Permanent magnet submersible pump working condition self-adaptive energy-saving control method and system

Abstract

The application discloses a permanent magnet submersible pump working condition self-adaptive energy-saving control method and system, and relates to the technical field of submersible pump control systems; the method comprises the following steps: screening a smooth running section in pre-collected operation alignment data, and statistically processing an active power sequence according to an output frequency sequence in the smooth running section to obtain a frequency power baseline table; a voltage fluctuation reference amount is obtained by extracting a voltage sequence in the smooth running section; the frequency power baseline table is compared with the active power sequence to obtain a power residual sequence; in combination with the voltage fluctuation reference amount, a continuous negative deviation interval is identified in the power residual sequence to obtain a load reduction event segment; a treatment action parameter set is calculated based on the operation alignment data and event fingerprint data, and the output frequency is adjusted according to the treatment action parameter set; the application can realize accurate differentiation of multiple sources of load reduction only by using existing available signals without additionally adding downhole sensors, and the reliability and economy of the operation of the permanent magnet submersible pump are improved.

Description

Technical Field

[0001] This invention relates to the field of submersible pump control system technology, and more specifically, to a method and system for adaptive energy-saving control of permanent magnet submersible pumps. Background Technology

[0002] Permanent magnet submersible pumps are widely used in fluid transportation scenarios such as deep well water extraction, mine drainage, oilfield extraction, and underground engineering construction. In these scenarios, the transported medium often contains gas, and the pumps may face challenges such as unstable fluid supply, complex pipeline layouts, and frequent valve operations. In these applications, the pumps are typically driven by frequency converters. Speed ​​control and energy-saving operation are achieved by adjusting the output frequency. The control system usually relies solely on the voltage, current, active power, frequency, and estimated speed values ​​that the frequency converter itself can collect. When necessary, it is supplemented with pressure detection elements installed on the ground to obtain pressure signals, enabling the monitoring of the pump's operating status and the handling of abnormal conditions.

[0003] In existing technologies, the handling logic for reduced load conditions during pump operation is based on threshold comparison and delay judgment of monitored parameters. When the control system detects that electrical parameters such as current and active power are lower than preset thresholds, and this state persists for a set duration, it determines that there is a risk of abnormal pump operation and triggers preset handling actions. Common handling methods include reducing the operating frequency, intermittent shutdown, and frequency sweeping. The core principle is to try to eliminate the impact of abnormal factors on the pump's water delivery function by changing the pump's speed or operating state, and to avoid damage to the equipment due to continuous abnormal operation. In particular, for handling airlock conditions, existing technologies believe that reducing the frequency can reduce the accumulation of gas in the pump body, and frequency sweeping can break the gas blockage, thereby restoring the pump's normal water delivery capacity.

[0004] However, in actual operation, the causes of load reduction are diverse, with airlock being only one of them. Cavitation, insufficient water supply, momentary valve closure, and transient fluctuations in pipelines are just some of the factors that can cause electrical parameters such as current and active power to exhibit similar trends to those caused by airlock, making it difficult to distinguish between load reduction phenomena caused by different factors. Current technologies lack effective means to decouple these multiple causes, relying solely on threshold comparisons of a single or a few monitoring parameters. This fails to extract characteristic information that accurately distinguishes different causes. Furthermore, factors such as interference from long cable transmissions, grid voltage fluctuations, and transient disturbances during pump operation further interfere with the stability of monitoring signals, reducing the characteristic identification of monitoring parameters and exacerbating the risk of confusion between different causes. Simultaneously, current technologies do not quantify and optimize the parameters for handling actions. Parameters such as the frequency reduction amplitude, holding duration, recovery slope, and frequency sweep range of handling actions are all fixed settings, failing to consider the severity and duration of abnormal conditions, and neglecting to establish a correlation between handling action parameters and abnormal condition characteristics.

[0005] The aforementioned technical defects cause the control system to easily misjudge non-airlock conditions as airlock conditions and incorrectly apply unlocking actions, leading to frequent adjustments in the pump's operating frequency and increased unnecessary energy consumption. Frequent switching between operating conditions exacerbates the mechanical stress on the motor and pump body, causing significant fluctuations in motor temperature rise, accelerating the aging of critical components such as seals, bearings, and impellers, reducing equipment stability and reliability, shortening equipment lifespan, and increasing maintenance costs. Furthermore, incorrect actions cannot effectively resolve the load reduction problem caused by non-airlock conditions, potentially leading to the continued development of abnormal conditions or even pump shutdown, affecting the continuity of fluid delivery and adversely impacting the overall production schedule.

[0006] In view of this, the present invention proposes an adaptive energy-saving control method and system for permanent magnet submersible pumps to solve the above problems. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art and achieve the above objectives, the present invention provides the following technical solution: an adaptive energy-saving control method for permanent magnet submersible pumps, comprising:

[0008] The electrical parameter sequence, consisting of voltage, current, active power, output frequency, and speed estimation sequences collected by the frequency converter, the pressure sequence collected by the pressure detection element, and the sampling timestamp sequence are aligned according to the sampling timestamp sequence to obtain the running aligned data.

[0009] In the running alignment data, a stable operating segment is selected, and within the stable operating segment, the active power sequence is statistically analyzed according to the output frequency sequence to obtain a frequency power baseline table. Within the stable operating segment, the voltage fluctuation reference value is extracted from the voltage sequence.

[0010] The frequency power baseline table is compared with the active power sequence to obtain the power residual sequence;

[0011] By combining voltage fluctuation reference values, continuous negative deviation intervals are identified in the power residual sequence to obtain load reduction event segments; based on the operation alignment data, event fingerprint data is generated in the load reduction event segments and a candidate cause set is determined;

[0012] Segmented perturbation detection commands are generated based on event fingerprint data, and perturbation detection response data is collected. Based on the perturbation detection response data, candidate cause sets are discriminated and calculated to obtain cause discrimination results and anomaly severity index. Based on the cause discrimination results, anomaly severity index and frequency power baseline table, the set of handling action parameters is calculated, and the output frequency is adjusted according to the set of handling action parameters.

[0013] Furthermore, the method for filtering stable operating segments from the aligned data and constructing a frequency power baseline table and voltage fluctuation reference values ​​within these stable operating segments includes:

[0014] The continuous analysis range is determined by the sampling timestamp sequence, and the voltage alignment value, active power alignment value, and output frequency alignment value are read using the alignment time point set as an index; the alignment time point set is generated by the main sampling timestamp sequence.

[0015] Within the continuous analysis range, the difference between the maximum and minimum values ​​of the output frequency alignment value is calculated according to a preset time window to determine the preset time window for the output frequency sequence to be stationary.

[0016] Within a preset time window where the output frequency sequence is stable, the difference between the maximum and minimum values ​​of the active power alignment value and the peak value of the change in active power alignment value at adjacent alignment time points are statistically analyzed to determine the candidate preset time window for the stable operation segment.

[0017] Merge candidate preset time windows for stable operation segments according to the aligned time point set, and record the start sampling timestamp and end sampling timestamp to obtain the stable operation segment;

[0018] During the stable operation period, the output frequency alignment values ​​are merged into frequency groups according to the preset frequency interval, and the active power alignment values ​​corresponding to each frequency group are collected to form a power sample set. The central trend and dispersion of the power sample set are statistically analyzed to obtain the frequency power baseline table.

[0019] The fluctuation amplitude is obtained by calculating the difference between the maximum and minimum values ​​of the voltage alignment value during the stable operation period, and the fluctuation rhythm is obtained by calculating the number of times the voltage alignment value switches between rising and falling. The central trend and dispersion of the fluctuation amplitude and fluctuation rhythm are calculated to obtain the voltage fluctuation reference value.

[0020] Furthermore, methods for obtaining the power residual sequence include:

[0021] The alignment data is traversed according to the alignment time point set. The output frequency alignment value and active power alignment value are read, and the output frequency alignment value is merged into the frequency group of the frequency power baseline table to obtain the frequency group label.

[0022] Locate the corresponding entry in the frequency power baseline table based on the frequency grouping mark, and read the center trend and dispersion to obtain the baseline power center trend and baseline power dispersion. When the output frequency alignment value is located at the boundary of the adjacent frequency group, interpolate the output frequency alignment value to obtain the baseline power center trend and baseline power dispersion.

[0023] For frequency groups with a sample size lower than a preset value, the baseline power center trend and baseline power dispersion are determined by merging the frequency group and adjacent frequency groups in the table.

[0024] For each alignment time point, the difference between the active power alignment value and the baseline power center trend is calculated to obtain the power residual value. The output frequency alignment value and the degree of dispersion of the baseline power are recorded to form a power residual record. The power residual records are spliced ​​according to the alignment time point set to obtain the power residual sequence.

[0025] Furthermore, the output frequency sequence and active power sequence in the running alignment data are read, and the central trend, dispersion and sample size of each frequency group in the frequency power baseline table are read to form a frequency power index relationship. After the frequency power index relationship is clear, the running alignment data can be quickly associated with the corresponding table entry in the frequency power baseline table at any alignment time point.

[0026] Furthermore, methods for obtaining load reduction event fragments include:

[0027] The negative deviation judgment condition is constructed based on the power residual value and the degree of dispersion of the baseline power, and the negative deviation points are marked. The negative deviation points are merged according to the set of aligned time points, and the continuous negative deviation interval is determined based on the preset duration and preset quantity.

[0028] Read the magnitude and direction of the output frequency alignment value change corresponding to the continuous negative deviation interval. When the output frequency alignment value is rapidly decreasing or rapidly increasing and the active power alignment value changes synchronously, mark the speed regulation related interval. When the output frequency alignment value is stable or changes slowly and the power residual value is continuously negative, determine the candidate interval of the load reduction event.

[0029] Read the voltage alignment value corresponding to the candidate interval of load reduction event, and count the fluctuation amplitude and fluctuation rhythm. When the fluctuation amplitude or fluctuation rhythm exceeds the typical fluctuation boundary of the voltage fluctuation reference, the voltage fluctuation interference interval is eliminated. When the fluctuation amplitude and fluctuation rhythm fall within the typical fluctuation boundary, the retention interval is determined.

[0030] The start and end times of the event are determined based on the reserved interval, and the reserved time windows before and after the event are expanded according to the set of aligned time points to obtain the load reduction event segments.

[0031] Furthermore, methods for generating event fingerprint data and determining a set of candidate causes include:

[0032] The start and end alignment time points of the load reduction event segment are located in the alignment time point set interval in the running alignment data, and the electrical parameter sequence and pressure sequence are extracted. The boundary alignment time points of the reserved time window before the event, the event duration window and the reserved time window after the event are marked to obtain the event segment boundary information.

[0033] Based on the event segment boundary information, the reserved time window before the event is used as the reference segment, the event duration window is used as the abnormal segment, and the reserved time window after the event is used as the recovery segment. The active power sequence, speed estimation sequence and output frequency sequence within the reference segment are statistically analyzed to obtain the reference power level, reference speed level and reference frequency level.

[0034] In the load reduction event segment, electrical parameter decline morphology features are generated based on active power sequence and current sequence, and electrical parameter fluctuation texture features are generated based on current sequence and active power sequence.

[0035] In the load reduction event segment, voltage disturbance stripping features are generated based on voltage sequence and voltage fluctuation reference, and pressure coupling response features are generated based on pressure sequence;

[0036] Speed ​​regulation correlation features are generated based on the output frequency sequence, speed estimation sequence, and active power sequence;

[0037] Event fingerprint data is obtained by combining the electrical parameter descent morphology features, electrical parameter fluctuation texture features, voltage interference stripping features, pressure coupling response features, and speed regulation correlation features.

[0038] The candidate cause set was determined to be airlock, cavitation, insufficient water supply, instantaneous valve closure, and transient pipeline fluctuation. Based on the pressure coupling response characteristics, electrical parameter fluctuation texture characteristics, and speed regulation correlation characteristics, the candidate cause set was retained and eliminated to obtain the candidate cause set.

[0039] Furthermore, the method for generating segmented perturbation detection commands based on event fingerprint data and collecting perturbation detection response data includes:

[0040] Based on the load reduction event segments, the alignment time point set interval is located in the running alignment data, and the running segment data is obtained by reading the electrical parameter sequence and pressure sequence, and the correspondence between the sampling timestamp sequence and the alignment time point set is retained;

[0041] Extract electrical parameter morphology features, electrical parameter fluctuation texture features, voltage interference stripping features, pressure coupling response features, and speed regulation correlation features from event fingerprint data, and establish correspondence between them and the retained items in the candidate cause set;

[0042] Determine the starting alignment time point in the running fragment data, and determine the instruction reference frequency using the corresponding output frequency alignment value;

[0043] The frequency change scale is determined based on the characteristics of the electrical parameter descent morphology and the electrical parameter fluctuation texture, and the frequency change boundary is constrained based on the frequency power baseline table;

[0044] Based on the pressure coupling response characteristics, the segment sequence and duration of the constant frequency holding segment are determined, and the frequency up-adjustment segment, frequency down-adjustment segment, and constant frequency holding segment are generated;

[0045] The segmented perturbation detection command is applied to the aligned time point set to generate the output frequency target sequence, and the output frequency target sequence is truncated at the boundary based on the frequency power baseline table; a transition constraint is added to the output frequency target sequence based on the speed regulation correlation feature;

[0046] Drive the frequency converter to adjust the output frequency sequence according to the target output frequency sequence and collect the perturbation detection response data;

[0047] Alignment results are obtained by aligning the perturbation detection response data according to the sampling timestamp sequence and segment boundary information is recorded. Based on the segment boundary information, the perturbation detection response data is segmented to obtain a set of response segments.

[0048] Furthermore, the method of discriminating and calculating the candidate cause set based on the perturbation detection response data to obtain the cause discrimination result and the anomaly severity index, and calculating the action parameter set based on the cause discrimination result, the anomaly severity index and the frequency power baseline table, and adjusting the output frequency according to the action parameter set includes:

[0049] Based on the segment boundary information of the response segment set, the voltage alignment value, active power alignment value, output frequency alignment value, speed estimation alignment value, and pressure alignment value of the up-frequency segment, down-frequency segment, and constant frequency holding segment are extracted from the perturbation detection response data. The statistical results within each segment are obtained by statistically analyzing the power center level, power fluctuation amplitude, pressure center level, pressure fluctuation amplitude, and frequency level within the segment.

[0050] The statistical results within the segment are correlated with the event fingerprint data according to the start and end alignment time points of the load reduction event segment, and the voltage alignment value fluctuation amplitude is marked based on the voltage interference stripping feature.

[0051] Based on the statistical results within the segment, evidence of recovery, pressure increase, wave reduction, slow return, and loss of return are constructed. Based on the evidence of recovery, pressure increase, wave reduction, slow return, and loss of return, the candidate cause set is retained and eliminated to obtain a converged candidate cause set. Based on the converged candidate cause set, the target cause name is determined to obtain the cause discrimination result.

[0052] Based on the perturbation detection response data and frequency power baseline table, the baseline power center trend is located, and the degree of power deviation, pressure deviation and fluctuation deviation are calculated and combined to obtain the anomaly severity index.

[0053] Based on the cause identification results, the severity index of the abnormality and the frequency power baseline table, the set of action parameters for handling actions is calculated. The set of action parameters for handling actions includes the frequency downsampling amplitude, the frequency downsampling duration, the frequency recovery slope, the frequency sweep range, the frequency sweep step amplitude and the frequency sweep dwell time.

[0054] The output frequency target sequence is generated according to the action parameter set, and the output frequency is adjusted. The output frequency target sequence is limited to the range of the lowest and highest frequency groups covered by the frequency power baseline table.

[0055] Furthermore, methods for obtaining the alignment data include:

[0056] The electrical parameter sequence and pressure sequence are time-stamped by adding time stamps to the sampling timestamp sequence, resulting in time-stamped electrical parameter sequence and time-stamped pressure sequence;

[0057] Construct an alignment time point set, and use the alignment time point set as a reference to interpolate and align the time-stamped electrical parameter sequence and the time-stamped pressure sequence to obtain voltage alignment value, active power alignment value, output frequency alignment value, speed estimation alignment value, and pressure alignment value.

[0058] The interpolated and aligned electrical parameter sequence and pressure sequence are subjected to consistent trimming and gap filling to obtain the running aligned data;

[0059] The voltage alignment value, active power alignment value, output frequency alignment value, speed estimation alignment value, and pressure alignment value are combined according to the alignment time point set to obtain the operation alignment data.

[0060] The permanent magnet submersible pump operating condition adaptive energy-saving control system includes:

[0061] The data acquisition module is used to align the voltage sequence, current sequence, active power sequence, output frequency sequence, and speed estimation sequence collected by the frequency converter, the pressure sequence collected by the pressure detection element, and the sampling timestamp sequence to obtain running aligned data.

[0062] The reference quantity filtering module is used to filter the stable operating segment in the running alignment data, and to statistically analyze the active power sequence according to the output frequency sequence within the stable operating segment to obtain the frequency power baseline table. The voltage fluctuation reference quantity is extracted from the voltage sequence within the stable operating segment.

[0063] The power comparison module is used to compare the frequency power baseline table with the active power sequence to obtain the power residual sequence;

[0064] The cause determination module is used to identify continuous negative deviation intervals in the power residual sequence by combining voltage fluctuation reference values, and obtain load reduction event segments; based on the running alignment data, it generates event fingerprint data in the load reduction event segments and determines the candidate cause set;

[0065] The frequency adjustment module is used to generate segmented perturbation detection commands and collect perturbation detection response data based on event fingerprint data; to discriminate and calculate the candidate cause set based on the perturbation detection response data, to obtain the cause discrimination result and the anomaly severity index, to calculate the handling action parameter set based on the cause discrimination result, the anomaly severity index and the frequency power baseline table, and to adjust the output frequency according to the handling action parameter set.

[0066] Compared with the prior art, the technical effects and advantages of the permanent magnet submersible pump adaptive energy-saving control method and system of the present invention are as follows:

[0067] The permanent magnet submersible pump operating condition adaptive energy-saving control scheme of the present invention is based on the electrical parameter sequence of voltage, current, active power, output frequency, and speed estimation collected by the frequency converter, combined with the optional ground pressure signal and sampling timestamp sequence, and obtains operation alignment data through data alignment. Then, the stable operation segment is selected to generate a frequency power baseline table and voltage fluctuation reference quantity. The power residual sequence is obtained by power comparison. Combined with the voltage fluctuation reference quantity, load reduction event segments are identified and event fingerprint data is generated to determine the candidate cause set. Then, the cause and abnormality severity are determined by segmented perturbation detection and response data. Finally, the appropriate action parameter set is calculated to adjust the output frequency.

[0068] This invention eliminates the need for additional downhole sensors, achieving precise differentiation of multiple causes of load reduction using only existing acquireable signals. This allows for accurate matching of treatment actions with their causes and severity, avoiding frequent speed adjustments caused by accidental triggering of unlocking actions. Simultaneously, the quantitative optimization of treatment action parameters reduces additional energy consumption and temperature fluctuations, lowers mechanical stress and aging rates of key components in the motor and pump body, ensures continuous fluid delivery, and solves problems in existing technologies such as confusion regarding load reduction causes, fixed treatment actions, increased energy consumption, and insufficient equipment operational stability. This improves the reliability and economy of permanent magnet submersible pump operation and extends equipment lifespan. Attached Figure Description

[0069] Figure 1 This is a schematic diagram of the adaptive energy-saving control system for a permanent magnet submersible pump according to an embodiment of the present invention.

[0070] Figure 2 This is a flowchart of the adaptive energy-saving control method for permanent magnet submersible pumps according to an embodiment of the present invention.

[0071] Figure 3 This is a flowchart of a method for obtaining a power residual sequence according to an embodiment of the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be described in detail, clearly, and completely below with reference to the accompanying drawings. It should be particularly noted that the specific embodiments described below are only for better illustrating and explaining the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and substance of the present invention, those skilled in the art can modify, adjust, or make equivalent substitutions based on the content disclosed in the present invention, and these should all be considered within the scope of protection of the present invention.

[0073] Example 1:

[0074] Please see Figure 1 As shown, this embodiment discloses an adaptive energy-saving control system for permanent magnet submersible pumps, including a data acquisition module, a reference quantity screening module, a power comparison module, a cause determination module, and a frequency adjustment module. Each module is connected via wired or wireless means to achieve data transmission.

[0075] The data acquisition module is used to align the voltage sequence, current sequence, active power sequence, output frequency sequence, and speed estimation sequence collected by the frequency converter into an electrical parameter sequence, the pressure sequence collected by the pressure detection element, and the sampling timestamp sequence according to the sampling timestamp sequence to obtain running aligned data.

[0076] The specific implementation method for aligning the electrical parameter sequence (composed of voltage, current, active power, output frequency, and speed estimation sequences collected by the frequency converter), the pressure sequence collected by the pressure detection element, and the sampling timestamp sequence according to the sampling timestamp sequence to obtain the operational aligned data is as follows:

[0077] The acquisition methods and sampling intervals for the voltage, current, active power, output frequency, and speed estimation sequences collected by the frequency converter are set, and the acquisition methods and sampling intervals for the pressure sequence collected by the pressure detection element are set simultaneously. The voltage, current, active power, output frequency, and speed estimation sequences collected by the frequency converter all originate from the frequency converter's operation monitoring interface. The collected content covers power input status, load status, and speed regulation status. The voltage sequence reflects amplitude changes introduced by power supply fluctuations and long cable transmission; the current sequence reflects changes in electromagnetic load on the motor side and, together with the active power sequence, characterizes the underload phenomenon; the active power sequence reflects the energy consumption level of the pump under the current operating conditions; the output frequency sequence reflects changes in the frequency converter's adjustment commands for pump speed; the speed estimation sequence reflects the mapping result of the electrical parameter sequences to the mechanical side rotation status; and the pressure sequence collected by the pressure detection element reflects the fluid transport status at the pipeline end. The sampling interval is configured in a fixed manner, so that the voltage sequence, current sequence, active power sequence, output frequency sequence, speed estimation sequence, and pressure sequence are all output in the form of continuous time series, and a corresponding sampling timestamp sequence is generated for each sampling. The time continuity of the electrical parameter sequence and pressure sequence is constrained by the uniform sampling interval.

[0078] The electrical parameter sequence and pressure sequence are time-stamped according to the sampling timestamp sequence, forming time-stamped electrical parameter sequence and time-stamped pressure sequence. Specifically, when the frequency converter collects voltage sequence, current sequence, active power sequence, output frequency sequence, and speed estimation sequence, a corresponding sampling timestamp is recorded for each sampled value, so that each voltage sampled value in the voltage sequence corresponds to a unique sampling timestamp, each current sampled value in the current sequence corresponds to a unique sampling timestamp, each active power sampled value in the active power sequence corresponds to a unique sampling timestamp, each output frequency sampled value in the output frequency sequence corresponds to a unique sampling timestamp, and each speed estimation sampled value in the speed estimation sequence corresponds to a unique sampling timestamp; when the pressure detection element collects the pressure sequence, a corresponding sampling timestamp is recorded for each pressure sampled value, so that each pressure sampled value in the pressure sequence corresponds to a unique sampling timestamp. If the sampling timestamp sequence output by the frequency converter and the sampling timestamp sequence output by the pressure sensing element come from different timing sources, the sampling timestamp sequence output by the frequency converter is used as the master sampling timestamp sequence. The sampling timestamp sequence output by the pressure sensing element is then converted to the time scale of the master sampling timestamp sequence. This allows the time-stamped pressure sequence to share the same sampling timestamp sequence representation as the time-stamped electrical parameter sequence. Introducing a unified sampling timestamp sequence for both the electrical parameter and pressure sequences enables data from different acquisition channels to be combined on the same time axis, reducing the accumulation of time misalignments caused by differences in timing sources.

[0079] An aligned time point set is constructed, and the time-stamped electrical parameter sequence and the time-stamped pressure sequence are interpolated and aligned based on this set. The aligned time point set is generated from the master sampling timestamp sequence. When the intervals between adjacent timestamps in the master sampling timestamp sequence are consistent, the master sampling timestamp sequence is used as the aligned time point set. When there are missing timestamps in the master sampling timestamp sequence or the intervals between adjacent timestamps are inconsistent, a continuous aligned time point set is regenerated according to a fixed sampling interval, so that the aligned time point set covers the common time range of the electrical parameter sequence and the pressure sequence. Subsequently, taking each alignment time point as the target time point, the nearest voltage sample values ​​on both sides of the target time point are selected from the time-stamped voltage sequence and interpolated according to the time position to obtain the voltage alignment value of the target time point; the nearest current sample values ​​on both sides of the target time point are selected from the time-stamped current sequence and interpolated according to the time position to obtain the current alignment value of the target time point; the nearest active power sample values ​​on both sides of the target time point are selected from the time-stamped active power sequence and interpolated according to the time position to obtain the active power alignment value of the target time point; the nearest output frequency sample values ​​on both sides of the target time point are selected from the time-stamped output frequency sequence and interpolated according to the time position to obtain the output frequency alignment value of the target time point; the nearest speed estimation sample values ​​on both sides of the target time point are selected from the time-stamped speed estimation sequence and interpolated according to the time position to obtain the speed estimation alignment value of the target time point; and the nearest pressure sample values ​​on both sides of the target time point are selected from the time-stamped pressure sequence and interpolated according to the time position to obtain the pressure alignment value of the target time point. Interpolation alignment uses the same rule across both the electrical parameter and pressure sequences to ensure that the aligned values ​​for voltage, current, active power, output frequency, speed estimation, and pressure at the same target time point come from the same set of aligned time points. This converts the electrical parameter and pressure sequences into aligned data indexed by the same set of aligned time points, reducing the impact of asynchronous acquisition and inconsistent sampling intervals on multi-source data fusion.

[0080] The interpolated and aligned electrical parameter and pressure sequences undergo consistency trimming and gap filling to obtain operational aligned data suitable for continuous analysis. Consistency trimming primarily focuses on the alignment time point set, eliminating time points where voltage, current, active power, output frequency, speed estimation, or pressure alignment values ​​cannot be generated at a given alignment time point, thus preventing inconsistencies in length between the electrical parameter and pressure sequences within the same alignment time point set. When the missing duration does not exceed a preset missing duration, the missing segment is interpolated and filled according to the alignment time point set, ensuring continuous voltage, current, active power, output frequency, speed estimation, and pressure alignment values ​​for the missing segment. When the missing duration exceeds the preset missing duration, the missing segment is discarded, and the segments before and after the discarded segment are retained as two consecutive segments, maintaining the integrity of the operational aligned data within each consecutive segment. The preset missing duration is determined based on a fixed sampling interval; when the sampling interval is 100 milliseconds, the preset missing duration is 500 milliseconds, used to distinguish between short-term communication fluctuations and long-term acquisition unavailability.

[0081] The aligned voltage, current, active power, output frequency, speed estimation, and pressure values ​​are combined according to the alignment time point set to form operational alignment data, which is then output in a structured manner. Specifically, for each alignment time point, the corresponding voltage, current, active power, output frequency, speed estimation, and pressure values ​​are combined into an operational record, with the alignment time point serving as the index. The set of operational records constitutes the operational alignment data. The operational alignment data retains the correspondence between the sampling timestamp sequence and the alignment time point set, enabling subsequent steps to trace the sampling timestamp corresponding to each operational record when referencing the operational alignment data. When outputting the operational alignment data, the names of the electrical parameter sequences and pressure sequences are kept consistent, explicitly including the voltage sequence, current sequence, active power sequence, output frequency sequence, speed estimation sequence, pressure sequence, and sampling timestamp sequence, and ensuring that all of these sequences use the alignment time point set as a unified index.

[0082] The reference quantity filtering module is used to filter stable operating segments in the running alignment data, and to statistically analyze the active power sequence according to the output frequency sequence within the stable operating segment to obtain a frequency power baseline table. Within the stable operating segment, the voltage fluctuation reference quantity is extracted from the voltage sequence.

[0083] After obtaining the operational alignment data, a stable operating period is selected based on the output frequency sequence, active power sequence, and voltage sequence in the operational alignment data. Within the stable operating period, a frequency power baseline table and voltage fluctuation reference values ​​are constructed. The specific implementation method is as follows:

[0084] In the running alignment data, the continuous analysis range is determined according to the sampling timestamp sequence, and the voltage alignment value, active power alignment value, output frequency alignment value and speed estimation alignment value are read point by point using the alignment time point set as an index to form a time sequence record consistent with the running alignment data. When the running alignment data contains multiple continuous segments, each continuous segment is processed independently to avoid time discontinuity between different continuous segments interfering with the interpretation of changes in the output frequency sequence.

[0085] The stationary candidate interval is determined based on the amplitude of the output frequency sequence change within a preset time window. The preset time window continuously takes values ​​along the set of aligned time points. The duration of the preset time window ranges from 0.5 seconds to 5 seconds. The number of aligned time points included in the preset time window is calculated from the fixed sampling interval corresponding to the sampling timestamp sequence, ensuring that the preset time window can cover short-term fluctuations in the output frequency sequence during one adjustment process without excessively spanning different operating stages. The selection of the preset time window follows a rule based on the adjustment rhythm of the output frequency sequence: when the output frequency sequence changes more densely between adjacent aligned time points in the running aligned data, the preset time window is closer to 0.5 seconds in length to isolate the frequency change interval earlier; when the output frequency sequence in the running aligned data remains nearly unchanged for a longer period of time, the preset time window is closer to 5 seconds in length to obtain a more sufficient sample size at the same operating frequency level for subsequent frequency power baseline table statistics. To avoid the difference between the maximum and minimum values ​​of the output frequency alignment value being affected by a single sampling point due to an excessively short preset time window, the number of alignment time points within the preset time window shall not be less than 5. To avoid the same preset time window spanning multiple adjustments across the output frequency sequence due to an excessively long preset time window, the duration of the preset time window shall not be greater than twice the typical adjustment interval covered by the set of alignment time points. The typical adjustment interval is obtained by statistically analyzing the interval time of the step change in the output frequency alignment value in the running alignment data.

[0086] Within each preset time window, the difference between the maximum and minimum values ​​of the output frequency alignment value is calculated, and this difference is compared with a preset value. Preset time windows where the difference does not exceed the preset value are designated as preset time windows with stable output frequency sequences. The preset value is determined by the quantization resolution of the output frequency alignment value in the running alignment data and the minimum adjustment step of the inverter's output frequency sequence. The preset value is ensured to be no less than twice the quantization resolution of the output frequency alignment value and no more than once the minimum adjustment step, thus avoiding the misinterpretation of quantization jitter as frequency changes and preventing a single minimum step adjustment from covering the preset time window of a stable output frequency sequence. Simultaneously, the center level of the output frequency alignment value within this preset time window is recorded. The center level is defined using the median; the output frequency alignment values ​​within the preset time window are sorted, and the median is taken as the center level to characterize the operating frequency level corresponding to this preset time window. After the preset time windows with stable output frequency sequences are marked, stable operating segments are selected based on clear frequency change amplitude and reproducible operating frequency level characteristics.

[0087] Based on a preset time window where the output frequency sequence is stable, non-stable intervals are further eliminated according to sudden jumps in the active power sequence. Within each marked preset time window, the difference between the maximum and minimum values ​​of the active power alignment value is calculated, and the peak value of the change in the active power alignment value at adjacent alignment time points is also calculated. When any of the above statistics exceeds a preset value, the preset time window is removed from the preset time window for stable output frequency sequences. When none of the above statistics exceed the preset value, the preset time window is retained as a candidate preset time window for stable operation. After both the output frequency sequence and the active power sequence satisfy the stability constraint, the representativeness of the stable operation section for the stable load state of the pump is improved.

[0088] Adjacent candidate time windows for stable operation segments are merged to form stable operation segments with the required duration. The merging process follows the alignment of time point sets. If two candidate time windows for stable operation segments are consecutive in time or the interval does not exceed a preset interval, they are merged into the same stable operation segment. The duration of the merged stable operation segments is calculated. Stable operation segments with durations lower than a preset value are discarded, while those with durations not lower than the preset value are retained. The start and end sampling timestamps of each stable operation segment are recorded for direct location in the alignment data. After obtaining a set of stable segments through duration constraints, the statistical sample source for the frequency power baseline table becomes more reliable.

[0089] Within a stable operating period, the active power sequence is statistically analyzed according to the output frequency sequence to obtain a frequency power baseline table. Specifically, for each stable operating period, the output frequency alignment value and active power alignment value corresponding to all alignment time points within that stable operating period are read, and the output frequency alignment values ​​are merged according to a preset frequency interval. Each merged group corresponds to a frequency group; for each frequency group, the active power alignment values ​​under that frequency group are collected as the power sample set of that frequency group. The preset frequency interval is between 0.1 Hz and 2 Hz. The selection of the preset frequency interval is based on the quantization resolution of the output frequency sequence, the minimum adjustment step of the inverter executing the output frequency sequence, and the distribution density of the output frequency alignment values ​​within the stable operating segment. When the quantization resolution and minimum adjustment step of the output frequency alignment values ​​in the operating alignment data are small and the frequency points covered by the output frequency alignment values ​​within the stable operating segment are relatively dense, the preset frequency interval is between 0.1 Hz and 0.5 Hz. This is used to form finer frequency groups near the same operating frequency level, thereby improving the accuracy of the frequency power baseline table in depicting frequency changes. When the quantization resolution or minimum adjustment step of the output frequency alignment values ​​in the operating alignment data is large and the output frequency alignment values ​​within the stable operating segment are concentrated in a small number of frequency points, the preset frequency interval is between 0.5 Hz and 2 Hz. This is used to reduce the number of frequency groups and increase the power sample set size of each frequency group. The preset frequency interval is no less than twice the quantization resolution of the output frequency alignment value and no less than once the minimum adjustment step size to avoid the frequency group boundaries falling into the quantization jitter range, causing the output frequency alignment values ​​to be repeatedly merged between adjacent frequency groups. The preset frequency interval is no greater than 1 / 10 of the coverage range of the output frequency alignment values ​​within the stable operation period to avoid excessively wide frequency groups causing active power alignment values ​​at different operating frequency levels to be grouped into the same power sample set, thus amplifying the inherent differences in the power sample set. After establishing the correspondence between frequency groups and power sample sets for the output frequency sequence and active power sequence within the stable operation period, the frequency power baseline table can organize data based on frequency groups and provide a unified frequency index for subsequent point-by-point comparison of the power residual sequence.

[0090] Central tendency and dispersion statistics are performed on the power sample set for each frequency group to form the entries in the frequency power baseline table. Central tendency is defined using the median. The power sample set is sorted by power value from smallest to largest. When the number of samples is odd, the middle sample in the sorted sequence is taken as the central tendency; when the number of samples is even, the average of the two middle samples in the sorted sequence is taken as the central tendency. Dispersion is defined using the absolute deviation from the median. The absolute value of the difference between each power sample in the power sample set and the central tendency is taken to form a deviation set. The deviation sets are then sorted, and the median is taken as the dispersion. The number of samples is the number of power samples in the power sample set, and this value, along with the frequency group, central tendency, and dispersion, is entered into the entries of the frequency power baseline table.

[0091] When the number of samples in a certain frequency group is lower than a preset value, the frequency group is merged with the adjacent frequency group to form a merged frequency group, and the power sample sets corresponding to the two frequency groups are merged to form a merged power sample set. Then, the merged power sample set is re-statistically analyzed and the contents of the frequency power baseline table are updated according to the aforementioned definitions of central tendency and dispersion. The number of samples is recorded according to the number of power samples in the merged power sample set to avoid random deviations in the frequency power baseline table caused by too few samples. The preset values ​​range from 20 to 200, and are selected based on the sampling interval, the minimum duration of the stable operation segment, and the density of the output frequency alignment value within the frequency group. With a sampling interval of 100 milliseconds and a minimum duration of 10 seconds for the stable operation segment, a single stable operation segment can generate 100 power samples within a single frequency group. The preset value of 50 ensures at least 5 seconds of power samples are covered within the frequency group. With a sampling interval of 200 milliseconds and a minimum duration of 6 seconds for the stable operation segment, a single stable operation segment can generate 30 power samples within a single frequency group. The preset value of 20 ensures at least 4 seconds of power samples are covered within the frequency group. The preset value should not be less than the number of power samples corresponding to one preset time window to ensure that the statistical coverage of the central trend and dispersion is at least within the natural fluctuations of the output frequency sequence and active power sequence within one preset time window. The preset value should not be greater than half the total number of power samples within the stable operation segment to avoid frequent merging of frequency groups under sample quantity constraints, which would lead to a decrease in the frequency resolution of the frequency power baseline table. Each entry in the frequency power baseline table includes frequency grouping, central tendency, dispersion, and sample size, allowing subsequent comparison processes to take into account both baseline level and baseline fluctuation boundaries.

[0092] A unified summary of the frequency power baseline table is performed across multiple stable operating periods to obtain a frequency power baseline table covering the commonly used output frequency range. When summarizing entries for the same frequency group, the power sample sets of that frequency group in each stable operating period are merged into a larger power sample set, and then the entries for that frequency group are updated according to the aforementioned central trend and dispersion statistical methods. When the output frequency sequence exhibits slight drift across different stable operating periods, the merging rules for frequency groups still ensure that samples at the same frequency level converge to the same frequency group. After being summarized across stable operating periods, the frequency power baseline table can reflect the typical active power level of the pump body when it is at the same output frequency sequence level at different times.

[0093] Within the stable operating range, voltage fluctuation reference values ​​are extracted from the voltage sequence. For each stable operating range, the voltage alignment value within that range is read, and the difference between the maximum and minimum values ​​is calculated as the fluctuation amplitude for that range. Simultaneously, the number of times the voltage alignment value switches between rising and falling values ​​is calculated as the fluctuation rhythm for that range. The fluctuation amplitudes and rhythms of all stable operating ranges are summarized, and the voltage fluctuation reference values ​​are obtained by statistically analyzing their central trend and dispersion. These reference values ​​include the central trend of the fluctuation amplitude, the dispersion of the fluctuation amplitude, the central trend of the fluctuation rhythm, and the dispersion of the fluctuation rhythm, while maintaining a traceable correlation with the sampling timestamp sequence of the operating alignment data. Once extracted from the stable operating range, the voltage fluctuation reference values ​​can characterize the typical fluctuation pattern of the voltage sequence under non-abnormal load conditions, providing a basis for subsequently eliminating false underloads caused by pure voltage fluctuations in the power residual sequence.

[0094] Output stable operating segment, frequency power baseline table and voltage fluctuation reference quantity; the stable operating segment is characterized by the start sampling time stamp and the end sampling time stamp, the frequency power baseline table is indexed by frequency group and includes the corresponding active power central trend and dispersion, and the voltage fluctuation reference quantity includes the voltage sequence fluctuation amplitude and the central trend and dispersion of the fluctuation rhythm.

[0095] The power comparison module is used to compare the frequency power baseline table with the active power sequence to obtain the power residual sequence.

[0096] Please see Figure 3 As shown, after obtaining the frequency power baseline table, the frequency power baseline table is compared with the active power sequence in the operational alignment data to form a power residual sequence that can characterize the load reduction phenomenon. The specific implementation method is as follows:

[0097] The output frequency and active power sequences from the alignment data are read, and the central tendency, dispersion, and sample size of each frequency group in the frequency power baseline table are retrieved. The frequency power baseline table is then arranged according to the order of the frequency groups, preserving the adjacency relationships between them to form a frequency power index relationship for point-by-point comparison. The calculation of the power residual sequence at each subsequent alignment time point is based on this frequency power index relationship. Once the frequency power index relationship is clear, the alignment data can be quickly associated with the corresponding entry in the frequency power baseline table at any alignment time point.

[0098] The alignment data is traversed sequentially according to the time sequence of the alignment time point set. For each alignment time point, the corresponding output frequency alignment value and active power alignment value are read, and the output frequency alignment value is merged into the frequency group of the frequency power baseline table. The merging rule is consistent with the preset frequency interval when constructing the frequency power baseline table, thus ensuring the consistency of the output frequency alignment value and the frequency group on the numerical boundaries. If the output frequency alignment value falls near the boundary of two frequency groups, the two frequency groups are recorded as adjacent frequency groups for subsequent interpolation to determine the baseline power level. After the output frequency alignment values ​​are merged, each alignment time point receives a frequency group label consistent with the frequency power baseline table.

[0099] For each alignment time point, the corresponding entry in the frequency power baseline table is located according to the frequency grouping mark, and the central trend and dispersion of that entry are read to obtain the baseline power central trend and baseline power dispersion for that alignment time point. When there are adjacent frequency groups at the alignment time point, the central trend and dispersion of the adjacent frequency groups are read respectively, and interpolation is performed according to the position of the output frequency alignment value between the two frequency groups to obtain the baseline power central trend and baseline power dispersion for that alignment time point. The interpolation process follows the interpolation method used during the alignment of the running alignment data to ensure that the baseline power central trend transitions continuously with the output frequency sequence. After the baseline power central trend is obtained at each alignment time point, the calculation of the power residual sequence has a unified reference.

[0100] For frequency groups with limited sample sizes in the frequency power baseline table, a sample size constraint is introduced when locating entries: when the sample size of the current frequency group is lower than a preset value, the merged entries of this frequency group and its adjacent frequency groups are preferentially used as the comparison benchmark. The central tendency and dispersion of the merged entries are derived from the merging rules used when constructing the frequency power baseline table, and they maintain the same name and synonymy as the frequency power baseline table. When both the current frequency group and its adjacent frequency groups meet the sample size requirement, the comparison is still performed based on the current frequency group. The sample size constraint does not change the content of the frequency power baseline table, but only changes the order in which entries are selected during comparison. After the sample size constraint is introduced, random fluctuations in the power residual sequence at sparse sample frequencies are suppressed.

[0101] For each alignment time point, a power residual value is calculated. This residual value is obtained by subtracting the active power alignment value from the baseline power center trend at that time point, retaining the sign of the difference. When the active power alignment value is lower than the baseline power center trend, the power residual value is negative with a corresponding negative deviation; when the active power alignment value is higher than the baseline power center trend, the power residual value is positive with a corresponding positive deviation. Simultaneously, the dispersion of the output frequency alignment value from the baseline power at that alignment time point is recorded as background information for the power residual value, forming the power residual record together with the power residual value. After the power residual values ​​are continuously generated according to the set of alignment time points, the power residual sequence possesses temporal continuity and sign interpretability.

[0102] The power residual records at each alignment time point are concatenated in the order of the alignment time point set to form a power residual sequence. The power residual sequence retains the sampling timestamp sequence index relationship consistent with the running alignment data. The power residual sequence includes alignment time points, output frequency alignment values, active power alignment values, baseline power center trend, baseline power dispersion, and power residual values. The names of each item are consistent with the names already used in the running alignment data and frequency power baseline table.

[0103] During the generation of the power residual sequence, the processing method for the boundary of the comparison range is recorded synchronously: when the output frequency alignment value is lower than the lowest frequency group covered by the frequency power baseline table, the central trend and dispersion corresponding to the lowest frequency group are used as the baseline power central trend and baseline power dispersion; when the output frequency alignment value is higher than the highest frequency group covered by the frequency power baseline table, the central trend and dispersion corresponding to the highest frequency group are used as the baseline power central trend and baseline power dispersion; when the output frequency alignment value is within the coverage range, the baseline power central trend and baseline power dispersion are determined according to the aforementioned frequency group positioning and interpolation method. The boundary processing method is consistent with the frequency grouping of the frequency power baseline table, and no new frequency grouping names are introduced. After the boundary processing method is clarified, the power residual sequence can still be continuously output when the output frequency sequence drifts outside the boundary for a short time.

[0104] The output power residual sequence maintains a parallel reference relationship with the voltage fluctuation reference. The power residual sequence is derived from the comparison between the frequency power baseline table and the active power sequence, while the voltage fluctuation reference is derived from the statistical results of the voltage sequence during the stable operating period. In subsequent steps, when identifying continuous negative deviation intervals in the power residual sequence, the voltage fluctuation reference can be used to eliminate false underloads caused by voltage sequence fluctuations. Thus, the power residual sequence and the voltage fluctuation reference can be used together to determine the load reduction event segment. After the power residual sequence is output, the triggering of the load reduction event segment has a stable quantization input source.

[0105] The cause determination module is used to identify continuous negative deviation intervals in the power residual sequence by combining voltage fluctuation reference values, and obtain load reduction event segments; based on the running alignment data, it generates event fingerprint data in the load reduction event segments and determines the candidate cause set.

[0106] After obtaining the power residual sequence and voltage fluctuation reference value, the continuous negative deviation intervals in the power residual sequence are identified by combining the voltage fluctuation reference value, and the specific implementation method of obtaining the load reduction event segment is as follows:

[0107] The system reads the output frequency alignment value, active power alignment value, baseline power center trend, baseline power dispersion, and power residual value corresponding to each alignment time point in the power residual sequence. Simultaneously, it reads the voltage fluctuation reference quantity, including the fluctuation amplitude center trend, fluctuation amplitude dispersion, fluctuation rhythm center trend, and fluctuation rhythm dispersion related to the voltage sequence. An index relationship is established between the voltage fluctuation reference quantity and the power residual sequence based on the same set of alignment time points, allowing access to both the power residual value and the typical fluctuation boundaries of the voltage sequence within that time period at each alignment time point. After the power residual sequence and voltage fluctuation reference quantity form a unified index, subsequent identification processes can simultaneously consider the load change phenomenon and the impact of power supply side fluctuations on the same time axis.

[0108] A negative deviation judgment condition is constructed based on the power residual sequence. This condition utilizes both the power residual value and the baseline power dispersion to avoid inconsistencies in thresholds caused by differences in baseline fluctuation amplitude at different output frequency levels. Specifically, for each alignment time point, if the power residual value is negative and its absolute value exceeds a preset multiple of the baseline power dispersion at that time point, then that time point is marked as a negative deviation point. If the power residual value is negative but its absolute value does not exceed the preset multiple of the baseline power dispersion at that time point, then it is not marked as a negative deviation point. The baseline power dispersion comes from the frequency power baseline table and reflects the natural power fluctuation range at that output frequency level within the stable operating range. Therefore, introducing the baseline power dispersion into the negative deviation judgment condition ensures consistent judgment across different output frequency levels. After the negative deviation point is marked, the natural fluctuations in the power residual sequence consistent with the stable operating range are eliminated, and the load reduction phenomenon is more prominently highlighted.

[0109] Based on the negative deviation points, a persistence constraint is introduced to identify continuous negative deviation intervals. Specifically, the power residual sequence is scanned in the order of the aligned time point set. Intervals with consecutive negative deviation points at adjacent aligned time points are merged into candidate intervals, and the duration and number of negative deviation points of each candidate interval are calculated. Candidate intervals with a duration shorter than a preset duration and a number of negative deviation points shorter than a preset number are eliminated. Candidate intervals that simultaneously satisfy the preset duration and preset number are retained as continuous negative deviation intervals. The preset duration and preset number can be set according to the sampling interval of the aligned time point set. For example, when the sampling interval is 100 milliseconds, the preset duration can be 800 milliseconds, and the preset number can be 8, to exclude false intervals formed by short-term random fluctuations. After the introduction of the persistence constraint, the continuous negative deviation intervals better correspond to the continuous load reduction process in actual working conditions.

[0110] To address short-term power disturbances introduced during inverter speed regulation, speed regulation-related constraints are added to the identification of continuous negative deviation intervals to avoid misjudging speed regulation transients as load reduction events. Specifically, for each continuous negative deviation interval, the output frequency alignment values ​​corresponding to the start and end alignment time points of the interval are read, and the frequency change amplitude and frequency change rate are calculated based on the sampling timestamp sequence. The frequency change amplitude is determined by the difference between the output frequency alignment value at the end alignment time point and the output frequency alignment value at the start alignment time point, and the frequency change rate is jointly determined by the frequency change amplitude and the duration from the start sampling timestamp to the end sampling timestamp. A threshold for rapid frequency change amplitude and a threshold for rapid frequency change rate are set. Continuous negative deviation intervals where the frequency change amplitude exceeds the threshold for rapid frequency change amplitude or the frequency change rate exceeds the threshold for rapid frequency change rate are marked as having a rapid downward or upward adjustment of the output frequency alignment value. The direction of the rapid downward or upward adjustment is determined by the sign of the frequency change amplitude. The threshold values ​​for the rapid frequency change amplitude and the rapid frequency change rate are determined according to the output frequency change step and sampling interval of the inverter. The value range of the rapid frequency change amplitude threshold is 1 Hz to 5 Hz, and the value range of the rapid frequency change rate threshold is 0.5 Hz to 5 Hz. When the sampling interval is 100 milliseconds, the larger rapid frequency change rate threshold is preferred, and when the sampling interval is 500 milliseconds, the smaller rapid frequency change rate threshold is preferred. This ensures that changes in the output frequency alignment value during speed regulation can be identified, while minor changes caused by operating noise do not trigger the judgment.

[0111] Within a continuous negative deviation interval where the output frequency alignment value rapidly decreases or increases, the active power alignment value of the same continuous negative deviation interval is read, and the power change amplitude and direction are calculated. The power change amplitude is determined by the difference between the active power alignment value at the end alignment time point and the active power alignment value at the start alignment time point, and the power change direction is determined by the sign of the power change amplitude. When the change direction of the output frequency alignment value is consistent with the change direction of the active power alignment value, and the power change amplitude exceeds a preset multiple of the baseline power dispersion within the continuous negative deviation interval, the continuous negative deviation interval is marked as a speed regulation-related interval and is temporarily not considered as a load reduction event segment. The speed regulation-related interval is isolated in the power residual sequence, and short-term power disturbances caused by inverter speed regulation in the continuous negative deviation interval are suppressed.

[0112] When the frequency change amplitude of a continuous negative deviation interval does not exceed the threshold for rapid frequency change amplitude, and the frequency change rate does not exceed the threshold for rapid frequency change rate, the power residual value within this continuous negative deviation interval is read and the persistence constraint is checked. Continuous negative deviation intervals where the power residual value is continuously negative and the persistence constraint is satisfied are recorded as candidate intervals for load reduction events. The candidate intervals for load reduction events are retained in the power residual sequence, and subsequent load reduction event segment pruning is carried out using the candidate intervals for load reduction events as the input range.

[0113] Voltage fluctuation reference values ​​are incorporated into the screening of candidate intervals for load reduction events to eliminate false underload caused by voltage sequence fluctuations. Specifically, for each candidate interval of a load reduction event, the voltage alignment value corresponding to that interval is read from the operational alignment data, and the fluctuation amplitude and fluctuation rhythm of the voltage alignment value are statistically analyzed within that interval. The fluctuation amplitude is the difference between the maximum and minimum values ​​of the voltage alignment value within that interval, and the fluctuation rhythm is the number of times the voltage alignment value switches between rising and falling within that interval. The fluctuation amplitude of that interval is compared with the central trend and dispersion of the fluctuation amplitude of the voltage fluctuation reference value, and the fluctuation rhythm of that interval is compared with the central trend and dispersion of the fluctuation rhythm of the voltage fluctuation reference value. When the fluctuation amplitude or fluctuation rhythm exceeds the typical fluctuation boundary represented by the voltage fluctuation reference value, the candidate interval of the load reduction event is marked as a voltage fluctuation interference interval and eliminated. When both the fluctuation amplitude and fluctuation rhythm fall within the typical fluctuation boundary, the candidate interval of the load reduction event is retained. The voltage fluctuation reference is derived from voltage sequence statistics during stable operating periods, representing typical voltage fluctuations under non-abnormal load conditions. Therefore, using it as a criterion for elimination can suppress false underload caused by grid voltage fluctuations and long cable transmission disturbances. After the introduction of the voltage fluctuation reference, the determination of load reduction event segments is more focused on power residual changes caused by load-side factors.

[0114] Load reduction event segments are generated based on the retained candidate intervals for load reduction events. Pre-event and post-event reserved time windows are configured for each segment to preserve dynamic information before and after triggering. Specifically, for each retained interval, the start alignment time point of the interval is used as the event start time point, and the end alignment time point is used as the event end time point. A pre-event reserved time window is obtained by extending the alignment time point set forward by a preset number of alignment time points before the event start time point. A post-event reserved time window is obtained by extending the alignment time point set backward by a preset number of alignment time points after the event end time point. The pre-event reserved time window, the event duration window, and the post-event reserved time window are merged to form the load reduction event segment. The preset number is also set according to the sampling interval. For example, when the sampling interval is 100 milliseconds, the pre-event reserved time window can be 2 seconds, and the post-event reserved time window can be 2 seconds, ensuring that the load reduction event segment contains sufficient transition information to facilitate the subsequent generation of event fingerprint data and determination of the candidate cause set within the load reduction event segment.

[0115] The load reduction event segments undergo segment boundary consistency processing to prevent event segments from crossing the continuous segment boundaries of the running alignment data. When the reserved time window before or after the event exceeds the continuous segment boundaries of the running alignment data, the excess portion is truncated, and the corresponding continuous segment identifier is recorded, ensuring that the set of aligned time points remains continuous within the load reduction event segment. Segment boundary consistency processing does not change the names and meanings of the power residual value, output frequency alignment value, active power alignment value, and voltage alignment value within the load reduction event segment; it only adjusts the start and end ranges of the segment.

[0116] The load reduction event segment includes the power residual value, output frequency alignment value, active power alignment value, voltage alignment value, speed estimation alignment value, and pressure alignment value corresponding to each aligned time point within the event segment, and retains the index relationship between the sampling timestamp sequence and the set of aligned time points.

[0117] After obtaining the runtime alignment data and load reduction event segments, the specific implementation method for generating event fingerprint data and determining the candidate cause set based on the runtime alignment data in the load reduction event segments is as follows:

[0118] Based on the start and end alignment time points of the load reduction event segment, the corresponding alignment time point set interval is located in the operational alignment data. Within this interval, an electrical parameter sequence consisting of voltage, active power, output frequency, and speed estimation sequences, as well as a pressure sequence collected by the pressure detection element, are extracted. The correspondence between the sampling timestamp sequence and the alignment time point set is preserved. When the load reduction event segment includes a pre-event reserved time window, an event duration window, and a post-event reserved time window, the boundary alignment time points of these three time windows are marked to form event segment boundary information. This processing ensures that the event fingerprint data comes only from the operational alignment data interval corresponding to the load reduction event segment and retains complete temporal sequence information.

[0119] Based on event segment boundary information, load reduction event segments are segmented and organized. A reserved time window before the event is used as the baseline segment, the event duration window as the anomaly segment, and a reserved time window after the event as the recovery segment. Within the baseline segment, the center levels of the active power sequence and the estimated speed sequence are statistically analyzed, along with the center level of the output frequency sequence, forming baseline power level, baseline speed level, and baseline frequency level. Within the anomaly segment and the recovery segment, the center levels of the active power sequence and the estimated speed sequence are statistically analyzed and compared with the baseline power level and the baseline speed level. This processing can establish comparable baseline and anomaly segments within the same load reduction event segment, providing a unified reference for event fingerprint data.

[0120] In load reduction event segments, electrical parameter decline morphology features are generated based on the changes in the active power and current sequences within these segments. At the beginning of the abnormal segment, multiple consecutive aligned time points are selected to statistically analyze the decline amplitude of the active power sequence, calculated as the difference between the baseline power level and the lowest active power alignment value within that segment. The decline rate of the active power sequence is also statistically analyzed, derived from the decline amplitude and the duration from the first appearance of the baseline power level to the appearance of the lowest active power alignment value. In the recovery segment, the recovery rate of the active power sequence is statistically analyzed, calculated from the change amplitude from the lowest active power alignment value to the center level of the recovery segment and the corresponding duration. Simultaneously, the current sequence is statistically analyzed for current decline amplitude, current decline rate, and current recovery rate over the same time intervals, forming a morphological description consistent with the active power sequence. This process transforms the underload degree and underload change rhythm of load reduction event segments into reproducible event fingerprint data.

[0121] Electrical parameter fluctuation texture features are generated in load reduction event segments, based on short-period fluctuations in the current and active power sequences within the abnormal segment. Multiple continuous statistical windows of fixed length are divided within the abnormal segment. The difference between the maximum and minimum values ​​of the current alignment value is calculated for each window as the fluctuation amplitude, and the persistence of the fluctuation amplitude across multiple statistical windows is recorded to form fluctuation persistence. The number of times the current alignment value rises and falls within each statistical window is counted to form the fluctuation rhythm. Simultaneously, the power fluctuation amplitude and power fluctuation rhythm are calculated for the active power sequence using the same statistical windows, and the synchronization relationship between the current fluctuation amplitude and the power fluctuation amplitude is recorded. This processing can solidify the texture differences of gas entrainment, cavitation fluctuations, and valve transient fluctuations in the electrical parameter sequence into event fingerprint data content.

[0122] Voltage interference stripping features are generated in load reduction event segments, based on the comparison between the voltage sequence and voltage fluctuation reference values. Within the abnormal segment, the fluctuation amplitude and rhythm of the voltage alignment value are statistically analyzed. The fluctuation amplitude is the difference between the maximum and minimum values ​​of the voltage alignment value within the abnormal segment, and the fluctuation rhythm is the number of times the voltage alignment value rises and falls within the abnormal segment. The fluctuation amplitude of the abnormal segment is compared with the central trend and dispersion of the fluctuation amplitude in the voltage fluctuation reference values, and the fluctuation rhythm of the abnormal segment is compared with the central trend and dispersion of the fluctuation rhythm in the voltage fluctuation reference values. Based on this, it is recorded whether the voltage sequence fluctuation is within the typical fluctuation boundary, forming voltage interference marking information. This processing can separately mark the influence strength of grid voltage fluctuations and long cable transmission disturbances in event fingerprint data, reducing the risk of misreading electrical parameter sequences.

[0123] Pressure coupling response features are generated within load reduction event segments, based on the changes in the pressure sequence across the baseline, abnormal, and recovery segments. At the start of the abnormal segment, the direction of pressure alignment value change is statistically analyzed, determined by the increase or decrease of the pressure alignment value relative to the center level of the baseline segment. The rate of change of the pressure alignment value is also statistically analyzed, derived from the pressure change amplitude and corresponding duration. Within the abnormal segment, the fluctuation amplitude and rhythm of the pressure alignment value are statistically analyzed and compared with the fluctuation amplitude and rhythm in the electrical parameter fluctuation texture features. When the pressure sequence is missing in the alignment data or continuous pressure alignment values ​​cannot be formed within the corresponding interval of the load reduction event segment, the pressure coupling response features are recorded as indicating an unusable pressure sequence, while retaining other event fingerprint data based on the electrical parameter sequence. This process establishes a correspondence between the fluid response at the pipeline end and the changes in the electrical parameter sequence on the same time axis, providing a direct basis for the initial screening of candidate cause sets.

[0124] Speed ​​regulation correlation features are generated within load reduction event segments, based on the synchronization relationship between the output frequency sequence, speed estimation sequence, and active power sequence within these segments. The amplitude and direction of change of the output frequency alignment value are statistically analyzed during both the abnormal and recovery segments, as are the amplitude and direction of change of the speed estimation alignment value. The starting point of the output frequency sequence change is compared with the starting point of the active power sequence change, recording synchronization or lag relationships. The direction of change of the output frequency sequence is compared with the direction of change of the active power sequence, recording consistent or opposite relationships. Simultaneously, any abrupt changes in the morphological characteristics of electrical parameter decline during the output frequency sequence change are recorded. This processing can distinguish between transient disturbances introduced by the speed regulation process and underload manifestations introduced by abnormal operating conditions, reducing the confusion range of the candidate cause set.

[0125] The electrical parameter decrease morphology characteristics, electrical parameter fluctuation texture characteristics, voltage interference stripping characteristics, pressure coupling response characteristics, and speed regulation correlation characteristics are combined according to the same load reduction event segment to form event fingerprint data. The event fingerprint data retains the start alignment time point, end alignment time point, sampling timestamp sequence interval, and reference power level, reference frequency level, and reference speed level of the load reduction event segment. When there are multiple load reduction event segments in the same running alignment data, corresponding event fingerprint data are generated for each segment, and the event fingerprint data set is formed according to the time order of the load reduction event segments.

[0126] The initial range of the candidate cause set is determined, limited to five categories: airlock, cavitation, insufficient water supply, momentary valve closure, and transient pipeline fluctuations. These five categories are used as the initial content of the candidate cause set. Subsequently, preliminary screening is conducted based on the pressure coupling response characteristics in the event fingerprint data. When the pressure coupling response characteristics indicate a rapid increase in the pressure alignment value at the beginning of the anomaly segment and a negative deviation in the active power sequence, the candidate cause set retains momentary valve closure and transient pipeline fluctuations while eliminating airlock, cavitation, and insufficient water supply. When the pressure coupling response characteristics indicate a rapid decrease or persistently low pressure alignment value at the beginning of the anomaly segment and a negative deviation in the active power sequence, the candidate cause set retains airlock, cavitation, and insufficient water supply while eliminating momentary valve closure and transient pipeline fluctuations. When the pressure sequence is unavailable, the candidate cause set retains all five categories without elimination, and a "pressure sequence unavailable" marker is recorded in the candidate cause set. This process can narrow down multi-source causes to a range consistent with the pressure response without adding new data, reducing the burden of subsequent discrimination.

[0127] By combining electrical parameter fluctuation texture features and speed regulation correlation features, the candidate cause set is further narrowed down. When the electrical parameter fluctuation texture features show continuous irregular fluctuations and the voltage interference stripping feature indicates that the voltage sequence is within the typical fluctuation boundary, the candidate cause set prioritizes cavitation and airlock while eliminating insufficient water supply. When the electrical parameter decline pattern features a large decrease in the active power sequence but a slow recovery in the recovery phase, and the speed regulation correlation feature indicates that the output frequency sequence is basically stable, the candidate cause set prioritizes insufficient water supply while eliminating instantaneous valve closure and pipeline transient fluctuations. When the speed regulation correlation feature indicates that the output frequency sequence changes rapidly and the direction of change in the active power sequence is consistent with the output frequency sequence, the candidate cause set retains pipeline transient fluctuations while eliminating airlock and cavitation. This processing can transform the texture differences of the electrical parameter sequence and the differences in the speed regulation process into elimination conditions for the candidate cause set, providing a more focused basis for the generation of subsequent segmented perturbation detection commands.

[0128] The system outputs event fingerprint data and a set of candidate causes, ensuring that both the event fingerprint data and the candidate cause set maintain a traceable correspondence with the load reduction event segments and operational alignment data. The event fingerprint data includes electrical parameter decrease morphology features, electrical parameter fluctuation texture features, voltage interference stripping features, pressure coupling response features, and speed regulation correlation features. The candidate cause set includes retained and rejected entries from airlock, cavitation, insufficient water supply, instantaneous valve closure, and pipeline transient fluctuations. This processing provides a unified event description input for the calculation of subsequent action parameter sets and reduces the probability of erroneous triggering of action actions under non-airlock conditions.

[0129] The frequency adjustment module is used to generate segmented perturbation detection commands and collect perturbation detection response data based on event fingerprint data; to discriminate and calculate the candidate cause set based on the perturbation detection response data, to obtain the cause discrimination result and the anomaly severity index, to calculate the handling action parameter set based on the cause discrimination result, the anomaly severity index and the frequency power baseline table, and to adjust the output frequency according to the handling action parameter set.

[0130] After obtaining the event fingerprint data and the candidate cause set, the specific implementation method for generating segmented perturbation detection commands and collecting perturbation detection response data based on the event fingerprint data is as follows:

[0131] Based on the load reduction event segment, the corresponding alignment time point set interval is located in the operational alignment data. Within this interval, the electrical parameter sequence composed of the voltage sequence, active power sequence, output frequency sequence, and speed estimation sequence collected by the frequency converter, as well as the pressure sequence collected by the pressure detection element, are simultaneously read to form operational segment data that corresponds one-to-one with the load reduction event segment. The sampling timestamp sequence is retained in the operational segment data, and the correspondence between the sampling timestamp sequence and the alignment time point set remains unchanged. This ensures that when generating segmented perturbation detection commands, the adjustment time point of the output frequency sequence can fall within the traceable position of the alignment time point set. After the time boundary of the operational segment data is consistent with the load reduction event segment, the segmented perturbation detection commands can be carried out around the same load reduction event segment, avoiding command scale mismatch caused by cross-event mixing.

[0132] The event fingerprint data extracts electrical parameter decline morphology features, electrical parameter fluctuation texture features, voltage interference stripping features, pressure coupling response features, and speed regulation correlation features, and establishes a correspondence between these features and the retained items in the candidate cause set. When the candidate cause set retains airlock, cavitation, and insufficient water supply, priority is given to the description of low or declining pressure in the electrical parameter fluctuation texture features and pressure coupling response features. When the candidate cause set retains instantaneous valve closure and pipeline transient fluctuations, priority is given to the description of pressure rise in the pressure coupling response features and the synchronous relationship between output frequency sequence changes and active power sequence changes in the speed regulation correlation features. When the voltage interference stripping feature marks the voltage sequence fluctuation deviating from the typical fluctuation boundary of the voltage fluctuation reference, this mark is used as the basis for limiting the command amplitude and holding time, avoiding the introduction of excessively short segment boundaries during periods of strong voltage sequence fluctuations. After the event fingerprint data and the candidate cause set form a unified reference relationship, the construction of segmented perturbation detection commands has a clear input basis, reducing the probability of blindly applying the same detection action across different causes.

[0133] The starting alignment time point of the segmented perturbation detection command is determined in the runtime data, and the output frequency alignment value at that time, associated with this starting alignment time point, is used as the command reference frequency. The starting alignment time point is selected within the event duration window of the load reduction event segment and within the time range where the power residual value is continuously negative, to ensure that the detection action occurs while the load reduction phenomenon is still in effect. When the load reduction event segment includes a pre-event reserved time window, an event duration window, and a post-event reserved time window, the starting alignment time point does not fall within the pre-event reserved time window to avoid introducing the detection action before the load reduction phenomenon appears. After the starting alignment time point and the command reference frequency are determined, the change amplitude of each subsequent output frequency sequence can be referenced to the same reference frequency, facilitating the comparison of changes before and after detection in the perturbation detection response data.

[0134] The frequency variation scale of the segmented perturbation detection command is determined based on the electrical parameter descent morphology and fluctuation texture characteristics. The frequency variation boundary is constrained by the lowest and highest frequency groups covered by the frequency-power baseline table. When the electrical parameter descent morphology indicates a large and rapid decrease in the active power sequence, a smaller frequency variation scale is used to prioritize obtaining the coupling difference between the pressure and active power sequences through shorter variation amplitudes, reducing disturbances to the pump's operating state. When the electrical parameter fluctuation texture indicates continuous irregular fluctuations in the current and active power sequences, a larger frequency variation scale is used to enhance the discriminability of the fluctuation texture, facilitating subsequent differentiation between airlock and cavitation. When the voltage disturbance stripping characteristics indicate that the voltage sequence fluctuation deviates from the typical fluctuation boundary, a smaller frequency variation scale is used and the holding time is extended to prevent voltage sequence fluctuations from masking the active power sequence changes caused by the detection. After the frequency variation scale is determined under the constraint of event fingerprint data, the segmented perturbation detection command can adaptively adjust the perturbation intensity between different load reduction event segments, balancing discriminability and additional perturbation control.

[0135] Based on the pressure coupling response characteristics, the segment sequence and intra-segment hold arrangement of the segmented perturbation detection command are determined. The segmented perturbation detection command includes an up-frequency segment, a down-frequency segment, and a constant-frequency hold segment. The up-frequency and down-frequency segments are used to generate output frequency sequence changes in two directions within the same load reduction event segment, and form a comparable intra-segment frequency level with the constant-frequency hold segment. This creates segmented boundary information in the perturbation detection response data and generates a set of response segments. The existence of the up-frequency and down-frequency segments supports the segmented response comparison of the candidate cause set among airlock, cavitation, insufficient water supply, instantaneous valve closure, and pipeline transient fluctuations, avoiding insufficient evidence for identifiable causes due to reliance solely on the constant-frequency hold segment. The frequency change intensity of the up-frequency and down-frequency segments is constrained by the frequency change scale. The cumulative upward adjustment of the output frequency target sequence relative to the command reference frequency in the upward adjustment frequency band does not exceed the frequency change scale, and the cumulative downward adjustment of the output frequency target sequence relative to the command reference frequency in the downward adjustment frequency band does not exceed the frequency change scale. The preset step size is consistent with the frequency change scale and falls between adjacent alignment time points. When the output frequency target sequence touches the output frequency range boundary corresponding to the lowest or highest frequency group covered by the frequency power baseline table, the output frequency target sequence takes a value at the boundary and maintains the sequence continuity.

[0136] When the pressure coupling response characteristics indicate that the pressure sequence begins to rise at the start of an abnormal segment and the active power sequence shows a negative deviation, the segment sequence should prioritize a downward frequency adjustment segment followed by a constant frequency holding segment, and then an upward frequency adjustment segment. This is to reduce the rate of pressure change in the event of potential instantaneous valve closure or pipeline transient fluctuations, and then observe the synchronization relationship between the pressure sequence and the active power sequence. When the pressure coupling response characteristics indicate that the pressure sequence is low or declining, the segment sequence should prioritize an upward frequency adjustment segment followed by a constant frequency holding segment, and then a downward frequency adjustment segment. This is to observe whether the pressure sequence recovers following the upward adjustment in the event of potential airlock, cavitation, or insufficient water supply, and then observe whether the electrical parameter fluctuation texture weakens after the downward adjustment. The duration of the constant frequency holding segment is set according to the characteristics of the electrical parameter fluctuation texture; the stronger the fluctuation texture, the longer the constant frequency holding segment, used to cover multiple fluctuation cycles of the electrical parameter fluctuation texture. After the segment sequence and constant frequency holding segment are arranged, the segmented perturbation detection command has a detection link with a fixed structure but variable order, which can quickly form a distinguishable response around the directional differences in the pressure sequence.

[0137] The segmented perturbation detection command is applied to the aligned time point set to generate an output frequency target sequence that corresponds one-to-one with the sampling timestamp sequence. Specifically, starting from the command reference frequency, the output frequency target sequence is gradually increased by a preset step size within the upward frequency range, gradually decreased by a preset step size within the downward frequency range, and kept unchanged within the constant frequency range. The preset step size is consistent with the aforementioned frequency change scale, and each step change occurs between adjacent aligned time points, ensuring that the output frequency target sequence is aligned with the running aligned data within the same aligned time point set. When the calculated output frequency target sequence exceeds the output frequency range corresponding to the lowest or highest frequency group covered by the frequency power baseline table, the excess portion is truncated to the boundary value, while preserving the continuity of the truncated output frequency target sequence. After the output frequency target sequence is bound to the aligned time point set, the segmented perturbation detection command can be executed by the frequency converter in the form of a continuous time sequence, while facilitating subsequent segmentation of the perturbation detection response data using the same aligned time point set.

[0138] Transitional constraints are added to the segmented perturbation detection command based on speed regulation correlation characteristics to reduce transient perturbations caused by changes in the output frequency sequence. When the speed regulation correlation characteristics indicate that the output frequency sequence has already undergone continuous changes within the load reduction event segment, the step size of the up and down frequency segments of the segmented perturbation detection command is reduced to a smaller value, and the constant frequency holding segment is extended to avoid superimposing excessively rapid changes on top of the original output frequency sequence changes. When the speed regulation correlation characteristics indicate that the output frequency sequence is basically stable, the step size is allowed to take the standard value corresponding to the aforementioned frequency change scale to improve the distinguishability of the detection action. After the transitional constraints are executed under the constraints of event fingerprint data, the perturbation of the output frequency sequence by the segmented perturbation detection command is closer to the existing change rhythm of the running segment data, which is beneficial for subsequently separating the detection-induced response from the perturbation detection response data.

[0139] The inverter is driven to adjust its output frequency sequence according to the segmented disturbance detection command, and disturbance detection response data is collected synchronously during the execution of the segmented disturbance detection command. The source of the disturbance detection response data is consistent with the operational alignment data, including the electrical parameter sequence composed of the voltage sequence, active power sequence, output frequency sequence, and speed estimation sequence collected by the inverter, the pressure sequence collected by the pressure detection element, and the sampling timestamp sequence. During the acquisition process, a fixed sampling interval is maintained, and a corresponding sampling timestamp is generated at each sampling point, so that the disturbance detection response data can completely cover the frequency adjustment segment, the frequency reduction segment, and the constant frequency holding segment. After the disturbance detection response data is collected synchronously with the segmented disturbance detection command, subsequent discrimination can compare the changes in the output frequency sequence with the response changes in the electrical parameter sequence and the pressure sequence on the same time axis.

[0140] The perturbation detection response data is aligned according to the sampling timestamp sequence to form an alignment result with a structure consistent with the operational alignment data. The alignment result retains the segment boundary alignment time points of the segmented perturbation detection command. Specifically, the electrical parameter sequence and pressure sequence of the perturbation detection response data are interpolated and aligned to the alignment time point set according to the sampling timestamp sequence, obtaining the voltage alignment value, active power alignment value, output frequency alignment value, speed estimation alignment value, and pressure alignment value corresponding to each alignment time point. The start and end alignment time points of the up-frequency segment, the down-frequency segment, and the constant frequency holding segment are recorded as segment boundary information. After the perturbation detection response data is aligned, subsequent discrimination of the candidate cause set can directly segment the response of different frequency segments according to the segment boundary information, avoiding response misalignment caused by asynchronous sampling.

[0141] Based on segment boundary information, the perturbation detection response data is divided into a set of response segments, each corresponding one-to-one with the segment structure of the segmented perturbation detection command. Each response segment contains voltage alignment values, active power alignment values, output frequency alignment values, speed estimation alignment values, and pressure alignment values ​​within that segment, while retaining the index relationship between the sampling timestamp sequence and the set of alignment time points. When a short-term gap occurs in the pressure sequence within the perturbation detection response data, it is uniformly pruned and gap-filled according to the processing rules for running alignment data, ensuring the continuity within each response segment. After the set of response segments is formed, subsequent steps can calculate the response differences between the pressure sequence and the electrical parameter sequence for the up-frequency segment, down-frequency segment, and constant-frequency holding segment, respectively, providing structured input for the convergence of the candidate cause set.

[0142] After obtaining the perturbation detection response data, response fragment set, event fingerprint data, and candidate cause set, the candidate cause set is discriminated and calculated based on the perturbation detection response data to obtain the cause discrimination result and anomaly severity index. The specific implementation method for calculating the handling action parameter set based on the cause discrimination result, anomaly severity index, and frequency power baseline table, and adjusting the output frequency according to the handling action parameter set, is as follows:

[0143] Based on the segment boundary information of the response segment set, voltage alignment values, active power alignment values, output frequency alignment values, speed estimation alignment values, and pressure alignment values ​​corresponding to the up-frequency segment, down-frequency segment, and constant-frequency holding segment are extracted from the perturbation detection response data, while maintaining the index relationship between the sampling timestamp sequence and the alignment time point set. Within each segment, the active power alignment values ​​are statistically analyzed to obtain the power center level and power fluctuation amplitude of that segment. Simultaneously, the pressure alignment values ​​are statistically analyzed to obtain the pressure center level and pressure fluctuation amplitude of that segment. Within the same segment, the intra-segment center level of the output frequency alignment value is recorded as the intra-segment frequency level. The above segmented statistical analysis does not change the naming of the perturbation detection response data; it only forms directly comparable intra-segment statistical results within the response segment set. After the intra-segment statistical results are formed, the power and pressure changes among the up-frequency segment, down-frequency segment, and constant-frequency holding segment have a unified quantitative description.

[0144] The intra-segment statistical results are correlated with the event fingerprint data. The correlation method uses the start and end alignment time points of the load reduction event segment as boundaries. The electrical parameter decrease morphology, electrical parameter fluctuation texture, voltage interference stripping, pressure coupling response, and speed regulation correlation features in the event fingerprint data are arranged chronologically alongside the intra-segment statistical results in the response segment set. When the voltage interference stripping feature indicates that the voltage sequence fluctuation deviates from the typical fluctuation boundary of the voltage fluctuation reference value, the voltage alignment value fluctuation amplitude within that time range is simultaneously marked in the intra-segment statistical results, avoiding directly using the power center level change during periods of strong voltage sequence fluctuations as the discrimination criterion. The above-mentioned correlated parallel information serves as the input source for causal discriminable evidence, while maintaining the limitation that the candidate causal set consists of five categories: airlock, cavitation, insufficient water supply, instantaneous valve closure, and pipeline transient fluctuations. With a unified input source for causal discriminable evidence, the subsequent discrimination process can simultaneously utilize the segmented response differences between the event fingerprint data and the perturbation detection response data.

[0145] A comparative relationship of identifiable evidence of causes is constructed among the five types of causes in the candidate cause set, and the evidence content is extracted item by item according to the segment order of the response segment set. When the power center level of the up-frequency band rebounds relative to the power center level of the constant frequency holding band, and the pressure center level rebounds synchronously, and the rebound occurs within the time range where the voltage interference stripping feature is not marked as deviating, it is recorded as rebound evidence. When the pressure center level of the up-frequency band rises rapidly while the power center level does not rebound, and the pressure fluctuation amplitude expands, it is recorded as voltage increase evidence. When the power fluctuation amplitude of the down-frequency band decreases significantly relative to the constant frequency holding band, while the pressure center level remains low or continues to decline, it is recorded as wave reduction evidence. When the pressure center level in the constant frequency holding band shows a slow rebound trend and the power center level rebounds synchronously and slowly, and the center level in the output frequency alignment value segment is basically stable, it is recorded as slow recovery evidence. When neither the up-frequency band nor the down-frequency band forms an effective rebound in the pressure center level, and the power fluctuation amplitude and pressure fluctuation amplitude change intermittently at multiple adjacent alignment time points, it is recorded as loss evidence. Causally distinguishable evidence is mapped to each load reduction event segment using the same nomenclature system, allowing for consistent processing of the convergence of the candidate cause set based on the evidence content. Once causally distinguishable evidence is identified, the differences between different causes in the perturbation detection response data are expressed as reproducible evidence.

[0146] Based on evidence of distinguishable causes, the candidate cause set is converged. The convergence process maintains consistent terminology across the candidate cause sets and retains records of the elimination process. When evidence of recovery is established, the candidate cause set prioritizes retaining pipeline transient fluctuations and eliminates airlock, cavitation, insufficient water supply, and momentary valve closure. This is because the synchronous recovery of the power center level and pressure center level after the output frequency sequence is upregulated better matches the short-term recovery characteristics of pipeline transient fluctuations. When evidence of pressure increase is established, the candidate cause set prioritizes retaining valve momentary closure and eliminates airlock, cavitation, insufficient water supply, and pipeline transient fluctuations. This is because the rise in pressure center level without a recovery in power center level better matches the restricted flow state caused by momentary valve closure. When evidence of wave reduction is established, the candidate cause set prioritizes retaining cavitation and eliminates valve momentary closure. For cavitation-related fluctuations, transient fluctuations in the pipeline are considered, while airlock and insufficient water supply are retained for further differentiation. This is because the reduced power fluctuation amplitude at lower frequencies, coupled with a still low pressure center level, better reflects the weakening of cavitation-related fluctuations as frequency decreases. When evidence of slow recovery is established, the candidate causative set prioritizes insufficient water supply and eliminates instantaneous valve closure and transient fluctuations in the pipeline, while retaining airlock and cavitation for further differentiation. This is because the slow recovery of the pressure center level when the output frequency sequence stabilizes better reflects the gradual recovery of insufficient water supply. When evidence of loss of recovery is established and the electrical parameter fluctuation texture indicates intermittent changes in power fluctuation amplitude, the candidate causative set prioritizes airlock and eliminates instantaneous valve closure and transient fluctuations in the pipeline, while retaining cavitation and insufficient water supply for further differentiation. After the candidate causative set converges, the candidate range is compressed to a factor set consistent with the perturbation detection response data.

[0147] The causal discrimination result is determined from the converged candidate causal set. The causal discrimination result includes at least the target causal name and the converged candidate causal set. When the converged candidate causal set retains only a single causal cause, that single causal cause is directly used as the target causal name. When the converged candidate causal set retains both airlock and cavitation, the electrical parameter fluctuation texture characteristics and wave reduction evidence are further used. If the power fluctuation amplitude decreases significantly in the down-frequency band and power fluctuation amplitude still exists in the constant frequency band, then cavitation is used as the target causal name. If the power fluctuation amplitude decreases only slightly in the down-frequency band and the loss of evidence remains valid, then airlock is used as the target causal name. The target cause name is determined as follows: When the converged candidate cause set retains both airlock and insufficient water supply, further evidence of slow recovery is used. If the pressure center level during the constant frequency maintenance segment shows a slow upward trend and the power center level rises synchronously and slowly, then insufficient water supply is taken as the target cause name. If the pressure center level during the constant frequency maintenance segment remains low and the evidence of loss of recovery continues to hold, then airlock is taken as the target cause name. When the converged candidate cause set still contains more than two types of causes and the voltage interference stripping feature marker deviates, the target cause name is retained as the list of retained items in the candidate cause set, and this list is used as the target cause name for the cause discrimination result. After the cause discrimination result is determined, the calculation of the subsequent action parameter set can be carried out around the target cause name.

[0148] An anomaly severity index is calculated based on perturbation detection response data and a frequency-power baseline table. This index characterizes the degree of deviation in energy consumption and water delivery response during a load reduction event segment. First, in the operational alignment data corresponding to the load reduction event segment, the corresponding frequency group is located in the frequency-power baseline table based on the output frequency alignment value, and the baseline power center trend is read to obtain the baseline power center trend sequence for that load reduction event segment. Then, the power center level and pressure center level are statistically analyzed in the constant-frequency holding segment and the up-frequency segment of the perturbation detection response data, respectively. The difference between the power center level and the baseline power center trend of the corresponding frequency group is compared to obtain the power deviation degree. Simultaneously, the difference between the pressure center level and the baseline pressure center level of the load reduction event segment is compared to obtain the pressure deviation degree. Power fluctuation amplitude and pressure fluctuation amplitude are also read to form the fluctuation deviation degree. The power deviation degree, pressure deviation degree, and fluctuation deviation degree are combined in a preset order to obtain the anomaly severity index. The combination method is to increase the anomaly severity index when the power deviation degree is large, increase the anomaly severity index when the pressure deviation degree is large, and increase the anomaly severity index when the fluctuation deviation degree is large and the voltage interference stripping feature is not marked as deviated. Once the severity index is established, the severity of load reduction event segments can be uniformly quantified and directly used for the selection of graded action parameter sets.

[0149] Based on the cause identification results, anomaly severity index, and frequency power baseline table, a set of action parameters is calculated. This set includes at least the frequency downsampling amplitude, frequency downsampling duration, frequency recovery slope, sweep range, sweep step amplitude, and sweep dwell time. When the target cause is airlock, the action parameter set uses a method where the sweep range and sweep dwell time increase with the anomaly severity index. A smaller sweep range and shorter dwell time are used when the anomaly severity index is low, and a larger sweep range and longer dwell time are used when the anomaly severity index is high. Simultaneously, the sweep range is constrained by the frequency power baseline table to not exceed the coverage of the lowest and highest frequency groups. When the target cause is cavitation, the action parameter set uses a smaller frequency downsampling amplitude and a gentler frequency recovery slope, and the frequency downsampling duration is matched to the duration of the load reduction event segment. Furthermore, the frequency power baseline table is selected... The frequency range with a lower baseline power center trend corresponding to the frequency group is used as the reference for the downgraded operating frequency. When the target cause is insufficient water supply, the action parameter set uses a larger frequency downgrade duration, and the frequency downgrade amplitude is selected to reduce the baseline power center trend of the corresponding frequency group in the frequency power baseline table to below the baseline power center trend of the benchmark segment. The frequency recovery slope is taken as a relatively gentle value to reduce frequent changes in the output frequency sequence. When the target cause is instantaneous valve closure, the action parameter set uses a smaller frequency downgrade amplitude and a relatively gentle frequency recovery slope, and the frequency downgrade duration is taken to cover the duration of the larger pressure fluctuation amplitude to reduce short-term surges in the pressure center level. When the target cause is transient pipeline fluctuation, the action parameter set uses the smallest frequency downgrade amplitude and a longer constant frequency maintenance arrangement, and the frequency sweep range is taken as zero or the minimum value to avoid introducing additional changes in the output frequency sequence. After the action parameter set is calculated, the action corresponding to different causes can form an interpretable parameter correlation with the anomaly severity index.

[0150] The output frequency is adjusted according to the set of action parameters. The adjustment process is organized based on the output frequency sequence and the set of alignment time points. When the action parameter set includes the frequency reduction magnitude and the frequency reduction hold duration, a new output frequency target sequence is generated starting from the current output frequency alignment value and changing at a rhythm determined by the frequency recovery slope. The output frequency target sequence is kept stable within the set of alignment time points covered by the frequency reduction hold duration. When the action parameter set includes the frequency sweep range, the frequency sweep step magnitude, and the frequency sweep dwell time, a segmented output frequency target sequence is generated within the frequency sweep range according to the frequency sweep step magnitude. Within each segment, the output frequency target sequence is kept unchanged according to the frequency sweep dwell time. Each value of the output frequency target sequence is limited to the lowest and highest frequency groups covered by the frequency power baseline table and is aligned with the sampling timestamp sequence. After the output frequency target sequence is generated, the inverter performs output frequency sequence adjustment according to the output frequency target sequence and continues to collect voltage, active power, output frequency, speed estimation, and pressure sequences at fixed sampling intervals and align them with the sampling timestamp sequence to form operating alignment data. After the action parameter set is executed, the output frequency sequence adjustment has a clear parameter source and time organization method and maintains consistency with the running alignment data.

[0151] Example 2:

[0152] Please see Figure 2 As shown, this embodiment provides an adaptive energy-saving control method for permanent magnet submersible pumps, including:

[0153] The electrical parameter sequence, consisting of voltage, current, active power, output frequency, and speed estimation sequences collected by the frequency converter, the pressure sequence collected by the pressure detection element, and the sampling timestamp sequence are aligned according to the sampling timestamp sequence to obtain the running aligned data.

[0154] In the running alignment data, a stable operating segment is selected, and within the stable operating segment, the active power sequence is statistically analyzed according to the output frequency sequence to obtain a frequency power baseline table. Within the stable operating segment, the voltage fluctuation reference value is extracted from the voltage sequence.

[0155] The frequency power baseline table is compared with the active power sequence to obtain the power residual sequence;

[0156] By combining voltage fluctuation reference values, continuous negative deviation intervals are identified in the power residual sequence to obtain load reduction event segments; based on the operation alignment data, event fingerprint data is generated in the load reduction event segments and a candidate cause set is determined;

[0157] Segmented perturbation detection commands are generated based on event fingerprint data, and perturbation detection response data is collected. Based on the perturbation detection response data, candidate cause sets are discriminated and calculated to obtain cause discrimination results and anomaly severity index. Based on the cause discrimination results, anomaly severity index and frequency power baseline table, the set of handling action parameters is calculated, and the output frequency is adjusted according to the set of handling action parameters.

[0158] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0159] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive energy-saving control method for permanent magnet submersible pumps, characterized in that, include: The electrical parameter sequence, consisting of voltage, current, active power, output frequency, and speed estimation sequences collected by the frequency converter, the pressure sequence collected by the pressure detection element, and the sampling timestamp sequence are aligned according to the sampling timestamp sequence to obtain the running aligned data. In the running alignment data, a stable operating segment is selected, and within the stable operating segment, the active power sequence is statistically analyzed according to the output frequency sequence to obtain a frequency power baseline table. Within the stable operating segment, the voltage fluctuation reference value is extracted from the voltage sequence. The frequency power baseline table is compared with the active power sequence to obtain the power residual sequence; By combining voltage fluctuation reference values, continuous negative deviation intervals are identified in the power residual sequence to obtain load reduction event segments; Based on runtime alignment data, generate event fingerprint data in load reduction event segments and determine a set of candidate causes; Based on the event fingerprint data, segmented perturbation detection instructions are generated and perturbation detection response data is collected. Specifically, this includes: locating the set of alignment time points in the running alignment data based on the load reduction event segment, reading the electrical parameter sequence and pressure sequence to obtain the running segment data, and retaining the correspondence between the sampling timestamp sequence and the set of alignment time points. Extract electrical parameter morphology features, electrical parameter fluctuation texture features, voltage interference stripping features, pressure coupling response features, and speed regulation correlation features from event fingerprint data, and establish correspondence between them and the retained items in the candidate cause set; Determine the starting alignment time point in the running fragment data, and determine the instruction reference frequency using the corresponding output frequency alignment value; The frequency change scale is determined based on the characteristics of the electrical parameter descent morphology and the electrical parameter fluctuation texture, and the frequency change boundary is constrained based on the frequency power baseline table; Based on the pressure coupling response characteristics, the segment sequence and duration of the constant frequency holding segment are determined, and the frequency up-adjustment segment, frequency down-adjustment segment, and constant frequency holding segment are generated; The segmented perturbation detection command is applied to the aligned time point set to generate the output frequency target sequence, and the output frequency target sequence is truncated at the boundary based on the frequency power baseline table; a transition constraint is added to the output frequency target sequence based on the speed regulation correlation feature; Drive the frequency converter to adjust the output frequency sequence according to the target output frequency sequence and collect the perturbation detection response data; Alignment results are obtained by aligning the perturbation detection response data according to the sampling timestamp sequence and recording the segment boundary information. Based on the segment boundary information, the perturbation detection response data is segmented to obtain a set of response segments. Based on the perturbation detection response data, the candidate cause set is discriminated and calculated to obtain the cause discrimination result and the anomaly severity index. Based on the cause discrimination result, the anomaly severity index and the frequency power baseline table, the set of handling action parameters is calculated, and the output frequency is adjusted according to the set of handling action parameters.

2. The adaptive energy-saving control method for permanent magnet submersible pumps according to claim 1, characterized in that, Methods for selecting stable operating segments from aligned data and constructing frequency power baseline tables and voltage fluctuation references within these stable operating segments include: The continuous analysis range is determined by the sampling timestamp sequence, and the voltage alignment value, active power alignment value, and output frequency alignment value are read using the alignment time point set as an index; the alignment time point set is generated by the main sampling timestamp sequence. Within the continuous analysis range, the difference between the maximum and minimum values ​​of the output frequency alignment value is calculated according to a preset time window to determine the preset time window for the output frequency sequence to be stationary. Within a preset time window where the output frequency sequence is stable, the difference between the maximum and minimum values ​​of the active power alignment value and the peak value of the change in active power alignment value at adjacent alignment time points are statistically analyzed to determine the candidate preset time window for the stable operation segment. Merge candidate preset time windows for stable operation segments according to the aligned time point set, and record the start sampling timestamp and end sampling timestamp to obtain the stable operation segment; During the stable operation period, the output frequency alignment values ​​are merged into frequency groups according to the preset frequency interval, and the active power alignment values ​​corresponding to each frequency group are collected to form a power sample set. The central trend and dispersion of the power sample set are statistically analyzed to obtain the frequency power baseline table. The fluctuation amplitude is obtained by calculating the difference between the maximum and minimum values ​​of the voltage alignment value during the stable operation period, and the fluctuation rhythm is obtained by calculating the number of times the voltage alignment value switches between rising and falling. The central trend and dispersion of the fluctuation amplitude and fluctuation rhythm are calculated to obtain the voltage fluctuation reference value.

3. The adaptive energy-saving control method for permanent magnet submersible pumps according to claim 1, characterized in that, Methods for obtaining power residual sequences include: The alignment data is traversed according to the alignment time point set. The output frequency alignment value and active power alignment value are read, and the output frequency alignment value is merged into the frequency group of the frequency power baseline table to obtain the frequency group label. Locate the corresponding entry in the frequency power baseline table based on the frequency grouping mark, and read the center trend and dispersion to obtain the baseline power center trend and baseline power dispersion. When the output frequency alignment value is located at the boundary of the adjacent frequency group, interpolate the output frequency alignment value to obtain the baseline power center trend and baseline power dispersion. For frequency groups with a sample size lower than a preset value, the baseline power center trend and baseline power dispersion are determined by merging the frequency group and adjacent frequency groups in the table. For each alignment time point, the difference between the active power alignment value and the baseline power center trend is calculated to obtain the power residual value. The output frequency alignment value and the degree of dispersion of the baseline power are recorded to form a power residual record. The power residual records are spliced ​​according to the alignment time point set to obtain the power residual sequence.

4. The adaptive energy-saving control method for permanent magnet submersible pumps according to claim 3, characterized in that, Read the output frequency sequence and active power sequence from the running alignment data, and read the central trend, dispersion and sample size of each frequency group in the frequency power baseline table to form a frequency power index relationship; once the frequency power index relationship is clear, the running alignment data can be quickly associated with the corresponding table entry in the frequency power baseline table at any alignment time point.

5. The adaptive energy-saving control method for permanent magnet submersible pumps according to claim 1, characterized in that, Methods for obtaining load reduction event fragments include: The negative deviation judgment condition is constructed based on the power residual value and the degree of dispersion of the baseline power, and the negative deviation points are marked. The negative deviation points are merged according to the set of aligned time points, and the continuous negative deviation interval is determined based on the preset duration and preset quantity. Read the magnitude and direction of the output frequency alignment value change corresponding to the continuous negative deviation interval. When the output frequency alignment value is rapidly decreasing or rapidly increasing and the active power alignment value changes synchronously, mark the speed regulation related interval. When the output frequency alignment value is stable or changes slowly and the power residual value is continuously negative, determine the candidate interval of the load reduction event. Read the voltage alignment value corresponding to the candidate interval of load reduction event, and count the fluctuation amplitude and fluctuation rhythm. When the fluctuation amplitude or fluctuation rhythm exceeds the typical fluctuation boundary of the voltage fluctuation reference, the voltage fluctuation interference interval is eliminated. When the fluctuation amplitude and fluctuation rhythm fall within the typical fluctuation boundary, the retention interval is determined. The start and end times of the event are determined based on the reserved interval, and the reserved time windows before and after the event are expanded according to the set of aligned time points to obtain the load reduction event segments.

6. The adaptive energy-saving control method for permanent magnet submersible pumps according to claim 1, characterized in that, Methods for generating event fingerprint data and determining the candidate cause set include: The start and end alignment time points of the load reduction event segment are located in the alignment time point set interval in the running alignment data, and the electrical parameter sequence and pressure sequence are extracted. The boundary alignment time points of the reserved time window before the event, the event duration window and the reserved time window after the event are marked to obtain the event segment boundary information. Based on the event segment boundary information, the reserved time window before the event is used as the reference segment, the event duration window is used as the abnormal segment, and the reserved time window after the event is used as the recovery segment. The active power sequence, speed estimation sequence and output frequency sequence within the reference segment are statistically analyzed to obtain the reference power level, reference speed level and reference frequency level. In the load reduction event segment, electrical parameter decline morphology features are generated based on active power sequence and current sequence, and electrical parameter fluctuation texture features are generated based on current sequence and active power sequence. In the load reduction event segment, voltage disturbance stripping features are generated based on voltage sequence and voltage fluctuation reference, and pressure coupling response features are generated based on pressure sequence; Speed ​​regulation correlation features are generated based on the output frequency sequence, speed estimation sequence, and active power sequence; Event fingerprint data is obtained by combining the electrical parameter descent morphology features, electrical parameter fluctuation texture features, voltage interference stripping features, pressure coupling response features, and speed regulation correlation features. The candidate cause set was determined to be airlock, cavitation, insufficient water supply, instantaneous valve closure, and transient pipeline fluctuation. Based on the pressure coupling response characteristics, electrical parameter fluctuation texture characteristics, and speed regulation correlation characteristics, the candidate cause set was retained and eliminated to obtain the candidate cause set.

7. The adaptive energy-saving control method for permanent magnet submersible pumps according to claim 1, characterized in that, The method of determining and calculating the candidate cause set based on perturbation detection response data to obtain the cause determination result and anomaly severity index, and calculating the action parameter set based on the cause determination result, anomaly severity index and frequency power baseline table, and adjusting the output frequency according to the action parameter set includes: Based on the segment boundary information of the response segment set, the voltage alignment value, active power alignment value, output frequency alignment value, speed estimation alignment value, and pressure alignment value of the up-frequency segment, down-frequency segment, and constant frequency holding segment are extracted from the perturbation detection response data. The statistical results within each segment are obtained by statistically analyzing the power center level, power fluctuation amplitude, pressure center level, pressure fluctuation amplitude, and frequency level within the segment. The statistical results within the segment are correlated with the event fingerprint data according to the start and end alignment time points of the load reduction event segment, and the voltage alignment value fluctuation amplitude is marked based on the voltage interference stripping feature. Based on the statistical results within the segment, evidence of recovery, pressure increase, wave reduction, slow return, and loss of return are constructed. Based on the evidence of recovery, pressure increase, wave reduction, slow return, and loss of return, the candidate cause set is retained and eliminated to obtain a converged candidate cause set. Based on the converged candidate cause set, the target cause name is determined to obtain the cause discrimination result. Based on the perturbation detection response data and frequency power baseline table, the baseline power center trend is located, and the degree of power deviation, pressure deviation and fluctuation deviation are calculated and combined to obtain the anomaly severity index. Based on the cause identification results, the severity index of the abnormality and the frequency power baseline table, the set of action parameters for handling actions is calculated. The set of action parameters for handling actions includes the frequency downsampling amplitude, the frequency downsampling duration, the frequency recovery slope, the frequency sweep range, the frequency sweep step amplitude and the frequency sweep dwell time. The output frequency target sequence is generated according to the action parameter set, and the output frequency is adjusted. The output frequency target sequence is limited to the range of the lowest and highest frequency groups covered by the frequency power baseline table.

8. The adaptive energy-saving control method for permanent magnet submersible pumps according to claim 1, characterized in that, Methods for obtaining runtime alignment data include: The electrical parameter sequence and pressure sequence are time-stamped by adding time stamps to the sampling timestamp sequence, resulting in time-stamped electrical parameter sequence and time-stamped pressure sequence; Construct an alignment time point set, and use the alignment time point set as a reference to interpolate and align the time-stamped electrical parameter sequence and the time-stamped pressure sequence to obtain voltage alignment value, active power alignment value, output frequency alignment value, speed estimation alignment value, and pressure alignment value. The interpolated and aligned electrical parameter sequence and pressure sequence are subjected to consistent trimming and gap filling to obtain the running aligned data. The voltage alignment value, active power alignment value, output frequency alignment value, speed estimation alignment value, and pressure alignment value are combined according to the alignment time point set to obtain the operation alignment data.

9. A permanent magnet submersible pump operating condition adaptive energy-saving control system, used to implement the permanent magnet submersible pump operating condition adaptive energy-saving control method according to any one of claims 1-8, characterized in that, include: The data acquisition module is used to align the voltage sequence, current sequence, active power sequence, output frequency sequence, and speed estimation sequence collected by the frequency converter, the pressure sequence collected by the pressure detection element, and the sampling timestamp sequence to obtain running aligned data. The reference quantity filtering module is used to filter the stable operating segment in the running alignment data, and to statistically analyze the active power sequence according to the output frequency sequence within the stable operating segment to obtain the frequency power baseline table. The voltage fluctuation reference quantity is extracted from the voltage sequence within the stable operating segment. The power comparison module is used to compare the frequency power baseline table with the active power sequence to obtain the power residual sequence; The cause determination module is used to identify continuous negative deviation intervals in the power residual sequence by combining voltage fluctuation reference values, and obtain load reduction event segments. Based on runtime alignment data, generate event fingerprint data in load reduction event segments and determine a set of candidate causes; The frequency adjustment module is used to generate segmented perturbation detection commands based on event fingerprint data and to collect perturbation detection response data. Based on the perturbation detection response data, the candidate cause set is discriminated and calculated to obtain the cause discrimination result and the anomaly severity index. Based on the cause discrimination result, the anomaly severity index and the frequency power baseline table, the set of handling action parameters is calculated, and the output frequency is adjusted according to the set of handling action parameters.

Images