A real-time energy consumption analysis system for pump station units

Abstract

This invention relates to the field of power consumption management technology, specifically to a real-time energy consumption analysis system for pump station units. The system includes a load feature extraction module, a frequency stability identification module, a start / stop window identification module, a voltage response control module, and a signal filtering module. In this invention, a sliding trend sequence is constructed using load state data and current increment information. Abrupt slope changes are identified and trend directions are classified. Combined with frequency variation amplitude, accurate identification of frequency operation stability segments is achieved. A control time window is constructed by judging the trend consistency of energy consumption and flow in the same time domain. A voltage input strategy is set based on the trend change rate and a mapping table is formed. In the confidence assessment of energy consumption labels, trend continuity judgment is used to achieve signal filtering and recovery, effectively enhancing the accuracy of behavior segmentation, the adaptability of power input control, and the continuity and controllability of overall operating status identification during energy consumption monitoring.

Description

Technical Field

[0001] This invention relates to the field of power consumption management technology, and in particular to a real-time energy consumption analysis system for pump station units. Background Technology

[0002] The field of power consumption management technology involves monitoring, analyzing, and regulating the energy consumption of various electronic and electrical equipment and systems during operation. Its core aspects include energy consumption data acquisition, power change trend analysis, power resource scheduling optimization, and energy consumption behavior assessment and diagnosis. It comprehensively utilizes measurement technology, power control strategies, and information processing methods to achieve refined management of equipment operating power consumption status. This technology is widely applied in industrial automation, data center management, and smart grid control to improve energy efficiency and support decision-making. Among these, the traditional real-time energy consumption analysis system for pumping station units refers to the system that collects key parameters such as active power, reactive power, water level changes, and flow rates of pumping units in real time by deploying electrical parameter acquisition terminals, water level instruments, and flow monitoring devices. It then uses a computing platform to perform basic statistical and graphical visualization operations to assist managers in understanding the energy consumption distribution and change characteristics of the units during operation. This system is based on the power monitoring methods within the power management sub-category of power consumption management technology, encompassing a specific combination of electrical parameter acquisition, real-time data communication, and data archiving functions.

[0003] Existing technologies only acquire data such as power, electricity, water level, and flow rate in real time through electrical parameter acquisition terminals and hydraulic monitoring equipment, and rely on computing platforms for static display and basic statistical analysis. However, they lack the ability to deeply explore the trend relationship when faced with frequent fluctuations in operating status or multi-parameter linkage. Especially in scenarios with drastic load changes, frequent frequency fluctuations, or unstable energy consumption, it is difficult to achieve accurate time period identification and strategic control. This leads to the operation scheduling strategy relying on manual judgment or delayed response, making it difficult to achieve dynamic coupling between load behavior and control behavior, which is not conducive to maintaining efficient and stable operation of the pumping station system under complex operating conditions. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a real-time energy consumption analysis system for pump station units.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a real-time energy consumption analysis system for pump station units includes:

[0006] The load feature extraction module acquires head change data and load current increment data during pump station operation, constructs head-current data pairs of sliding sequence, identifies abrupt change segments in slope value direction, classifies and marks the trend before and after, and generates load response variation segment labels.

[0007] Based on the load response variation segment marker, the frequency stability identification module extracts the pump group frequency record sequence in the corresponding segment, constructs the sliding time frequency sequence and divides it into segments, identifies the frequency value change amplitude of the segment, determines whether the frequency is stable, and generates a frequency stable segment marker.

[0008] The start-stop window identification module uses the frequency operation stable section mark to extract the total flow data and energy consumption change data within the same time period, constructs the energy consumption trend sequence and flow fluctuation sequence, determines whether the energy consumption trend direction and the flow stable state simultaneously meet the conditions, marks the corresponding time period as the control zone, and generates the pump group start-stop coordinated control time window.

[0009] The voltage response control module acquires head and current sequence data within the section based on the pump group start-stop coordinated control time window, establishes a sliding window slope sequence, determines the slope trend type as hold, increase or accelerate, and generates a voltage control distribution mapping table.

[0010] As a further embodiment of the present invention, the load response variation segment marker includes the location of the slope direction variation point, the trend sequence classification result, and the correspondence between the head and current sequence; the frequency operation stable segment marker includes the frequency fluctuation amplitude status, the continuous segment identification result, and the stable frequency interval distribution; the pump group start-stop coordinated control time window includes the energy consumption trend direction determination result, the flow stability analysis result, and the control zone time range; and the voltage control distribution mapping table includes the trend type classification result, the category-corresponding voltage input strategy, and the control time period integration result.

[0011] As a further aspect of the present invention, the load feature extraction module includes:

[0012] The data sequence construction submodule acquires head change data and load current increment data during pump station operation, constructs a fixed-length head-current data pair sequence according to the time sliding window method, and combines each data pair in pairs according to the time sequence to obtain the slope trend data sequence.

[0013] Based on the slope trend data sequence, the trend slope recognition submodule detects the sign change between adjacent slope values, identifies the sign abrupt change point, determines the direction of slope change before and after, and marks the abrupt change segment as a differentiated trend category segment, generating a trend classification label sequence.

[0014] The response variation segment marking submodule calls the trend classification label sequence to jointly compare the head change value and load current increment value in adjacent differentiated trend segments, calculate the load response change intensity value, extract the segment boundary position where the load response change intensity value exceeds the threshold, and generate load response variation segment marking.

[0015] As a further aspect of the present invention, the load response change intensity value is expressed by the formula:

[0016] ;

[0017] in, This represents the intensity of the load response change. Indicates the number of points before the trend inflection point The average head of the section, Indicates the number of points after the trend inflection point. The average head of the section, Indicates the first Average load current within the segment Indicates the first Average load current within the segment This represents the mean squared deviation of load current fluctuation between the two segments before and after the point of trend inflection. This is a dimensionless coupling adjustment coefficient. It is a constant.

[0018] As a further aspect of the present invention, the frequency stability identification module includes:

[0019] The segmented reassembly submodule extracts the time synchronization frequency record data of the corresponding pump group within the segment based on the load response variation segment marker, segments and reassembles the frequency data according to a fixed sliding window, and assigns time index labels to the sliding window to obtain the sliding frequency time series set.

[0020] The fluctuation amplitude judgment submodule calls the sliding frequency time series set to perform a difference detection operation on the frequency data within the time segment. It combines the difference between the peak frequency and the trough frequency within the sliding window to calculate the frequency dynamic stability index value and compares it with the frequency stability judgment threshold. It then filters out the time intervals where the frequency dynamic stability index value is lower than the threshold to obtain the frequency stable segment index sequence.

[0021] The continuous identification submodule performs clustering and recombination operations on the continuous window numbers on the time axis based on the frequency stable segment index sequence, removes isolated window numbers, retains continuous distribution segments, marks the corresponding original time period range, and establishes frequency stable segment markers.

[0022] As a further aspect of the present invention, the start / stop window recognition module includes:

[0023] Based on the frequency stable segment marking results, the data stream receiving submodule extracts the total flow data and energy consumption change data within the corresponding time period, serializes the energy consumption change data according to the time order, divides the energy consumption values ​​within adjacent time steps, and divides and merges the total flow data according to the same time step to obtain the energy consumption change sequence and the flow change sequence.

[0024] The trend feature construction submodule calls the energy consumption change sequence and the flow change sequence, performs difference calculation on the values ​​between adjacent time points in the energy consumption change sequence and establishes a direction vector, takes the sign value of the direction vector of the time period and performs continuous statistics to identify the energy consumption trend change status, detects the difference range between the peak and trough values ​​of the time period in the flow change sequence, compares the difference range with the set flow stability judgment benchmark range, judges whether the flow is stable, and generates the flow stability judgment result.

[0025] The collaborative section marking submodule uses the flow stability determination result to filter time periods that simultaneously satisfy the upward or downward energy consumption trend and the stable flow state. It then pairs and marks the time periods with their corresponding start and end times to generate a pump start-stop collaborative control time window.

[0026] As a further aspect of the present invention, the voltage response regulation module includes:

[0027] The parameter sequence extraction submodule, based on the pump group start-stop coordinated control time window, combines the head sampling data and current sampling data in the section, performs synchronous pairing processing on the head and current according to the set time interval, arranges the pairing results into a continuous time sequence structure, performs window division on the continuous time sequence structure, establishes a sliding calculation section on the sequence according to a fixed sliding step size, and generates a head-current sliding window sequence.

[0028] The trend slope determination submodule calls the head current sliding window sequence, performs linear fitting on the head and current values ​​within the sliding window according to the time dimension, calculates the slope value of the corresponding segment, and determines whether the trend is maintained, increasing or accelerating according to the increase or decrease relationship of the slope values ​​between adjacent windows, and generates the sliding window trend classification result.

[0029] The voltage strategy mapping submodule matches the trend categories with the preset voltage strategy input methods based on the sliding window trend classification results, integrates and groups the time periods corresponding to the same trend categories, sets a voltage regulation strategy identifier for each time period, and performs joint mapping between the time period identifier and the voltage input method to generate a voltage regulation distribution mapping table.

[0030] As a further aspect of the present invention, the system also includes a signal filtering module:

[0031] The signal filtering module uses the voltage regulation distribution mapping table to extract energy consumption label data within the marked time period, construct a trend sequence and identify directional continuity, set confidence and rejection flags, record the trend change direction of the rejected data, restore the flag state when the continuous trend is reconstructed, and generate an energy consumption label confidence participation list.

[0032] The energy consumption label confidence participation list includes trend continuity judgment status, confidence label marking, removal trend records and recovery indication indicators.

[0033] As a further aspect of the present invention, the signal filtering module includes:

[0034] The energy consumption tag extraction submodule extracts energy consumption tag data within the marked time period based on the voltage regulation distribution mapping table, collects paired energy consumption values ​​under the time index corresponding to the time period, arranges them in chronological order, performs time linear interpolation to fill in missing positions, and generates a time series energy consumption tag array.

[0035] The trend direction identification submodule calls the time series energy consumption tag array, divides the sliding window according to a fixed step size and calculates the difference between adjacent points, counts the number of difference directions with consecutive identical signs, identifies the continuous state, and generates a trend direction continuous state set.

[0036] The confidence removal tag submodule determines whether there is a sudden change in direction or the direction length is lower than the set trend continuity threshold in the time period based on the trend direction continuity state set. It assigns a removal tag to the time period that does not meet the conditions and records the corresponding trend change direction. It assigns a confidence tag to the remaining part and generates a confidence and removal tag sequence.

[0037] The tag allocation submodule calls the confidence and rejection tag sequence. For the data that has been assigned rejection tags, it checks whether the time period forms a continuous sequence consistent with the original trend direction. If the trend reconstruction trigger length threshold is reached, the original tag is restored to the confidence state. The confidence state data index is sorted and output to generate an energy consumption tag confidence participation list.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0039] In this invention, a sliding trend sequence is constructed by combining load status data and current increment information. The slope abrupt change points are identified and trend direction classification is performed. Combined with frequency variation amplitude, accurate identification of frequency operation stability segments is achieved. A control time window is constructed by judging the trend consistency of energy consumption and flow in the same time domain. A voltage input strategy is set based on the trend change rate, and a mapping table is formed. In the confidence assessment of energy consumption tags, trend continuity judgment is used to achieve signal filtering and recovery. The overall processing chain enhances the multi-dimensional coupling between pump station operation data and load response, frequency fluctuation, start-stop control, and voltage regulation. The obtained time period markers have the capabilities of trend direction identification, variation trend linkage, and dynamic strategy response, effectively enhancing the accuracy of behavior segmentation, the adaptability of power input control, and the continuity and controllability of overall operation status identification during energy consumption monitoring. Attached Figure Description

[0040] Figure 1 This is a system flowchart of the present invention;

[0041] Figure 2 This is a flowchart of the load feature extraction module in this invention;

[0042] Figure 3 This is a flowchart of the frequency stability identification module in this invention;

[0043] Figure 4 This is a flowchart of the start / stop window recognition module in this invention;

[0044] Figure 5 This is a flowchart of the voltage response control module in this invention;

[0045] Figure 6 This is a flowchart of the signal filtering module in this invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0047] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0048] Please see Figure 1 A real-time energy consumption analysis system for pump station units includes:

[0049] The load feature extraction module acquires head change data and load current increment data during pump station operation, constructs head-current data pairs in a sliding sequence, calculates the slope values ​​of the data pairs to form a trend sequence, identifies abrupt changes in slope value direction, classifies and marks the sections before and after the trend, performs pump group load change correlation section identification operation based on trend classification, and generates load response variation section labels.

[0050] The frequency stability identification module extracts the pump group frequency record sequence within the corresponding segment based on the load response variation segment marker, constructs the sliding time frequency sequence and divides it into segments, identifies the frequency value change amplitude of the segment, determines whether the frequency is stable, performs continuous segment screening and unified identification, and generates frequency stable segment markers.

[0051] The start-stop window identification module uses frequency operation stable section marking, extracts total flow data and energy consumption change data within the same time period, constructs energy consumption trend sequence and flow fluctuation sequence, determines whether the energy consumption trend direction and flow stability state simultaneously meet the conditions, marks the corresponding time period as the control zone, and generates pump group start-stop coordinated control time window.

[0052] The voltage response control module acquires head and current sequence data within the section based on the pump group start-stop coordinated control time window, establishes a sliding window slope sequence, determines whether the slope trend type is maintained, increasing or accelerating, sets the corresponding voltage strategy input method according to the trend category, classifies and integrates the time periods, and generates a voltage control distribution mapping table.

[0053] The signal filtering module uses a voltage regulation distribution mapping table to extract energy consumption label data within a marked time period, constructs a trend sequence and identifies directional continuity, sets confidence and rejection flags, records the trend change direction of the rejected data, restores the flag state when the continuous trend is reconstructed, and generates an energy consumption label confidence participation list.

[0054] The load response variation segment markers include the location of slope direction variation points, trend sequence classification results, and the correspondence between head and current sequences. The frequency operation stable segment markers include the frequency fluctuation amplitude status, continuous segment identification results, and stable frequency interval distribution. The pump start-stop coordinated control time window includes the energy consumption trend direction determination results, flow stability analysis results, and control zone time range. The voltage control distribution mapping table includes the trend type classification results, category-corresponding voltage input strategies, and control time period integration results. The energy consumption label confidence participation list includes the trend continuity judgment status, confidence label markings, trend record removal, and recovery indication indicators.

[0055] Please see Figure 2 The load feature extraction module includes:

[0056] The data sequence construction submodule acquires head change data and load current increment data during pump station operation, constructs a fixed-length head-current data pair sequence according to the time sliding window method, and combines each data pair in pairs according to the time sequence to obtain the slope trend data sequence.

[0057] The original data sequence was divided into sliding time windows, with each sliding time window width set to 20 minutes and a sliding step size set to 5 minutes. After normalizing the time series data within each window, they were combined into head change subsequences and load current subsequences. A bidirectional combination method was used so that each set of data included the head and load current values ​​at two consecutive time points. This combination method is suitable for scenarios where load and head fluctuate frequently in a short period of time. For example, a water pumping station experienced frequent start-stop cycles during the rainy season. In this case, each set of time window data reflected the characteristics of load changes. The above slicing method can accurately capture abrupt change points. For example, from 14:00 to 14:20 on August 12, 2023, the load current changed from 8.4A to 12.7A, and the head increased from 3.5m to 6.1m. The combined subsequence is [(3.5, 8.4), (6.1, 12.7)], which yields the slope trend data sequence.

[0058] The trend slope recognition submodule detects the sign change between adjacent slope values ​​based on the slope trend data sequence, identifies the sign abrupt change point, determines the direction of slope change before and after, and marks the abrupt change segment as a differentiated trend category segment, generating a trend classification label sequence.

[0059] To determine the directional change of adjacent slope trend values ​​within a continuous time window, the process involves calculating the slope of head change between two adjacent time windows using the following formula: If the direction of the head slope in the previous window changes from positive to negative or from negative to positive, it is recorded as a "sign change". At this time, it is determined that the current moment is at a trend change point, and the head change slope of the previous window is set to... The slope of the second window is If a sudden change in direction occurs, the difference between the standard deviation and the mean is calculated for the five sets of data before and after the change point. This difference is used to classify the slope trend into five categories: "sharp rise", "gradual rise", "stable", "gradual fall" and "sharp fall". These correspond to five intervals: head slope greater than 0.25, between 0.10 and 0.25, -0.10 and 0.10, -0.25 and -0.10, and less than -0.25, respectively. By classifying the slope before and after the change point, a trend classification label sequence is generated.

[0060] The response submodule for the variable segment marking calls the trend classification label sequence to jointly compare the head change value and the load current increment value in adjacent differential trend segments, using the following formula:

[0061] ;

[0062] Calculate the load response change intensity value, extract the boundary locations of the segments where the load response change intensity value exceeds the threshold, and generate load response variation segment markers;

[0063] in, This represents the intensity of the load response change. Indicates the number of points before the trend inflection point The average head of the section, Indicates the number of points after the trend inflection point. The average head of the section, Indicates the first Average load current within the segment Indicates the first Average load current within the segment This represents the mean squared deviation of load current fluctuation between the two segments before and after the point of trend inflection. This is a dimensionless coupling adjustment coefficient. It is a constant;

[0064] Formula calculation logic: The head change is calculated by comprehensively evaluating the load response change intensity value composed of three parts. As a measure of external operating condition disturbances, it reflects the magnitude of changes in the physical environment before and after the abrupt change point, and is calculated by squared the change in load current. with mean squared deviation Weighted summation, as a core indicator of pump station load response, is the adjustment coefficient. To balance the effects of average change and instantaneous fluctuation, the product term is divided by... Normalization is performed to ensure the comparability of values ​​under different operating conditions and to suppress abnormal peaks, resulting in a comprehensive scoring index for the coupling degree of abrupt change segments. Used to identify key variant regions;

[0065] The load response change intensity value represents the comprehensive change amplitude and fluctuation degree of the pump station load current when the head changes abruptly. It integrates the change and fluctuation of the load current before and after the change point, and takes into account the influence of the head change on the load response. The higher the value, the more severe the load response of the pump station to the sudden event. It can be used to mark abnormal operating sections.

[0066] Parameter meaning and calculation process:

[0067] : These represent the average head within the time window before and after the mutation point, respectively;

[0068] : These represent the average load current within the time window before and after the abrupt change point, respectively;

[0069] : The result of the mean square error calculation of the load current in the front and rear windows;

[0070] : This is an adjustment coefficient, and its value fluctuates between [0.3, 1.2] based on engineering experience;

[0071] This is a fine-tuning constant, set to 0.01 to avoid the denominator being zero.

[0072] Let's illustrate the calculation with an example:

[0073] Parameter input:

[0074] ;

[0075] Substitute into the formula to calculate:

[0076] ;

[0077] Step-by-step calculation:

[0078] ;

[0079] ;

[0080] ;

[0081] Molecular part: ;

[0082] Denominator: ;

[0083] Comprehensive calculation:

[0084] ;

[0085] The results show that the load response change intensity value is 5.18, indicating a medium-to-high intensity coupling influence between head change and load current variation before and after the abrupt change. Based on actual engineering data, an influence intensity threshold of 3.0 can be set. If the calculated value is greater than the threshold, it is determined to be an affected variation segment; otherwise, it is not marked. The threshold of 3.0 is based on a 30-day pump station operation sample, and is determined by comparing more than 95% of the normally operating samples. The value distribution is obtained by setting a confidence interval;

[0086] Table 1: Example Data Table for Influence Intensity Calculation

[0087] ;

[0088] As shown in Table 1, actual pump station operation data was selected for analysis. After calculation using the formula, the influence intensity value was found to be 5.18, which exceeded the preset threshold of 3.0. Therefore, the data segment was marked as the influence variation segment.

[0089] The innovation of the formula lies in incorporating the change in head. By combining the squared term of the load current change and the mean squared term of the fluctuation, a multi-factor fusion coupling strength index is constructed through the product and the normalized denominator. This effectively avoids misjudgment caused by a single variable change. It is particularly robust in complex operating environments where the signal reverses or delays in response near the change point, reflecting the enhancement mechanism of the combined effect of the change rate and fluctuation amplitude of the load current on the judgment index.

[0090] Please see Figure 3 The frequency stability identification module includes:

[0091] The segmented reassembly submodule extracts the time synchronization frequency record data of the corresponding pump group within the segment based on the load response variation segment marker, segments and reassembles the frequency data according to a fixed sliding window, and assigns time index labels to the sliding window to obtain the sliding frequency time series set.

[0092] The variation segment markers are mapped to the actual pump unit operation data. Pump unit frequency records matching the marked time period are retrieved by querying the data. This process relies on timestamp alignment. The start and end times of the variation markers are compared with the frequency sequences in the pump unit control records to extract frequency change records during the period. A fixed sliding window method is applied to segment the extracted data. With the sliding window set to 30 seconds and a step size of 10 seconds, the entire extracted segment is divided into multiple overlapping data fragments. Each window contains a data sequence of the same length. During processing, a unique time index label is assigned to each window, generated based on the window's start time, for subsequent retrieval and analysis. After each set of sliding windows is generated, it serves as the input for subsequent analysis and is set in the pump station's automatic diagnostic equipment. After a load fluctuation event is marked, the frequency data is slid segmented according to the above rules, forming several structurally consistent information units with clear time positioning, facilitating fluctuation identification and fault tracing. The frequency response characteristics of different segments in different time periods are analyzed to obtain a sliding frequency time series set.

[0093] The fluctuation amplitude judgment submodule calls the sliding frequency time series set to perform difference detection on the frequency data within the time segment. Combining the difference between the peak and trough frequencies within the sliding window, it uses the following formula:

[0094] ;

[0095] Calculate the frequency dynamic stability index value and compare it with the frequency stability judgment threshold. Filter the time intervals where the frequency dynamic stability index value is lower than the threshold to obtain the frequency stable segment index sequence.

[0096] in, This represents the value of the frequency dynamic stability index. Represents a sliding window Inner Frequency values ​​at each time point Indicates the first With the Frequency difference between time points Represents a sliding window The maximum frequency value within, Represents a sliding window The minimum frequency value within, Display window Internal frequency mean This represents the number of frequency sampling points.

[0097] Formula calculation logic: Calculate the frequency difference between adjacent time points within a sliding window. The values ​​are summed item by item to represent the cumulative change in frequency over time; the maximum frequency within the window is extracted. With minimum frequency The difference between the two values ​​reflects the overall fluctuation range within the entire window period. This value is added to the sum of the frequency changes to form the numerator, which measures the total frequency fluctuation. The denominator is calculated by comparing the frequency values ​​at each time point with the window mean. The sum of squared deviations is averaged and squared to obtain the standard deviation, which reflects the volatility dispersion. The ratio is obtained by dividing the numerator by the denominator. If the frequency changes drastically but the standard deviation is small, then... A high value indicates concentrated fluctuations and strong changes, and vice versa. A low value indicates that the fluctuations are dispersed and stable;

[0098] The frequency dynamic stability index is used to quantify the intensity and stability of the pump station's operating frequency within a short time window. The ratio between the frequency change amplitude and the mean deviation reflects the instantaneous dynamic stability of the frequency. The larger the value, the more concentrated and violent the frequency fluctuations are, while the smaller the value, the more stable the frequency is in the operating range.

[0099] The meaning of the parameters and the calculation process are explained below:

[0100] : The frequency difference between two adjacent time points within the current sliding window;

[0101] : Maximum frequency within the sliding window;

[0102] Minimum frequency within the sliding window;

[0103] : The average frequency within the sliding window;

[0104] : Number of frequency samples within the current sliding window;

[0105] The frequency sampling data for a certain pumping station is shown in Table 2 below:

[0106] Table 2: Sample Parameters of Sliding Window Frequency

[0107] ;

[0108] As shown in Table 2, a total of 7 sets of frequency data were collected within the window, which were distributed in the interval from 0s to 30s. When identifying frequency fluctuation stability, the data will be used as a set of samples of the sliding window to participate in the subsequent calculation of frequency fluctuation stability index.

[0109] Actual example (using data from Table 2): ;

[0110] The frequency change is calculated as follows:

[0111] ;

[0112] The sum of the absolute values ​​of the frequency changes is:

[0113] ;

[0114] The maximum frequency is: ;

[0115] The minimum frequency is: ;

[0116] Frequency difference: ;

[0117] The molecular results are as follows: ;

[0118] Frequency mean:

[0119] ;

[0120] Standard deviation calculation:

[0121] ;

[0122] The normalized standard deviation is:

[0123] ;

[0124] Substitute into the formula to calculate:

[0125] ;

[0126] The results show that the frequency dynamic stability index is 7.23, and the sliding window frequency fluctuates drastically, so it needs to be included in the stability assessment index and participate in the time axis clustering numbering.

[0127] Threshold setting instructions and judgment process:

[0128] Frequency stability determination threshold The value was selected as 5.0 based on statistical analysis. The setting was based on collecting 7 days of pump station operating frequency data and comparing it with 3000 sets of sliding windows, where over 95% were in a stable state. The value distribution range shows that the vast majority of values ​​are less than 5.0, therefore 5.0 is set as the upper limit of stability. If the value is greater than 5.0, it is determined to be an unstable segment. In this example, SK=7.23, which is greater than the set threshold of 5.0, so the current sliding window is determined to be an unstable segment with fluctuating frequency.

[0129] The continuous identification submodule performs clustering and recombination operations on the continuous window numbers on the time axis based on the frequency stable segment index sequence, removes isolated window numbers, retains the continuously distributed segments, marks the corresponding original time period range, and establishes frequency stable segment markings.

[0130] Based on the time number interval difference, it is determined whether there is a continuous number sequence. If the interval between a number and the previous number is less than 2 window lengths, it is identified as the same continuous stable segment. The continuous number segment is assigned a unique number and registered as a valid frequency operating segment. Conversely, if the interval between a number and the previous number is greater than or equal to 2 window lengths, it is classified as an independent isolated number and does not participate in clustering and recombination. If the frequency stability index sequence contains the numbers [1, 2, 3, 7, 8, 15], then the continuous segments [1, 2, 3] and [7, 8] are identified and numbered FZ1 and FZ2 respectively. The number 15 is excluded because it does not meet the continuity requirement. A frequency operating stable segment mark is established.

[0131] Please see Figure 4 The start / stop window recognition module includes:

[0132] Based on the frequency stable segment marking results, the data stream receiving submodule extracts the total flow data and energy consumption change data within the corresponding time period. It serializes the energy consumption change data according to the time order, divides the energy consumption values ​​within adjacent time steps, and divides and merges the total flow data according to the same time step to obtain the energy consumption change sequence and the flow change sequence.

[0133] The process involves extracting frequency data sequences of the pump unit from the equipment monitoring equipment. This includes time-series processing of the data, binding the frequency at each point in time with its corresponding timestamp, statistically analyzing the frequency fluctuations within each operating period, and identifying intervals with smaller frequency fluctuations within a set time period as stable operating segments. Total flow and energy consumption data within these stable segments are then extracted. Total flow data is typically collected periodically using a flow meter, while energy consumption data can be provided by an electricity metering device. The resulting energy consumption data forms a continuous time series. The data is then divided into multiple equal-length time segments according to a set time step. The energy consumption changes within each segment are statistically analyzed to form an energy consumption change sequence. Simultaneously, the total flow data is also divided according to the same time step, and the flow rates within each time segment are merged to obtain the flow rate changes within each time segment, resulting in both the energy consumption change sequence and the flow rate change sequence.

[0134] The trend feature construction submodule calls the energy consumption change sequence and the flow change sequence. It performs difference calculation on the values ​​between adjacent time points in the energy consumption change sequence and establishes a direction vector. It takes the sign value of the direction vector of the time period and performs continuous statistics to identify the energy consumption trend change status. It detects the difference range between the peak and trough values ​​of the time period in the flow change sequence. It compares the difference range with the set flow stability judgment benchmark range to determine whether the flow is stable and generates the flow stability judgment result.

[0135] The numerical differences between adjacent time periods in the energy consumption sequence need to be processed. The increase or decrease of energy consumption value in each time period is analyzed one by one in chronological order. A direction vector representing the trend of energy consumption change is established. By comparing the energy consumption values ​​of two adjacent time periods, it is determined whether energy consumption is continuously increasing or decreasing. The trend direction vector is uniformly marked with symbols to indicate an upward, downward or flat state, and arranged in chronological order to form a trend symbol sequence. The system statistically analyzes the continuity of trend symbols in time, identifies the interval where energy consumption is in a relatively single direction of change over a long period of time, analyzes the flow change sequence, identifies the peak and trough values ​​in the sequence, calculates the difference interval between the two, compares the difference with the pre-set flow stability benchmark interval, and marks the time period as a flow stability state when the difference is within a stable range, generating a flow stability state determination result.

[0136] The collaborative section marking submodule uses the flow stability state determination result to filter the time period that simultaneously satisfies the upward or downward energy consumption trend and the stable flow state, and pairs and marks the time period with the corresponding start and end time to generate the pump group start-stop collaborative control time window.

[0137] The system iterates through time periods one by one, filtering out those periods that simultaneously meet the criteria of continuously increasing or decreasing energy consumption and stable flow rate. It then identifies collaborative operating time periods through this filtering process, lists these time periods separately, and uses their actual start and end times as identifiers. In practical applications, this operation compares multiple data sources using set rules to find and mark time periods that meet the joint judgment conditions. Each eligible time period is recorded and paired with its start and end times to form a collaborative operating time window, generating a collaborative control time window for pump start-stop.

[0138] Please see Figure 5 The voltage response control module includes:

[0139] The parameter sequence extraction submodule is based on the pump group start-stop coordinated control time window. It combines the head sampling data and current sampling data in the section, performs synchronous pairing processing on the head and current according to the set time interval, arranges the pairing results into a continuous time sequence structure, performs window division on the continuous time sequence structure, establishes a sliding calculation section on the sequence according to a fixed sliding step size, and generates a head-current sliding window sequence.

[0140] The system acquires head and current sampling data for each time period. During operation, the system reads the head and current values ​​of the pump unit through sensors, ensuring that these two types of data are collected at the same sampling frequency. In actual engineering, a sampling interval of 5 or 10 seconds is used. Each sampled data is accompanied by a timestamp. The system compares and calibrates the timestamps, and pairs the head and current data collected at the same or closest time to form a binary data point sequence. The paired data are arranged in chronological order to form a continuous head-current time series. The system performs sliding window division processing on the time series. During the division process, the window length and sliding step size are set. If the window length is set to 60 seconds and the step size is set to 15 seconds, then each window includes 12 sets of data. Each time the system slides forward, it moves 3 sets of data, generating a head-current sliding window sequence.

[0141] The trend slope determination submodule calls the head and current sliding window sequence, performs linear fitting on the head and current values ​​within the sliding window according to the time dimension, calculates the slope value of the corresponding segment, and judges whether the trend is maintained, increasing or accelerating according to the increase or decrease relationship of the slope values ​​between adjacent windows, and generates the sliding window trend classification result.

[0142] The system sequentially calls the data sequence within each sliding window, processing the head and current values. Within each window, the system performs linear fitting on the head and current data in chronological order, constructing fitting curves using the least squares method or correlation regression, and extracting the fitting slope values ​​to represent the rate of change of head or current over time within the window. The system then compares the slope values ​​of two adjacent sliding windows. If the slope of the current window is similar to that of the previous window, it is judged as "maintaining". If the current slope is higher than that of the previous window, it is judged as "increasing". If the slope of the current window is higher than that of the previous window and the difference further widens, it is judged as "accelerating". The trend state of each window is classified and labeled, generating sliding window trend classification results.

[0143] The voltage strategy mapping submodule matches the trend categories with the preset voltage strategy input methods based on the sliding window trend classification results, integrates and groups the time periods corresponding to the same trend categories, sets a voltage regulation strategy identifier for each time period, and jointly maps the time period identifier with the voltage input method to generate a voltage regulation distribution mapping table.

[0144] The system maps the preset voltage input control strategies to each trend category, setting "hold" state to maintain the current voltage, "increase" state to gradually increase voltage, and "accelerate" state to rapidly increase voltage. The system iterates and classifies the trend result sequence according to the mapping rules, groups time periods with the same trend type into the same group, and establishes the correspondence between trend categories and time periods. Then, it sets a voltage regulation strategy identifier for each time period group, setting the identifier as V1, V2, V3, etc., and binds each time period identifier to the specific voltage input method to generate a voltage regulation distribution mapping table.

[0145] Please see Figure 6 The signal filtering module includes:

[0146] The energy consumption tag extraction submodule extracts energy consumption tag data within the marked time period based on the voltage regulation distribution mapping table, collects paired energy consumption values ​​under the time index corresponding to the time period, arranges them in chronological order, performs time linear interpolation to fill in missing positions, and generates a time series energy consumption tag array.

[0147] During actual processing, the system iterates through each time period marked in the mapping table. For each time period, it extracts energy consumption monitoring data based on the start and end time indices. The energy consumption data comes from the energy consumption sensors of the pump station's power distribution equipment and is recorded at a fixed sampling frequency. Each record contains a corresponding timestamp and energy consumption value. After extraction, the data is arranged into a complete time series array in chronological order. For missing data points, the system calls a time linear interpolation algorithm to calculate interpolation points based on the energy consumption values ​​at two consecutive valid time points. The system sets the energy consumption values ​​for two adjacent time points to be 2.5 kWh and 3.0 kWh, respectively. With two sampling points in between, the system fills in the missing values ​​at 2.625 kWh and 2.75 kWh to ensure the continuity and smoothness of the overall time series data, generating a time series energy consumption tag array.

[0148] The trend direction identification submodule calls the time series energy consumption label array, divides the sliding window according to a fixed step size and calculates the difference between adjacent points, counts the number of difference directions with consecutive identical signs, identifies the continuous state, and generates a set of trend direction continuous states.

[0149] The entire array is divided into sliding windows according to a set time step. Each sliding window contains 5 sampling points. The system performs adjacent point difference calculations on the data in each window in turn. That is, the energy consumption value of the current point is subtracted from the value of the previous point to obtain the difference. The difference result is used to determine the direction of energy consumption change. If the difference is positive, it indicates an increase; if it is negative, it indicates a decrease; and if it is zero, it indicates no change. The system counts the signs of the differences in the entire sliding window to determine whether they are continuous in the same direction. If 5 consecutive differences are all positive, it is set as "continuous upward state". If there are alternating positive and negative differences or only 1 or 2 consecutive consistent directions, it is considered as discontinuous direction. The system marks the continuity state of each window and generates a trend direction continuity state set.

[0150] The confidence removal label submodule determines whether there is a sudden change in direction or the direction length is lower than the set trend continuity threshold based on the trend direction continuity state set. It assigns a removal label to time periods that do not meet the conditions and records the corresponding trend change direction. It assigns a confidence label to the remaining part and generates a confidence and removal label sequence.

[0151] The system iterates through the trend direction of each time period in chronological order, determining whether there are abrupt changes in direction. If the previous period's direction was upward and the next period abruptly changed to downward, or if the current period shows frequent changes in direction without a stable trend, it also needs to determine whether the continuous length of a certain direction is lower than a preset trend continuity threshold. If a trend is considered valid only if it lasts for more than 3 consecutive windows, and a trend direction only lasts for 1 to 2 windows before terminating, then the condition is not met. The system assigns a rejection mark to such time periods and records the corresponding trend change direction, indicating that there is an unstable change. For time periods that meet the requirements of direction continuity and stability, a confidence mark is assigned, indicating that the change trend is reliable. The rejection or confidence status of the time periods is arranged in order to generate a confidence and rejection mark sequence.

[0152] The tag allocation submodule calls the confidence and elimination tag sequence. For data that has been assigned elimination tags, it checks whether the time period forms a continuous sequence consistent with the original trend direction. If the trend reconstruction trigger length threshold is reached, the original tag is restored to the confidence state. The confidence state data is indexed, organized, and output to generate an energy consumption tag confidence participation list.

[0153] For time periods that have been marked for removal, the system performs a continuity analysis again to determine whether the segment is consistent with the original trend direction. If the direction is consistent, it checks whether there are multiple consecutive time periods that reach the length threshold for trend reconstruction. The system sets a condition that if five consecutive data points in the removal segment show the same directional change, a reconstruction judgment can be triggered. If the condition is met, the original removal status is changed back to a confidence status. The system then organizes the data segments marked as confidence status after the repair, extracts the corresponding time index, and generates a list of confidence participants for energy consumption labels.

[0154] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A real-time energy consumption analysis system for pump station units, characterized in that, The system includes: The load feature extraction module acquires head change data and load current increment data during pump station operation, constructs head-current data pairs of sliding sequence, identifies abrupt change segments in slope value direction, classifies and marks the trend before and after, and generates load response variation segment labels. Based on the load response variation segment marker, the frequency stability identification module extracts the pump group frequency record sequence in the corresponding segment, constructs the sliding time frequency sequence and divides it into segments, identifies the frequency value change amplitude of the segment, determines whether the frequency is stable, and generates a frequency stable segment marker. The start-stop window identification module uses the frequency operation stable section mark to extract the total flow data and energy consumption change data within the same time period, constructs the energy consumption trend sequence and flow fluctuation sequence, determines whether the energy consumption trend direction and the flow stable state simultaneously meet the conditions, marks the corresponding time period as the control zone, and generates the pump group start-stop coordinated control time window. The voltage response control module obtains the head and current sequence data within the section according to the pump group start-stop coordinated control time window, establishes a sliding window slope sequence, determines the slope trend type as maintaining, increasing or accelerating, and generates a voltage control distribution mapping table. The load feature extraction module includes: The data sequence construction submodule acquires head change data and load current increment data during pump station operation, constructs a fixed-length head-current data pair sequence according to the time sliding window method, and combines each data pair in pairs according to the time sequence to obtain the slope trend data sequence. Based on the slope trend data sequence, the trend slope recognition submodule detects the sign change between adjacent slope values, identifies the sign abrupt change point, determines the direction of slope change before and after, and marks the abrupt change segment as a differentiated trend category segment, generating a trend classification label sequence. The response variation segment marking submodule calls the trend classification label sequence, performs joint comparison of head change value and load current increment value in adjacent differential trend segments, calculates load response change intensity value, extracts the segment boundary position where the load response change intensity value exceeds the threshold, and generates load response variation segment marking; The load response change intensity value is expressed by the formula: ； in, This represents the intensity of the load response change. Indicates the number of points before the trend inflection point The average head of the section, Indicates the number of points after the trend inflection point. The average head of the section, Indicates the first Average load current within the segment Indicates the first Average load current within the segment This represents the mean squared deviation of load current fluctuation between the two segments before and after the point of trend inflection. This is a dimensionless coupling adjustment coefficient. It is a constant; The system also includes a signal filtering module: The signal filtering module uses the voltage regulation distribution mapping table to extract energy consumption label data within the marked time period, construct a trend sequence and identify directional continuity, set confidence and rejection flags, record the trend change direction of the rejected data, restore the flag state when the continuous trend is reconstructed, and generate an energy consumption label confidence participation list. The energy consumption label confidence participation list includes trend continuity judgment status, confidence label marking, removal trend records and recovery indication.

2. The real-time energy consumption analysis system for pump station units according to claim 1, characterized in that, The load response variation segment markers include the location of slope direction variation points, trend sequence classification results, and the correspondence between head and current sequences. The frequency operation stable segment markers include the frequency fluctuation amplitude status, continuous segment identification results, and stable frequency interval distribution. The pump group start-stop coordinated control time window includes the energy consumption trend direction determination results, flow stability analysis results, and control zone time range. The voltage control distribution mapping table includes the trend type classification results, category-corresponding voltage input strategies, and control time period integration results.

3. The real-time energy consumption analysis system for pump station units according to claim 1, characterized in that, The frequency stability identification module includes: The segmented reassembly submodule extracts the time synchronization frequency record data of the corresponding pump group within the segment based on the load response variation segment marker, segments and reassembles the frequency data according to a fixed sliding window, and assigns time index labels to the sliding window to obtain the sliding frequency time series set. The fluctuation amplitude judgment submodule calls the sliding frequency time series set to perform a difference detection operation on the frequency data within the time segment. It combines the difference between the peak frequency and the trough frequency within the sliding window to calculate the frequency dynamic stability index value and compares it with the frequency stability judgment threshold. It then filters out the time intervals where the frequency dynamic stability index value is lower than the threshold to obtain the frequency stable segment index sequence. The continuous identification submodule performs clustering and recombination operations on the continuous window numbers on the time axis based on the frequency stable segment index sequence, removes isolated window numbers, retains continuous distribution segments, marks the corresponding original time period range, and establishes frequency stable segment markers.

4. The real-time energy consumption analysis system for pump station units according to claim 3, characterized in that, The start / stop window recognition module includes: Based on the frequency stable segment marking results, the data stream receiving submodule extracts the total flow data and energy consumption change data within the corresponding time period, serializes the energy consumption change data according to the time order, divides the energy consumption values ​​within adjacent time steps, and divides and merges the total flow data according to the same time step to obtain the energy consumption change sequence and the flow change sequence. The trend feature construction submodule calls the energy consumption change sequence and the flow change sequence, performs difference calculation on the values ​​between adjacent time points in the energy consumption change sequence and establishes a direction vector, takes the sign value of the direction vector of the time period and performs continuous statistics to identify the energy consumption trend change status, detects the difference range between the peak and trough values ​​of the time period in the flow change sequence, compares the difference range with the set flow stability judgment benchmark range, judges whether the flow is stable, and generates the flow stability judgment result. The collaborative section marking submodule uses the flow stability determination result to filter time periods that simultaneously satisfy the upward or downward energy consumption trend and the stable flow state. It then pairs and marks the time periods with their corresponding start and end times to generate a pump start-stop collaborative control time window.

5. The real-time energy consumption analysis system for pump station units according to claim 4, characterized in that, The voltage response control module includes: The parameter sequence extraction submodule, based on the pump group start-stop coordinated control time window, combines the head sampling data and current sampling data in the section, performs synchronous pairing processing on the head and current according to the set time interval, arranges the pairing results into a continuous time sequence structure, performs window division on the continuous time sequence structure, establishes a sliding calculation section on the sequence according to a fixed sliding step size, and generates a head-current sliding window sequence. The trend slope determination submodule calls the head current sliding window sequence, performs linear fitting on the head and current values ​​within the sliding window according to the time dimension, calculates the slope value of the corresponding segment, and determines whether the trend is maintained, increasing or accelerating according to the increase or decrease relationship of the slope values ​​between adjacent windows, and generates the sliding window trend classification result. The voltage strategy mapping submodule matches the trend categories with the preset voltage strategy input methods based on the sliding window trend classification results, integrates and groups the time periods corresponding to the same trend categories, sets a voltage regulation strategy identifier for each time period, and performs joint mapping between the time period identifier and the voltage input method to generate a voltage regulation distribution mapping table.

6. The real-time energy consumption analysis system for pump station units according to claim 1, characterized in that, The signal filtering module includes: The energy consumption tag extraction submodule extracts energy consumption tag data within the marked time period based on the voltage regulation distribution mapping table, collects paired energy consumption values ​​under the time index corresponding to the time period, arranges them in chronological order, performs time linear interpolation to fill in missing positions, and generates a time series energy consumption tag array. The trend direction identification submodule calls the time series energy consumption tag array, divides the sliding window according to a fixed step size and calculates the difference between adjacent points, counts the number of difference directions with consecutive identical signs, identifies the continuous state, and generates a trend direction continuous state set. The confidence removal tag submodule determines whether there is a sudden change in direction or the direction length is lower than the set trend continuity threshold in the time period based on the trend direction continuity state set. It assigns a removal tag to the time period that does not meet the conditions and records the corresponding trend change direction. It assigns a confidence tag to the remaining part and generates a confidence and removal tag sequence. The tag allocation submodule calls the confidence and rejection tag sequence. For the data that has been assigned rejection tags, it checks whether the time period forms a continuous sequence consistent with the original trend direction. If the trend reconstruction trigger length threshold is reached, the original tag is restored to the confidence state. The confidence state data index is sorted and output to generate an energy consumption tag confidence participation list.

Images