A coastal sluice tide level prediction method and system

Abstract

The present invention discloses a method and system for predicting tidal levels of coastal sluices and pumping stations. The present invention breaks away from the traditional hydraulic modeling idea, constructs an overall prediction model for sea tides and inland river water levels by using historical inland river water level sequences and historical sea tide water level sequences, predicts future inland river water level sequences and future sea tide water level sequences by using the model, is fast in solution and accurate in calculation, and can meet the requirements of real-time optimal scheduling of coastal sluices and pumping stations.

Description

Technical Field

[0001] The present invention relates to a method and system for predicting tidal levels of coastal floodgates and pumping stations, belonging to the technical field of estuary and coastal data processing. Background Art

[0002] The real-time scheduling of coastal floodgates and pumping stations is the final link in urban flood control and drainage, the most important part of preventing urban flooding and river network waterlogging, and determines when floods can be completely drained from the plain river network system. The opening and closing of coastal floodgates are mainly affected by the high and low tide points. When the sea tide rises to the same level as the water level of the drainage river in the floodgate, the floodgate needs to be closed in time to prevent seawater backflow; when the sea tide drops to the same level as the water level of the drainage river in the floodgate, the floodgate needs to be opened in time for drainage. Therefore, accurate prediction of tidal levels and inland river water levels can provide a reasonable plan for the operation of floodgates and pumping stations, guide the staff to open the floodgates for drainage and close the floodgates to block the tide in time, and improve the timeliness of scheduling.

[0003] The existing prediction schemes currently use hydrodynamics for runoff prediction and combine short-term forecast data for tidal level prediction. However, this scheme requires a large amount of measured data and needs to perform hydrodynamic modeling on the river network, which not only has a high cost but also a slow calculation speed. At the same time, considering the influence of seasons on tides, the deviation of the prediction results is relatively large. Most of these schemes are preliminary schemes before rainfall and cannot meet the real-time optimization scheduling requirements of coastal floodgates and pumping stations. Summary of the Invention

[0004] The present invention provides a method and system for predicting tidal levels of coastal floodgates and pumping stations, which solves the problems disclosed in the background art.

[0005] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows:

[0006] A method for predicting tidal levels of coastal floodgates and pumping stations includes:

[0007] Obtaining the current inland river water level sequence according to the current inland river water level and the historical inland river water levels;

[0008] Obtaining the current sea tide water level sequence according to the current sea tide water level and the historical sea tide water levels;

[0009] Inputting the current inland river water level sequence and the current sea tide water level sequence into a pre-constructed overall prediction model of sea tide and inland river water levels to obtain a future inland river water level sequence and a future sea tide water level sequence; wherein, the overall prediction model of sea tide and inland river water levels is constructed according to the historical inland river water level sequence and the historical sea tide water level sequence.

[0010] Constructing an overall prediction model of sea tide and inland river water levels according to the historical inland river water level sequence and the historical sea tide water level sequence, including:

[0011] Obtain the historical inland river water level sequence and the historical tidal water level sequence;

[0012] Convert the historical inland river water level sequence into a historical inland river water level high-frequency sequence and a historical inland river water level low-frequency sequence;

[0013] Construct an inland river water level high-frequency sequence prediction model based on the historical inland river water level high-frequency sequence;

[0014] Construct an inland river water level low-frequency sequence prediction model based on the historical inland river water level low-frequency sequence;

[0015] Obtain the historical inland river water level residual sequence based on the historical inland river water level high-frequency sequence predicted by the inland river water level high-frequency sequence prediction model and the corresponding true value, and the historical inland river water level low-frequency sequence predicted by the inland river water level low-frequency sequence prediction model and the corresponding true value, and construct an inland river water level residual sequence prediction model based on the historical inland river water level residual sequence;

[0016] Convert the historical tidal water level sequence into a historical tidal water level high-frequency sequence and a historical tidal water level low-frequency sequence;

[0017] Construct a tidal water level high-frequency sequence prediction model based on the historical tidal water level high-frequency sequence;

[0018] Construct a tidal water level low-frequency sequence prediction model based on the historical tidal water level low-frequency sequence;

[0019] Obtain the historical tidal water level residual sequence based on the historical tidal water level high-frequency sequence predicted by the tidal water level high-frequency sequence prediction model and the corresponding true value, and the historical tidal water level low-frequency sequence predicted by the tidal water level low-frequency sequence prediction model and the corresponding true value, and construct a tidal water level residual sequence prediction model based on the historical tidal water level residual sequence;

[0020] Perform signal combination on the inland river water level high-frequency sequence prediction model, the inland river water level low-frequency sequence prediction model, the inland river water level residual sequence prediction model, the tidal water level high-frequency sequence prediction model, the tidal water level low-frequency sequence prediction model, and the tidal water level residual sequence prediction model to construct an overall prediction model for tides and inland river water levels.

[0021] The low-frequency sequence prediction model is an autoregressive integrated moving average model;

[0022] When modeling, perform hypothesis testing on the differences between the predicted values of the low-frequency sequence model at different times and the corresponding true values, and the predicted values of the low-frequency sequence model at the same time under different orders of differencing and the corresponding true values, and optimize the order of differencing for different hydrological parameters;

[0023] When modeling, according to the historical low-frequency sequences, calculate the autocorrelation parameters and partial autocorrelation parameters, determine the value ranges of the autoregressive order and the moving average order for different low-frequency sequences, and use the first evaluation function to optimize the autoregressive order and the moving average order; among them, the first evaluation function, based on the AIC criterion and the BIC criterion, introduces the influence of the differencing order on the sum of the autoregressive order and the moving average order.

[0024] Suppose the test formula is:

[0025]

[0026]

[0027] Where α is the constant term, β is the trend term, d is the differencing order, t represents the time, δ is the test term, ΔX t is the predicted difference at time t, ΔX t-1 is the predicted difference at time t - 1, ε t is the high-order residual term after differenced autoregressive moving average processing at time t, m′ is the number of differencing processes that have been carried out, ΔX d is the predicted difference of the d-th differencing, β i is the trend term coefficient of the predicted difference, the predicted difference is the difference between the predicted value and the corresponding true value, ε d is the high-order residual term after the d-th differencing, X d-1 is the predicted value of the (d - 1)-th differencing, X t-1 is the predicted value of the differencing at time t - 1; the differencing order when the test term δ is not 0 is the preferred differencing order.

[0028] The formula of the first evaluation function is:

[0029]

[0030]

[0031] Where S(m, d) is the first evaluation function, m is the sum of the autoregressive order p and the moving average order q, d is the differencing order, σ 2 (m) is the intermediate process term of the first evaluation function, N is the number of historical low-frequency sequence samples participating in the calculation, T″ represents the number of times of the samples participating in the modeling, ΔX t is the predicted difference at time t; the autoregressive order and the moving average order when the evaluation function value is the smallest are the preferred autoregressive order and the preferred moving average order.

[0032] The residual sequence prediction model is an SVM model; when modeling, a residual sequence is constructed based on the residual modeled by the low-frequency sequence model and the residual modeled by the high-frequency sequence model. A second evaluation function is constructed using the residual sequence and the penalty coefficient. The value range of the penalty coefficient is determined using the second evaluation function. Based on the value range of the penalty coefficient and the value range of the kernel parameter determined using the LIBSVM tool, the grid search method is used to determine the optimal penalty coefficient and kernel parameter.

[0033] The second evaluation function is:

[0034]

[0035] where, f(ΔW i ,C) is the second evaluation function, ΔW i is the residual sequence, C is the penalty coefficient, k is the residual sequence fitting coefficient, and N′ is the number of samples of the residual sequence.

[0036] A coastal sluice tide level prediction system includes:

[0037] An inland river water level sequence acquisition module, which obtains the inland river water level sequence at the current moment according to the current inland river water level and the historical inland river water levels;

[0038] A sea tide water level sequence acquisition module, which obtains the sea tide water level sequence at the current moment according to the current sea tide water level and the historical sea tide water levels;

[0039] A prediction module that inputs the inland river water level sequence and the sea tide water level sequence at the current moment into a pre-constructed overall prediction model of sea tide and inland river water levels to obtain a future inland river water level sequence and a future sea tide water level sequence; among them, the overall prediction model of sea tide and inland river water levels is constructed based on the historical inland river water level sequence and the historical sea tide water level sequence.

[0040] A computer-readable storage medium storing one or more programs, where the one or more programs include instructions that, when executed by a computing device, cause the computing device to execute the coastal sluice tide level prediction method.

[0041] A computing device includes one or more processors, one or more memories, and one or more programs, where the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the coastal sluice tide level prediction method.

[0042] Advantages achieved by the present invention: The present invention breaks away from the traditional hydraulic modeling idea, constructs an overall prediction model for tidal and inland river water levels by using historical inland river water level sequences and historical tidal water level sequences, predicts future inland river water level sequences and future tidal water level sequences by using the model, has fast solution and accurate calculation, and can meet the requirements of real-time optimal scheduling of coastal sluices and pumping stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 FIG. is a flow chart of a tidal level prediction method for coastal sluices and pumping stations;

[0044] Figure 2 FIG. is a flow chart of constructing an overall prediction model for tidal and inland river water levels;

[0045] Figure 3 FIG. is a flow chart of a method for determining a gate opening and closing plan. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] The present invention will be further described below with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention and should not be used to limit the protection scope of the present invention.

[0047] As Figure 1 shown, a tidal level prediction method for coastal sluices and pumping stations includes the following steps:

[0048] Step 1: Obtain the inland river water level sequence at the current moment according to the inland river water level and the historical inland river water level at the current moment;

[0049] Step 2: Obtain the tidal water level sequence at the current moment according to the tidal water level and the historical tidal water level at the current moment;

[0050] Step 3: Input the inland river water level sequence and the tidal water level sequence at the current moment into the pre-constructed overall prediction model for tidal and inland river water levels to obtain the future inland river water level sequence and the future tidal water level sequence; wherein, the overall prediction model for tidal and inland river water levels is constructed according to the historical inland river water level sequence and the historical tidal water level sequence.

[0051] The above method breaks away from the traditional hydraulic modeling idea, constructs an overall prediction model for tidal and inland river water levels by using historical inland river water level sequences and historical tidal water level sequences, predicts future inland river water level sequences and future tidal water level sequences by using the model, has fast solution and accurate calculation, and can meet the requirements of real-time optimal scheduling of coastal sluices and pumping stations.

[0052] Before implementing the above method, it is necessary to pre-construct an overall prediction model for tidal and inland river water levels. The specific process can be as Figure 2 shown and includes:

[0053] 11) Obtain the historical inland river water level sequence and the historical tidal water level sequence.

[0054] 12) Convert the historical inland river water level series into a historical inland river water level high-frequency series and a historical inland river water level low-frequency series; convert the historical tidal water level series into a historical tidal water level high-frequency series and a historical tidal water level low-frequency series.

[0055] Use wavelet transform in signal processing to perform noise reduction processing on the dataset, and process the original data into a smooth and equally spaced time series with a time interval of 5 minutes.

[0056] Wavelet transform is a time-scale (time-frequency) analysis method for signals. It has the characteristics of multi-resolution analysis, and has the ability to represent the local characteristics of signals in both the time and frequency domains. It is a time-frequency localization analysis method with a fixed window size but a changeable shape, and both the time window and the frequency window can be changed. That is, it has a lower time resolution and a higher frequency resolution in the low-frequency part, and a higher time resolution and a lower frequency resolution in the high-frequency part, and can effectively extract the characteristics of signals and reduce data noise.

[0057] The specific principle is as follows:

[0058] After shifting a function called the mother wavelet, and then at different scales, taking the inner product of the mother wavelet function and the signal to be analyzed (historical inland river water level series and historical tidal water level series), the formula is as follows:

[0059]

[0060] Among them, a>0 is the scale factor, whose function is to stretch the mother wavelet function, τ is the displacement, and its value can be positive or negative. Both a and τ are continuous variables. At different scales, the duration of the wavelet increases with the increase of the value, and the amplitude decreases in inverse proportion to but the shape of the wave remains unchanged. WT x (a,τ) is the fitted sequence, and x(t) represents the original tidal water level or inland river water level sequence. is the mother wavelet function.

[0061] Select the wavelet basis function according to the change characteristics of the inland river water level and the tidal water level. Use the db4 wavelet with better regularity to decompose the signal. Fix the scale factor a = 5min, compare it with the original segment of the signal, calculate the wavelet coefficients through the calculation formula, and then change the translation factor and the scale factor, and continuously decompose the signal until the sequence no longer has periodicity, and obtain the low-frequency series and high-frequency series of the two water levels.

[0062] 13) Construct a high-frequency sequence prediction model for the inland river water level based on the historical high-frequency sequence of the inland river water level; construct a low-frequency sequence prediction model for the inland river water level based on the historical low-frequency sequence of the inland river water level; construct a high-frequency sequence prediction model for the sea tide water level based on the historical high-frequency sequence of the sea tide water level; construct a low-frequency sequence prediction model for the sea tide water level based on the historical low-frequency sequence of the sea tide water level.

[0063] The low-frequency sequence prediction model adopts the autoregressive integrated moving average (ARIMA) model. The ARIMA model is a regression analysis model based on a stationary time series or a stable one after differencing, denoted as ARIMA(p,d,q), and the formula is as follows:

[0064]

[0065] Y t =β1Y t-1 +…+β p Y t-p +e t +a1e t-1 +…+a q e t-q

[0066] Among them, d is the differencing order, p is the number of autoregressive terms, q is the number of moving average terms, is the d-order difference, B is the differencing parameter, β1,...β p 、a1,...a q are fitting coefficients, X t ′ represents the low-frequency sequence, e represents the moving average prediction error, Y t 、Y t-1 、Y t-p represent the predicted values at different times. In the present invention, the number of autoregressive terms p is used to limit the time interval traced back from the current moment, reducing the calculation length of the prediction sequence and avoiding overfitting at the same time.

[0067] When constructing the ARIMA model, optimize d, p, and q:

[0068] The optimization process of d is as follows:

[0069] Based on the ADF (unit root test) principle, during modeling, perform hypothesis testing on the differences between the predicted values and the corresponding true values of the low-frequency sequence models at different times, and the predicted values and the corresponding true values of the low-frequency sequence models at the same time under different orders of differencing, so as to optimize the differencing order of different hydrological parameters;

[0070] Among them, the hypothesis testing formula is:

[0071]

[0072]

[0073] Among them, α is the constant term, β is the trend term, t represents the time, δ is the test term, and ΔX t is the predicted difference at time t, and ΔX t-1 is the predicted difference at time t - 1, and ε t is the high-order residual term after the autoregressive integrated moving average (ARIMA) processing at time t, m′ is the number of differencing processes that have been performed, and ΔX d is the predicted difference of the d-th order differencing, and β i is the trend term coefficient of the predicted difference. The predicted difference is the difference between the predicted value and the corresponding true value, and ε d is the high-order residual term after the d-th order differencing; X d-1 is the predicted value of the (d - 1)-th order differencing, and X t-1 is the predicted value of the differencing at time t - 1.

[0074] Continuously perform the autoregressive integrated moving average of the original historical low-frequency sequence, compare the difference between the true value and the predicted value, and substitute it into the above test formula for testing. The differencing order when the test term δ is not 0 is the preferred differencing order.

[0075] The optimization process of p and q is as follows:

[0076] p affects the accuracy of future predictions, and q determines the influence of noise. Based on the historical low-frequency sequence, the autocorrelation parameter and partial autocorrelation parameter can be calculated using general statistical methods to determine the value range of the autoregressive term number and the moving average term number for different low-frequency sequences. The first evaluation function is used to optimize the autoregressive term number and the moving average term number; among them, the first evaluation function introduces the influence of the differencing order on the sum of the autoregressive term number and the moving average term number based on the AIC criterion and the BIC criterion.

[0077] The formula of the first evaluation function is:

[0078]

[0079]

[0080] Among them, S(m, d) is the first evaluation function, m is the sum of the autoregressive term number p and the moving average term number q, d is the differencing order, σ 2 (m) is the intermediate process term of the first evaluation function, which plays a role in optimizing the function structure. N is the number of historical low-frequency sequence samples participating in the calculation (the tide level and the inland river level are processed separately), T″ represents the number of times of the samples participating in the current ARIMA modeling, and ΔX t is the predicted difference at time t;

[0081] The autoregressive term number and moving average term number when the evaluation function value is minimized are the preferred autoregressive term number and moving average term number.

[0082] The first evaluation function solves the problems of local optimization and overfitting that only evaluate p and q by introducing the influence of the difference order on the sum of the autoregressive term number and the moving average term number.

[0083] The high-frequency sequence prediction model uses a multiple regression model. For high-frequency sequences, the wavelet transform is still used to convert the sequence into a random sequence and a periodic sequence. According to the historical water level change situation, a hard threshold T (T is composed of outliers, maximum values, and minimum values in proportion) is set for the sequence, and the periodic sequence is denoised, and then the random sequence and the periodic sequence are combined to reconstruct the high-frequency sequence.

[0084] At the same time, based on the principles of water conservancy, a multiple regression model is constructed, with the gate opening v, the coastal wind speed f, and the tide time t c as influencing factors, and the rainfall s on the previous day as a pre-factor. According to the correlation analysis, the influencing factors with a multiple correlation coefficient above 0.85 are (v - v0) 2 、 t c -t c0 、s; where, v0 is the initial gate opening, and t c0 is the tide start time.

[0085] The finally obtained high-frequency sequence prediction model is:

[0086]

[0087] where, W t 、W t-1 represent the high-frequency sequence prediction results at times t and t - 1, and β′, a′, b, c, d′ represent the fitting coefficients of different terms.

[0088] 14) According to the historical high-frequency sequence of the inland river water level predicted by the high-frequency sequence prediction model of the inland river water level and the corresponding true value, and the historical low-frequency sequence of the inland river water level predicted by the low-frequency sequence prediction model of the inland river water level and the corresponding true value, obtain the historical residual sequence of the inland river water level, and construct a prediction model for the residual sequence of the inland river water level according to the historical residual sequence of the inland river water level;

[0089] According to the historical high-frequency sequence of the sea tide water level predicted by the high-frequency sequence prediction model of the sea tide water level and the corresponding true value, and the historical low-frequency sequence of the sea tide water level predicted by the low-frequency sequence prediction model of the sea tide water level and the corresponding true value, obtain the historical residual sequence of the sea tide water level, and construct a prediction model for the residual sequence of the sea tide water level according to the historical residual sequence of the sea tide water level.

[0090] The residual sequence prediction model uses an SVM model, and the Gaussian radial basis function is selected as the SVM kernel function.

[0091] Specifically as follows:

[0092] K(ΔX i ,ΔY i )=exp{-g||ΔX i -ΔY i ||}

[0093] Among them, ΔX i represents the low- and high-frequency residual sequences of the historical sample prediction value and the true value after ARIMA modeling and multiple regression modeling, ΔY i is the predicted value of the residual sequence, g is the kernel parameter, and K(ΔX i ,ΔY i ) is the kernel function.

[0094] When constructing the SVM model, optimize g and the penalty coefficient C to avoid overfitting and underfitting, and reduce manual participation.

[0095] Specifically, construct the residual sequence according to the residuals of the low-frequency sequence model and the high-frequency sequence model, construct the second evaluation function with the residual sequence and the penalty coefficient, determine the value range of the penalty coefficient using the second evaluation function, and use the grid search method according to the value range of the penalty coefficient and the value range of the kernel parameter determined by the LIBSVM tool to determine the optimal penalty coefficient and kernel parameter, that is, the optimal penalty coefficient and kernel parameter.

[0096] According to the correlation analysis, it is known that the residual sequence of the drainage volume has a strong correlation with the penalty coefficient C. After actual calculation, the second evaluation function can be constructed as follows:

[0097]

[0098] Among them, f(ΔW i ,C) is the second evaluation function, ΔW i is the residual sequence, C is the penalty coefficient, k is the residual sequence fitting coefficient, and N′ is the sample number of the residual sequence.

[0099] According to the actual residual sequence and the modeling situation, when it is determined that the value of the second evaluation function is less than the threshold, the threshold is generally 1.05, and the C taken can improve the prediction accuracy of the SVM modeling in the present invention, so as to determine the value range of C.

[0100] Meanwhile, the LIBSVM tool is used to determine the value range of \(g\), and then a search grid of \(g\) and \(C\) is constructed. A set of \((g, C)\) is selected sequentially for calculation. Then, the training sample set predicted by the residual sequence (i.e., the set of real residual sequences and predicted residual sequences) is evenly divided into \(M\) subsets. Any one of the training sample subsets is selected as the test set, and the remaining \(M - 1\) training sample subsets are used as the training set. \(M\)-fold cross-validation is performed on \((C, g)\) in the selected test set, and then the training set and the test set are replaced until each subset has been used as the test set. Calculate the average value of the classification accuracy of \(M\) groups; repeat the above process of selecting parameters for calculation until all grid parameter combinations are selected. Then, sort the average accuracy values under each parameter combination from largest to smallest, and select the \((C, g)\) corresponding to the largest average accuracy value as the required optimal parameter, that is, the best parameter.

[0101] 15) Combine the high-frequency sequence prediction model of the inland river water level, the low-frequency sequence prediction model of the inland river water level, the residual sequence prediction model of the inland river water level, the high-frequency sequence prediction model of the sea tide water level, the low-frequency sequence prediction model of the sea tide water level, and the residual sequence prediction model of the sea tide water level to construct an overall prediction model of the sea tide and the inland river water level.

[0102] When implementing the above method, first, according to the current inland river water level and the historical inland river water levels, obtain the inland river water level sequence at the current moment. According to the current sea tide water level and the historical sea tide water levels, obtain the sea tide water level sequence at the current moment.

[0103] Ultrasonic water level gauges can be set within a reasonable range upstream and downstream of coastal sluice stations, and the 4G signal card and RTU module are used to obtain real-time data of the inland river water level and the sea tide water level, and write the data into the database. Use kettle data synchronization to synchronize historical hydrological data to the same database, and the inland river water level sequence and the sea tide water level sequence at the current moment can be obtained.

[0104] In addition to the above water level sequences, parameters such as tide time, gate opening, wind speed, antecedent rainfall, and drainage volume are also obtained. The obtained data is input into the overall prediction model of the sea tide and the inland river water level to obtain the future inland river water level sequence and the future sea tide water level sequence.

[0105] The above method breaks away from the traditional hydraulic modeling idea. Through wavelet transform, ARIMA modeling, multiple regression modeling, and residual SVM modeling, a future change prediction model of the inland river water level and the sea tide water level in the sluice is obtained, and the future water level sequence is calculated, which solves the problems of long calculation time, large time granularity, and the inability of the prediction results to guide the actual operation of the gate scheduling in the traditional scheme, and plays a major role in practical engineering applications, improving the gate drainage time and alleviating the urban waterlogging problem.

[0106] Based on the same technical solution, the present invention also discloses a coastal sluice tide level prediction system, including:

[0107] An inland river water level sequence acquisition module, which obtains the inland river water level sequence at the current moment according to the current inland river water level and the historical inland river water level;

[0108] A sea tide water level sequence acquisition module, which obtains the sea tide water level sequence at the current moment according to the current sea tide water level and the historical sea tide water level;

[0109] A prediction module, which inputs the inland river water level sequence and the sea tide water level sequence at the current moment into a pre-constructed overall prediction model of sea tide and inland river water levels to obtain a future inland river water level sequence and a future sea tide water level sequence; wherein, the overall prediction model of sea tide and inland river water levels is constructed according to the historical inland river water level sequence and the historical sea tide water level sequence.

[0110] The data processing flow of each module in the above system is consistent with that of the method, and will not be described repeatedly here.

[0111] Based on the above method, a method for determining a gate opening and closing plan is further disclosed, see Figure 3 , including:

[0112] 1) Sampling the above-mentioned coastal sluice tide level prediction method to obtain a future inland river water level sequence and a future sea tide water level sequence;

[0113] 2) Determining the slack tide points of rising and falling according to the future inland river water level sequence and the future sea tide water level sequence;

[0114] 3) Determining the gate opening and closing plan according to the slack tide points of rising and falling, the water level rising and falling trends.

[0115] The above method uses wavelet transform to process low-frequency and high-frequency sequences, performs ARIMA autoregressive modeling on the low-frequency sequences, performs multiple regression modeling on the high-frequency sequences after noise reduction processing, and finally introduces SVM for non-linear modeling of residuals to establish a water level and tide level prediction model, analyzes the future changes of tide levels and inland river water levels, predicts future time series, calculates the time of slack tide points of rising and falling, optimizes the operation scheduling of opening and closing gates, effectively extends the drainage time, with reasonable results and fast calculation, and can well meet the optimization scheduling operation requirements of coastal sluices.

[0116] A software system corresponding to the method for determining the gate opening and closing plan, a gate opening and closing plan determination system, including:

[0117] A future sequence acquisition module, which samples the above-mentioned coastal sluice tide level prediction method to obtain a future inland river water level sequence and a future sea tide water level sequence;

[0118] The still tide point determination module determines the still tide points of ebb and flow according to the future inland river water level sequence and the future sea tide water level sequence;

[0119] The plan determination module determines the gate opening and closing plan according to the still tide points of ebb and flow, the rising and falling trends of the water level.

[0120] To verify the present invention, taking a certain sea lock in Zhejiang as an example, through data acquisition, signal processing, ARIMA modeling, multiple regression modeling and residual SVM modeling, the change prediction process of the sea tide water level and the inland river water level of the coastal sea lock is finally obtained, the time of the future still tide point is obtained, and the scheduling operation plan of the gate is given, which plays an important role during the typhoon drainage period such as "In-Fa" and "Chanthu". The prediction gap with the actual operation still tide point is within 10 minutes, and the still tide point water level error is within 10 cm, which greatly improves the drainage time, alleviates the urban waterlogging problem, and avoids the occurrence of phenomena such as road flooding, house waterlogging, and river water backflow, effectively protecting the lives and property safety of residents.

[0121] Based on the same technical solution, the present invention also discloses a computer-readable storage medium storing one or more programs, the one or more programs including instructions which, when executed by a computing device, cause the computing device to execute the coastal lock tide level prediction method or the gate opening and closing plan determination method.

[0122] Based on the same technical solution, the present invention also discloses a computing device, including one or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for executing the coastal lock tide level prediction method or the gate opening and closing plan determination method.

[0123] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0124] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0125] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0126] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0127] The above are only embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A method for predicting the tidal level of coastal sluices and pumping stations, characterized in that, Including: Obtain the inland river water level sequence at the current moment according to the current inland river water level and the historical inland river water level; Obtain the sea tide water level sequence at the current moment according to the current sea tide water level and the historical sea tide water level; Input the inland river water level sequence and the sea tide water level sequence at the current moment into the pre-constructed overall prediction model of sea tide and inland river water levels to obtain the future inland river water level sequence and the future sea tide water level sequence; Among them, the overall prediction model of sea tide and inland river water levels is constructed based on the historical inland river water level sequence and the historical sea tide water level sequence. The construction process includes: Obtain the historical inland river water level sequence and the historical sea tide water level sequence; Convert the historical inland river water level sequence into a historical inland river water level high-frequency sequence and a historical inland river water level low-frequency sequence; Construct an inland river water level high-frequency sequence prediction model based on the historical inland river water level high-frequency sequence; Construct an inland river water level low-frequency sequence prediction model based on the historical inland river water level low-frequency sequence; Obtain the historical inland river water level residual sequence according to the historical inland river water level high-frequency sequence predicted by the inland river water level high-frequency sequence prediction model and the corresponding true value, the historical inland river water level low-frequency sequence predicted by the inland river water level low-frequency sequence prediction model and the corresponding true value, and construct an inland river water level residual sequence prediction model according to the historical inland river water level residual sequence; Convert the historical sea tide water level sequence into a historical sea tide water level high-frequency sequence and a historical sea tide water level low-frequency sequence; Construct a sea tide water level high-frequency sequence prediction model based on the historical sea tide water level high-frequency sequence; Construct a sea tide water level low-frequency sequence prediction model based on the historical sea tide water level low-frequency sequence; Obtain the historical sea tide water level residual sequence according to the historical sea tide water level high-frequency sequence predicted by the sea tide water level high-frequency sequence prediction model and the corresponding true value, the historical sea tide water level low-frequency sequence predicted by the sea tide water level low-frequency sequence prediction model and the corresponding true value, and construct a sea tide water level residual sequence prediction model according to the historical sea tide water level residual sequence; Perform signal combination on the inland river water level high-frequency sequence prediction model, the inland river water level low-frequency sequence prediction model, the inland river water level residual sequence prediction model, the sea tide water level high-frequency sequence prediction model, the sea tide water level low-frequency sequence prediction model, and the sea tide water level residual sequence prediction model to construct an overall prediction model of sea tide and inland river water levels.

2. The coastal sluice tidal level prediction method according to claim 1, wherein The low-frequency sequence prediction model is an autoregressive integrated moving average model; When modeling, perform hypothesis testing on the differences between the predicted values of the low-frequency sequence models at different moments and the corresponding true values, and the predicted values of the low-frequency sequence models at the same moment under different orders of differencing and the corresponding true values, and optimize the orders of differencing for different hydrological parameters; When modeling, calculate the autocorrelation parameter and the partial autocorrelation parameter according to the historical low-frequency sequence, determine the value range of the number of autoregressive terms and the number of moving average terms for different low-frequency sequences, and use the first evaluation function to optimize the number of autoregressive terms and the number of moving average terms; among them, the first evaluation function introduces the influence of the order of differencing on the sum of the number of autoregressive terms and the number of moving average terms based on the AIC criterion and the BIC criterion.

3. The coastal sluice tide level prediction method according to claim 2, characterized in that The hypothesis testing formula is: Among them, α is the constant term, β is the trend term, d is the order of differencing, t represents the time, δ is the test term, and ΔX t is the predicted difference at time t, and ΔX t-1 is the predicted difference at time t - 1, and ε t is the high-order residual term after autoregressive integrated moving average (ARIMA) processing at time t, m′ is the number of differencing processes that have been carried out, and ΔX d is the predicted difference of the d-th order differencing, and β i is the trend term coefficient of the predicted difference. The predicted difference is the difference between the predicted value and the corresponding true value, and ε d is the high-order residual term after the d-th order differencing, and X d-1 is the predicted value of the (d - 1)-th order differencing, and X t-1 is the predicted value of the differencing at time t - 1; the order of differencing when the test term δ is not 0 is the optimal order of differencing.

4. A coastal sluice tide level prediction method according to claim 2, wherein The formula of the first evaluation function is: Among them, S(·) is the first evaluation function, m is the sum of the autoregressive term number p and the moving average term number q, d is the difference order, and σ 2 (m) is the intermediate process term of the first evaluation function, N is the number of historical low-frequency sequence samples participating in the calculation, T″ represents the number of moments of the samples participating in the modeling, and ΔX t is the prediction difference at time t; the autoregressive term number and the moving average term number when the evaluation function value is the smallest are the optimal autoregressive term number and the optimal moving average term number.

5. A coastal sluice tidal level prediction method according to claim 1, characterized in that The residual sequence prediction model is an SVM model; when modeling, a residual sequence is constructed based on the residual of the low-frequency sequence model and the residual of the high-frequency sequence model. A second evaluation function is constructed with the residual sequence and the penalty coefficient. The value range of the penalty coefficient is determined using the second evaluation function. Based on the value range of the penalty coefficient and the value range of the kernel parameter determined using the LIBSVM tool, the grid search method is used to determine the optimal penalty coefficient and kernel parameter.

6. The tidal level prediction method for coastal sluice stations according to claim 5, characterized in that The second evaluation function is: where f(·) is the second evaluation function, ΔW i is the residual sequence, C is the penalty coefficient, k is the fitting coefficient of the residual sequence, and N′ is the number of samples of the residual sequence.

7. A coastal sluice tide level prediction system, characterized in that, It includes: An inland water level sequence acquisition module that obtains the inland water level sequence at the current moment based on the current inland water level and the historical inland water level. A tidal water level sequence acquisition module that obtains the tidal water level sequence at the current moment based on the current tidal water level and the historical tidal water level. A prediction module that inputs the inland water level sequence and the tidal water level sequence at the current moment into a pre-constructed overall prediction model of tidal and inland water levels to obtain the future inland water level sequence and the future tidal water level sequence. Among them, the overall prediction model of tidal and inland water levels is constructed based on the historical inland water level sequence and the historical tidal water level sequence. The construction process includes: Obtain the historical inland water level sequence and the historical tidal water level sequence. Convert the historical inland water level sequence into a historical inland water level high-frequency sequence and a historical inland water level low-frequency sequence. Construct a prediction model for the historical inland water level high-frequency sequence based on the historical inland water level high-frequency sequence. Construct a prediction model for the historical inland water level low-frequency sequence based on the historical inland water level low-frequency sequence. Based on the historical inland water level high-frequency sequence predicted by the inland water level high-frequency sequence prediction model and the corresponding true value, and the historical inland water level low-frequency sequence predicted by the inland water level low-frequency sequence prediction model and the corresponding true value, obtain the historical inland water level residual sequence, and construct an inland water level residual sequence prediction model based on the historical inland water level residual sequence. Convert the historical tidal water level sequence into a historical tidal water level high-frequency sequence and a historical tidal water level low-frequency sequence. Construct a prediction model for the historical tidal water level high-frequency sequence based on the historical tidal water level high-frequency sequence. Construct a prediction model for the historical tidal water level low-frequency sequence based on the historical tidal water level low-frequency sequence. Based on the historical tidal water level high-frequency sequence predicted by the tidal water level high-frequency sequence prediction model and the corresponding true value, and the historical tidal water level low-frequency sequence predicted by the tidal water level low-frequency sequence prediction model and the corresponding true value, obtain the historical tidal water level residual sequence, and construct a tidal water level residual sequence prediction model based on the historical tidal water level residual sequence. Perform signal combination on the inland water level high-frequency sequence prediction model, the inland water level low-frequency sequence prediction model, the inland water level residual sequence prediction model, the tidal water level high-frequency sequence prediction model, the tidal water level low-frequency sequence prediction model, and the tidal water level residual sequence prediction model to construct an overall prediction model of tidal and inland water levels.

8. A computer-readable storage medium storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to execute any of the methods described in claims 1 to 6.

9. A computing device, characterized in that, It includes: One or more processors, one or more memories, and one or more programs, where the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods according to claims 1 to 6.

Images