Reservoir inflow smoothing method based on dynamic parameter adaptive extended Kalman filtering

Abstract

A reservoir inflow flow smoothing method based on dynamic parameter adaptive extended Kalman filtering is characterized by comprising the following steps: firstly, obtaining reservoir dam front and reservoir area water level processes, reservoir capacity curves and outflow flow data, and then performing smoothing calculation processing on the water level processes by adopting a median filtering method and a moving average method; then, on the basis of the smoothed water level process and ex-reservoir flow data, according to reservoir capacity characteristics, a water balance principle is adopted to calculate the in-reservoir flow, a Kalman filtering model is constructed, self-adaptive extended Kalman filtering calculation is carried out, and a process noise covariance matrix and an observation noise covariance matrix are dynamically updated; according to the method, the algorithm is reasonable, the dynamic change in the operation process of the reservoir can be responded in real time, data noise is accurately restrained, higher-precision flow estimation is achieved under the complex hydrological condition, and the method can well adapt to the dynamic change of the reservoir.

Description

TECHNICAL FIELD

[0001] The application relates to a reservoir inflow smoothing method, in particular to a reservoir inflow smoothing method based on dynamic parameter adaptive extended Kalman filtering, and belongs to the technical field of hydrological measurement. BACKGROUND

[0002] The reservoir inflow is a core quantitative index in the reservoir operation and management system, is key basic data for scientific allocation of water resources, guarantee of flood control safety in a basin and implementation of water ecosystem protection, and the accuracy thereof directly determines the rationality and effectiveness of reservoir dispatching decisions.

[0003] In current engineering practice, the conventional calculation method of the reservoir inflow mainly relies on the dam front water level data, and is specifically implemented through two paths: one is reverse calculation based on the water level-storage capacity curve, and the other is derivation according to the water balance principle. However, both the two traditional methods have significant technical limitations, resulting in that the calculation results are difficult to meet the high-precision application requirements, and the main problems are as follows: one is time scale limitation: the calculation process has strong dependence on the time period length, and cannot realize dynamic flow capture in a short time period; two is observation error transmission: a small error generated in the dam front water level observation process is amplified through the calculation model, and directly affects the accuracy of the flow result; three is dynamic storage capacity interference: the dynamic storage effect (such as dynamic adjustment of the storage capacity caused by the water level change) in the actual operation of the reservoir is not effectively included in the calculation model, resulting in that there is a deviation between the theoretical calculation value and the actual storage capacity.

[0004] The above problems jointly act, so that the process line of the inflow calculated by the traditional method presents obvious sawtooth fluctuation, the flow amplitude is severe, and the error has a cumulative effect, and the real inflow process of the reservoir cannot be accurately reflected. Especially in the flood period (the flow peak value needs to be accurately captured to guarantee the flood control safety) or the dry period (the discharge flow needs to be accurately controlled to guarantee the water resource supply), the calculation accuracy sharply decreases, and the accuracy and scientificity of the reservoir dispatching decision are seriously affected.

[0005] In order to solve the defects of the traditional method, in recent years, the emerging data assimilation technology and filtering algorithm are gradually applied to the field of reservoir inflow calculation. Such a method can theoretically improve the dynamic adaptability of flow calculation by introducing the dynamic model concept. However, the existing technology still has key bottlenecks: on the one hand, the algorithm mostly adopts fixed parameter setting, and cannot adaptively adjust the parameters according to the dynamic factors such as the climate conditions (such as changes of precipitation and evaporation) and the topographic features (such as differences of the basin confluence path) in the reservoir basin; on the other hand, the uncertainty processing mode of the measurement data (such as water level and rainfall observation data) is relatively rough, and a specific error correction mechanism is not established.

[0006] The technical bottleneck causes the existing data assimilation and filtering algorithm to have time lag or numerical deviation in the estimation result of the reservoir inflow in actual application, and it is difficult to meet the high precision and high timeliness requirement of the reservoir operation management on the flow data. Therefore, developing a reservoir inflow calculation method which can effectively suppress noise interference, smooth flow fluctuation and adaptively match the dynamic change of the reservoir basin has become an urgent practical requirement in the field of reservoir operation management. SUMMARY

[0007] The purpose of the present application is to overcome the defects and shortcomings of the conventional reservoir inflow calculation method, which mostly uses fixed parameter setting and cannot adapt to the dynamic change caused by factors such as climate and terrain of the reservoir basin, and the processing of measurement data uncertainty is also relatively simple and extensive. The estimation result obtained by using the conventional method often has lag or deviation, which leads to unsatisfactory effect in actual application and cannot meet the hydrological measurement requirement. The present application provides a reservoir inflow smoothing method based on dynamic parameter adaptive extended Kalman filtering, which has reasonable algorithm, can respond to the dynamic change in the reservoir operation process in real time, accurately suppresses data noise, realizes higher precision flow estimation under complex hydrological conditions, and can well adapt to the dynamic change of the reservoir.

[0008] To achieve the above-mentioned purpose of the application, the technical solution of the present application is as follows: a reservoir inflow smoothing method based on dynamic parameter adaptive extended Kalman filtering, characterized by comprising the following steps:

[0009] S1, first, the water level process before the dam and in the reservoir area, the reservoir capacity curve and the outflow data are obtained;

[0010] S2, then the median filtering method and the moving average method are used to perform smoothing calculation and processing on the water level process;

[0011] S3, then based on the smoothed water level process and the outflow data, the reservoir inflow is calculated according to the water balance principle based on the reservoir capacity characteristics;

[0012] S4, a Kalman filtering model is constructed, and the process noise covariance matrix and the observation noise covariance matrix are initialized;

[0013] S5, then the adaptive extended Kalman filtering calculation is performed, and the process noise covariance matrix and the observation noise covariance matrix are dynamically updated, and the time prediction and measurement update calculation are repeated to obtain the optimal estimation value of the reservoir inflow;

[0014] S6, finally, the result output and feedback optimization are performed.

[0015] Further, in the S1 step, the original observation data with a sampling interval of 5 minutes of the water level process before the dam and in the reservoir area of the target reservoir are collected, and the abnormal data are screened;

[0016] Collect the target reservoir storage curve or segmented storage curve to obtain the relationship between water level and storage capacity;

[0017] Collect the target reservoir outflow, including the sum of power generation, flood discharge, and water transfer engineering outflow, with a time period of 1 hour.

[0018] Further, the S3 step performs median filtering on the 5-minute water level process of each station along the reservoir area to remove noise, and the data sequence in the window is filtered , and the filtered value is

[0019]

[0020] In the formula: is the median filtered water level, n is the current time, k is the window size, is the original observed water level.

[0021] Then, the median filtered water level process is further processed by moving average processing by hour, and 4 decimal places are retained, and the calculation formula is:

[0022]

[0023] In the formula: is the moving average water level, t is the current time, n is the average moving step.

[0024] Further, the calculation method of the inflow in the S3 step is as follows:

[0025] For a lake-type reservoir, the static storage capacity method is used to calculate the inflow, and the formula is as follows:

[0026]

[0027] In the formula, and are the average inflow and outflow of the period, respectively; is the water loss of the reservoir, including water surface evaporation, seepage loss in the reservoir area, etc., is the change value of the reservoir storage capacity at the beginning and end of the period, which is calculated by interpolation of the water level-storage capacity curve; is the calculation period.

[0028] For a river-type reservoir, the dynamic storage capacity method is used to calculate the inflow, and the calculation steps of the total storage capacity are as follows:

[0029] (1) According to the setting of the water level stations along the reservoir area, the reservoir is divided into independent reservoir sections (a-b, b-c, c-d,...) between adjacent water level stations.

[0030] (2) Calculate the actual storage capacity of each reservoir section

[0031] For each reservoir section, use the average water level at both ends of the section as the representative water level, and then query the reservoir capacity curve for that section:

[0032]

[0033] (3) Total storage capacity = Sum of storage capacity of all storage sections

[0034]

[0035] Therefore, the time period storage capacity difference

[0036] (4) Calculate the instantaneous inflow rate based on the water balance method.

[0037]

[0038] In the formula, This is a series of hourly outbound flow data. This represents the difference in storage capacity within a unit of time period.

[0039] Furthermore, the method for calculating the inbound flow rate in step S4 is as follows:

[0040] Step S4a: Define the state vector to be estimated, where the parameter to be estimated is the reservoir inflow. (k) and its rate of change (k):

[0041]

[0042] Step S4b: Establish the state model:

[0043] =

[0044] in The state transition matrix describes the transition of the system state from... Time's up The time-transition relationship is determined based on the reservoir's hydrodynamic characteristics and time intervals. Let N be the process noise matrix, which follows the order N(0, 1) and ... respectively. ).

[0045] Step S4c: Establish the observation model:

[0046] Inflow rate calculated from water level-reservoir capacity curve The observation vector for the observed variable:

[0047]

[0048] The observation equation is established as follows:

[0049]

[0050] in The observation matrix is ​​obtained from the relationship between inflow and outflow rates in the water level-reservoir capacity curve. To measure noise and assess the uncertainty of the estimation, the expression follows N(0, ).

[0051] Furthermore, the specific method for step S5 is as follows:

[0052] Introducing adaptive factors and The process covariance matrix and the observation noise covariance matrix are dynamically updated using the following formula:

[0053]

[0054]

[0055] in and The calculation formula is as follows:

[0056]

[0057] In the formula: The trace of the prediction error covariance matrix (reflecting prediction uncertainty); Threshold (calibrated according to the actual system); : Scaling factor (default 1.0, used to adjust sensitivity)

[0058]

[0059] This is the innovation vector, representing the deviation between observation and prediction; The baseline variance of the observed noise is determined based on the reservoir's capacity characteristics; This is the attenuation coefficient, which is typically taken as 1;

[0060] The extended Kalman filter algorithm consists of two computational processes: time prediction and measurement update time prediction. The formulas for calculating inflow and covariance state prediction are as follows:

[0061]

[0062]

[0063] During the measurement update phase, the Kalman gain is calculated using the following formula:

[0064]

[0065] The modified Kalman gain is used for state variable updates to obtain the optimal estimator. :

[0066] = +

[0067] Simultaneously update the covariance. ,

[0068] By iteratively executing time prediction and measurement updates, and dynamically adjusting... This allows us to obtain the filtered results of the inflow rate over a time series.

[0069] Furthermore, the specific method for step S6 is as follows:

[0070] First, the inflow data calculated using the adaptive extended Kalman filter is... By comparing the inflow rate reported during the same period with that of a large reservoir, the smoothness of the smoothed correction curve of the inflow rate after Kalman filtering is calculated. The peak value and peak occurrence time difference of the two methods are compared to determine whether the smoothing effect meets the requirements.

[0071] The formula for calculating the smoothness index is as follows:

[0072]

[0073] In the formula:

[0074] QFi is the smoothed correction value of the inflow to the reservoir for time period i, in m³ / s, where i = 1, 2, ... (n-1).

[0075] Qi represents the flood season inflow value of reservoir i for time period i, in m³ / s, where i = 1, 2, … (n-1).

[0076] Esm—smoothness, with a value range of -1 to 1. The larger the value, the smoother the curve and the better the correction effect. When it is less than 0, it means that the smoothing effect of this method is worse than the current flood reporting method. When it is greater than 0, it means that it is better than the current flood reporting method.

[0077] relative error of flood peak (RE) f This indicator can clearly show the error change pattern of the flood peak before and after correction, preventing the flood peak from flattening. The closer its value is to 0, the better the flood peak information is preserved. The calculation formula is as follows:

[0078]

[0079] in This represents the maximum value among the smoothed inbound flow values. This represents the maximum inflow rate reported during the flood season.

[0080] Then, a parameter optimization database is created to store each parameter optimization step. , The adjusted values ​​reflect the corresponding function model and stochastic model indices; the least squares method is used to fit the data using this database. , With the initial prediction covariance matrix and observation noise variance Based on the functional relationship, the adaptive factor adjustment formula for a large reservoir is obtained, so that the initial state... , The settings are more targeted.

[0081] The beneficial effects of this invention are:

[0082] 1. In the water level processing method of this invention, the median filtering noise reduction is first performed on the 5-minute original water level process of each water level station in the reservoir area, and then the moving average processing is performed on it. Finally, the arithmetic average of the processed water level processes of each station is used as the reservoir water level, which effectively reduces the fluctuation error of using the water level of a single station in front of the dam as the reservoir water level in the previous inflow calculation.

[0083] 2. This invention preprocesses the acquired data, constructs a state-space model to comprehensively describe the dynamic characteristics of the reservoir inflow system, and uses a dynamic parameter adaptive extended Kalman filter to achieve smooth processing of the reservoir inflow, which can accurately restore the reservoir inflow process.

[0084] 3. The algorithm of this invention is reasonable and can respond to the dynamic changes in the reservoir operation process in real time, accurately suppress data noise, and achieve higher accuracy flow estimation under complex hydrological conditions. It well meets the requirements for high accuracy and high timeliness of flow data, and can provide more accurate data support for reservoir scheduling, flood and drought disaster prevention, and water resource allocation. Attached Figure Description

[0085] Figure 1 This is a flowchart of the present invention.

[0086] Figure 2 This is a schematic diagram of the static and dynamic storage capacity of the reservoir according to the present invention.

[0087] Figure 3 This is a comparison chart of the inflow process of a large reservoir in 2024 and its application before and after the flood in a certain embodiment of the present invention. Detailed Implementation

[0088] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0089] See Figures 1 to 3 The present invention provides a method for smoothing reservoir inflow based on dynamic parameter adaptive extended Kalman filtering, comprising the following steps:

[0090] S1. First, obtain data on water level processes, reservoir capacity curves, and outflow from the reservoir dam and reservoir area.

[0091] S2. Subsequently, the median filtering method and the moving average method were used to perform smoothing calculations on the water level process.

[0092] S3. Then, based on the smoothed water level process and outflow data, the inflow is calculated according to the reservoir capacity characteristics using the water balance principle.

[0093] S4. Construct a Kalman filter model, initialize the process noise covariance matrix, and observe the noise covariance matrix;

[0094] S5. Then perform adaptive extended Kalman filtering calculation, and dynamically update the process noise covariance matrix and observation noise covariance matrix. Repeat time prediction and measurement update calculation are performed to obtain the optimal estimate of the inflow rate.

[0095] S6. Finally, optimize the output and feedback of the results.

[0096] In step S1, the water level process in front of the target reservoir dam and in the reservoir area is collected, and raw observation data are sampled at 5-minute intervals, and abnormal data are screened.

[0097] Collect the reservoir capacity curve or segmented reservoir capacity curve of the target reservoir to obtain the relationship between water level and water storage.

[0098] Collect the outflow from the target reservoir, including the sum of outflows from power generation, flood discharge, and water diversion projects, for a period of 1 hour.

[0099] Step S3 involves median filtering and noise reduction of the 5-minute water level data at each station along the reservoir area, and then applying this to the window. Data sequence within The filtered value is

[0100]

[0101] In the formula: The water level is the result of median filtering, where n is the current time and k is the window size. This is the original observed water level.

[0102] Then, the water level process after median filtering is processed by hourly moving average, and four decimal places are retained. The calculation formula is as follows:

[0103]

[0104] In the formula: Let t be the water level after the moving average, t be the current time, and n be the average moving step size.

[0105] The method for calculating the inbound flow rate in step S3 is as follows:

[0106] For lake-type reservoirs, the inflow is calculated using the static storage capacity method, as shown in the following formula:

[0107]

[0108] In the formula, and These represent the average inbound and outbound flow rates for the respective time periods; Water loss from the reservoir includes losses from evaporation from the reservoir surface and seepage within the reservoir area. The change in reservoir water storage at the beginning and end of the time period is calculated by interpolation of the water level and reservoir capacity curve. This is the calculation period.

[0109] For river-type reservoirs, the inflow is calculated using the dynamic storage capacity method. The steps for calculating the total storage capacity using the dynamic storage capacity method are as follows:

[0110] (1) Based on the establishment of water level stations along the reservoir area, the reservoir is divided into independent reservoir sections (ab, bc, cd, ...) between adjacent water level stations.

[0111] (2) Calculate the actual storage capacity of each section of the reservoir.

[0112] For each reservoir section, use the average water level at both ends of the section as the representative water level, and then query the reservoir capacity curve for that section:

[0113]

[0114] (3) Total storage capacity = Sum of storage capacity of all storage sections

[0115]

[0116] Therefore, the time period storage capacity difference

[0117] (4) Calculate the instantaneous inflow rate based on the water balance method.

[0118]

[0119] In the formula, This is a series of hourly outbound flow data. This represents the difference in storage capacity within a unit of time period.

[0120] The method for calculating the inbound flow rate in step S4 is as follows:

[0121] Step S4a: Define the state vector to be estimated, where the parameter to be estimated is the reservoir inflow. (k) and its rate of change (k):

[0122]

[0123] Step S4b: Establish the state model:

[0124] =

[0125] in The state transition matrix describes the transition of the system state from... Time's up The time-transition relationship is determined based on the reservoir's hydrodynamic characteristics and time intervals. Let N be the process noise matrix, which follows the order N(0, 1) and ... respectively. ).

[0126] Step S4c: Establish the observation model:

[0127] Inflow rate calculated from water level-reservoir capacity curve The observation vector for the observed variable:

[0128]

[0129] The observation equation is established as follows:

[0130]

[0131] in The observation matrix is ​​obtained from the relationship between inflow and outflow rates in the water level-reservoir capacity curve. To measure noise and assess the uncertainty of the estimation, the expression follows N(0, ).

[0132] The specific method for step S5 is as follows:

[0133] Introducing adaptive factors and The process covariance matrix and the observation noise covariance matrix are dynamically updated using the following formula:

[0134]

[0135]

[0136] in and The calculation formula is as follows:

[0137]

[0138] In the formula: The trace of the prediction error covariance matrix (reflecting prediction uncertainty); Threshold (calibrated according to the actual system); : Scaling factor (default 1.0, used to adjust sensitivity)

[0139]

[0140] This is the innovation vector, representing the deviation between observation and prediction; The baseline variance of the observed noise is determined based on the reservoir's capacity characteristics; This is the attenuation coefficient, which is typically taken as 1;

[0141] The extended Kalman filter algorithm consists of two computational processes: time prediction and measurement update time prediction. The formulas for calculating inflow and covariance state prediction are as follows:

[0142]

[0143]

[0144] During the measurement update phase, the Kalman gain is calculated using the following formula:

[0145]

[0146] The modified Kalman gain is used for state variable updates to obtain the optimal estimator. :

[0147] = +

[0148] Simultaneously update the covariance. ,

[0149] By iteratively executing time prediction and measurement updates, and dynamically adjusting... This allows us to obtain the filtered results of the inflow rate over a time series.

[0150] The specific method for step S6 is as follows:

[0151] First, the inflow data calculated using the adaptive extended Kalman filter is... By comparing the inflow rate reported during the same period with that of a large reservoir, the smoothness of the smoothed correction curve of the inflow rate after Kalman filtering is calculated. The peak value and peak occurrence time difference of the two methods are compared to determine whether the smoothing effect meets the requirements.

[0152] The formula for calculating the smoothness index is as follows:

[0153]

[0154] In the formula:

[0155] QFi is the smoothed correction value of the inflow to the reservoir for time period i, in m³ / s, where i = 1, 2, ... (n-1).

[0156] Qi represents the flood season inflow value of reservoir i for time period i, in m³ / s, where i = 1, 2, … (n-1).

[0157] Esm—smoothness, with a value range of -1 to 1. The larger the value, the smoother the curve and the better the correction effect. When it is less than 0, it means that the smoothing effect of this method is worse than the current flood reporting method. When it is greater than 0, it means that it is better than the current flood reporting method.

[0158] relative error of flood peak (RE) f This indicator can clearly show the error change pattern of the flood peak before and after correction, preventing the flood peak from flattening. The closer its value is to 0, the better the flood peak information is preserved. The calculation formula is as follows:

[0159]

[0160] in This represents the maximum value among the smoothed inbound flow values. This represents the maximum inflow rate reported during the flood season.

[0161] Then, a parameter optimization database is created to store each parameter optimization step. , The adjusted values ​​reflect the corresponding function model and stochastic model indices; the least squares method is used to fit the data using this database. , With the initial prediction covariance matrix and observation noise variance Based on the functional relationship, the adaptive factor adjustment formula for a large reservoir is obtained, so that the initial state... , The settings are more targeted.

[0162] The following detailed description is provided in conjunction with specific embodiments:

[0163] In the specific implementation process, the hourly water level process of the reservoir is calculated by averaging the water levels of multiple water level stations. The storage volume is interpolated based on the water level and reservoir capacity curves, and the storage variable is calculated to inversely deduce the hourly flow process. When data from some stations is missing, the stations participating in the calculation can be removed as appropriate during the calculation process.

[0164] In this embodiment, the core idea of ​​median filtering is to use a sliding window to traverse the data, sort all the data within the window, and then take the median value to replace the center point value, thus eliminating outlier noise points. In actual production processes, because the inflow rate needs to be reported in real time, the water level median filtering calculation method is adjusted to forward calculation.

[0165] The inflow of the reservoir during the major flood events of 2024 was recalculated using multi-station water level filtering combined with dynamic parameter adaptive Kalman filtering. The inflow was then compared with the reported inflow provided by the reservoir management unit, as shown in the table below.

[0166] Calculation table of inflow smoothing index using Kalman method after multi-station water level filtering

[0167] Field flood Smoothness Peak (error) Peak difference Negative value 7.15~7.25 0.675 16100(2.4%) 0h 0 7.25~8.5 0.819 7430(-4.4%) 0h 0

[0168] As can be seen from the table above, the smoothness of the inflow calculated using multi-station water level filtering combined with the dynamic Kalman method is as high as 0.67 relative to the reported flood flow, while there is no significant deviation between the peak value and the peak occurrence. Furthermore, from the attached table... Figure 3 The comparison shows that the smoothness of the technical solution proposed in this invention is significantly higher than that of the flood reporting flow and the hourly calculated flow.

[0169] The above description is a further detailed explanation of the present invention in conjunction with specific embodiments. It should not be considered that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, any simple modifications and substitutions made without departing from the concept of the present invention should be considered to fall within the protection scope of the present invention.

Claims

1. A method for smoothing reservoir inflow based on dynamic parameter adaptive extended Kalman filtering, characterized in that... Includes the following steps: S1. First, obtain data on water level processes, reservoir capacity curves, and outflow from the reservoir dam and reservoir area. S2. Subsequently, the median filtering method and the moving average method were used to perform smoothing calculations on the water level process. S3. Then, based on the smoothed water level process and outflow data, the inflow is calculated according to the reservoir capacity characteristics using the water balance principle. S4. Construct a Kalman filter model, initialize the process noise covariance matrix, and observe the noise covariance matrix; S5. Then perform adaptive extended Kalman filtering calculation, and dynamically update the process noise covariance matrix and observation noise covariance matrix. Repeat time prediction and measurement update calculation are performed to obtain the optimal estimate of the inflow rate. S6. Finally, optimize the output and feedback of the results.

2. The method for smoothing reservoir inflow based on dynamic parameter adaptive extended Kalman filtering according to claim 1, characterized in that: In step S1, the water level process in front of the target reservoir dam and in the reservoir area is collected, and raw observation data are sampled at 5-minute intervals, and abnormal data are screened. Collect the reservoir capacity curve or segmented reservoir capacity curve of the target reservoir to obtain the relationship between water level and water storage. Collect the outflow from the target reservoir, including the sum of outflows from power generation, flood discharge, and water diversion projects, for a period of 1 hour.

3. The method for smoothing reservoir inflow based on dynamic parameter adaptive extended Kalman filtering according to claim 1, characterized in that: Step S3 involves median filtering and noise reduction of the 5-minute water level data at each station along the reservoir area, and then applying this to the window. Data sequence within The filtered value is In the formula: The water level is the result of median filtering, where n is the current time and k is the window size. This is the original observed water level. Then, the water level process after median filtering is processed by hourly moving average, and four decimal places are retained. The calculation formula is as follows: In the formula: Let t be the water level after the moving average, t be the current time, and n be the average moving step size. The method for calculating the inbound flow rate in step S3 is as follows: For lake-type reservoirs, the inflow is calculated using the static storage capacity method, as shown in the following formula: In the formula, and These represent the average inbound and outbound flow rates for the respective time periods; Water loss from the reservoir includes losses from evaporation from the reservoir surface and seepage within the reservoir area. The change in reservoir water storage at the beginning and end of the time period is calculated by interpolation of the water level and reservoir capacity curve. This is the calculation period. For river-type reservoirs, the inflow is calculated using the dynamic storage capacity method. The steps for calculating the total storage capacity using the dynamic storage capacity method are as follows: (1) Based on the establishment of water level stations along the reservoir area, the reservoir is divided into independent reservoir sections (ab, bc, cd, ...) between adjacent water level stations. (2) Calculate the actual storage capacity of each section of the reservoir. For each reservoir section, use the average water level at both ends of the section as the representative water level, and then query the reservoir capacity curve for that section: (3) Total storage capacity = Sum of storage capacity of all storage sections Therefore, the time period storage capacity difference (4) Calculate the instantaneous inflow rate based on the water balance method. In the formula, This is a series of hourly outbound flow data. This represents the difference in storage capacity within a unit of time period.

4. The method for smoothing reservoir inflow based on dynamic parameter adaptive extended Kalman filtering according to claim 1, characterized in that: The method for calculating the inbound flow rate in step S4 is as follows: Step S4a: Define the state vector to be estimated, where the parameter to be estimated is the reservoir inflow. (k) and its rate of change (k): Step S4b: Establish the state model: = in The state transition matrix describes the transition of the system state from... Time's up The time-transition relationship is determined based on the reservoir's hydrodynamic characteristics and time intervals. Let N be the process noise matrix, which follows the order N(0, 1) and ... respectively. ). Step S4c: Establish the observation model: Inflow rate calculated from water level-reservoir capacity curve The observation vector for the observed variable: The observation equation is established as follows: in The observation matrix is ​​obtained from the relationship between inflow and outflow rates in the water level-reservoir capacity curve. To measure noise and assess the uncertainty of the estimation, the expression follows N(0, ).

5. The method for smoothing reservoir inflow based on dynamic parameter adaptive extended Kalman filtering according to claim 1, characterized in that: The specific method for step S5 is as follows: Introducing adaptive factors and The process covariance matrix and the observation noise covariance matrix are dynamically updated using the following formula: in and The calculation formula is as follows: In the formula: The trace of the prediction error covariance matrix (reflecting prediction uncertainty); Threshold (calibrated according to the actual system); : Scaling factor (default 1.0, used to adjust sensitivity) This is the innovation vector, representing the deviation between observation and prediction; The baseline variance of the observed noise is determined based on the reservoir's capacity characteristics; This is the attenuation coefficient, which is typically taken as 1; The extended Kalman filter algorithm consists of two computational processes: time prediction and measurement update time prediction. The formulas for calculating inflow and covariance state prediction are as follows: During the measurement update phase, the Kalman gain is calculated using the following formula: The modified Kalman gain is used for state variable updates to obtain the optimal estimator. : = + Simultaneously update the covariance. , By iteratively executing time prediction and measurement updates, and dynamically adjusting... This allows us to obtain the filtered results of the inflow rate over a time series.

6. The method for smoothing reservoir inflow based on dynamic parameter adaptive extended Kalman filtering according to claim 1, characterized in that: The specific method for step S6 is as follows: First, the inflow data calculated using the adaptive extended Kalman filter is... By comparing the inflow rate reported during the same period with that of a large reservoir, the smoothness of the smoothed correction curve of the inflow rate after Kalman filtering is calculated. The peak value and peak occurrence time difference of the two methods are compared to determine whether the smoothing effect meets the requirements. The formula for calculating the smoothness index is as follows: In the formula: QFi is the smoothed correction value of the inflow rate to the reservoir in time period i, m3 / s, i=1,2…(n-1); Qi is the flood discharge value of reservoir i during the time period i, m3 / s, i=1,2 …(n-1); Esm—Smoothness, its value ranges from -1 to 1. The larger the value, the smoother the curve and the better the correction effect. When it is less than 0, it means that the smoothing effect of this method is worse than the current flood reporting method. When it is greater than 0, it means that it is better than the current flood reporting method. relative error of flood peak (RE) f This indicator can clearly show the error change pattern of the flood peak before and after correction, preventing the flood peak from flattening. The closer its value is to 0, the better the flood peak information is preserved. The calculation formula is as follows: in This represents the maximum value among the smoothed inbound flow values. This represents the maximum inflow rate reported during the flood season. Then, a parameter optimization database is created to store each parameter optimization step. , The adjusted values ​​reflect the corresponding function model and stochastic model indices; the least squares method is used to fit the data using this database. , With the initial prediction covariance matrix and observation noise variance Based on the functional relationship, the adaptive factor adjustment formula for a large reservoir is obtained, so that the initial state... , The settings are more targeted.

Images