Smooth warehousing flow calculation method and system, storage medium and electronic equipment

Abstract

The invention belongs to the technical field of reservoir scheduling, and provides a smooth reservoir inflow calculation method and system, a storage medium and electronic equipment, and the method comprises the steps: collecting reservoir hydrological data, processing the reservoir hydrological data, and determining original data; generating sample data according to the original data, constructing a sample set, and dividing the sample set into a training set and a test set; constructing a traffic prediction model, training the traffic prediction model by using the training set, and determining an optimal parameter model; and obtaining analysis flow data and future forecast flow data, and generating simulated smooth flow according to the analysis flow data and the future forecast flow data by using the optimal parameter model. In the process of predicting the storage flow, the manual smoothing work of the storage flow is simulated, the prediction accuracy and stability of the storage flow are improved, and the labor cost is reduced.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of reservoir operation, and particularly relates to a method, a system, a storage medium, and an electronic device for calculating smooth inflow. Background Art

[0002] The calculation of inflow is one of the most important basic tasks in reservoir operation. The compilation of reservoir flood forecasting and operation plans, the formulation of reservoir operation curves, the economic evaluation of reservoir operation, and the reservoir flood routing calculation all require the calculation of inflow.

[0003] In the prior art, when using the water balance method to obtain the calculated inflow, a smoothing method is usually adopted for processing; the smoothing method often weakens the extreme values in the data. Therefore, this method may cause some flow values to be distorted, ultimately resulting in the calculated inflow being difficult to truthfully reflect important information such as flood peaks, and extreme hydrological events such as flood peaks often have an important impact on reservoir operation and safe operation. At the same time, due to the large error of measured water levels and the influence of dynamic storage, abnormal values often appear in the calculated inflow, and it is often necessary to manually smooth the inflow according to the trend and magnitude of the forecast flow to ensure the rationality of the inflow. Summary of the Invention

[0004] To solve the above problems, the present disclosure provides a method, a system, a storage medium, and an electronic device for calculating smooth inflow. By collecting reservoir hydrological data, constructing sample data, forming a training set, training a flow prediction model, and obtaining an optimal parameter model; constructing an analytical optimization algorithm to generate an analytical inflow for subsequent generation of simulated smooth flow; the optimal parameter model generates a simulated smooth flow according to the analytical flow data and future forecast flow data. In the process of determining the optimal parameter model, artificially smoothed inflow data is used, which can include artificial experience when the optimal parameter model generates a simulated smooth flow, improving the prediction accuracy and reducing the labor cost at the same time.

[0005] The present invention is realized through the following technical solutions:

[0006] In a first aspect, an embodiment of the present disclosure provides a method for calculating smooth inflow, the method comprising:

[0007] Collect reservoir hydrological data, process the reservoir hydrological data, and determine the original data;

[0008] Generate sample data according to the original data, construct a sample set, and divide the sample set into a training set and a test set;

[0009] Construct a flow prediction model, use the training set to train the flow prediction model, and determine the optimal parameter model;

[0010] Obtain the parsed flow data and the future predicted flow data, and use the optimal parameter model to generate the simulated smooth flow based on the parsed flow data and the future predicted flow data.

[0011] Furthermore,

[0012] Collect the inflow data into the reservoir, the historical water level data, and the water level - storage capacity relationship. The inflow data into the reservoir is associated with the historical water level data and the water level - storage capacity relationship.

[0013] Integrate the inflow data into the reservoir to determine the original data, which includes the calculated inflow data into the reservoir, the predicted inflow data into the reservoir, and the artificially smoothed inflow data into the reservoir.

[0014] Furthermore,

[0015] Pre - process the original data to supplement the missing values in the original data.

[0016] Determine the time steps of the calculated inflow data into the reservoir, the predicted inflow data into the reservoir, and the artificially smoothed inflow data into the reservoir.

[0017] According to the time steps, divide the calculated inflow data into multiple first data segments, divide the predicted inflow data into second data segments, and divide the artificially smoothed inflow data into third data segments.

[0018] Use the first data segment and the corresponding second data segment as inputs, and the third data segment corresponding to the first data segment and the second data segment as outputs to construct sample data, and multiple sample data generate a sample set.

[0019] Divide the sample set into a training set and a test set according to the time sequence.

[0020] Furthermore,

[0021] The time step of the calculated inflow data into the reservoir is the first time step, the time step of the predicted inflow data into the reservoir is the second time step, and the time step of the artificially smoothed inflow data into the reservoir is the third time step; the starting time point of the predicted inflow data into the reservoir lags behind the starting time points of the calculated inflow data into the reservoir and the artificially smoothed inflow data into the reservoir; the starting time point of the artificially smoothed inflow data into the reservoir lags behind the starting time point of the calculated inflow data into the reservoir.

[0022] Furthermore,

[0023] Based on the calculated inflow data into the reservoir with the first time step, divide the first data segment. Each first data segment corresponds to the calculated inflow data into the reservoir at the first time, and adjacent first data segments are continuous in time.

[0024] Based on the predicted inflow data at the second time step, divide the second data segment. Each second data segment corresponds to the predicted inflow data at the second time, and adjacent second data segments are continuous in time;

[0025] Based on the artificially smoothed inflow data at the third time step, divide the third data segment. Each third data segment corresponds to the artificially smoothed inflow data at the third time, and adjacent third data segments are continuous in time.

[0026] Furthermore,

[0027] Take 3 first data segments that are continuous in time as the first data set, 2 second data segments that are continuous in time as the second data set. The first data set and the second data set are used as inputs, and the third data segment is used as the output; among them,

[0028] The first data set is continuous in time; for two adjacent second data sets, the calculated inflow data corresponding to the later second data set is delayed by the second time compared to the calculated inflow data corresponding to the earlier second data set.

[0029] Furthermore,

[0030] In the process of generating sample data, one second data set corresponds to two first data sets, and the two first data sets are continuous in time and are continuous in time with the corresponding second data set; one first data set corresponds to one third data segment; the nth sample data is specifically expressed as:

[0031] Input: (X n , Y round (n / 2)); Output [q n ;

[0032] Among them, X n is the first data set, n takes non-zero positive integers, and each first data set includes 3 first data segments that are continuous in time; Y round(n / 2) is the second data set, round(n / 2) represents rounding up, and each second data set includes 2 second data segments that are continuous in time; q n is the third data segment, and n takes non-zero positive integers.

[0033] Furthermore,

[0034] Construct an LSTM model as a flow prediction model, set the initial parameters of the LSTM model, and use the training set to train the LSTM model to generate simulated flows;

[0035] Combine the penalty term determined according to the simulated flow and the given flow to construct a loss function; among them, the given flow is the measured data or the standard data calculated actually;

[0036] Calculate the output error of the LSTM model layer by layer, update the initial parameters of the LSTM model according to the output error, and combine with the loss function to determine the optimal parameter model.

[0037] Furthermore,

[0038] Construct a loss function by combining the penalty term determined according to the simulated flow rate and the given flow rate, which is specifically expressed as:

[0039] L(θ) = λ data loss data + λ smoothing loss smoothing ;

[0040] where L(θ) is the loss function of the LSTM model; loss data is the root mean square error between the simulated flow rate and the given flow rate; loss smoothing is the penalty term used to penalize the part of the simulated flow rate that does not meet the flow continuity constraint; λ data , λ smoothing is the weight coefficient;

[0041] The formula for the penalty term loss smoothing is specifically expressed as:

[0042]

[0043] where q t is the simulated flow rate determined by the LSTM model at the t-th time period, and q t+1 is the simulated flow rate determined by the LSTM model at the (t + 1)-th time period; Q t is the artificially smoothed inflow rate into the reservoir at the t-th time period, and Q t+1 is the artificially smoothed inflow rate into the reservoir at the (t + 1)-th time period, and Q t and Q t+1 are used as the given flow rate; N takes a non-zero positive integer or positive infinity.

[0044] Furthermore,

[0045] Construct an objective function and solve it to obtain the calculation formula for the analytical inflow rate related to the real-time data of the reservoir; the real-time data of the reservoir includes the observed value of the real-time reservoir water storage and the reservoir outflow rate;

[0046] Generate the analytical inflow rate based on the observed value of the real-time reservoir water storage and the reservoir outflow rate.

[0047] In a second aspect, based on the same inventive concept, embodiments of the present disclosure further provide a smooth inflow calculation system, which includes a data collection module, a data processing module, a model generation module, an algorithm generation module, and a model calculation module;

[0048] The data collection module is configured to collect reservoir hydrological data, process the reservoir hydrological data, and determine the original data;

[0049] The data processing module is configured to generate sample data according to the original data, construct a sample set, and divide the sample set into a training set and a test set;

[0050] The model generation module is configured to construct a flow prediction model, train the flow prediction model using the training set, and determine an optimal parameter model;

[0051] The algorithm generation module is configured to construct an analytical optimization algorithm, and generate an analytical inflow by combining the real-time data of the reservoir;

[0052] The model calculation module is configured to determine the analytical flow data and the future forecast flow data, and generate a simulated smooth flow using the optimal parameter model according to the analytical flow data and the future forecast flow data.

[0053] In a third aspect, based on the same inventive concept, embodiments of the present disclosure further provide a computer-readable storage medium storing one or more programs, which can implement the foregoing smooth inflow calculation method when the one or more programs are executed.

[0054] In a fourth aspect, based on the same inventive concept, embodiments of the present disclosure further provide an electronic device, which includes a processor, a communication interface, the foregoing computer-readable storage medium, and a communication bus. Among them, the processor, the communication interface, and the computer-readable storage medium communicate with each other through the communication bus. Among them, the processor is configured to execute the program stored in the foregoing computer-readable storage medium.

[0055] Compared with the prior art, the present disclosure has the following advantages:

[0056] During the process of predicting the inflow, the smooth work of the inflow is simulated manually, which improves the prediction accuracy and stability of the inflow, and reduces the labor cost.

[0057] Other features and advantages of the present disclosure will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present disclosure. The objectives and other advantages of the present disclosure can be achieved and obtained through the structures pointed out in the specification, the claims, and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] To more clearly illustrate the technical solutions in the embodiments of the present disclosure or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0059] Figure 1 Flowchart of a smooth incoming flow calculation method provided by an embodiment of the present disclosure;

[0060] Figure 2 Schematic diagram of an LSTM model provided by an embodiment of the present disclosure;

[0061] Figure 3 Block diagram of a smooth incoming flow calculation system provided by an embodiment of the present disclosure;

[0062] Figure 4 Schematic diagram of the structure of an electronic device provided by an embodiment of the present disclosure. Specific embodiments

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present disclosure with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are some, but not all, of the embodiments of the present disclosure. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present disclosure without creative efforts fall within the scope of protection of the present disclosure.

[0064] First aspect, Figure 1 Flowchart of a smooth incoming flow calculation method provided by an embodiment of the present disclosure. As Figure 1 shown, an embodiment of the present disclosure provides a smooth incoming flow calculation method, including:

[0065] S1: Collect reservoir hydrological data, process the reservoir hydrological data, and determine the original data.

[0066] S2: Generate sample data based on the original data, construct a sample set, and divide the sample set into a training set and a test set.

[0067] S3: Construct a flow prediction model, use the training set to train the flow prediction model, and determine the optimal parameter model.

[0068] S4: Obtain the parsed flow data and future forecast flow data, and use the optimal parameter model to generate the simulated smooth flow according to the parsed flow data and the future forecast flow data.

[0069] In the embodiments of the present disclosure, reservoir hydrological data is collected, sample data is constructed to form a training set, and a flow prediction model is trained to obtain an optimal parameter model; by analyzing historical data, observation data, forecast data, and through methods such as prediction models or software, analytical flow data and future forecast flow data are obtained. Alternatively, an analytical optimization algorithm is constructed to generate analytical inflow for subsequent generation of simulated smooth flow; the optimal parameter model generates simulated smooth flow based on the analytical flow data and future forecast flow data. In the process of determining the optimal parameter model, artificial smooth inflow data is used, so the simulated smooth flow generated by the optimal parameter model contains artificial experience, improving the prediction accuracy and reducing the labor cost at the same time.

[0070] In some examples, reservoir hydrological data is collected, the reservoir hydrological data is processed, and the original data is determined, including:

[0071] S11: Collect inflow data, historical water level data, and water level-storage capacity relationship. The inflow data is associated with the historical water level data and the water level-storage capacity relationship.

[0072] S12: Integrate the inflow data to determine the original data. The original data includes calculated inflow data, forecast inflow data, and artificially smoothed inflow data.

[0073] In some examples, sample data is generated based on the original data, a sample set is constructed, and the sample set is divided into a training set and a test set, including:

[0074] S21: Preprocess the original data to supplement the missing values in the original data.

[0075] Specifically, it is necessary to check for missing values in the original data, and methods such as interpolation and filling can be used to supplement the missing values, which may specifically include linear interpolation method, previous value filling method, and mean filling method. The calculated inflow data, forecast inflow data, and artificially smoothed inflow data are preprocessed into temporally continuous data for subsequent division of the calculated inflow data, forecast inflow data, and artificially smoothed inflow data into data segments according to the time step.

[0076] S22: Determine the time step of the calculated inflow data, forecast inflow data, and artificially smoothed inflow data.

[0077] S23: According to the time step, divide the calculated inflow data into multiple first data segments, divide the forecast inflow data into second data segments, and divide the artificially smoothed inflow data into third data segments.

[0078] S24: Using the first data segment and the corresponding second data segment as inputs, and the third data segment corresponding to the first and second data segments as the output, construct sample data, and multiple sample data generate a sample set.

[0079] S25: Divide the sample set into a training set and a test set according to time series.

[0080] In some examples, determining the time steps for calculating the incoming reservoir flow data, predicting the incoming reservoir flow data, and manually smoothing the incoming reservoir flow data includes:

[0081] The time step for calculating the incoming reservoir flow data is the first time step, the time step for predicting the incoming reservoir flow data is the second time step, and the time step for manually smoothing the incoming reservoir flow data is the third time step; the starting time point of the predicted incoming reservoir flow data lags behind the starting time points of the calculated incoming reservoir flow data and the manually smoothed incoming reservoir flow data; the starting time point of the manually smoothed incoming reservoir flow data lags behind the starting time point of the calculated incoming reservoir flow data.

[0082] In the embodiments of the present disclosure, the first time is 2 hours, the second time is 12 hours, and the third time is 6 hours. The time step for calculating the incoming reservoir flow data is 2 hours, the time step for predicting the incoming reservoir flow data is 12 hours, and the time step for manually smoothing the incoming reservoir flow data is 6 hours; the starting time point of the predicted incoming reservoir flow data lags behind the starting time point of the calculated incoming reservoir flow data by 12 hours, and the starting time point of the manually smoothed incoming reservoir flow data lags behind the starting time point of the calculated incoming reservoir flow data by 6 hours.

[0083] In some examples, according to the time steps, dividing the calculated incoming reservoir flow data into multiple first data segments, dividing the predicted incoming reservoir flow data into second data segments, and dividing the manually smoothed incoming reservoir flow data into third data segments includes:

[0084] Based on the calculated incoming reservoir flow data with the first time step, divide the first data segments. Each first data segment corresponds to the calculated incoming reservoir flow data for the first time, and adjacent first data segments are continuous in time;

[0085] Based on the predicted incoming reservoir flow data with the second time step, divide the second data segments. Each second data segment corresponds to the predicted incoming reservoir flow data for the second time, and adjacent second data segments are continuous in time;

[0086] Based on the manually smoothed incoming reservoir flow data with the third time step, divide the third data segments. Each third data segment corresponds to the manually smoothed incoming reservoir flow data for the third time, and adjacent third data segments are continuous in time. The specific representation methods of the first data segment, the second data segment, and the third data segment are as follows:

[0087] Each first data segment is represented as: [x 1 ,x2 ,x 3 ......,x 8 ,......];

[0088] Each second data segment is expressed as: [y 1 ,y 2 ,......];

[0089] Each third data segment is expressed as: [q 1 ,q 2 ,......];

[0090] Among them, [x 1 ,x 2 ,x 3 ......,x 8 ,......] the adjacent first data segments are continuous in time, [y 1 ,y 2 ,......] the adjacent second data segments are continuous in time, [q 1 ,q 2 ,......] the adjacent third data segments are continuous in time.

[0091] In the embodiments of the present disclosure, since the starting time point of the predicted inflow data into the reservoir lags behind the starting time point of the calculated inflow data into the reservoir by 12 hours, and the starting time point of the artificially smoothed inflow data into the reservoir lags behind the starting time point of the calculated inflow data into the reservoir by 6 hours; and because the step size of the first data segment is 2 hours, the step size of the second data segment is 12 hours, and the step size of the third data segment is 6 hours; therefore, it can be understood that the end time point of x 6 is the starting time point of y 1 ,the end time point of x 3 is the starting time point of q 1 ,and so on.

[0092] In some examples, using the first data segment and the corresponding second data segment as inputs, and the third data segment corresponding to the first data segment and the second data segment as outputs, sample data is constructed, and multiple sample data generate a sample set, including:

[0093] Using 3 first data segments that are continuous in time as the first data set, 2 second data segments that are continuous in time as the second data set, the first data set and the second data set as inputs, and the third data segment as the output; among them,

[0094] The first data set is continuous in time; for two adjacent second data sets, the calculated incoming flow data corresponding to the later second data set is delayed by a second time compared to the calculated incoming flow data corresponding to the earlier second data set; therefore, there are duplicate second data segments in two adjacent second data sets.

[0095] Further, in the process of generating sample data, one second data set corresponds to two first data sets, and the two first data sets are continuous in time and are continuous in time with the corresponding second data set; one first data set corresponds to one third data segment; the nth sample data can be specifically expressed as:

[0096] Input: (X n , Y round(n / 2) ); Output [q n ;

[0097] wherein, X n is the first data set, n is a non-zero positive integer, and each first data set includes 3 first data segments that are continuous in time; Y round(n / 2 ) is the second data set, round(n / 2) represents rounding up, and each second data set includes 2 second data segments that are continuous in time; q n is the third data segment, and n is a non-zero positive integer.

[0098] It should be understood that the time step of the calculated incoming flow data, the predicted incoming flow data, and the artificially smoothed incoming flow data provided in the embodiments of the present disclosure, as well as the construction method of the sample data, are only used as a specific example, and can be adjusted according to the actual flow data situation, the time step of the calculated incoming flow data, the predicted incoming flow data, and the artificially smoothed incoming flow data, and the construction method of the sample data is correspondingly adjusted.

[0099] In some examples, a flow prediction model is constructed, and the flow prediction model is trained using a training set to determine an optimal parameter model, including:

[0100] S31: Construct an LSTM model as the flow prediction model, set the initial parameters of the LSTM model, and train the LSTM model using the training set to generate simulated flow.

[0101] S32: Combine a penalty term determined according to the simulated flow and the given flow to construct a loss function; wherein, the given flow is measured data or standard data calculated actually.

[0102] S33 Calculate the output error of the LSTM model layer by layer, update the initial parameters of the LSTM model according to the output error, and combine the loss function to determine the optimal parameter model.

[0103] Specifically, according to the output error calculated layer by layer, the initial parameters of the LSTM model are updated using the gradient descent method to obtain an LSTM model with a loss function value as low as possible. The LSTM model at this time is tested using the test set. If it passes the test of the test set, the optimal parameter model is finally determined. The optimal parameter model can generate artificial smooth inflow data with artificial experience.

[0104] In some examples, a loss function is constructed by combining a penalty term determined according to the simulated flow and the given flow, and is specifically expressed as:

[0105] L(θ) = λ data loss data + λ smoothing loss smoothing ;

[0106] where L(θ) is the loss function of the LSTM model; loss data is the root mean square error between the simulated flow and the given flow, and the root mean square error can be normalized; loss smoothing is the penalty term used to penalize the part of the simulated flow that does not meet the flow continuity constraint; λ data , λ smoothing is the weight coefficient.

[0107] The formula for the penalty term loss smoothing is specifically expressed as:

[0108]

[0109] where q t is the simulated flow determined by the LSTM model at the t-th time period, and q t+1 is the simulated flow determined by the LSTM model at the (t + 1)-th time period; Q t is the artificial smooth inflow at the t-th time period, Q t+1 is the artificial smooth inflow at the (t + 1)-th time period, Q t and Q t+1 are used as the given flow; N takes a non-zero positive integer or positive infinity. It can be understood that Q t and Q t+1 correspondingly serve as the output values in the sample data, that is, the artificial smooth inflow with artificial experience, and thus are regarded as the standard data actually calculated.

[0110] In some examples, before obtaining the parsed flow data and the future forecast flow data, using the optimal parameter model, and generating the simulated smooth flow according to the parsed flow data and the future forecast flow data, it also includes: constructing a parsing optimization algorithm, and generating the parsed inflow by combining the real-time data of the reservoir.

[0111] Furthermore, construct an analytical optimization algorithm, and combine it with the real-time data of the reservoir to generate the analytical inflow, including:

[0112] S41’: Construct an objective function and solve it to obtain the calculation formula of the analytical inflow related to the real-time data of the reservoir; the real-time data of the reservoir includes the observed value of the real-time reservoir water storage and the reservoir outflow.

[0113] S42’: Generate the analytical inflow according to the observed value of the real-time reservoir water storage and the reservoir outflow.

[0114] It can be understood that by analyzing historical data, observed data, forecast data, and through prediction models or software, etc., obtain the analytical flow data and future forecast flow data. Also, an analytical optimization algorithm can be constructed to generate the analytical inflow for subsequent generation of simulated smooth flow.

[0115] Example 1:

[0116] S101: Collect the reservoir hydrological data, process the reservoir hydrological data, and determine the original data.

[0117] The reservoir hydrological data includes inflow data, historical water level data, and water level-storage relationship. Collect the historical water level data and water level-storage relationship according to actual research needs. The historical water level data includes the measured water level data with high time resolution obtained from the water level measuring station in the reservoir area. The water level-storage relationship includes the water level-storage relationship curve of the reservoir area; the inflow data is associated with the historical water level data and the water level-storage relationship. For example, in the process of determining the inflow data, the historical water level data and the water level-storage relationship can be used. The inflow data specifically includes calculated inflow data, forecast inflow data, and artificially smoothed inflow data, etc. Among them, the calculated inflow data is determined by the eight-segment method.

[0118] Process the collected reservoir hydrological data. The processing methods include integrating and collating the collected historical water level data and water level-storage relationship; it also includes integrating the calculated inflow data, forecast inflow data, and artificially smoothed inflow data according to time series as the original data.

[0119] S102: Generate a sample set according to the original data, and divide the sample set into a training set and a test set.

[0120] First, it is necessary to check for missing values in the original data. Methods such as interpolation and filling can be used to supplement the missing values, specifically including linear interpolation, previous value filling, and mean filling. The calculated incoming flow data, predicted incoming flow data, and artificially smoothed incoming flow data are preprocessed into temporally continuous data for subsequent division of data segments based on time steps for the calculated incoming flow data, predicted incoming flow data, and artificially smoothed incoming flow data.

[0121] Determine the time steps for the calculated incoming flow data, predicted incoming flow data, and artificially smoothed incoming flow data, and divide the calculated incoming flow data, predicted incoming flow data, and artificially smoothed incoming flow data into data segments according to the time steps. The calculated incoming flow data has a 2-hour time step, the predicted incoming flow data has a 12-hour time step, and the artificially smoothed incoming flow data has a 6-hour time step. The starting time point of the predicted incoming flow data lags 12 hours behind the starting time point of the calculated incoming flow data, and the starting time point of the artificially smoothed incoming flow data lags 6 hours behind the starting time point of the calculated incoming flow data. Based on the calculated incoming flow data with a 2-hour time step, divide the first data segments, where each first data segment corresponds to 2 hours of calculated incoming flow data and adjacent first data segments are temporally continuous; based on the predicted incoming flow data with a 12-hour time step, divide the second data segments, where each second data segment corresponds to 12 hours of predicted incoming flow data and adjacent second data segments are temporally continuous; based on the artificially smoothed incoming flow data with a 6-hour time step, divide the third data segments, where each third data segment corresponds to 6 hours of artificially smoothed incoming flow data and adjacent third data segments are temporally continuous. The specific representations of the first data segments, second data segments, and third data segments are as follows:

[0122] Each first data segment is represented as: [x 1 ,x 2 ,x 3 ......,x 8 ,......];

[0123] Each second data segment is represented as: [y 1 ,y 2 ,......];

[0124] Each third data segment is represented as: [q 1 ,q 2 ,......];

[0125] Among them, the adjacent first data segments in [x 1 ,x 2 ,x 3 ......,x 8 ,......] are temporally continuous, and [y 1, y 2 ,......] The adjacent second data segments in [q 1 , q 2 ,......] The adjacent third data segments in are continuous in time.

[0126] Since the starting time point of the predicted incoming flow data lags behind the starting time point of the calculated incoming flow data by 12 hours, and the starting time point of the artificially smoothed incoming flow data lags behind the starting time point of the calculated incoming flow data by 6 hours; also because the step size of the first data segment is 2 hours, the step size of the second data segment is 12 hours, and the step size of the third data segment is 6 hours; therefore, it can be understood that the end time point of x 6 is the starting time point of y 1 , the end time point of x 3 is the starting time point of q 1 , and so on for the subsequent ones.

[0127] In the embodiments of the present disclosure, according to the time step sizes of the calculated incoming flow data, the predicted incoming flow data, and the artificially smoothed incoming flow data, the corresponding first data segment, second data segment, and third data segment, sample data is constructed. All the sample data is used as a sample set, and the sample data is continuous in time. In the process of constructing the sample data, three consecutive first data segments in time are used as the first data set, two consecutive second data segments in time are used as the second data set, the first data set and the second data set are used as inputs, and the third data segment is used as the output; the first data set is continuous in time, and for two adjacent first data sets, the calculated incoming flow data corresponding to the later first data set is 6 hours later than the calculated incoming flow data corresponding to the earlier first data set; the second data set is continuous in time, and for two adjacent second data sets, the calculated incoming flow data corresponding to the later second data set is 12 hours later than the calculated incoming flow data corresponding to the earlier second data set; since the second data set includes two adjacent second data segments, and the second data segment corresponds to 12 hours of predicted incoming flow data, there are duplicate second data segments in two adjacent second data sets; the sample data is constructed as follows:

[0128] Four sets of sample data are given below for example:

[0129] Input: ([x 1 , x 2 , x 3 , [y 1 , y 2 ); Output [q 1 ;

[0130] Input: ([x 4 , x 5 , x6 , [y 1 , y 2 );Output [q 2 ;

[0131] Input: ([x 7 , x 8 , x 9 , [y 2 , y 3 );Output [q 3 ;

[0132] Input: ([x 10 , x 11 , x 12 , [y 2 , y 3 );Output [q 4 ;

[0133] Furthermore, during the process of generating sample data, one second data set corresponds to two first data sets, and the two first data sets are continuous in time and are continuous in time with the corresponding second data set; one first data set corresponds to one third data segment; the nth sample data can be specifically expressed as:

[0134] Input: (X n , Y round(n / 2) );Output [q n ;

[0135] Among them, X n is the first data set, n takes non-zero positive integers, each first data set includes 3 first data segments that are continuous in time, the 1st first data set is [x 1 , x 2 , x 3 , the 2nd first data set is [x 4 , x 5 , x 6 , and so on; Y round(n / 2) is the second data set, round(n / 2) represents rounding, which means rounding up here, each second data set includes 2 second data segments that are continuous in time, the 1st second data set is [y 1 , y 2 , the 2nd second data set is [y 2 , y 3 , and so on; q n is the third data segment, and n takes non-zero positive integers.

[0136] In the embodiments of the present disclosure, a training set and a test set are determined according to the calibration period and the verification period. For example, the sample data in the sample set are arranged in chronological order. According to the time sequence, the first 80% of the sample data in the sample set are used as the training set, and the last 20% of the sample data are used as the test set. The values of the traffic data in the sample data are usually relatively large. In order to prevent problems such as numerical explosion or gradient disappearance, a normalization method needs to be used for preprocessing.

[0137] S103: Construct a traffic prediction model, use the training set to train the traffic prediction model, and determine the optimal parameter model.

[0138] In the embodiments of the present disclosure, when constructing a traffic prediction model, an LSTM (Long Short-Term Memory) model is selected as the traffic prediction model.

[0139] Set the initial parameters of the LSTM model. The initial parameters include the weight of the forget gate, the bias term of the forget gate, the weight matrix of the input gate, the bias term of the input gate, the weight matrix of the candidate memory cell, the bias term of the candidate memory cell, the weight of the output gate, the bias term of the output gate, as well as the learning rate, the number of layers, etc.

[0140] As Figure 2 shown, each layer of the LSTM model calculates in the order of the forget gate, the input gate, the update cell state, and the output gate, and finally obtains an output value. The output value is used as the input of the next layer for the calculation of the next layer.

[0141] The forget gate determines the information to be retained and forgotten in the previous cell state C t-1 . Among them, the cell state is the long-term memory in the LSTM model. It can retain information for a long time without being forgotten. It is adjusted by the forget gate and the input gate. The cell state can selectively retain or forget information to ensure that useful information can be retained for a long time. The calculation formula of the forget gate is:

[0142] f t = σ(W f [x t , h t-1 +b f ) (1)

[0143] Among them, f t is the output of the forget gate, and the range is (0, 1), indicating the proportion of retaining the previous state information; σ is the sigmoid activation function; W f is the weight of the forget gate; h t-1 is the previous hidden state, which contains the memory information of the previous moment; x t is the input data; b fIt is the bias term of the forget gate, used to adjust the flexibility of the model.

[0144] Input gate, which determines the information to be stored in the cell state; candidate memory unit, used to update the cell state. The calculation formula of the input gate is:

[0145] i t = σ(W i [h t-1 , x t +b i ) (2)

[0146]

[0147] Among them, i t is the output of the input gate, representing the retention ratio of the current input information; W i is the weight matrix of the input gate; b i is the bias term of the input gate; is the candidate memory unit; W c is the weight matrix of the candidate memory unit; b c is the bias term of the candidate memory unit; h t-1 is the hidden state at the previous moment, containing the memory information at the previous moment; x t is the input data.

[0148] Update the cell state:

[0149]

[0150] In the formula, C t is the cell state at the current moment; f t is the output of the forget gate, with a range of (0, 1), representing the ratio of retaining the information of the previous state; C t-1 represents the cell state at the previous moment; i t is the output of the input gate; is the candidate memory unit.

[0151] Output gate, which determines the final output value and is passed to the next layer or the next moment. The calculation formula of the output gate is:

[0152] o t = σ(W o [h t-1 , x t +b o ) (5)

[0153] h t = o t tanh(C t ) (6)

[0154] Among them, ot is the output of the output gate, with a range of (0, 1), representing the proportion of information output in the cell state; W o is the weight of the output gate; b o is the bias term of the output gate; C t is the cell state at the current time; h t is the hidden state at the current time, which is passed as the output value to the next layer or the next time; h t-1 is the hidden state at the previous time, which contains the memory information of the previous time; x t is the input data.

[0155] The training set includes multiple sample data. The sample data takes the calculated incoming reservoir flow data and the predicted incoming reservoir flow data as inputs, and takes the artificially smoothed incoming reservoir flow data as the output. During the training process of the LSTM model, the LSTM model performs forward propagation. For each set of calculated incoming reservoir flow data and predicted incoming reservoir flow data, the model will try to calculate and output a predicted value close to the artificially smoothed incoming reservoir flow data. After multiple iterations of the LSTM model, the simulated flow corresponding to the training set is generated.

[0156] In order to further optimize the LSTM model to determine the optimal parameter model. Combining the penalty term determined according to the simulated flow and the given flow, a loss function is constructed. Among them, the given flow is the measured data or the standard data calculated actually. The loss function is used to evaluate the LSTM model. The smaller the value of the loss function, the more accurate the output of the LSTM model; evaluate the LSTM model according to the value of the loss function and adjust the initial parameters to determine the optimal parameter model.

[0157] The loss function is specifically expressed as:

[0158] L(θ) = λ data loss data + λ smoothing loss smoothing (7)

[0159] Among them, L(θ) is the loss function of the LSTM model; loss data is the root mean square error between the simulated flow and the given flow, and the root mean square error can be normalized; loss smoothing is the penalty term, which is used to penalize the part of the simulated flow that does not meet the flow continuity constraint; λ data , λ smoothing is the weight coefficient.

[0160] The formula of the penalty term loss smoothing is specifically expressed as:

[0161]

[0162] Among them, q t is the simulated flow determined by the LSTM model in the t-th time period, and q t+1 is the simulated flow determined by the LSTM model in the (t + 1)-th time period; Q t is the artificially smoothed inflow in the t-th time period, and Q t+1 is the artificially smoothed inflow in the (t + 1)-th time period, and Q t and Q t+1 are used as the given flow; N takes a non-zero positive integer or positive infinity.

[0163] Furthermore, the loss function is extended to:

[0164]

[0165] Among them, L(θ) is the loss function of the LSTM model; is the extended formula of loss data , q t is the simulated flow determined by the LSTM model in the t-th time period, is the average value of the artificially smoothed inflow.

[0166] After completing the construction of the loss function, for each layer in the LSTM model, from the output layer to the input layer, calculate the output error layer by layer:

[0167] δ (l) = δ (l+1) ·W (l+1) ·σ′ (l) (z (l) ) (10)

[0168] Among them, δ (l) is the output error of the l-th layer, and δ (l+1) is the output error of the (l + 1)-th layer, W (l+1) represents the weight matrix from the l-th layer to the (l + 1)-th layer, σ is the sigmoid activation function, and σ′ (l) is the derivative of the activation function, and z (l) is the linear combination of the l-th layer, for example, the product of the input and the weight matrix of the l-th layer of the LSTM model, plus the bias term; the specific formula of z (l) is adjusted according to the actual situation and is not further specifically limited here.

[0169] According to the output error calculated layer by layer, use the gradient descent method to update the initial parameters of the LSTM model to obtain an LSTM model with as low a loss function value as possible. Use the test set to test the LSTM model at this time. If it passes the test of the test set, then finally determine the optimal parameter model. The optimal parameter model can generate artificially smoothed inflow data with artificial experience.

[0170] S104: Construct an analytical optimization algorithm, and generate the analytical inflow by combining with the real-time data of the reservoir.

[0171] In constructing the analytical optimization algorithm, the continuity of the inflow and the water level estimation error are considered simultaneously, which is a bi-objective optimization problem. Thus, the objective function is constructed and solved.

[0172] First, determine the objective function and the corresponding constraint conditions. The objective function is expressed as:

[0173]

[0174] The constraint conditions are expressed as:

[0175]

[0176] Among them, Z i，j and are the estimated water level value and the observed water level value at time i and position j respectively. Q i is the average inflow at time i. O i is the average outflow at time i. Δf is the time interval from time i to time i + 1. I i is the inflow at time i. R i is the reservoir outflow at time i. is the observed value of the reservoir water storage at time i, which is obtained from the observed water level and the water level-storage relationship f(·). α is a weight factor ranging from 0 to 1. n and m are the lengths of the inflow sequence and the number of measured water levels respectively.

[0177] Furthermore, Equation (11) simultaneously minimizes the difference in flow rates between adjacent times and the difference between the observed water level and the estimated water level. When α = 0, Equations (11) and (12) are equivalent to the traditional water balance method (SWB). Substituting Equation (12) into Equation (11) gives:

[0178]

[0179] Among them, the reservoir outflow R i and the observed value of the reservoir water storage are known. V i is used as the decision variable for inflow inversion.

[0180] After completing the construction of the objective function, it is necessary to solve the objective function to obtain the optimal analytical inflow.

[0181] Using the Lagrange method to analyze the objective function, the specific process is as follows:

[0182]

[0183]

[0184] Among them, W is a coefficient matrix of dimension (n + 1)×(n + 1); V = {V i , i = 1, 2,..., n + 1} is a state variable vector of dimension (n + 1)×1, representing the estimated value of the reservoir water storage; is a constant vector of dimension 1×(n + 1), and its value is related to the reservoir outflow R i and the observed value of the reservoir water storage .

[0185] The estimated value of the reservoir water storage can be determined by formulas (14a)-(14c) as follows:

[0186]

[0187] Furthermore, determine the calculation formula for the inflow and outflow, and complete the construction of the analytical optimization algorithm. The calculation formula for the inflow is:

[0188]

[0189] Among them, ΔT is the time interval from the i-th moment to the (i + 1)-th moment, I i is the inflow at the i-th moment, V i as the decision variable for inflow inversion can represent the reservoir capacity at the i-th moment, Q i is used to analyze the inflow, and thus obtain the calculation formula for the analytical inflow related to the real-time reservoir data.

[0190] In the analytical optimization algorithm of this embodiment of the present disclosure, the weight value α represents the weight of the inflow change term in the objective function, and its value range is from 0 to 1. When α takes 1, it means only considering the change in the inflow; when α takes 0, it means only considering the water level error, and at this time the analytical method is equivalent to the water balance method. The value of α is determined by the competition relationship between the continuity of the inflow and the estimation error of the water storage. Under the condition of determining the water level observation accuracy and the non-negative inflow of the reservoir, the trial method is used for estimation.

[0191] The real-time reservoir data includes the observed value of the real-time reservoir water storage and the reservoir outflow. Collect the observed value of the real-time reservoir water storage and the reservoir outflow, and use the analytical optimization algorithm to calculate the analytical inflow.

[0192] S105: Determine the parsed flow data and the future predicted flow data, and use the optimal parameter model to generate the simulated smooth flow based on the parsed flow data and the future predicted flow data.

[0193] Collect sufficient parsed incoming reservoir flow and future predicted flow. For example, there are three sets of parsed incoming reservoir flow data with a two-hour time step and two sets of future predicted flow data with a 12-hour time step. After determining the parsed flow data and the future predicted flow data, use the optimal parameter model to generate the simulated smooth flow based on the parsed flow data and the future predicted flow data.

[0194] It can be understood that the method for obtaining the future predicted flow data is different from that of the aforementioned predicted incoming reservoir flow data in terms of corresponding different time periods, and the acquisition methods can be the same or different.

[0195] In a second aspect, Figure 3 The following is a block diagram of a smooth incoming reservoir flow calculation system provided by an embodiment of the present disclosure. As Figure 3 shown, based on the same inventive concept, an embodiment of the present disclosure also provides a smooth incoming reservoir flow calculation system. The system includes: a data collection module, a data processing module, a model generation module, an algorithm generation module, and a model calculation module. The data collection module is used to collect reservoir hydrological data, process the reservoir hydrological data, and determine the original data. The data processing module is used to generate sample data based on the original data, construct a sample set, and divide the sample set into a training set and a test set. The model generation module is used to construct a flow prediction model, train the flow prediction model using the training set, and determine the optimal parameter model. The algorithm generation module is used to construct a parsing optimization algorithm, and combine the real-time reservoir data to generate the parsed incoming reservoir flow. The model calculation module is used to determine the parsed flow data and the future predicted flow data, and use the optimal parameter model to generate the simulated smooth flow based on the parsed flow data and the future predicted flow data.

[0196] In a third aspect, based on the same inventive concept, an embodiment of the present disclosure also provides a computer-readable storage medium storing one or more programs, which can implement the aforementioned smooth incoming reservoir flow calculation method when the one or more programs are executed.

[0197] In a fourth aspect, based on the same inventive concept, an embodiment of the present disclosure also provides an electronic device, including a processor, a communication interface, the aforementioned computer-readable storage medium, and a communication bus. Among them, the processor, the communication interface, and the computer-readable storage medium communicate with each other through the communication bus. Among them, the processor is used to execute the program stored in the aforementioned computer-readable storage medium.

[0198] It should be noted that the electrical connections between the above-mentioned units do not necessarily represent the connections between the circuits. Indirect connection methods, as long as they achieve the purpose of the present disclosure, can be applied to the embodiments of the present disclosure.

[0199] Although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A method for calculating smooth inbound flow, characterized in that: The method comprises: Collecting reservoir hydrological data, processing the reservoir hydrological data, and determining original data; Generate sample data according to the original data, construct a sample set, and divide the sample set into a training set and a test set; Constructing a traffic prediction model, using the training set to train the traffic prediction model, and determining an optimal parameter model; The analyzed flow data and the future predicted flow data are obtained, and the simulated smoothed flow is generated according to the analyzed flow data and the future predicted flow data using the optimal parameter model.

2. The method according to claim 1, characterized in that: The collecting of reservoir hydrological data, processing of the reservoir hydrological data, and determination of original data include: Collect inflow data, historical water level data and water level-reservoir capacity relationship; The inflow flow data are integrated to determine the original data, wherein the original data includes calculated inflow flow data, forecasted inflow flow data and artificially smoothed inflow flow data.

3. The method according to claim 1, characterized in that The method of generating sample data according to the original data, constructing a sample set, and dividing the sample set into a training set and a test set includes: Preprocessing the original data to supplement missing values ​​in the original data; Determine the time step of the calculated inflow flow data, the forecasted inflow flow data and the artificially smoothed inflow flow data; According to the time step, the calculated inflow flow data is divided into a plurality of first data segments, the forecasted inflow flow data is divided into a second data segment, and the artificially smoothed inflow flow data is divided into a third data segment; Taking a first data segment and a corresponding second data segment as input and a third data segment corresponding to the first data segment and the second data segment as output, constructing sample data, and a plurality of the sample data generate a sample set; The sample set is divided into a training set and a test set according to time sequence.

4. The method according to claim 3, characterized in that The step of determining the time step of calculating the inflow flow data, the forecast inflow flow data and the artificially smoothed inflow flow data includes: The calculated inbound flow data is the first time step, the forecasted inbound flow data is the second time step, and the artificially smoothed inbound flow data is the third time step; the starting time point of the forecasted inbound flow data lags behind the starting time point of the calculated inbound flow data and the starting time point of the artificially smoothed inbound flow data; the starting time point of the artificially smoothed inbound flow data lags behind the starting time point of the calculated inbound flow data.

5. The method according to claim 3, characterized in that: The method of dividing the calculated inflow flow data into a plurality of first data segments, dividing the forecasted inflow flow data into a second data segment, and dividing the artificially smoothed inflow flow data into a third data segment according to the time step includes: Divide the calculated inbound flow data based on the first time step into first data segments, each first data segment corresponds to the calculated inbound flow data at the first time, and adjacent first data segments are continuous in time; Divide the forecasted inflow flow data of the second time step into second data segments, each second data segment corresponds to the forecasted inflow flow data of the second time, and adjacent second data segments are continuous in time; Based on the artificially smoothed inbound flow data of the third time step, third data segments are divided, each third data segment corresponds to the artificially smoothed inbound flow data of the third time, and adjacent third data segments are continuous in time.

6. The method according to claim 3, characterized in that The method uses the first data segment and the corresponding second data segment as input, and the third data segment corresponding to the first data segment and the second data segment as output, constructs sample data, and generates a sample set with multiple sample data, including: Three first data segments that are consecutive in time are used as the first data set, two second data segments that are consecutive in time are used as the second data set, the first data set and the second data set are used as input, and the third data segment is used as output; wherein, The first data sets are continuous in time; for two adjacent second data sets, the calculated inflow flow data corresponding to the later second data set is delayed by a second time compared to the calculated inflow flow data corresponding to the earlier second data set.

7. The method according to claim 6, characterized in that The method uses the first data segment and the corresponding second data segment as input, and the third data segment corresponding to the first data segment and the second data segment as output, constructs sample data, and generates a sample set with multiple sample data, and also includes: In the process of generating sample data, one second data set corresponds to two first data sets, and the two first data sets are continuous in time and continuous in time with the corresponding second data sets; one first data set corresponds to one third data segment; the nth sample data is specifically expressed as: Input: (X n , Y round(n / 2) ); output [q n ]; Among them, X n is a first data set, n is a non-zero positive integer, and each first data set includes three first data segments that are continuous in time; Y round(n / 2) is the second data set, round(n / 2) means rounding up, and each second data set includes two second data segments that are continuous in time; q n is the third data segment, and n is a non-zero positive integer.

8. The method according to claim 1, characterized in that: The method of constructing a traffic prediction model, training the traffic prediction model using a training set, and determining an optimal parameter model includes: Constructing an LSTM model as the traffic prediction model, setting initial parameters of the LSTM model, training the LSTM model using a training set, and generating simulated traffic; A loss function is constructed by combining the penalty term determined according to the simulated flow and the given flow, wherein the given flow is measured data or standard data actually calculated; The output error of the LSTM model is calculated layer by layer, the initial parameters of the LSTM model are updated according to the output error, and the optimal parameter model is determined in combination with the loss function.

9. The method according to claim 8, characterized in that The loss function is constructed by combining the penalty term determined according to the simulated traffic and the given traffic, which is specifically expressed as: L(θ)=λ data loss data +λ smoothing loss smoothing ; Among them, L(θ) is the loss function of the LSTM model; loss data is the root mean square error between the simulated flow and the given flow; loss smoothing is a penalty term, which is used to penalize the part of the simulated traffic that does not meet the traffic continuity constraint; data ,λ smoothing is the weight coefficient; Penalty term loss smoothing The formula is specifically expressed as: Among them, q t is the simulated traffic volume determined by the LSTM model in the tth time period, q t+1 is the simulated traffic volume determined by the LSTM model in the t+1 period; Q t is the artificial smoothed inflow flow in the tth time period, Q t+1 is the artificial smoothed inflow flow in the t+1 period, Q t and Q t+1 Used as a given flow rate; N is a non-zero positive integer or positive infinity.

10. The method according to claim 1, characterized in that The smoothed inbound flow calculation method further includes: Construct an objective function and solve it to obtain a calculation formula for the analytical inflow flow related to the real-time data of the reservoir; the real-time data of the reservoir includes the observed value of the real-time reservoir water storage and the outflow flow of the reservoir; The inflow is analyzed based on the observed values ​​of the real-time reservoir storage and the outflow from the reservoir.

11. A smooth inbound flow calculation system, characterized in that: The system includes a data collection module, a data processing module, a model generation module, an algorithm generation module and a model calculation module; The data collection module is used to collect reservoir hydrological data, process the reservoir hydrological data, and determine the original data; The data processing module is used to generate sample data based on the original data, construct a sample set, and divide the sample set into a training set and a test set; The model generation module is used to build a traffic prediction model, train the traffic prediction model using a training set, and determine the optimal parameter model; The algorithm generation module is used to build an analytical optimization algorithm and generate analytical inflow flow in combination with real-time reservoir data; The model calculation module is used to determine the analyzed flow data and the future forecast flow data, and use the optimal parameter model to generate simulated smooth flow based on the analyzed flow data and the future forecast flow data.

12. A computer-readable storage medium storing one or more programs, characterized in that: When the one or more programs are executed, the smoothed inflow flow calculation method described in any one of claims 1-10 can be implemented.

13. An electronic device comprising a processor, a communication interface, the computer-readable storage medium of claim 12, and a communication bus; wherein: The processor, the communication interface, and the computer-readable storage medium communicate with each other via a communication bus; It is characterized in that The processor is configured to execute a program stored in a computer-readable storage medium.

Images