A method for inbound flow estimation based on real-time error control

Abstract

The present invention discloses a method for inflow reverse calculation based on real-time error control, which comprises the following steps: first, collecting the reservoir water level and outflow data of each hour within 24 hours before the forecast time, and setting the initial value of the reverse calculation period to 1 hour; then, automatically calculating the reverse calculation period and the calculated inflow corresponding to the forecast time that meet the real-time error control requirements; finally, performing water volume correction on the calculated inflow process from 8:00 the day before the forecast time to the forecast time, and finally obtaining a corrected inflow process that meets the water volume balance constraint. The method can improve the sawtooth fluctuation problem of the inflow process, ensure the water volume balance of the inflow process, obtain an inflow process that is closer to the real process, thereby improving the accuracy of the hydrological forecast accuracy assessment, and providing a reliable basis for the evaluation of the forecast model results.

Description

Technical Field

[0001] The invention relates to the technical field of hydrological forecasting, and in particular to a method for inflow reverse estimation based on real-time error control. Background Art

[0002] Inflow is the "criterion" for calibrating the parameters of the hydrological model and the "yardstick" for evaluating the accuracy level of the hydrological forecasting scheme. At present, the most commonly used method for calculating inflow is the water balance method, which inverts the inflow based on data such as reservoir water level and outflow. The error in inflow calculation mainly comes from the amplification effect of reservoir water level record errors. When the reservoir capacity is large and the inversion period is short, even a small reservoir water level record error is enough to produce a large inflow calculation error, resulting in sawtooth fluctuations in the inflow process, and even unreasonable phenomena such as negative values. In order to overcome the sawtooth fluctuation phenomenon of inflow, the traditional method is to select a longer inversion period to reduce the amplification effect of reservoir water level record errors, but a long period will flatten the flood process, resulting in the phenomenon of equalization and flat head of the flood peak, which is difficult to reflect the flood peak information truthfully. Therefore, the selection of the inversion period has a significant impact on the shape of the inflow process and the flood peak. Selecting a reasonable inversion period is the key to ensuring the accuracy of the inflow process. Most real-time flood forecasting systems for reservoirs set the backpropagation period to a fixed 1-hour period, without considering whether the backpropagation period corresponding to the forecast time meets the requirements of real-time error control, resulting in a large deviation between the calculated inflow process and the actual flow process. Summary of the invention

[0003] The purpose of the present invention is to provide an inflow flow reverse estimation method based on real-time error control, to improve the sawtooth fluctuation phenomenon of the inflow flow process, to ensure the water balance of the inflow flow process, and to obtain an inflow flow process that is closer to the real process.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A method for inbound flow inference based on real-time error control comprises the following steps:

[0006] (1) Collect the reservoir water level and outflow flow data for each hour within 24 hours before the forecast time, and set the initial value of the reverse calculation period Δt to 1 hour.

[0007] (2) Using 1 hour before the forecast time as the initial calculation period, reverse the inflow flow I1 and calculate the Δt corresponding to the forecast time. 优 If at this time |Δt 优 -Δt|<Δt, then the current Δt meets the error control requirements, 1h and I1 are selected as the reverse period corresponding to the forecast time and the calculated inflow flow; if at this time |Δt 优-Δt|≥Δt, the current Δt does not meet the error control requirements, and Δt needs to be increased by 1h to 2h. Use 2h before the forecast time as the calculation period to reverse the inflow flow I2 and calculate the Δt corresponding to the forecast time at this time 优 If at this time |Δt 优 -Δt|<Δt, then the current Δt meets the error control requirements, 2h and I2 are selected as the reverse period corresponding to the forecast time and the calculated inflow flow; if at this time |Δt 优 -Δt|≥Δt, the current Δt does not meet the error control requirements, and Δt needs to be increased by 1h to 3h. Repeat the above trial calculation steps until the reverse period and calculated inflow flow corresponding to the current forecast time are found. Among them, Δt is an integer period between 1h and 24h. If the error control requirements are still not met when Δt is 24h, 24h and the corresponding calculated inflow flow are used as the final reverse period and calculated inflow flow.

[0008] (3) The period before the forecast time is divided into two periods (the front period and the back period) with 8:00 a.m. the day before the forecast time as the boundary. First, the front period is ensured to meet the water balance condition, and the calculated inflow flow process of the back period is corrected to meet the water balance constraint; then the outflow and storage change of the back period are counted and the real inflow water volume of the period is calculated. The corresponding calculated inflow water volume is obtained according to the calculated inflow flow process of the back period. The unbalanced water volume is recorded as the difference between the calculated inflow water volume and the real inflow water volume. Then, according to the proportion of the calculated inflow water volume of each hour in the back period in the calculated inflow water volume of the period, the unbalanced water volume is allocated to each hour according to a certain proportion until the water volume error is eliminated, and the corrected inflow flow process that meets the water balance constraint is obtained. From the perspective of flood control risk, if the unbalanced water volume value is positive, the unbalanced water volume is allocated in the inverse proportion of the water volume proportion of each hour; if the unbalanced water volume value is negative, the unbalanced water volume is allocated in the positive proportion of the water volume proportion of each hour.

[0009] In the above technical scheme, further, in the step (3), the time period before the forecast time is divided into two time periods before and after the forecast time with 8:00 of the day before the forecast time as the boundary, specifically: for a specific date, the forecast time is between 1:00 and 24:00 of the day, and the front and back time periods are divided with 8:00 of the day as the boundary; when the forecast time is transferred to 1:00 of the next day, the front and back time periods should be divided with 8:00 of the day as the boundary, and the inflow flow process from 8:00 of the day to 8:00 of the day needs to be corrected for water volume, that is, to ensure that the historical time period before 8:00 of the day meets the water balance condition.

[0010] The inventive principle of the present invention is:

[0011] The principle of the water balance method is the conservation of water volume in the reservoir during a certain period of time, that is, the difference between the total amount of water inflow and outflow in the reservoir during a certain period of time is equal to the change in the water storage volume of the reservoir during that period of time. This method uses the water balance equation to reverse the inflow flow based on the water level of the reservoir in front of the dam, the water level storage capacity curve, and the outflow flow. The equation is as follows:

[0012]

[0013] Where: Δt is the length of the backpropagation period, h; is the average inflow flow during the period, m 3 / s; is the average outbound flow rate during the period, m 3 / s; V1 and V2 are the storage capacities at the beginning and end of the time period, m 3 ; H1 and H2 are the reservoir water levels at the beginning and end of the time period, m; V(H1) and V(H2) indicate that V1 and V2 are obtained from the corresponding reservoir water levels H1 and H2 according to the water level and storage capacity curve. Suppose the reservoir water level difference at the beginning and end of the time period is ΔH, m; the average area of ​​the reservoir cross section between H1 and H2 is km 2 The water balance equation can be expressed as:

[0014]

[0015] Due to the current limitations of water level observation technology and the influence of wind and waves on the water surface, there is always a certain error in the reservoir water level recording data. Assuming that there is an error ζ in ΔH, the resulting water storage error W 误 for The corresponding inflow flow calculation error ΔQ is deduced as:

[0016]

[0017] From the above formula, it can be seen that ΔQ is affected by the amplification effect of the reservoir water level recording error, which is inversely proportional to Δt. In most reservoir real-time flood forecasting systems, Δt is 1h. When the reservoir surface area is large, even a small reservoir water level recording error is enough to produce a large inflow calculation error, which leads to the inflow process being prone to sawtooth fluctuations, and even unreasonable phenomena such as negative values. The choice of Δt has a significant impact on the shape of the inflow process and the flood peak, so choosing a reasonable Δt is the key to ensuring the accuracy of the inflow process.

[0018] How to select a reasonable Δt to reflect the actual inflow process involves calculating the optimal Δt value Δt under real-time error control. 优Most reservoir real-time flood forecasting systems set Δt to a fixed 1-hour period. In real-time operational forecasting, the inflow is calculated based on the 1-hour period before the forecast time. Whether the current Δt meets the error control requirements is not considered, resulting in a large deviation between the calculated inflow process and the actual process. In order to obtain a Δt that meets the error control requirements, it is necessary to calculate the Δt under error control. 优 Calculate the allowable error Q of the incoming flow according to the 5% flow test error standard requirements 允 :

[0019] Q 允 =0.05I

[0020] Where: I represents the inflow calculated by the current Δt. The minimum record of reservoir water level is 0.01m. Under normal circumstances, the reservoir water level record error ζ is within the range of 0 to 0.01m. The water storage error W caused by ζ being 0.01m 误 for:

[0021] W 误 =1000000×0.01A

[0022] Where: A is the water surface area corresponding to the current water level. 误 , Q 允 The relationship between and Δt is used to calculate Δt under real-time error control. 优 for:

[0023]

[0024] Get Δt 优 After that, compare |Δt 优 -Δt| and the size of Δt: If |Δt 优 -Δt|<Δt, then the current Δt meets the error control requirements, Δt and I are the reverse period and calculated inflow flow corresponding to the forecast time; if |Δt 优 -Δt|≥Δt, the current Δt does not meet the error control requirements, and Δt needs to be increased by 1h. Repeat the above trial calculation process until a reasonable reverse calculation period and calculated inflow flow are obtained.

[0025] The beneficial effects of the present invention are:

[0026] The method of the present invention constructs a dynamic adjustment method for the reverse period based on the relationship between the water storage error, the allowable error of the inflow flow and the optimized reverse period, based on the flow test error specification and the reservoir water level record error, and automatically calculates the reverse period and calculated inflow flow corresponding to the forecast time that meet the real-time error control requirements. By correcting the water volume of the calculated inflow flow process from 8:00 the day before the forecast time to the forecast time, the corrected inflow flow process that meets the water balance constraint is finally obtained. This method can effectively improve the sawtooth fluctuation phenomenon of the inflow flow process and overcome the defect that the traditional method cannot truthfully reflect the flood peak information, while ensuring the water balance of the inflow flow process, and obtaining an inflow flow process that is closer to the real process, thereby improving the accuracy of the hydrological forecast accuracy assessment and providing a reliable basis for the evaluation of the forecast model results. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flow chart of the inbound flow reverse estimation method based on real-time error control of the present application;

[0028] Figure 2 It is the original inflow flow process from 8:00 on June 1 to 13:00 on October 18;

[0029] Figure 3 It is the original inflow flow process from 8 / 25 8:00 to 8 / 26 20:00;

[0030] Figure 4 It is the process of calculating the inbound flow from 8 / 25 8:00 to 8 / 26 20:00;

[0031] Figure 5 It is the process of correcting the inbound flow from 8 / 25 8:00 to 8 / 26 20:00;

[0032] Figure 6 This is the process of correcting the incoming flow from 8:00 on June 1 to 13:00 on October 18. DETAILED DESCRIPTION

[0033] The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0034] A method for inbound flow inference based on real-time error control, the calculation method is as follows:

[0035] Jinping I Hydropower Station is located in Yanyuan County and Muli County of Liangshan Yi Autonomous Prefecture, Sichuan Province. It is a control reservoir cascade power station in the lower reaches of the Yalong River (from Kala to Jiangkou), about 358 kilometers from the river mouth. The basin area above the dam site is 103,000 km 2 , accounting for 75.4% of the Yalong River basin area. The average annual flow at the dam site is 1220m 3 / s, with an average annual runoff of 38.5 billion m 3 The reservoir water level, outflow flow, and water level-storage capacity relationship data were collected from 8:00 on June 1, 2021 to 13:00 on October 18, 2021. The inflow flow data at each forecast time was collected in the real-time flood forecasting system (the reverse period was 1 hour) and recorded as the original inflow flow, such as Figure 2 As shown in the figure, the original inflow flow process shows obvious sawtooth fluctuation phenomenon. Figure 3 It shows the original inflow flow process from 8:00 on 8 / 25 to 20:00 on 8 / 26, and shows that the peak flow process fluctuated greatly from 17:00 on 8 / 26 to 20:00 on 8 / 26.

[0036] An example of inbound flow inference based on error control is as follows:

[0037] (1) Calculate the reverse time period and inflow flow corresponding to each forecast time. The current forecast time is 8 / 26 20:00, and the corresponding original inflow flow I1 is 1489m 3 / s, calculate the corresponding optimized back-stepping period Δt 优 About 2.42h. At this time |Δt 优 -Δt|≥Δt, the current Δt does not meet the error control requirements, and Δt needs to be increased by 1h to 2h. Taking 2h before the forecast time as the calculation period, the inflow flow I2 is reversed to 6598m 3 / s, calculate the corresponding optimized back-stepping period Δt 优 About 0.55h. At this time |Δt 优 -Δt|<Δt, the current Δt meets the error control requirements, 2h and I2 are defined as the current reverse thrust period and calculated inflow flow, respectively.

[0038] Figure 4 It represents the calculated inflow flow process from 8:00 on August 25 to 20:00 on August 26. Compared with the original inflow flow process, the calculated inflow flow process effectively retains the peak information while reducing the abnormal fluctuation of the flow process before and after the peak, making the inflow flow process more stable and effectively improving the sawtooth fluctuation phenomenon of the original inflow flow process.

[0039] (2) The water volume correction is performed on the calculation process of the inflow flow from 8:00 the day before the forecast time to the forecast time. The current forecast time is 8 / 26 20:00. The period before and after 8 / 25 8:00 is divided into two periods (6 / 1 0:00~8 / 25 8:00, 8 / 25 8:00~8 / 26 20:00). On the basis that the water volume balance condition is met in the former period, the outflow volume and water storage change in the latter period are counted and added up to calculate the actual inflow volume of the period, which is 415144800m 3According to the calculation process of the inflow flow in the later period, the corresponding calculated inflow water volume is 437542200m 3 , the calculated unbalanced water volume is 22397400m 3 The unbalanced water volume is apportioned according to the inverse proportional relationship of the water volume proportion in each hour during the period from 8:00 on August 25 to 20:00 on August 26, and the modified inflow flow process that meets the water balance constraint during this period is obtained, such as Figure 5 shown.

[0040] (3) According to the time sequence, refer to the above calculation steps to calculate the modified inflow flow process during the period of 6 / 1 8:00 to 10 / 18 13:00, as shown in the following figure: Figure 6 shown.

[0041] Figure 6 It shows that the modified inflow process can effectively improve the sawtooth fluctuation phenomenon of the original inflow process, while meeting the water balance conditions of the inflow process, which is closer to the real process.

[0042] Although the present invention has been described in detail in general and in specific embodiments in the specification of the present invention, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all fall within the scope of protection claimed by the present invention.

Claims

1. A method for inbound flow inference based on real-time error control, characterized in that: The following steps are involved: (1) Collect the reservoir water level and outflow data for each hour within 24 hours before the forecast time, and set the initial value of the reverse calculation period Δt to 1 hour; (2) Using 1 hour before the forecast time as the initial calculation period, reverse the inflow flow I1 and calculate the Δt corresponding to the forecast time. 优 , Among them, I represents the inflow calculated at the current Δt, and A is the water surface area corresponding to the current water level; if |Δt 优 -Δt|<Δt, then the current Δt meets the error control requirements, 1h and I1 are selected as the reverse period corresponding to the forecast time and the calculated inflow flow; if at this time |Δt 优 -Δt|≥Δt, the current Δt does not meet the error control requirements, and Δt needs to be increased by 1h; use 2h before the forecast time as the calculation period to reverse the inflow flow I2 and calculate the Δt corresponding to the forecast time at this time 优 ; If at this time |Δt 优 -Δt|<Δt, then the current Δt meets the error control requirements, 2h and I2 are selected as the reverse period corresponding to the forecast time and the calculated inflow flow; if at this time |Δt 优 -Δt|≥Δt, the current Δt does not meet the error control requirements, and Δt needs to be increased by 1h; repeat the above steps until the reverse calculation period and calculated inflow flow corresponding to the current forecast time are found; Δt is an integer period between 1h and 24h. If the error control requirements are still not met when Δt is 24h, 24h and the corresponding calculated inflow flow are used as the final reverse calculation period and calculated inflow flow; (3) The period before the forecast time is divided into two periods, the former and the latter, with 8:00 a.m. on the day before the forecast time as the boundary. First, the former period is ensured to meet the water balance condition, and the water volume correction is performed on the calculated inflow process of the latter period to meet the water balance constraint; Then, the outflow and storage changes in the later period are counted and the actual inflow of water in this period is calculated. The corresponding calculated inflow of water is obtained according to the calculated inflow flow process of the later period, and the unbalanced water is recorded as the difference between the calculated inflow of water and the actual inflow of water. Then, according to the proportion of the calculated inflow of water in each hour in the later period in the calculated inflow of water in the later period, the unbalanced water is allocated to each hour according to a certain proportion until the water volume error is eliminated, and the corrected inflow flow process that meets the water balance constraint is obtained. From the perspective of flood control risk, if the unbalanced water value is positive, the unbalanced water is allocated in the inverse proportional relationship of the proportion of water volume in each hour; if the unbalanced water value is negative, the unbalanced water is allocated in the positive proportional relationship of the proportion of water volume in each hour.

2. The method for inbound flow inference based on real-time error control according to claim 1 is characterized in that: In the step (3), the time period before the forecast time is divided into two time periods, namely, when the forecast time is between 1:00 and 24:00 on the day, the time period before and after ...

Images