Reactive power compensation energy efficiency optimization method and system based on big data interval analysis

Abstract

This application relates to the field of power grid distribution technology, and in particular to a method and system for optimizing reactive power compensation efficiency based on big data interval analysis. The method includes: dividing the distribution system into intervals by selecting at least one feature dimension based on historical operating data to obtain multiple typical operating condition intervals; setting different target power factors for the reactive power compensation device and collecting operating data of the distribution system under different target power factors as optimization samples; for any typical operating condition interval, obtaining the active power on the power supply side in the corresponding optimization sample, and obtaining the representative power consumption value of the active power on the power supply side under the target power factor; determining the long-term optimal value for the typical operating condition interval; and using the target power factor corresponding to the long-term optimal value of each typical operating condition interval as the optimal power factor to control the reactive power compensation device. This application has the effect of maintaining optimal energy consumption of the distribution system during long-term operation.

Description

Technical Field

[0001] This application relates to the field of power grid distribution technology, and in particular to a method and system for optimizing reactive power compensation efficiency based on big data interval analysis. Background Technology

[0002] During the operation of a power distribution system, the load typically includes electromagnetic equipment such as motors and transformers. When these devices operate, the current phase lags behind the voltage, resulting in reactive power in the system. To reduce line current, minimize transmission losses, and meet the power factor requirements of the power grid, reactive power compensation devices (such as parallel capacitor banks and static var generators) are usually configured in engineering projects to compensate and control the system.

[0003] In existing technologies, reactive power compensation control mostly adopts a fixed target power factor control method, that is, a target power factor value (such as 0.90, 0.95, or 0.98) is preset, and the power factor during system operation is maintained near the set value by controlling the switching or continuous adjustment of the reactive power compensation device. This type of method is simple to implement and can reduce line current to a certain extent and meet the power grid assessment requirements.

[0004] However, control methods with a fixed target power factor are primarily based on empirical settings or performance indicators, focusing on the power factor itself rather than the total active power of the system. In actual operation, while increased reactive power compensation may reduce line current, it can lead to increased voltage at the line's end, altering the operating state of motors and other electrical equipment and increasing their losses. In this situation, the total active power of the system may not necessarily decrease with an improved power factor, and could even increase. Therefore, a control strategy solely focused on improving the power factor is insufficient to guarantee optimal overall system energy efficiency. Summary of the Invention

[0005] To address the problem that reactive power compensation control methods in related technologies cannot guarantee optimal overall energy consumption of the power distribution system, this application provides a reactive power compensation energy efficiency optimization method and system based on big data interval analysis.

[0006] Firstly, this application provides a reactive power compensation efficiency optimization method based on big data interval analysis, employing the following technical solution: A reactive power compensation efficiency optimization method based on big data interval analysis includes: selecting at least one feature dimension to divide the power distribution system into intervals based on historical operating data of the power distribution system, thereby obtaining multiple typical operating condition intervals; wherein the feature dimension includes at least one of time dimension, current dimension or power dimension; In the typical operating condition ranges during the actual operation of the power distribution system, different target power factors are set for the reactive power compensation device, and the operating data of the power distribution system under different target power factors are collected as optimization samples; the operating data includes at least: the active power on the power supply side of the power distribution system; For any typical operating condition range, obtain the active power on the power supply side in the corresponding optimization sample; for any target power factor, obtain the power consumption representative value that can represent the active power level on the power supply side under the target power factor; determine the long-term optimal value of the typical operating condition range based on the minimum value of the power consumption representative value corresponding to different target power factors. The target power factor corresponding to the long-term optimal value of each typical operating condition range is used as the optimal power factor of each typical operating condition range, and is used to control the reactive power compensation device.

[0007] The load structure of a power distribution system varies significantly across different time periods. For example, the electrical characteristics differ greatly between start-up / shutdown phases and stable production phases, as well as during the start-up and shutdown of large equipment in production. Therefore, typical operating condition intervals are defined based on historical operating data. These intervals can be time periods or the electrical characteristics of different operating conditions (e.g., different current ranges). This interval division ensures that each optimization process is conducted within a relatively stable load environment, improving the reliability of the optimization results. Secondly, different target power factors are set within each typical operating condition interval, and operating data is collected. By actively scanning different power factors, the actual operating state of the system is used as feedback in the optimization process, thereby avoiding deviations caused by empirical settings.

[0008] Furthermore, using active power on the power supply side as the optimization basis is designed because active power directly reflects the total energy consumption level of the system, truly demonstrating the energy efficiency optimization effect, thus shifting the optimization goal from meeting performance targets to achieving optimal energy efficiency. A representative power consumption value is constructed, and the minimum value is selected to determine the long-term optimal value, enabling a comparison of energy consumption corresponding to different power factors, thereby finding the operating point with the lowest energy consumption. Finally, the optimal power factor is used to control the reactive power compensation device, ensuring the system operates in a state of optimal energy consumption over the long term. This scheme replaces experience-based settings with a data-driven approach, reducing the increase in energy consumption caused by over-compensation or under-compensation, and optimizing the overall energy consumption level of the power distribution system.

[0009] Optionally, based on the historical operating data of the power distribution system, at least one feature dimension is selected for interval division to obtain multiple typical operating condition intervals, including: selecting the feature dimension for interval division, dividing the feature dimension into multiple candidate intervals representing different operating conditions of the power distribution system based on historical operating data; calculating the relative rate of change of the historical operating data of the power distribution system in each candidate interval; and taking the candidate interval with a relative rate of change less than a preset change threshold as the typical operating condition interval.

[0010] Based on a certain feature dimension, the data of this feature dimension is divided into multiple candidate intervals, providing an initial division criterion. Then, by calculating the relative rate of change of historical operating data within each candidate interval, the fluctuation degree of parameters such as voltage, current, and active power is quantified, thereby reflecting the operating condition stability within that interval. Candidate intervals with a relative rate of change less than a preset threshold are designated as typical operating condition intervals, achieving automatic screening of stable intervals. The technical effect is to eliminate intervals with drastic load fluctuations, avoid optimization calculations under non-steady-state conditions, and thus improve the reliability of subsequent optimization results.

[0011] Optionally, different target power factors are set for the reactive power compensation device, and the operating data of the power distribution system under different target power factors are collected as optimization samples, including: making the target power factor monotonically increase or monotonically decrease within a preset scanning interval with a preset fixed step size.

[0012] A scanning interval is set to limit the optimization range, avoiding meaningless exploration of extreme values ​​while ensuring that the requirements of power grid operation are met. A fixed step size is used to monotonically increase or decrease, ensuring the continuity and controllability of the scanning process, so that the operating data corresponding to different power factors can form an ordered sequence, which is convenient for subsequent comparison and analysis.

[0013] Optionally, different target power factors are set for the reactive power compensation device, and the operating data of the power distribution system under different target power factors are collected as optimization samples. This includes: making the target power factor monotonically increase or decrease within a preset scanning interval according to a preset coarse step size, while collecting the operating data of the power distribution system; constructing a coarse interval based on the time when the minimum active power on the power supply side is located in the operating data; and making the target power factor monotonically increase or decrease within the coarse interval according to a preset fine step size, where the coarse step size is greater than the fine step size.

[0014] By using coarse-step scanning to quickly locate approximate regions with low energy consumption, high-precision traversal of the entire region is avoided, thus reducing time costs. A coarse region is constructed based on the minimum active power, and the region is narrowed using the preliminary scan results, focusing subsequent calculations on the region most likely to contain the optimal solution, thereby improving search efficiency.

[0015] Optionally, it also includes: filtering the sampling points in the optimization sample based on preset conditions, and removing the sampling point if the sampling point does not meet the preset conditions; Among them, the selection of sampling points in the optimization sample based on preset conditions includes: calculating multiple evaluation indicators of the sampling points, comparing the evaluation indicators with preset conditions to determine whether the sampling points meet the preset conditions.

[0016] In actual data collection, noise, outliers, and external interference are unavoidable. Directly using these in subsequent calculations can lead to biased results. By filtering sampling points based on preset conditions and removing data that does not meet the conditions, data preprocessing is achieved, thus eliminating outliers.

[0017] Optionally, the evaluation metrics include at least one of the following: voltage deviation, which characterizes the degree of deviation of the sampling point voltage from the nominal voltage of the distribution system; current deviation, which characterizes the degree of fluctuation of the sampling point current; voltage distortion rate; or current distortion rate.

[0018] Voltage deviation reflects voltage quality, current deviation reflects load fluctuation, and distortion rate reflects power quality. By introducing these indicators, a comprehensive assessment of the operating status of sampling points can be achieved, accurately identifying various abnormal conditions such as voltage anomalies, load surges, and harmonic interference, thereby improving the precision of data screening.

[0019] Optionally, for any target power factor, obtain a power consumption representative value that can represent the active power level of the power supply side under the target power factor, including: dividing the multiple active power of the power supply side corresponding to the same target power factor into multiple boxes, taking the box containing the most active power of the power supply side as the mode box; and taking the median of all active power of the power supply side contained in the mode box as the power consumption representative value.

[0020] Binning discretizes continuous data to identify densely distributed data regions. Secondly, the bin with the largest sample size is selected as the mode bin, which best represents the stable operating state of the system. Furthermore, using the median as the representative value avoids the influence of extreme values ​​on the results.

[0021] Optionally, it also includes: during the operation of the power distribution system, when any two typical operating condition intervals switch, using a linear transition method to control the target power factor of the reactive power compensation device to smoothly transition from the set value of the current typical operating condition interval to the set value of the next typical operating condition interval.

[0022] The optimal power factor may vary significantly across different operating conditions. Direct switching could cause transient shocks to the system. A gradual transition using a linear transition method allows the target value to change smoothly.

[0023] Optionally, it also includes: during the operation of the power distribution system, when the relative rate of change of active power on the power supply side in any typical operating condition interval is greater than the preset correction threshold, re-acquiring the long-term optimal value corresponding to the typical operating condition interval; and using the optimal power factor corresponding to the newly acquired long-term optimal value to control the reactive power compensation device.

[0024] The load on a power distribution system changes with production fluctuations, and the original optimal values ​​may become invalid. By setting thresholds to trigger re-optimization, the system can respond promptly to load changes, maintain the system in a state of optimal energy consumption over the long term, avoid optimization failure due to changes in operating conditions, and thus improve the long-term effectiveness of the solution.

[0025] Secondly, this application provides a reactive power compensation efficiency optimization system based on big data interval analysis, which adopts the following technical solution: A reactive power compensation efficiency optimization system based on big data interval analysis includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned reactive power compensation efficiency optimization method based on big data interval analysis.

[0026] The aforementioned reactive power compensation efficiency optimization method based on big data interval analysis is generated into a computer program and stored in a memory for loading and execution by a processor. Thus, a system is built based on the memory and processor for convenient use.

[0027] This application has the following technical effects: By scientifically dividing stable typical operating condition ranges using historical data, and actively stepping scans and multiple sampling within each range to obtain the actual active power consumption under different power factors, the optimal power factor that minimizes the system's true total energy consumption is determined for closed-loop control. This enables long-term adaptive energy-saving operation of the distribution network under complex dynamic load scenarios, improving overall energy efficiency. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of a reactive power compensation efficiency optimization method based on big data interval analysis according to an embodiment of this application. Detailed Implementation

[0029] This application discloses a reactive power compensation efficiency optimization method based on big data interval analysis. The method involves dividing historical data into multiple independent typical operating condition intervals through stability assessment; secondly, actively adjusting the target power factor within each interval for step scanning and multiple sampling; next, cleaning invalid and abnormal data using multi-dimensional judgment conditions; then, using the mode method to obtain a representative daily power consumption value; subsequently, collecting representative power consumption values ​​from multiple days for statistical calculation to solidify them into standard values; finally, sending the standard value sequence to the reactive power compensation controller for timing control, and periodically polling and correcting during operation to adapt to long-term load changes.

[0030] Reference Figure 1 A reactive power compensation efficiency optimization method based on big data interval analysis includes steps S1-S4.

[0031] S1: Based on the historical operating data of the power distribution system, at least one feature dimension is selected to divide the intervals, resulting in multiple typical operating condition intervals; wherein, the feature dimension includes at least one of the time dimension, current dimension, or power dimension.

[0032] Collect data for a preset historical time period (preferably in this embodiment). The power distribution system operation data within (days) should include at least: active power, voltage, current, and load-side power factor (before reactive power compensation). Additionally, the total harmonic distortion (THD) of voltage (the ratio of the effective values ​​of all harmonic components in the voltage signal to the effective value of the fundamental frequency) and the total harmonic distortion (THD_I) of current (the ratio of the effective values ​​of all harmonic components in the current signal to the effective value of the fundamental frequency) can also be collected.

[0033] At least one characteristic dimension is selected for interval division, resulting in multiple typical operating condition intervals. The characteristic dimension includes at least one of time, current, or power dimensions. Taking the time dimension as an example: a preset historical period is divided into multiple time cycles. Based on characteristics such as the fluctuation range and trend of active power on the power supply side and the change in power factor on the load side, the time cycle is further divided into multiple candidate intervals. For example, the preset time cycle can be one day, i.e. It is divided into several candidate intervals.

[0034] The number and boundaries of candidate intervals are initially determined by staff based on experience. For example, if a factory operates on a three-shift, continuous production model, it can be divided into the following candidate intervals based on the actual production situation: Interval 1: Within this section of the workshop, only basic lighting and auxiliary power are available, resulting in an extremely low load rate.

[0035] Zone 2: (Early morning shift warming zone): This period is the production start-up period, during which large inductive loads (such as injection molding machine heating coils and air compressors) are started up in a concentrated manner, resulting in the largest current rise rate.

[0036] Zone 3: (Stable production zone for the early shift): Within this range, the load reaches its rated power, resulting in minimal power fluctuation.

[0037] Section 4: (Lunch break waiting area) In this section, some of the main equipment was unloaded due to workers having lunch.

[0038] Interval 5: (Afternoon stable production interval): .

[0039] Section 6 (Evening Production Section): .

[0040] Of course, the division of candidate intervals can be adjusted according to the actual production situation. For example, if no production activities are carried out on Saturdays and Sundays, the collected historical operational data can be divided into two categories: weekday data and holiday data. Then, candidate intervals can be divided for each of the two categories of data.

[0041] For example, the candidate intervals for holidays can be directly divided into: Interval 1 (early morning duty interval): ; Section 2 (Daytime Maintenance Section): ; Section 3 (Night Watch Section): .

[0042] To verify the rationality of the manually divided candidate intervals, for any candidate interval within any time period, the relative change rate of data in each dimension is used to screen the candidate interval as a typical working condition interval.

[0043] Specifically, for the operational stability of any candidate interval, the relative rate of change of each dimension of data within the candidate interval is first calculated.

[0044] In one embodiment, the formula for calculating the relative rate of change of any dimension of data within the candidate interval can be expressed as: ; Indicates the first in the running data The relative rate of change of the data items; Indicates the first candidate in the interval The maximum value of the data item; Indicates the first candidate in the interval The minimum value of the data item; Indicates the first candidate in the interval The average value of the data items.

[0045] In other embodiments, the relative rate of change of any dimension of data can also be represented by the coefficient of variation (i.e., the ratio of the standard deviation to the mean of the data in that dimension).

[0046] By following the steps above, the relative change rates of each dimension of the operational data in any candidate interval within any time period can be obtained. To reduce the error in calculating the relative change rate caused by random data disturbances, this embodiment uses the average of the relative change rates of the same dimension for the same candidate interval across multiple time periods as the standard change rate. Then, candidate intervals are selected as typical operating condition intervals based on the standard change rates of each dimension.

[0047] If the relative rate of change of data in each dimension is less than a preset threshold, the operating condition within the candidate interval is considered stable, and this candidate interval is designated as a typical operating condition interval. Otherwise, this candidate interval is designated as an interval to be adjusted. Subsequently, the boundaries of the interval to be adjusted are adjusted based on the experience of the staff (e.g., by surveying and generating actual production rhythms in the workshop). Simultaneously, the standard rate of change is recalculated to further filter the interval until it meets the requirement that the standard rate of change of each dimension is less than the preset threshold. The interval to be adjusted is then designated as a candidate interval. The preset threshold is set based on the staff's experience; for example, it can be set between 5% and 15%, such as 10%.

[0048] It is understandable that during the adjustment of the boundary of the adjustment range, the typical operating condition range does not need to cover the entire preset time period. During the adjustment process based on experience, there may be periods of short duration with drastic fluctuations in electrical quantities caused by production switching or the instantaneous switching on and off of large loads between two typical operating condition ranges.

[0049] For the current dimension, this refers to dividing the historical operating data into multiple current ranges as typical operating condition ranges based on the magnitude of the current during operation. During this division, operators can directly divide the current data into operating condition ranges by combining waveform diagrams, forming multiple current ranges, which are then considered typical operating condition ranges. Alternatively, unsupervised clustering methods (such as DBSCAN clustering) can be used to cluster the current data in the operating data, resulting in multiple clusters. The range formed by the maximum and minimum current values ​​within each cluster is then used as the typical operating condition range.

[0050] S2: In the typical operating conditions of the power distribution system during actual operation, different target power factors are set for the reactive power compensation device, and the operating data of the power distribution system under different target power factors are collected as optimization samples.

[0051] In one embodiment: In each typical operating condition range during the actual operation of the power distribution system, different target power factors are set for the reactive power compensation device (i.e., the target of the reactive power compensation device for power factor control). The target power factor starts from the preset lower limit of the scan and increases monotonically according to the preset fixed step size until it reaches the upper limit of the scan, so as to traverse the entire scan range formed by the upper and lower limits of the scan, or decreases monotonically from the upper limit of the scan to the lower limit of the scan until it reaches the lower limit of the scan. In actual implementation, it is often a cycle between the upper and lower limits.

[0052] It is understandable that, for typical operating condition intervals in the time dimension, when the current time period is within the typical operating condition interval, the power distribution system is considered to be operating within the typical operating condition interval. Similarly, for typical operating condition intervals in the current or other dimensions, when the current level of the current power distribution system is within the current interval corresponding to the typical operating condition interval, the current power distribution system is considered to be operating within the typical operating condition interval.

[0053] In this embodiment, the lower limit of the target power factor setting value is set to 0.920. The advantage of setting the lower limit to 0.920 is that this value is higher than the conventional power grid assessment standard of 0.90, ensuring that the system does not violate the power sector's compliance requirements at any scanning time. The upper limit of the target power factor setting value is set to 0.999. The reason for setting the upper limit to 0.999 (or 0.990 depending on the tolerance of specific equipment) is to reduce excessive reactive power compensation that could lead to abnormal voltage increases at the end of the distribution network, causing insulation aging or damage to electrical equipment.

[0054] Reactive power compensation devices can generally be divided into two categories. One type uses stepped capacitor banks to switch in combination via contactors or thyristors. Since the smallest adjustment unit (i.e., the capacity of a single capacitor bank) for this type of device typically corresponds to a power factor change of around 0.01, if the step size is set too small (e.g., 0.002), multiple consecutive commands may fail to trigger because the physical switching threshold is not reached.

[0055] For another type of reactive power compensation device with continuous adjustment capability (such as SVG (Static Var Generator) or MCR (Magnetically Controlled Reactor)), the preferred step size is... .

[0056] After each step (i.e., entering a new scanning range), the reactive power compensation device maintains stable operation at that scanning range for a preset duration. The preset duration can be set to 60 seconds. This preset duration ensures that electrical quantities such as voltage and current in the line return to a steady state after the transient fluctuations caused by the switching of the reactive power compensation device, thus improving the representativeness of the collected data.

[0057] For any given scanning range, during its stable operating period, to facilitate description, this period is considered a stable operating phase. Initial samples are obtained by repeatedly collecting and recording power distribution system operating data at a constant sampling rate. In one embodiment, multiple samples are taken at a fixed frequency within a preset duration (e.g., 10 seconds) before the end of the stable operating phase. The sampled operating data includes: active power on the power supply side, actual power factor, voltage, current, and other parameters, for subsequent cleaning and verification.

[0058] In another embodiment: Within each typical operating condition range of the power distribution system during actual operation, a target power factor is set for the reactive power compensation device. Starting from a pre-set lower limit of the scan, the factor is monotonically increased according to a pre-set coarse step size, while simultaneously collecting operating data of the power distribution system (the data collection method is also the last 10 seconds of the stable operation period of the reactive power compensation device), until the upper limit of the scan is reached. A coarse interval is determined based on the active power. Within the coarse interval, the process is traversed according to a pre-set fine step size, while simultaneously collecting operating data of the power distribution system. The operating data of the power distribution system obtained from the two collections are used as optimization samples. The coarse step size is larger than the fine step size; for example, the coarse step size can be set to 0.01, and the fine step size can be set to 0.001.

[0059] The steps for determining the coarse interval based on active power include: For any scanning range, obtain the minimum active power in the power distribution system operation data collected under that scanning range, and construct a coarse interval based on the time when the minimum active power is located. For example, the interval can be expanded by a coarse step length on both sides of the time when the minimum is located, and the expanded interval is used as the coarse interval.

[0060] While maintaining the same optimization accuracy, the total scanning time is significantly shortened, enabling the optimization process to be completed quickly within a typical operating range. For example, in a high-precision adjustment scenario (e.g., step size 0.001), a full-range traversal would result in an excessively long scanning range (e.g., 80 increments are needed to go from 0.920 to 0.999, and if each increment includes a 1-minute steady-state observation period, it would take 80 minutes). In this embodiment, the scanning range is first traversed with a coarse step size of 0.01 to determine the coarse range, and then traversed within the coarse range with a fine step size, thereby shortening the scanning cycle, improving scanning efficiency, and quickly finding the target power factor when the active power is at its lowest.

[0061] In some embodiments, the directly acquired running data may contain some noise, so abnormal data can be removed from the running data, and the running data after removing abnormal data can be used as the optimization sample.

[0062] Specifically, in this embodiment, abnormal data in each dimension of the optimization sample is removed according to preset rules to obtain the optimization sample.

[0063] Preset rules include the following categories: For any sampling point, calculate multiple evaluation indicators for that sampling point. If any evaluation indicator of the sampling point fails to meet the preset conditions, the sampling point is considered to be abnormal data.

[0064] Evaluation indicators include: Voltage deviation: For any sampling point, calculate the absolute value of the difference between the sampling point and the nominal voltage of the power distribution system, and use the ratio of this absolute value to the nominal voltage as the voltage deviation.

[0065] Current deviation: For any sampling point, obtain the mean value of the current of each sampling point during the stable operation period of the current scanning mode, calculate the absolute value of the difference between the current of the sampling point and the mean value of the current, and use the ratio of the absolute value to the current of the sampling point as the current deviation.

[0066] The voltage and current distortion rates at the sampling points are obtained through multi-functional network meters or power quality monitoring terminals installed on the power supply side of the power distribution system. Specifically, the system periodically reads the register values ​​inside the aforementioned meters via RS485 communication bus or Ethernet using Modbus-RTU or DL / T 645 protocol.

[0067] Tracking deviation: The absolute value of the difference between the target power factor of the reactive power compensation device corresponding to the sampling point and the actual power factor measured at the sampling point is taken as the tracking deviation.

[0068] The preset conditions are: voltage deviation is less than a preset voltage deviation threshold; current deviation is less than a preset current deviation threshold; voltage distortion rate is less than a preset voltage distortion threshold; current distortion rate is less than a preset current distortion threshold; the tracking deviation of a preset number of consecutive sampling points (e.g., 10) is less than a preset tracking threshold; and there are no missing data in any dimension of the sampling points. The voltage deviation threshold, current deviation threshold, voltage distortion threshold, current distortion threshold, and tracking threshold are all adjusted by those skilled in the art based on actual production conditions, and will not be elaborated upon here.

[0069] Multiple conditions in the preset conditions must be met simultaneously. If any of the above conditions are not met, it indicates that there is significant external interference at the sampling point. Such sampling points are treated as abnormal data and removed to obtain the optimal sample.

[0070] The above conditions can be adjusted according to the actual situation; one or more conditions can be selected. The thresholds corresponding to each condition can also be adjusted based on actual production conditions and staff experience.

[0071] S3: For any typical operating condition range, obtain the active power on the power supply side in the corresponding optimization sample. For any target power factor, obtain the power consumption representative value that can represent the active power level on the power supply side under the target power factor. Determine the long-term optimal value of the typical operating condition range based on the minimum value of the power consumption representative value corresponding to different target power factors.

[0072] This step aims to mathematically model the effective gear data retained after cleaning within a single typical operating condition range, use the mode method to eliminate small-amplitude random disturbances, extract the central tendency characteristics of each gear, and finally compare and obtain the transient optimal target power factor with the lowest energy consumption in that period.

[0073] Specifically, assuming there are a total of A typical operating condition range. For the first... A typical operating condition range ( Each of its corresponding scan levels corresponds to a specific target power factor. The corresponding optimization sample retains multiple valid active power sampling values ​​from the power supply side.

[0074] For any scanning range within any typical operating condition interval, the active power sequence is formed by multiple active power sample values. The active power sequence can be represented as: . Indicates the first Within a typical operating condition range, the first In the optimization sample under each scanning level, the first Active power at each sampling point.

[0075] First, obtain a representative power consumption value that can represent the active power level on the power supply side under the target power factor.

[0076] For any given scan range, multiple active power samples are binned. The purpose of binning is to discretize the continuous real-valued space in order to find the densest clustering region of data.

[0077] Suppose the system divides the data sequence from the current minimum value to the maximum value into several equal-width boxes. The box width is preferably 1% to 2% of the arithmetic mean of all sampled active power values ​​at this scanning range, or fixed at a small absolute physical quantity value (e.g., 0.5kW).

[0078] If the bin width is too large, it will obscure the true distribution details of the data; if the bin width is too small, the number of points in each bin will be too sparse, failing to reflect clustering (mode feature).

[0079] For any box formed after partitioning, its corresponding interval can be represented as: ; in, .

[0080] In the formula, Indicates the first Within a typical operating condition range, the first The first scan level corresponds to the 1st The interval of each box; Indicates the first Within a typical operating condition range, the first The minimum value in the active power sequence of each scanning range; For the first Within a typical operating condition range, the first The total number of boxes generated by each scanning setting is obtained by dividing the difference between the maximum and minimum values ​​by the box width and rounding up; the box width represents the width of the box body.

[0081] Count the absolute number of sampling points contained inside each box.

[0082] The more data points appear within a certain power range (box), the more likely the actual total active power of the power distribution system on the power supply side is to be stable within this power range during that period and at that scanning level.

[0083] The box with the most data points is designated as the mode box, and the median of the total active power on the power supply side of each sampling point in the mode box is taken as the representative power consumption value for that scanning range.

[0084] If calculations reveal that the mode bin contains only one data point, then the value of that point is directly taken as the representative power consumption value. The advantage of using the median of the mode bin instead of the average of all data is that it can strongly resist the pulling effect of individual extreme outliers (even if they are not cleaned up by regularization).

[0085] Then, the long-term optimal value for the typical operating condition range is determined based on the minimum value of the power consumption representative value corresponding to different target power factors.

[0086] In one embodiment, for any typical operating condition range, the minimum value of the power consumption representative value corresponding to each scanning level in the typical operating condition range is taken as the optimal value of the operating condition. The optimal value of the operating condition can be directly taken as the long-term optimal value, that is, without continuous observation and recording, the long-term optimal value is determined only by one day's data.

[0087] In another embodiment, for any typical operating condition interval, in order to overcome the random errors caused by the daily random load characteristic changes, continuous observation and recording can be performed. The system calculates the optimal operating value for each typical operating condition interval over a given day, thereby obtaining a sequence of optimal operating values ​​for that typical operating condition interval. Then, it selects the optimal operating value from the sequence as the long-term optimal value. In one embodiment, the median or mode of the optimal operating value sequence can be used as the long-term optimal value.

[0088] In this embodiment, the number of days parameter The preferred value is Three days; when the company's power consumption is relatively stable, three days can be selected to shorten the implementation cycle.

[0089] In some embodiments, to ensure the stability of data obtained from continuous observation over multiple days, and thus verify the accuracy of the selected long-term optimal value, the long-term stability of the power distribution system can be evaluated by the volatility of the optimal operating condition value sequence. For example, in this embodiment, the standard deviation of the optimal operating condition value sequence is calculated. If the standard deviation is greater than a preset volatility threshold, it indicates that the active power of the power distribution system fluctuates significantly, and the observation period can be appropriately extended, for example, to 10 days. Alternatively, in step S1, the typical operating condition interval can be readjusted and divided more finely.

[0090] S4: The target power factor corresponding to the long-term optimal value of each typical operating condition range is used as the optimal power factor to control the reactive power compensation device.

[0091] The scanning range (target power parameter) corresponding to the long-term optimal value is taken as the optimal power factor.

[0092] Associate all typical operating condition interval numbers, corresponding start and end times, and calculated standard power factor settings obtained in step S1 to form a typical operating condition interval-target power factor mapping table. Write the mapping table into the controller of the underlying reactive power compensation device.

[0093] In daily operation mode, based on the current time, current or other electrical parameters, i.e. the characteristic dimension corresponding to the typical operating condition range, the system actively retrieves the current typical operating condition range in the mapping table and calls the corresponding optimal power factor as the target value for the closed-loop adjustment of the reactive power compensation device, driving the reactive power compensation device to operate, so that the actual power factor of the power distribution system closely follows the target value.

[0094] Due to differences in load characteristics across different time periods, the target setpoint will undergo a step change when the system crosses a time boundary into an adjacent time period (for example, at the time of shift handover in a factory, the target value needs to jump instantaneously from 0.942 to 0.978). If the system is allowed to transiently track this large step change, it will cause a wide range of intensive operations of the reactive power compensation device, resulting in severe voltage fluctuations in the distribution network or high-intensity mechanical shock damage to the switching contactors.

[0095] To avoid this problem, this embodiment uses a linear transition method to smoothly transition the target power factor of the reactive power compensation device from the set value of the current typical operating condition range to the set value of the next typical operating condition range. When the current operating time reaches the switching point between adjacent time periods, the controller does not immediately switch the old target power factor value to the new target power factor value, but instead constructs a linear target guidance function.

[0096] The controller of the reactive power compensation device after switching trigger During the minute transition window, the real-time actual tracking target parameters are generated according to the following formula: ; In the formula, for The command value issued by the controller of the reactive power compensation device at all times; This represents the new target power factor; This represents the old target power factor; The preset total transition time (the value range is...) minute); This is the trigger time for the time period switch.

[0097] As time progresses, the target power factor gradually increases until a new target power factor is reached, thus avoiding sudden changes in the target power factor.

[0098] In some embodiments, during the operation of the power distribution system, when the relative rate of change of active power on the power supply side in any typical operating condition interval is greater than a preset correction threshold, the long-term optimal value corresponding to the typical operating condition interval is re-acquired; and the optimal power factor corresponding to the newly acquired long-term optimal value is used to control the reactive power compensation device.

[0099] Specifically, the total active power variation on the power supply side is continuously monitored across all typical operating conditions.

[0100] When the relative rate of change of total active power on the power supply side is found to be greater than the preset correction threshold within any typical operating condition range. When the value reaches %, it indicates a significant change in the process structure or production line equipment. At this point, the system triggers the anomaly correction mechanism. After an alarm event pops up and the operation interface receives manual authorization confirmation from the maintenance personnel, the system initiates a scanning and optimization process from step S2 to step S4 to obtain the long-term optimal value again.

[0101] Otherwise, the system can automatically perform steps S2 to S4 for optimization based on a preset frequency. At this point, the system executes an automatic correction process to obtain the long-term optimal value again. For example, this can be performed once per quarter.

[0102] The extracted new long-term optimal value is compared with the existing long-term optimal value, and the absolute deviation between the two is calculated. When the absolute deviation is greater than a preset value, the target power factor control reactive power compensation device corresponding to the newly acquired long-term optimal value is used for compensation. The absolute deviation is the absolute value of the difference between the new long-term optimal value and the existing long-term optimal value.

[0103] To explain the rationale behind the aforementioned continuous optimization process at the physics level, the underlying theoretical model upon which the system relies is: In a microgrid system with inductive loads, the total active power consumption... To achieve the optimal power factor for energy efficiency Nearby, its mathematical form can be approximated as a quadratic parabola function opening upwards, and its mathematical expression can be represented as: ; In this formula, Indicates the current target power factor of the power distribution system. The total active power response value on the power supply side; This represents the independent variable, which is the current actual operating power factor of the system. This represents the rigid foundation active power loss that cannot be eliminated under the most ideal compensation state of the physical system, i.e., the theoretical minimum loss. Corresponding to the comprehensive line characteristic proportional constant of the system, These are the coordinates of the point where the energy efficiency is optimal under theoretical conditions.

[0104] Logically, when the actual power factor of the system equals When the quadratic term is zero, the total active power of the system reaches its theoretical minimum. Any overcompensation or undercompensation that causes the power factor to deviate from this specific value will lead to an increase in overall energy consumption due to the rapid increase of the quadratic term. This parabolic characteristic ensures the uniqueness and certainty of the extreme point within the closed interval, providing theoretical support for the positioning and comparison optimization used in each step of this invention.

[0105] This application also discloses a reactive power compensation efficiency optimization system based on big data interval analysis, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a reactive power compensation efficiency optimization method based on big data interval analysis according to this application is implemented.

[0106] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0107] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A reactive power compensation energy efficiency optimization method based on big data interval analysis, characterized in that, Based on the historical operating data of the power distribution system, at least one feature dimension is selected to divide the system into intervals, resulting in multiple typical operating condition intervals; wherein, the feature dimension includes at least one of time dimension, current dimension, or power dimension; In the typical operating condition ranges during the actual operation of the power distribution system, different target power factors are set for the reactive power compensation device, and the operating data of the power distribution system under different target power factors are collected as optimization samples; the operating data includes at least: the active power on the power supply side of the power distribution system; For any typical operating condition range, obtain the active power on the power supply side in the corresponding optimization sample; for any target power factor, obtain the power consumption representative value that can represent the active power level on the power supply side under the target power factor; determine the long-term optimal value of the typical operating condition range based on the minimum value of the power consumption representative value corresponding to different target power factors. The target power factor corresponding to the long-term optimal value of each typical operating condition range is used as the optimal power factor of each typical operating condition range, and is used to control the reactive power compensation device. 2.The reactive power compensation energy efficiency optimization method based on big data interval analysis of claim 1, wherein, Based on the historical operating data of the power distribution system, at least one feature dimension is selected for interval division to obtain multiple typical operating condition intervals, including: selecting the feature dimension for interval division, dividing the feature dimension into multiple candidate intervals representing different operating conditions of the power distribution system based on historical operating data; calculating the relative rate of change of the historical operating data of the power distribution system in each candidate interval; and taking the candidate interval with a relative rate of change less than a preset change threshold as the typical operating condition interval. 3.The reactive power compensation energy efficiency optimization method based on big data interval analysis of claim 1, wherein, Different target power factors are set for reactive power compensation devices, and the operation data of the power distribution system under different target power factors are collected as optimization samples, including: making the target power factor monotonically increase or monotonically decrease within a preset scanning interval with a preset fixed step size. 4.The reactive power compensation energy efficiency optimization method based on big data interval analysis of claim 1, wherein, Different target power factors are set for the reactive power compensation device, and the operation data of the power distribution system under different target power factors are collected as optimization samples. This includes: making the target power factor monotonically increase or decrease within a preset scanning interval according to a preset coarse step size, while collecting the operation data of the power distribution system; constructing a coarse interval based on the time when the minimum active power on the power supply side is located in the operation data; making the target power factor monotonically increase or decrease within the coarse interval according to a preset fine step size; the coarse step size is greater than the fine step size.

5. The method for optimizing reactive power compensation efficiency based on big data interval analysis according to claim 1, characterized in that, Also includes: The sampling points in the optimization sample are filtered based on preset conditions. If a sampling point does not meet the preset conditions, the sampling point is removed. ； Among them, the selection of sampling points in the optimization sample based on preset conditions includes: calculating multiple evaluation indicators of the sampling points, comparing the evaluation indicators with preset conditions to determine whether the sampling points meet the preset conditions.

6. The reactive power compensation energy efficiency optimization method based on big data interval analysis according to claim 5, characterized in that, The evaluation metrics include at least one of the following: voltage deviation, which characterizes the degree of deviation of the sampling point voltage from the nominal voltage of the power distribution system; current deviation, which characterizes the degree of current fluctuation at the sampling point; voltage distortion rate; or current distortion rate. 7.The reactive power compensation energy efficiency optimization method based on big data interval analysis of claim 1, wherein, For any target power factor, obtain a power consumption representative value that can represent the active power level of the power supply side under the target power factor, including: dividing the multiple active power of the power supply side corresponding to the same target power factor into multiple boxes, taking the box containing the most active power of the power supply side as the mode box; and taking the median of all active power of the power supply side contained in the mode box as the power consumption representative value. 8.The reactive power compensation energy efficiency optimization method based on big data interval analysis of claim 1, wherein, Also includes: During the operation of the power distribution system, when any two typical operating condition intervals switch, the target power factor of the reactive power compensation device is smoothly transitioned from the set value of the current typical operating condition interval to the set value of the next typical operating condition interval using a linear transition method. 9.The reactive power compensation energy efficiency optimization method based on big data interval analysis of claim 1, wherein, Also includes: During the operation of the power distribution system, when the relative rate of change of active power on the power supply side in any typical operating condition interval is greater than the preset correction threshold, the long-term optimal value corresponding to that typical operating condition interval is re-acquired. The reactive power compensation device is then controlled using the optimal power factor corresponding to the newly acquired long-term optimal value.

10. A reactive power compensation energy efficiency optimization system based on big data interval analysis, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions, which, when executed by the processor, implement a reactive power compensation energy efficiency optimization method based on big data interval analysis according to any one of claims 1-9.

Images