Pump station forebay water level inflection point judgment method based on slope sliding window cooperation

Abstract

The invention discloses a pump station forebay water level inflection point judgment method based on slope sliding window cooperation, which comprises the following steps of: obtaining a forebay water level data set of a time sequence, calculating an average time interval of data acquisition, and determining a minimum data point number meeting an analysis requirement; if the data points are insufficient, prompting data insufficiency; setting the size of a sliding window and forward and backward time window parameters, and ensuring the stability and representativeness of candidate inflection points; setting the shortest allowable time interval between the candidate inflection points, and screening the inflection points with independence; according to the water level change characteristics, determining a slope threshold and a floating range thereof for identifying an inflection point of which the slope symbol is not changed; setting the longest allowable time interval between the inflection point and the latest data, and ensuring the timeliness of the inflection point; the system provides powerful technical support for intelligent development of the water conservancy industry.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy projects, and in particular to a method for determining a water level inflection point in a pump station forebay based on slope sliding window collaboration. Background Art

[0002] Traditional methods for managing water levels in pump station forebays at water conservancy projects face numerous challenges. Water level data is affected by environmental factors and equipment, resulting in frequent noise and outliers, which severely interfere with the assessment of true water level trends and make the extraction of reliable information difficult. Furthermore, key inflection points in water level changes are difficult to accurately locate, and traditional methods present several drawbacks. For example, many inflection point detection algorithms require a predefined number of inflection points, which limits their adaptability and flexibility, especially when the number of inflection points is unknown. Furthermore, algorithms often require a minimum segment length, which can lead to detection delays in streaming data processing. Furthermore, some algorithms employ brute-force search methods, which can be inefficient. Traditional autoregressive models such as AR and MA struggle with non-stationary data, requiring additional differencing processing. While machine learning methods, particularly deep learning models such as LSTM, have demonstrated excellent performance in processing complex time series data, they also carry the risk of becoming trapped in local optima and exhibit a certain degree of lag in their prediction results. Summary of the Invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a method for determining the inflection point of the water level in the forebay of a pumping station based on the coordination of a slope sliding window, in order to solve the limitations of streaming data processing delays and insufficient efficiency caused by the preset number of inflection points and minimum segment length setting of existing algorithms, the deficiency of traditional autoregressive models such as AR and MA requiring additional differential processing when processing non-stationary data, and the risk of deep learning models such as LSTM falling into local optimality and prediction lag when processing complex time series data, and facing data noise and outlier problems due to the influence of the environment and equipment.

[0004] Technical solution: The present invention provides a method for determining the inflection point of the water level in the forebay of a pumping station based on the coordination of a slope sliding window, comprising the following steps:

[0005] (1) Obtain a time series forebay water level dataset, calculate the average time interval for data collection, and determine the minimum number of data points required to meet the analysis requirements; if there are insufficient data points, indicate insufficient data;

[0006] (2) Setting the sliding window size, as well as the forward and backward time window parameters, to ensure the stability and representativeness of the candidate inflection points;

[0007] (3) Setting the shortest allowed time interval between candidate inflection points to screen independent inflection points;

[0008] (4) According to the water level change characteristics, the slope threshold and its floating range are determined to identify the inflection point where the slope sign does not change;

[0009] (5) Set the maximum allowable time interval between the inflection point and the latest data to ensure the timeliness of the inflection point.

[0010] Furthermore, in step (2), data smoothing and slope calculation include: applying the exponentially weighted moving average method to smooth the original water level data, retaining an appropriate number of decimal places; using the sliding window method to traverse the smoothed data, using the least squares method for linear fitting, calculating the slope of the data in each window, and proportionally amplifying the slope value based on the amplitude of the water level change; wherein the sliding window size, time interval parameter, slope threshold parameter and time validity parameter are all optimized and set according to the dynamic change characteristics of the actual water level data.

[0011] Furthermore, the weight coefficient of the exponentially weighted moving average method is expressed by the formula

[0012]

[0013] In the formula, span is the time span, that is, the amount of data to be processed.

[0014] Furthermore, in step (3), the candidate inflection point screening includes: slope sign change type: traversing the slope sequence, checking whether the slope signs in the window before and after the candidate point are completely opposite, and satisfying the time interval condition, and recording it as a candidate inflection point; slope sign no change type: checking the consistency of the slope signs in the window before and after the candidate point, and the current slope is within the threshold floating range, and if it meets the time interval condition, it is recorded as a candidate inflection point;

[0015] Furthermore, in the candidate inflection point screening step, the screening of slope sign change type and slope sign no change type are both implemented by an automated algorithm.

[0016] 6. A method for determining the inflection point of the water level in the forebay of a pumping station based on slope sliding window coordination according to claim 1, characterized in that, in step (4), the inflection point determination includes: a priority rule: candidate inflection points of the type with no slope sign change take precedence over those of the type with sign change; if only a single type of candidate inflection points exists, the last point is selected; if both types exist, the later point is selected;

[0017] Furthermore, the priority rule comprehensively evaluates the natural turning point of the water level and the start-up and shutdown factors of the unit, and determines that the type with no change in slope sign has a higher priority.

[0018] Furthermore, in step (5), time validity verification is performed: checking whether the time interval between the final inflection point and the latest data exceeds the maximum allowed time interval. If so, the data is marked as invalid.

[0019] Furthermore, the maximum allowed time interval is set based on the response time of the prediction model and the dynamic characteristics of the water level, ensuring that the inflection point is closely related to the current water level status.

[0020] The present invention provides a pump station forebay water level inflection point determination system based on slope sliding window collaboration, comprising:

[0021] Data integrity check and frequency determination module: used to obtain the time series forebay water level data set, calculate the average time interval of data collection, and determine the minimum number of data points to meet the analysis requirements; if there are insufficient data points, it will prompt insufficient data;

[0022] Key parameter initialization module: used to set the sliding window size, as well as the forward and backward time window parameters, to ensure the stability and representativeness of the candidate inflection points;

[0023] Time interval parameter module: used to set the shortest allowed time interval between candidate inflection points and screen independent inflection points;

[0024] Slope threshold parameter module: used to determine the slope threshold and its floating range according to the water level change characteristics, and to identify the inflection point where the slope sign does not change;

[0025] Time validity parameter module: used to set the maximum allowable time interval between the inflection point and the latest data to ensure the timeliness of the inflection point.

[0026] Beneficial Effects: Compared with existing technologies, this invention offers the following significant advantages: It utilizes slope feature analysis and sliding window technology, combined with intelligent screening and preprocessing algorithms, to effectively handle noise and outliers, accurately capture key inflection points, and optimize the algorithm to achieve both high precision and real-time performance. This provides strong support for the scientific management and optimized operation of water conservancy projects, and holds great practical value and promising prospects for determining water levels in pump station forebays. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flow chart of the present invention;

[0028] Figure 2 This is a schematic diagram of screening candidate inflection points of the present invention;

[0029] Figure 3 This is a schematic diagram of determining the final inflection point of the present invention. DETAILED DESCRIPTION

[0030] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0031] like Figure 1 As shown, the embodiment of the present invention provides a method for determining the inflection point of the water level in the forebay of a pumping station based on the coordination of a slope sliding window, as shown in FIG. Figure 1As shown, the method includes the following steps S101-S103:

[0032] Step S101: Basic Parameter Calculation. This step involves preprocessing the collected water level data to ensure that the data quality meets the analysis requirements. First, the average time interval for water level data collection is calculated. Then, based on this time interval, the minimum number of data points required for data analysis is determined. Next, the average time interval value is used to set a series of key parameters, including:

[0033] The sliding window size used for slope calculation should reflect the local variation trend of water level data;

[0034] The shortest allowed time interval between candidate inflection points to ensure the independence and practical significance of the inflection points;

[0035] The slope threshold and its floating range are used to identify candidate inflection points where the slope sign changes or does not change;

[0036] The maximum allowed time interval between the inflection point and the latest data to ensure the timeliness of the inflection point and its relevance to the current water level changes.

[0037] Step S102: Data Smoothing and Slope Calculation. The original water level data column is smoothed using the exponentially weighted moving average (EWA) method. Appropriate smoothing parameters are set to obtain a smoothed data column. The data is then rounded to an appropriate number of decimal places to reduce data fluctuations. The smoothed data is also rounded to an appropriate number of decimal places to further optimize data quality and provide a stable and reliable foundation for subsequent slope calculations.

[0038] The calculation formula of the EWA algorithm is:

[0039]

[0040] Where: EWA t is the EWA value at the tth time point; x t 、x t-1 is the value of the original data point; the denominator is the sum of a geometric progression, which ensures that the EWA value is a weighted average; (1-β) i It represents the weight of the data at the corresponding moment. Newer data points have larger weights, while older data points have smaller weights. β is an adjustable weighting parameter, which represents the rate of weighted decrease. The smaller the value, the faster the decrease. 0<β<1.

[0041]

[0042] In the formula, span is the time span, that is, the amount of data to be processed.

[0043] Then, when calculating the slope, if we calculate the slope every two points, the water level change trend will be extremely sensitive and changeable, and it will be impossible to accurately grasp the water level change trend. Therefore, we use the least squares method to perform linear fitting, calculate the slope of each window, and appropriately amplify the slope value. For example:

[0044] The current time is time t, and the existing data {(x1,y1),(x2,y2),…,(x t ,y t )},x t Represents the sample index in time order and interval; y t represents the treated water level at time t. It is expected that a linear equation will be fitted to these data points.

[0045] y=ax+b

[0046] To solve this problem, the least squares method is used to complete the linear fitting work.

[0047] Define the objective function, that is, find the parameters by minimizing the sum of squared errors:

[0048]

[0049] This function is usually solved by taking its partial derivatives and setting them equal to zero.

[0050] Find the partial derivatives:

[0051]

[0052] Solve the system of equations:

[0053]

[0054] Where ∑xy represents all x i y i The sum of ∑x represents all x i The sum of ∑y represents all y i The sum of ∑x 2 Indicates all The harmony.

[0055] Through the above calculation, the slope of the water level data at the corresponding moment can be obtained, reflecting the changing trend of the water level data at the corresponding moment.

[0056] Step S103: Candidate inflection point screening. Perform a detailed analysis of the calculated slope sequence to identify potential inflection point locations. The specific operations are as follows:

[0057] Candidate inflection point screening based on slope sign changes: The slope sequence is traversed, and for each potential candidate point, the slope signs within the specified window before and after it are checked to see if they are completely opposite. Specifically, for each candidate point, all slope values ​​within the window before it are analyzed to see if they have the same sign as the slope at that point, and all slope values ​​within the window after it are analyzed to see if they have the opposite sign. If a candidate point meets this condition, and the time interval between it and the recorded inflection points before and after it is greater than or equal to the preset minimum allowable time interval, it is recorded as a preliminary candidate inflection point.

[0058] Let S i is the slope of the i-th point, W1 is the size of the front window, W2 is the size of the back window, t i is the time of the i-th point, T update The time interval threshold for updating the inflection point.

[0059] (1) Check whether the slope signs in the front and back windows are completely opposite:

[0060]

[0061] Among them, ChangeSign i Is a Boolean value indicating whether the signs of the first W1 slopes and the last W2 slopes of the i-th point are completely opposite. means "for all", i.e. It means that the slope signs of the W1 points before the i-th candidate point are the same as those of the W2 points after it, which are opposite to the slope signs of the i-th candidate point.

[0062] (2) Candidate inflection point records under cold start conditions:

[0063] Cand i =ChangeSign i

[0064] Among them, Cand i It is a Boolean value indicating whether the i-th point meets the conditions for becoming a candidate inflection point in the cold start case.

[0065] (3) Update of candidate inflection points in non-cold start situations:

[0066] UpdateCand i =Cand i ·f(|t i -t prev |≥T update )

[0067] Among them, UpdateCand iIs a Boolean value, indicating that in the case of non-cold start, if the time interval between the new candidate point and the last recorded candidate point is greater than T update , then update the candidate inflection points.

[0068] (4) Update of the final candidate inflection point list:

[0069] If UpdateCand i If true, the time t of the i-th point i Add to the final candidate inflection point list and update the recorded candidate inflection point time list.

[0070] Candidate inflection point screening for unchanged slope signs: For cases where the slope sign remains unchanged, the system checks whether the slope of the current point falls within a preset threshold and whether the slope signs within a certain window before and after the current point are consistent with the slope sign of the current point. If the current point meets these conditions and the time interval between it and the recorded inflection point meets the preset requirements, the point is also recorded as a preliminary candidate inflection point.

[0071] Let S i is the slope of the i-th point, W is the window size, θ is the preset slope threshold, ∈ is the floating range of the slope threshold, t i is the time of the i-th point, T update The time interval threshold for updating the inflection point.

[0072] The candidate inflection point screening formula when the slope sign does not change is:

[0073] (1) Check the consistency of slope signs in the front and back windows:

[0074]

[0075] Among them, SameSign i Is a Boolean value indicating whether the slope of the i-th point is consistent with the slopes of the W points before and after it. means "for all", i.e. It means that the slopes of the W1 points before the i-th candidate point have the same sign as the slopes of the W2 points after it.

[0076] (2) Check whether the slope is within a certain threshold range:

[0077] InRange i =(||S i |-θ|<∈)

[0078] Among them, InRange i It is a Boolean value indicating whether the slope of the i-th point is within the range of the preset slope threshold θ.

[0079] (3) Candidate inflection point records under cold start conditions:

[0080] Candi′ i =SameSign i *InRange i

[0081] Among them, Candi′ i It is a Boolean value indicating whether the i-th point meets the conditions for becoming a candidate inflection point in the cold start case.

[0082] (4) Update of candidate inflection points in non-cold start situations:

[0083] UpdateCand′ i =Cand′ i ·f(|t i -t prev |≥T update )

[0084] Among them, UpdateCand′ i Is a Boolean value, indicating that in the case of non-cold start, if the time interval between the new candidate point and the last recorded candidate point is greater than T update , then update the candidate inflection points.

[0085] (5) Update of the final candidate inflection point list:

[0086] If UpdateCand′ i If true, the time t of the i-th point i Add to the final candidate inflection point list and update the recorded candidate inflection point time list.

[0087] Step S104: Determine the final inflection point.

[0088] 1. Priority rule application: Based on the pre-defined inflection point priority rule, selection is made from two categories of candidate inflection points: those with a slope sign change and those with no slope sign change. This priority rule takes into account that inflection points with no slope sign change generally better reflect the natural turning points of water level changes, while inflection points with a slope sign change may be affected by a variety of complex factors. Therefore, inflection points with no slope sign change have higher priority.

[0089] 2. Single-type candidate inflection point selection: If there is only one type of candidate inflection point, the last point of that type is directly selected as the final inflection point, provided that the candidate inflection point list of that type is not empty.

[0090] 3. Selection of two types of candidate inflection points: If both types of candidate inflection points exist, their time order is compared and the later inflection point is selected as the final inflection point. This selection is based on the fact that the later inflection point better reflects the current trend and status of water level changes.

[0091] 4. Time validity verification: The final inflection point is checked for time validity, that is, the time interval between the inflection point and the latest data is verified to ensure that it exceeds the maximum allowed time interval. If the time interval exceeds the threshold, the inflection point is considered outdated and cannot effectively reflect the current water level changes, so it is marked as invalid.

[0092] Set Cand change is the set of candidate inflection points where the slope sign changes, Cand same is the set of candidate inflection points whose slope signs do not change, T max is the preset maximum allowed time interval, t latest The time of the latest data.

[0093] (1) Determine the valid candidate inflection points:

[0094] ValidCand change ={t i ∈Cand change ||t i -t latest |≤T max}

[0095] ValidCand same ={t i ∈Cand same ||t i -t latest |≤T max}

[0096] ValidCand change is the set of valid candidate inflection points where the slope sign changes, ValidCand same is a set of valid candidate inflection points where the slope sign does not change.

[0097] (2) Select the final inflection point:

[0098]

[0099] In practical applications, the various parameters of the method (such as smoothing algorithm parameters, slope threshold, number of data screening copies, etc.) can be appropriately adjusted and optimized according to the specific conditions and needs of different water conservancy projects to achieve the best prediction results. At the same time, with the continuous accumulation of data and technological advancements, more advanced algorithms and models can be further explored, and this method can be continuously improved and perfected to adapt to the increasingly complex management needs of water conservancy projects, providing strong technical support for the intelligent development of the water conservancy industry.

Claims

1. A method for determining the inflection point of water level in the forebay of a pumping station based on slope sliding window coordination, characterized in that: The following steps are involved: (1) Obtain a time series forebay water level dataset, calculate the average time interval for data collection, and determine the minimum number of data points required to meet the analysis requirements; if there are insufficient data points, indicate insufficient data; (2) Setting the sliding window size, as well as the forward and backward time window parameters, to ensure the stability and representativeness of the candidate inflection points; (3) Setting the shortest allowed time interval between candidate inflection points to screen independent inflection points; (4) According to the water level change characteristics, the slope threshold and its floating range are determined to identify the inflection point where the slope sign does not change; (5) Set the maximum allowable time interval between the inflection point and the latest data to ensure the timeliness of the inflection point.

2. A method for determining the inflection point of water level in the forebay of a pumping station based on slope sliding window collaboration according to claim 1, characterized in that: In step (2), data smoothing and slope calculation include: applying the exponentially weighted moving average method to smooth the original water level data, retaining an appropriate number of decimal places; using the sliding window method to traverse the smoothed data, using the least squares method to perform linear fitting, calculating the slope of the data in each window, and proportionally amplifying the slope value based on the amplitude of the water level change; wherein the sliding window size, time interval parameter, slope threshold parameter and time validity parameter are all optimized and set according to the dynamic change characteristics of the actual water level data.

3. The method for determining the inflection point of the water level in the forebay of a pumping station based on slope sliding window coordination according to claim 2 is characterized in that: The weight coefficient of the exponentially weighted moving average method is expressed by the formula In the formula, span is the time span, that is, the amount of data to be processed.

4. The method for determining the inflection point of the water level in the forebay of a pumping station based on slope sliding window coordination according to claim 1 is characterized in that: In step (3), the screening of candidate inflection points includes: slope sign change type: traverse the slope sequence, check whether the slope signs in the windows before and after the candidate point are completely opposite, and meet the time interval conditions, and record it as a candidate inflection point; slope sign no change type: check the consistency of the slope signs in the windows before and after the candidate point, and the current slope is within the threshold floating range, and if it meets the time interval conditions, it is recorded as a candidate inflection point.

5. The method for determining the inflection point of the water level in the forebay of a pumping station based on slope sliding window coordination according to claim 4 is characterized in that: In the candidate inflection point screening step, the screening of slope sign change type and sign no change type are both achieved through automated algorithms.

6. The method for determining the inflection point of the water level in the forebay of a pumping station based on slope sliding window coordination according to claim 1 is characterized in that: In step (4), the inflection point determination includes: priority rule: candidate inflection points of the slope sign unchanged type take precedence over those of the sign changed type; if there is only a single type of candidate inflection point, the last point is selected; if both types exist, the later point is selected.

7. The method for determining the inflection point of the water level in the forebay of a pumping station based on slope sliding window coordination according to claim 6 is characterized in that: The priority rule comprehensively evaluates the natural turning point of the water level and the start-up and shutdown factors of the unit, and determines that the type with no change in slope sign has a higher priority.

8. The method for determining the inflection point of the water level in the forebay of a pumping station based on slope sliding window coordination according to claim 1 is characterized in that: In step (5), time validity verification: check whether the time interval between the final inflection point and the latest data exceeds the maximum allowed time interval. If so, mark it as invalid.

9. The method for determining the inflection point of the water level in the forebay of a pumping station based on slope sliding window coordination according to claim 8 is characterized in that: The maximum allowable time interval is set based on the response time of the prediction model and the dynamic characteristics of the water level to ensure that the inflection point is closely related to the current water level status.

10. A pump station forebay water level inflection point judgment system based on slope sliding window coordination, characterized in that: include: Data integrity check and frequency determination module: used to obtain the time series forebay water level data set, calculate the average time interval of data collection, and determine the minimum number of data points to meet the analysis requirements; if there are insufficient data points, it will prompt insufficient data; Key parameter initialization module: used to set the sliding window size, as well as the forward and backward time window parameters, to ensure the stability and representativeness of the candidate inflection points; Time interval parameter module: used to set the shortest allowed time interval between candidate inflection points and screen independent inflection points; Slope threshold parameter module: used to determine the slope threshold and its floating range according to the water level change characteristics, and to identify the inflection point where the slope sign does not change; Time validity parameter module: used to set the maximum allowable time interval between the inflection point and the latest data to ensure the timeliness of the inflection point.

Images