content of the nth harmonic current.

[0069] Step 2: Retrieve the steady-state harmonic fingerprint reference vector of the matched legitimate device from the spectrum library. Calculate the currently extracted vector H and... The weighted Euclidean distance between them. If this distance is greater than the preset harmonic fingerprint offset threshold, it is determined that although the currently operating device is the same model as the legitimate device, there are significant differences in its individual electrical characteristics. Therefore, an unknown device access mark is still generated and a suspected unregistered label of the same model is attached.

[0070] Step 3: In the method for constructing the device feature map library, for each legally registered device, in addition to recording its startup transient feature vector cluster, it is also necessary to record its steady-state harmonic feature vector after steady-state operation for no less than 5 minutes under rated operating conditions, as the unique electrical fingerprint of the individual device and enter it into the library.

[0071] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0072] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0073] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

[0074] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A big data-based intelligent enterprise management method, characterized in that, include: Obtain the multidimensional electrical parameter data sequence of the target circuit and the corresponding line parameter matrix. The multidimensional electrical parameter data sequence shall include at least the high-frequency active power sequence and the terminal voltage RMS value sequence. The high-frequency active power sequence is subjected to a sliding window iterative calculation with preset multi-time scales to extract the corresponding multi-scale autocorrelation function sequence, and the power consumption rhythm baseline data is generated by fusion and weighting. A forward differential operation is performed on the high-frequency active power sequence to generate a differential power sequence. When the value of the differential power sequence exceeds the preset dynamic jump threshold, the power envelope data block of the corresponding time interval is adaptively extracted. The power envelope data block is parsed to extract a set of transient feature vectors, and the set of transient feature vectors is compared with the device feature map library by similarity vector mapping. If no match is found, unknown device access marker data is generated. Calculate the total power increment data within the target monitoring period based on the high-frequency active power sequence; When an unknown device accesses the marked data, the steady-state power increment data corresponding to all power envelope data blocks within the target monitoring period are accumulated to obtain the cumulative unknown power increment data. Calculate the proportion of cumulative unknown power increment data to total power increment data. When the proportion is greater than a preset proportion threshold and the multi-scale autocorrelation function sequence at the current time node decays and deforms relative to the electricity consumption rhythm baseline data, generate evidence data of unauthorized capacity expansion. Based on the cumulative unknown power increment data and the line parameter matrix, thermal field boundary analysis is performed to generate safe boundary load data. The safe boundary load data is then structurally spliced ​​with the evidence data of unauthorized capacity expansion to output the management judgment result.

2. The enterprise intelligent management method based on big data according to claim 1, characterized in that, The high-frequency active power sequence is subjected to a sliding window iterative calculation with preset multi-time scales to extract the corresponding multi-scale autocorrelation function sequence, and the electricity consumption rhythm baseline data is generated by fusion and weighting, including: For high-frequency active power sequences, sampling interval sequences matching one natural day, one natural week, and one natural month are used as sliding windows to extract targeted first-scale autocorrelation function sequences, second-scale autocorrelation function sequences, and third-scale autocorrelation function sequences. Calculate the standard deviation of the load rate of the high-frequency active power sequence within the current sliding window; The standard deviation of the load rate is used as an independent variable to input into a preset nonlinear activation function, which inversely maps out the dynamic weight coefficient vectors corresponding to the first-scale autocorrelation function sequence, the second-scale autocorrelation function sequence, and the third-scale autocorrelation function sequence. By using dynamic weight coefficient vectors to perform dot product fitting on autocorrelation function sequences at three scales, baseline data of electricity consumption rhythm expressed by feature matrices are generated.

3. The enterprise intelligent management method based on big data according to claim 1, characterized in that, The power envelope data block is parsed to extract a set of transient feature vectors, including: In the power envelope data block, locate the data at the starting point of the mutation that triggers the threshold to exceed the limit, as well as the data at the new steady-state endpoint after the mutation that enters a range of numerical fluctuations smaller than the preset steady-state residual value. Extract the power value difference between the new steady-state endpoint data and the abrupt change initiation point data as a steady-state increment feature; Within the time span between the mutation initiation point data and the new steady-state endpoint data, the maximum power peak data is extracted, and the ratio of the maximum power peak data to the mutation initiation point data is calculated as the initiation shock multiple feature. Calculate the time difference between the data rising from the mutation initiation point to the numerical node where the corresponding 90% steady-state increment feature is located, and use it as the rising slope feature. By combining the steady-state incremental features, the initiation impact multiple features, and the rising slope features in a predetermined dimension order, a set of transient feature vectors for the current out-of-limit event is constructed.

4. The enterprise intelligent management method based on big data according to claim 1, characterized in that, The transient feature vector set is compared with the device feature map library using similarity vector mapping. If no match is found, unknown device access marker data is generated, including: Retrieve the registered and legitimate equipment feature vector clusters from the equipment feature map library; Calculate the multidimensional Euclidean distance between the transient feature vector set and each legitimate device feature vector cluster; Minimum value search is performed on all multidimensional Euclidean distance values ​​to obtain the minimum distance identifier value; When the minimum distance identifier value is greater than the preset safety boundary radius parameter, it is determined that a non-registered load merging event has occurred, and the current time point and transient feature vector set are encapsulated as unknown device access marker data.

5. The enterprise intelligent management method based on big data according to claim 1, characterized in that, The multidimensional electrical parameter data sequence also includes the sequence of effective values ​​of the first-terminal voltage, and before calculating the proportion of the cumulative unknown power increment data to the total power increment data, it also includes: Simultaneously extract the effective value sequence of the first-end voltage and the effective value sequence of the last-end voltage corresponding to the current time node, and perform quotient calculation on the voltage difference sequence between the two and the load current sequence calculated corresponding to the high-frequency active power sequence to generate the equivalent impedance sequence of the loop. Extract the historical equivalent impedance base value before the corresponding time period of the unknown device access tag data, and compare the deviation with the loop equivalent impedance sequence; When the equivalent impedance sequence of the loop remains in a preset stable state and the voltage difference sequence descent gradient and the cumulative unknown power increment data show positive covariance in the corresponding time period, an impedance backtracking verification pass mark is generated and superimposed on the private capacity expansion anomaly evidence data.

6. The enterprise intelligent management method based on big data according to claim 1, characterized in that, Thermal field boundary analysis is performed based on the cumulative unknown power increment data and the line parameter matrix to generate safe boundary load data, including: The cumulative unknown power increment data and the current high-frequency active power sequence are summed and converted into estimated current data. Extract conductor cross-section data, laying method coefficient, and rated permissible temperature data from the line parameter matrix; The estimated current data, conductor cross-section data, and laying method coefficient are substituted into the set equivalent thermal circuit differential equations for integration and solution, and the estimated temperature rise time series data within the future preset time window is derived. Extract the highest peak value from the estimated temperature rise time series data. When the highest peak value is greater than the extreme value after the rated permissible temperature data is attenuated according to a preset ratio, calculate the corresponding critical area limit value as the safety boundary load data.

7. The enterprise intelligent management method based on big data according to claim 6, characterized in that, By substituting the estimated current data, conductor cross-section data, and laying method coefficients into the set of equivalent thermal circuit differential equations and integrating them, the predicted temperature rise time series data for a future preset time window is derived, including: Equivalent thermal resistance parameters and equivalent heat capacity parameters are constructed based on conductor cross-section data and laying method coefficients; The transient temperature change rate is represented by the first-order differential term, and an equivalent thermal circuit differential equation system is constructed, which includes the equivalent thermal resistance parameter, the equivalent heat capacity parameter, and the square term of the estimated current data. Using the ambient reference temperature data as the initial boundary condition, Runge-Kutta numerical iteration is performed on the equivalent thermal circuit differential equations to output the estimated temperature rise time series data, which is composed of multiple discrete time points and corresponding temperature prediction values.

8. The enterprise intelligent management method based on big data according to claim 1, characterized in that, After generating evidence data of unauthorized capacity expansion, a verification step to prevent collaborative misjudgment is also included: Capture the high-frequency active power sequence of other parallel circuits under the same transformer area, and extract the negative power increment event sequence that matches the time period corresponding to the unknown device access marker data. The negative power increment event sequence is the transient power difference sequence in the parallel circuit where the active power drops sharply. Perform time-axis cross-correlation scans and algebraic sum operations on the absolute value sequences of the steady-state power increment sequence and the negative power increment event sequence representing the access of unknown devices; When the cross-correlation peak reaches the preset synchronization threshold, and the algebraic sum of the sum of steady-state power increment data and the sum of negative power increment event data approaches the preset zero-point band range, the physical phenomenon is defined as load transfer scheduling, and a clearing instruction is generated to block and intercept the current private capacity increase abnormal evidence data and return the value to zero.

9. The enterprise intelligent management method based on big data according to claim 1, characterized in that, The process of generating the preset dynamic transition threshold includes: Extract a fixed-length historical sliding sequence segment ahead of the current acquisition point of the high-frequency active power sequence; Calculate the numerical discrete variance of historical sliding sequence segments; The discrete variance of the data is multiplied by a preset sensitivity multiplier coefficient, and the resulting product is used as the preset dynamic jump threshold for the current acquisition period, so that the preset dynamic jump threshold is adaptively scaled according to the background fluctuation of the power grid.

10. A big data-based enterprise intelligent management system, characterized in that: It is based on the enterprise intelligent management method based on big data according to any one of claims 1-9, including a data acquisition module, a rhythm modeling module, a feature recognition module, an incremental judgment module, and a boundary output module.

Images