Long, medium and short term nested cascade hydropower station forecast scheduling system

Abstract

The invention provides a long-medium-short-term nested cascade hydropower station forecasting and dispatching system. The system comprises a long-medium-short-term nested power generation optimal dispatching module and a reservoir inflow forecasting module. The long-medium-short-term nested power generation optimization scheduling module comprises a cascade hydropower station long-term optimization scheduling module, a cascade hydropower station medium-term optimization scheduling module, a cascade hydropower station short-term optimization scheduling module, a cascade hydropower station in-plant economic operation module and a scheduling model heuristic optimization algorithm module; the reservoir inflow forecasting module comprises a monthly inflow forecasting module based on a self-made probability distribution model, a daily inflow LSTM forecasting module based on multi-dimensional feature data driving and an hourly inflow forecasting module. An economic operation module is embedded, operation of cascade hydropower stations is optimized, and efficient management of time and space scales is achieved; a novel algorithm is adopted to correct an initial flow value, water balance and final water level constraint are met, and optimization efficiency is improved; the water inflow is predicted through data driving, model calculation is simplified, and the forecast scheduling efficiency is improved.

Description

Technical Field

[0001] The invention relates to the technical field of hydropower generation, and in particular to a long-, medium- and short-term nested forecasting and dispatching system for cascade hydropower stations. Background Art

[0002] Affected by random factors such as basin meteorology and hydrology, the optimal operation and management of cascaded small hydropower groups are becoming increasingly complex. Considering the influence of uncertain factors such as water inflow and load, relevant scholars have made many achievements in the field of reservoir random optimal scheduling and forecast scheduling from the two aspects of scheduling information utilization and optimization scheduling model construction. Common methods include explicit random optimal scheduling, implicit random optimal scheduling, and hierarchical nested optimal scheduling.

[0003] In actual engineering applications, optimal dispatching must not only consider the runoff changes in the time period, take into account the short-term benefits, but also comprehensively consider the long-term evolution of runoff to improve long-term benefits. In addition to random runoff, uncertain factors such as load fluctuations also have a huge impact on dispatching decisions. Therefore, the actual operation process of the reservoir is a rolling optimization decision-making process of "forecasting, decision-making, implementation, re-forecasting, re-decision-making, and re-implementation". The existing explicit random and implicit random optimization dispatching methods are not enough to meet the actual needs of reservoir optimization operation and management, while the hierarchical nested optimization dispatching technology can make rolling optimization decisions and better guide the actual operation of the reservoir. However, the existing hierarchical nested optimization dispatching system needs to be organically combined with the economic operation module within the plant. Summary of the invention

[0004] The main purpose of the present invention is to provide a long-, medium- and short-term nested cascade hydropower station forecasting and dispatching system in response to the above-mentioned problems.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A long-, medium- and short-term nested cascade hydropower station forecasting and dispatching system, comprising a long-, medium- and short-term nested power generation optimization dispatching module and a reservoir water inflow forecasting module;

[0007] The long-, medium- and short-term nested power generation optimization scheduling module includes: a long-term optimization scheduling module for cascade hydropower stations, a medium-term optimization scheduling module for cascade hydropower stations, a short-term optimization scheduling module for cascade hydropower stations, an in-plant economic operation module for cascade hydropower stations and a scheduling model heuristic optimization algorithm module;

[0008] The reservoir water inflow forecast module includes: a monthly water inflow forecast module based on a self-made probability distribution model, a daily water inflow LSTM forecast module driven by multi-dimensional feature data, and an hourly water inflow forecast module.

[0009] While adopting the above technical solutions, the present invention may also adopt or combine the following technical solutions:

[0010] As a preferred technical solution of the present invention: the operation of the long-, medium- and short-term nested power generation optimization scheduling module includes the following steps:

[0011] S1: Initialize the target final water level of the 12th month of the kth year and use it as the final water level boundary constraint of the long-term optimal scheduling module of the cascade hydropower stations, k = current year, l = 1;

[0012] S2: Execute the long-term optimization scheduling module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, and obtain the corrected optimal long-term scheduling remaining period reference flow, as well as the end-of-month water levels of all months within the long-term scheduling remaining period, that is, the end-of-month water levels of the reservoir from the first month of the kth year to the 12th month of the kth year. The objective function of the long-term optimization scheduling module of the cascade hydropower station is:

[0013]

[0014] Where, T y is the remaining period of long-term scheduling, that is, from the current month to the 12th month, Δt y is the length of the long-term scheduling period, i.e., 1 month, n s and n re are the total number of cascade hydropower stations and reservoirs, respectively, i t , are the total power output and power station reference flow of the i-th power station in the t-th period, For the kth re The discharge flow of a reservoir in period t;

[0015] S3: The water level at the end of the first month of the long-term scheduling remaining period, that is, the reservoir water level at the end of the month of the kth year, is used as the final water level boundary constraint of the mid-term optimization scheduling module of the cascade hydropower stations;

[0016] S4: Initialize m=1;

[0017] S5: Execute the mid-term optimization scheduling module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, and obtain the corrected optimal mid-term scheduling remaining period reference flow, and all end-of-day water levels within the mid-term scheduling remaining period, that is, the end-of-day water levels of the reservoir from the mth day of the lth month of the kth year to the last day of the lth month of the kth year. The objective function of the mid-term optimization scheduling module of the cascade hydropower station is:

[0018]

[0019] Where, T m is the remaining period of mid-term scheduling, that is, from the current day to the end of the month, Δt m The duration of the medium-term scheduling period is 1 day;

[0020] S6: The final water level of the first day of the remaining period of the mid-term optimal dispatch, i.e., the final water level of the reservoir on the mth day of the lth month of the kth year, is used as the final water level boundary constraint of the short-term optimal dispatch module of the cascade hydropower stations;

[0021] S7: Execute the short-term optimization scheduling module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, and obtain the corrected optimal daily reference flow, daily reference flow plan, daily output plan, and all hourly water levels within the day, that is, the hourly water levels of the reservoir from 1:00 on the mth day of the lth month of the kth year to 24:00 on the mth day of the lth month of the kth year. The objective function of the short-term optimization scheduling module of the cascade hydropower station is:

[0022]

[0023] Where, T d is the short-term scheduling period, i.e. 1 day, Δt d is the duration of the short-term scheduling period, i.e. 1 hour, c t is the electricity price for the t-th period;

[0024] S8: Initialize n=1;

[0025] S9: Determine whether to adopt the "electricity determines water" method. If yes, execute the in-plant economic operation module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, carry out the "electricity determines water" in-plant economic operation of the cascade hydropower station, and obtain the optimal unit output. If no, execute the in-plant economic operation module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, carry out the "water determines electricity" in-plant economic operation of the cascade hydropower station, and obtain the optimal unit reference flow. If the "electricity determines water" method is adopted, the objective function of the in-plant economic operation module of the cascade hydropower station is:

[0026]

[0027] If the “water determines electricity” method is adopted, the objective function of the economic operation module in the cascade hydropower station is:

[0028]

[0029] In the formula, q ij , P ij are the unit reference flow and unit output of the jth unit in the i-th power station, Δt is the duration of the current period, i.e., 1 hour, n u is the total number of units in the power station;

[0030] S10: Check if n≥24? If yes, proceed to the next step; if no, go to S14;

[0031] S11: Check if m ≥ the total number of days in the first month. If yes, proceed to the next step; if no, go to S15;

[0032] S12: Check if l≥12? If yes, proceed to the next step; if no, go to S16;

[0033] S13: Determine whether to end, if yes, end the process, if not, go to S17;

[0034] S14: Determine whether the time is a new hour. If yes, n=n+1 and go to S9. If no, continue to determine the time.

[0035] S15: Determine whether the time is a new day, if yes, m=m+1 and go to S5, if no, continue to determine the time;

[0036] S16: Determine whether the time is a new month, if yes, then l=l+1 and go to S2, if no, continue to determine the time;

[0037] S17: Determine whether the time is a new year, if yes, go to S1, if no, continue to determine the time.

[0038] As a preferred technical solution of the present invention: the daily reference flow rate plan = the corrected optimal daily reference flow rate, and the daily output plan = the power station comprehensive output coefficient * the corrected optimal daily reference flow rate * the water head.

[0039] As a preferred technical solution of the present invention: the long-term scheduling remaining period reference flow population correction value, the medium-term scheduling remaining period reference flow population correction value and the daily reference flow population correction value, as well as the corrected optimal long-term scheduling remaining period reference flow, the corrected optimal medium-term scheduling remaining period reference flow and the corrected optimal daily reference flow are all obtained by a forward reservoir water level calculation method containing a reverse reference flow correction algorithm and include the following steps:

[0040] S101: Initialize the upper limit V of the reservoir capacity max , Reservoir capacity lower limit V min , Reference flow limit Q e_max , the upper limit of the reference flow Q in the first period e1_max and the total number of time periods NT, the upper limit of the reference flow Q e1_max The expression is:

[0041] Q e1_max =[SL L (1)-(V min -V b (1))] / DT(1)-Q s (1) (6)

[0042] In the formula, SL L (1) V b (1), DT(1) and Q s(1) are the reservoir inflow in the first period, the initial storage capacity of the reservoir period, the period duration and the ecological flow;

[0043] S102: Determine the reference flow Q in the first period e (1)>Q e1_max ? Yes, then Q e (1) = Q e1_max , and execute the next step, otherwise, directly execute the next step;

[0044] S103: Initialize t k =1;

[0045] S104: Determine t k <NT-1? If yes, proceed to the next step, otherwise execute the reverse reference flow correction algorithm and go to S108;

[0046] S105: Determine V b (t k )+SL L (t k )-(Q e (t k )+Q s (t k ))*DT(t k )>V max ? Yes, then

[0047]

[0048] Go to S107, if not, go to the next step;

[0049] Where V b (t k ), V e (t k ), V x (t k ), SL L (t k ), Q e (t k ), Q s (t k ) and DT(t k ) are respectively k The reservoir's initial storage capacity for the time period, the reservoir's storage capacity at the end of the time period, the reservoir's discharge, the reservoir's incoming water, the reference flow, the ecological flow and the time period duration;

[0050] S106: Determine V b (t k )+SL L (t k )-(Q e (t k)+Q s (t k ))*DT(t k )<V min ? Yes, then V e (t k )=V min , execute the reverse reference flow correction algorithm and proceed to the next step, otherwise,

[0051]

[0052] And execute the next step;

[0053] S107: No. k +1 period reservoir initial water level V b (t k +1)=V e (t k ), t k =t k +1, and transfer to S104;

[0054] S108: The final storage capacity V of the reservoir at all time periods is calculated by the water level-storage capacity relationship curve. e Convert to the final water level of the reservoir in all time periods and end the process.

[0055] As a preferred technical solution of the present invention: the reverse reference flow correction algorithm includes the following steps:

[0056] S111: Initialization t k2 =t k ;

[0057] S112: Determine t k2 ≥1? If yes, proceed to the next step; if no, go to S117.

[0058] S113: Determine V e (t k2 )-V b (t k2 )<SL L (t k2 )-Q s (t k2 )*DT(t k2 )? Yes, then

[0059]

[0060] Go to S117, if not, go to the next step;

[0061] Where V b (t k2 ), V e (t k2), V x (t k2 ), SL L (t k2 ), Q e (t k2 ), Q s (t k2 ) and DT(t k2 ) are respectively k2 The reservoir's initial storage capacity for the time period, the reservoir's storage capacity at the end of the time period, the reservoir's discharge, the reservoir's incoming water, the reference flow, the ecological flow and the time period duration;

[0062] S114: Determine t k2 =1? If yes, calculate the initial water level V of the reservoir in the first period according to the following formula: b1

[0063]

[0064] And execute the next step, if not, go to S116;

[0065] S115: Judgement | V b (t k2 )-V b1 | / V b (t k2 )<1E-5? Yes, then

[0066]

[0067] Go to S117. If not, it is an error, the reference flow cannot be corrected, and go to S117.

[0068]

[0069] t k2 =t k2 -1, and transfer to S112;

[0070] Where V e (t k2 -1) indicates the tth k2 -1 The reservoir water level at the end of the period;

[0071] S117: The reverse reference flow correction is terminated, and the process ends.

[0072] As a preferred technical solution of the present invention: the operation of the monthly water volume prediction module based on the self-made probability distribution model includes the following steps:

[0073] S201: Determine whether the time is a new year, if yes, proceed to the next step, if no, continue to determine the time;

[0074] S202: Count the monthly rainfall from January to December in the previous years before the current year;

[0075] S203: Calculate the rainfall probability distribution model from January to December respectively, where the rainfall probability distribution model includes the probability density function and probability distribution function of the monthly rainfall;

[0076] S204: Calculate the mathematical expectations of the rainfall probability distribution model from January to December respectively, and use them as the monthly rainfall prediction values ​​of the corresponding months of the current year;

[0077] S205: Calculate the 90% confidence interval of the rainfall probability distribution model from January to December respectively, and use it as the monthly rainfall prediction interval of the corresponding month of the current year;

[0078] S206: Calculate the monthly water inflow forecast value and monthly water inflow forecast interval for the corresponding month of the current year of the reservoir through the monthly water inflow-monthly rainfall relationship curve;

[0079] S207: Determine whether to end, if yes, end the process, if not, go to S201.

[0080] As a preferred technical solution of the present invention: the daily water volume LSTM prediction module driven by multi-dimensional feature data is executed once at the beginning of each month.

[0081] As a preferred technical solution of the present invention: the hourly water inflow prediction module is executed once at the beginning of each day.

[0082] The present invention provides a long-, medium- and short-term nested forecast and dispatch system for cascade hydropower stations, which has the following beneficial effects: an in-plant economic operation module is embedded in the forecast and dispatch, and efficient optimization of the operation of cascade hydropower stations on a monthly to hourly time scale and a power station to unit spatial scale is achieved; in the reference flow optimization process of the forecast and dispatch, a forward reservoir water level calculation method containing a reverse reference flow correction algorithm is used to correct the initial reference flow value randomly generated by the heuristic algorithm, so that the corrected reference flow can simultaneously meet the water balance constraint and the final water level boundary constraint, thereby effectively reducing the possibility of the occurrence of solutions that do not meet the constraints and improving the optimization efficiency; a data-driven method is used to predict the water inflow for the next year, month and day, and while ensuring the forecast accuracy, there is no need to construct a complex hydrological model and carry out water inflow process evolution calculation, thereby improving the implementation efficiency of the forecast and dispatch. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 This is a schematic diagram of the long-, medium- and short-term nested cascade hydropower station forecasting and dispatching system provided by the present invention.

[0084] Figure 2 This is the operation flow chart of the long-, medium- and short-term nested power generation optimization scheduling module.

[0085] Figure 3 Flow chart of the forward reservoir water level calculation method including the reverse reference flow correction algorithm.

[0086] Figure 4 Flowchart of the reverse reference traffic correction algorithm.

[0087] Figure 5 This is the operation flow chart of the monthly water volume prediction module based on the self-made probability distribution model.

[0088] Figure 6 This is a schematic diagram of the hydraulic connections of the Xiecunyuan cascade hydropower stations.

[0089] Figure 7 This is the relationship curve between monthly water inflow and monthly rainfall of Xiecunyuan Secondary Reservoir.

[0090] Figure 8 This is the predicted result of monthly rainfall in Xiecunyuan Secondary Reservoir.

[0091] Fig. 9 This is the forecast result of monthly water inflow to Xiecunyuan Secondary Reservoir.

[0092] Fig.10 This is the predicted result of daily water inflow to Xiecunyuan Secondary Reservoir.

[0093] Fig.11 This is the output curve of Unit 1 of Xiecunyuan Second-Level Hydropower Station before and after optimization.

[0094] Fig.12 This is the output curve of Unit 2 of Xiecunyuan Second-Level Hydropower Station before and after optimization.

[0095] Fig.13 This is the output curve of Unit 1 of Xiecunyuan Level 3 Hydropower Station before and after optimization.

[0096] Fig.14 This is the output curve of Unit 2 of Xiecunyuan Level 3 Hydropower Station before and after optimization.

[0097] In the figure: 1-long-, medium- and short-term nested power generation optimization scheduling module; 2-reservoir water inflow forecast module; 11-long-term optimization scheduling module of cascade hydropower stations; 12-medium-term optimization scheduling module of cascade hydropower stations; 13-short-term optimization scheduling module of cascade hydropower stations; 14-in-plant economic operation module of cascade hydropower stations; 15-scheduling model heuristic optimization algorithm module; 21-monthly water inflow prediction module based on self-made probability distribution model; 22-daily water inflow LSTM prediction module driven by multi-dimensional feature data; 23-hourly water inflow prediction module. DETAILED DESCRIPTION

[0098] The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

[0099] like Figure 1 As shown, a long-, medium- and short-term nested cascade hydropower station forecast dispatching system includes a long-, medium- and short-term nested power generation optimization dispatching module 1 and a reservoir water inflow forecasting module 2;

[0100] The long-, medium- and short-term nested power generation optimization scheduling module 1 includes: a long-term optimization scheduling module 11 for cascade hydropower stations, a medium-term optimization scheduling module 12 for cascade hydropower stations, a short-term optimization scheduling module 13 for cascade hydropower stations, an in-plant economic operation module 14 for cascade hydropower stations and a scheduling model heuristic optimization algorithm module 15;

[0101] The reservoir water inflow forecast module 2 includes: a monthly water inflow forecast module 21 based on a self-made probability distribution model, a daily water inflow LSTM forecast module 22 driven by multi-dimensional feature data, and an hourly water inflow forecast module 23.

[0102] The data interaction relationship between modules is as follows:

[0103] The long-term optimization scheduling module of cascade hydropower stations obtains the monthly water inflow forecast value from the monthly water inflow forecast module based on the self-made probability distribution model as input conditions, calculates the correction value of the long-term scheduling remaining period reference flow population according to the original value of the long-term scheduling remaining period reference flow population, and then calculates the long-term scheduling remaining period objective function value and inputs it into the scheduling model heuristic optimization algorithm module. The scheduling model heuristic optimization algorithm module iteratively updates the new original value of the long-term scheduling remaining period reference flow population, and returns it to the long-term optimization scheduling module of cascade hydropower stations for a new round of calculations, iterates repeatedly until the maximum number of iterations, and returns the optimal solution of the final generation of population to the long-term optimization scheduling module of cascade hydropower stations.

[0104] The medium-term optimization scheduling module of cascade hydropower stations obtains the daily water volume forecast value from the daily water volume LSTM prediction module driven by multi-dimensional feature data as input conditions, obtains the water level at the end of the first month of the long-term scheduling remaining period from the long-term optimization scheduling module of cascade hydropower stations as the boundary constraint of the final water level, calculates the correction value of the medium-term scheduling remaining period reference flow population according to the original value of the medium-term scheduling remaining period reference flow population, and then calculates the medium-term scheduling remaining period objective function value and inputs it into the scheduling model heuristic optimization algorithm module. The scheduling model heuristic optimization algorithm module iteratively updates the new original value of the medium-term scheduling remaining period reference flow population, and returns it to the medium-term optimization scheduling module of cascade hydropower stations for a new round of calculations, iterates repeatedly until the maximum number of iterations, and sends the optimal solution of the final generation of population to the medium-term optimization scheduling module of cascade hydropower stations.

[0105] The short-term optimization scheduling module of cascade hydropower stations obtains the hourly water inflow forecast value from the hourly water inflow forecast module as input conditions, obtains the final water level of the first day of the mid-term scheduling remaining period from the mid-term optimization scheduling module of cascade hydropower stations as the final water level boundary constraint, calculates the daily reference flow population correction value based on the original value of the daily reference flow population, and then calculates the daily objective function value and inputs it into the scheduling model heuristic optimization algorithm module. The scheduling model heuristic optimization algorithm module iteratively updates the new original value of the daily reference flow population, and returns it to the short-term optimization scheduling module of cascade hydropower stations for a new round of calculations, iterates repeatedly until the maximum number of iterations, and sends the optimal solution of the final generation of population to the short-term optimization scheduling module of cascade hydropower stations.

[0106] The in-plant economic operation module of the cascade hydropower station obtains the daily output plan and the daily reference flow plan from the short-term optimization scheduling module of the cascade hydropower station as input conditions, calculates the in-plant economic operation objective function value according to the unit output population or the unit reference flow population, and inputs it into the scheduling model heuristic optimization algorithm module. The scheduling model heuristic optimization algorithm module iteratively updates the new unit output population or the unit reference flow population, and returns it to the in-plant economic operation module of the cascade hydropower station for a new round of calculations, iterates repeatedly until the maximum number of iterations, and sends the optimal solution of the final generation of population to the in-plant economic operation module of the cascade hydropower station.

[0107] The hourly water inflow prediction module obtains the daily water inflow prediction value from the daily water inflow LSTM prediction module driven by multi-dimensional feature data to calculate the hourly water inflow prediction value.

[0108] like Figure 2 As shown in the figure, the operation of the long-, medium- and short-term nested power generation optimization scheduling module includes the following steps:

[0109] S1: Initialize the target final water level of the 12th month of the kth year and use it as the final water level boundary constraint of the long-term optimal scheduling module of the cascade hydropower stations, k = current year, l = 1;

[0110] S2: Execute the long-term optimal dispatching module of cascade hydropower stations, and call the dispatching model heuristic optimization algorithm module for optimization, and obtain the corrected optimal long-term dispatching remaining period reference flow, as well as the end-of-month water levels of all months within the long-term dispatching remaining period, that is, the end-of-month water levels of the reservoir from the first month of the kth year to the 12th month of the kth year. The objective function of the long-term optimal dispatching module of cascade hydropower stations is:

[0111]

[0112] Where, T y is the remaining period of long-term scheduling, that is, from the current month to the 12th month, Δt y is the length of the long-term scheduling period, i.e., 1 month, n s and n reare the total number of cascade hydropower stations and reservoirs, respectively, i t , are the total power output and power station reference flow of the i-th power station in the t-th period, For the kth re The discharge flow of a reservoir in period t;

[0113] S3: The water level at the end of the first month of the long-term scheduling remaining period, that is, the reservoir water level at the end of the month of the kth year, is used as the final water level boundary constraint of the mid-term optimization scheduling module of the cascade hydropower stations;

[0114] S4: Initialize m=1;

[0115] S5: Execute the mid-term optimization scheduling module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, and obtain the corrected optimal mid-term scheduling remaining period reference flow, as well as all end-of-day water levels within the mid-term scheduling remaining period, that is, the end-of-day water levels of the reservoir from the mth day of the lth month of the kth year to the last day of the lth month of the kth year. The objective function of the mid-term optimization scheduling module of the cascade hydropower station is:

[0116]

[0117] Where, T m is the remaining period of mid-term scheduling, that is, from the current day to the end of the month, Δt m The duration of the medium-term scheduling period is 1 day;

[0118] S6: The final water level of the first day of the remaining period of the mid-term optimal dispatch, i.e., the final water level of the reservoir on the mth day of the lth month of the kth year, is used as the final water level boundary constraint of the short-term optimal dispatch module of the cascade hydropower stations;

[0119] S7: Execute the short-term optimization scheduling module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, and obtain the corrected optimal daily reference flow, daily reference flow plan, daily output plan, and all hourly water levels within the day, that is, the hourly water levels of the reservoir from 1:00 on the mth day of the lth month of the kth year to 24:00 on the mth day of the lth month of the kth year. The objective function of the short-term optimization scheduling module of the cascade hydropower station is:

[0120]

[0121] Where, T d is the short-term scheduling period, i.e. 1 day, Δt d is the duration of the short-term scheduling period, i.e. 1 hour, c t is the electricity price for the t-th period;

[0122] S8: Initialize n=1;

[0123] S9: Determine whether to adopt the "electricity determines water" method. If yes, execute the in-plant economic operation module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, carry out the "electricity determines water" in-plant economic operation of the cascade hydropower station, and obtain the optimal unit output. If no, execute the in-plant economic operation module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, carry out the "water determines electricity" in-plant economic operation of the cascade hydropower station, and obtain the optimal unit reference flow. If the "electricity determines water" method is adopted, the objective function of the in-plant economic operation module of the cascade hydropower station is:

[0124]

[0125] If the “water determines electricity” method is adopted, the objective function of the economic operation module in the cascade hydropower station is:

[0126]

[0127] In the formula, q ij , P ij are the unit reference flow and unit output of the jth unit in the i-th power station, Δt is the duration of the current period, i.e., 1 hour, n u is the total number of units in the power station;

[0128] S10: Check if n≥24? If yes, proceed to the next step; if no, go to S14;

[0129] S11: Determine whether m ≥ the total number of days in the first month. If yes, proceed to the next step; if no, go to S15;

[0130] S12: Check if l≥12? If yes, proceed to the next step; if no, go to S16;

[0131] S13: Determine whether to end, if yes, end the process, if not, go to S17;

[0132] S14: Determine whether the time is a new hour. If yes, n=n+1 and go to S9. If no, continue to determine the time.

[0133] S15: Determine whether the time is a new day, if yes, m=m+1 and go to S5, if no, continue to determine the time;

[0134] S16: Determine whether the time is a new month, if yes, then l=l+1 and go to S2, if no, continue to determine the time;

[0135] S17: Determine whether the time is a new year, if yes, go to S1, if no, continue to determine the time.

[0136] Daily reference flow plan = corrected optimal daily reference flow, daily output plan = power station comprehensive output coefficient * corrected optimal daily reference flow * head.

[0137] The long-term optimization scheduling module of cascade hydropower stations uses 1 year as the scheduling cycle and 1 month as the scheduling period. It optimizes the power station reference flow with the goal of maximizing the total power generation and minimizing the total amount of abandoned water, and is executed once at the beginning of each month. The long-term optimization scheduling module of cascade hydropower stations contains the objective function, constraints, input conditions and decision variables for the long-term scheduling remaining period. The constraints include equality constraints and inequality constraints. Equality constraints include: water balance constraints, water level-reservoir capacity relationship curve, water level-flow relationship curve, initial water level boundary constraints and final water level boundary constraints; inequality constraints include: water level upper and lower limit constraints, water head upper and lower limit constraints, power station reference flow upper and lower limit constraints and power station total output upper and lower limit constraints. The initial water level boundary constraint is the actual measured initial water level of the month, and the final water level boundary constraint is the target final water level of the 12th month. The input condition is the monthly water inflow forecast value. The decision variable is the reference flow for the long-term scheduling remaining period. Long-term scheduling remaining period reference flow It is a column vector composed of the monthly reference flow of all cascade hydropower stations within the remaining period of long-term scheduling, and its expression is as follows:

[0138]

[0139] The mid-term optimization scheduling module of cascade hydropower stations takes one month as the scheduling cycle and one day as the scheduling period. It optimizes the power station reference flow with the goal of maximizing the total power generation and minimizing the total amount of abandoned water, and is executed once at the beginning of each day. The mid-term optimization scheduling module of cascade hydropower stations contains the objective function, constraints, input conditions and decision variables for the mid-term scheduling residual period.

[0140] The difference between the constraint conditions and the long-term optimization scheduling module of cascade hydropower stations is that the initial water level boundary constraint of the medium-term optimization scheduling module of cascade hydropower stations is the actual measured initial water level on the day, and the final water level boundary constraint is the water level at the end of the first month of the long-term scheduling remaining period.

[0141] The input condition is the predicted value of daily water inflow.

[0142] The decision variable is the reference flow rate of the remaining period of mid-term scheduling Medium-term scheduling remaining period reference flow It is a column vector composed of the daily reference flow of all cascade hydropower stations within the remaining period of medium-term scheduling, and its expression is as follows:

[0143]

[0144] The short-term optimization scheduling module of cascade hydropower stations uses 1 day as the scheduling cycle and 1 hour as the scheduling period. It optimizes the power station's reference flow with the goal of maximizing the total power generation revenue and minimizing the total amount of abandoned water, and is executed once a day. The short-term optimization scheduling module of cascade hydropower stations contains short-term scheduling objective functions, constraints, input conditions and decision variables. The difference between the constraints and the long-term optimization scheduling module of cascade hydropower stations is that the initial water level boundary constraint of the short-term optimization scheduling module of cascade hydropower stations is the actual measured initial water level of the day, and the final water level boundary constraint is the final water level of the first day of the mid-term scheduling period. The input condition is the hourly water inflow forecast value.

[0145] The decision variable is the daily citation flow Daily citation flow It is a column vector composed of the hourly reference flow of all power stations in the cascade hydropower station within a day, and the expression is as follows:

[0146]

[0147] The in-plant economic operation module of cascade hydropower stations adopts the optimization method of "determining water consumption by electricity" or "determining electricity by water", with the goal of minimizing the total water consumption or maximizing the total power generation of cascade hydropower stations in the current period, and is executed once every hour. The in-plant economic operation module of cascade hydropower stations contains the in-plant economic operation objective function, constraints, input conditions and decision variables.

[0148] The constraints include equality constraints and inequality constraints. Equality constraints include: water balance constraints, water level-reservoir capacity relationship curve, water level-flow relationship curve, and unit efficiency model; inequality constraints include: water level upper and lower limit constraints, water head upper and lower limit constraints, unit reference flow upper and lower limit constraints, unit output upper and lower limit constraints, and unit vibration zone constraints.

[0149] The input conditions are the measured water inflow in the current period, the daily output plan and the daily reference flow plan. For the in-plant economic operation of the cascade hydropower station that "determines water by electricity", the unit output allocation optimization is carried out according to the planned total output of the power station in the current period in the daily output plan, so that the total output of the power station = the planned total output of the power station in the current period, and the total water consumption of the cascade hydropower station in the current period is the lowest; for the in-plant economic operation of the cascade hydropower station that "determines electricity by water", the unit reference flow allocation optimization is carried out according to the planned reference flow of the power station in the current period in the daily reference flow plan, so that the total power generation of the cascade hydropower station in the current period is the largest when the power station reference flow = the planned reference flow of the power station in the current period.

[0150] If the "electricity determines water" method is adopted, the decision variable is the unit output; if the "water determines electricity" method is adopted, the decision variable is the unit reference flow.

[0151] Since the process of generating the original value of the long-term scheduling remaining period reference flow population, the original value of the medium-term scheduling remaining period reference flow population, and the original value of the daily reference flow population by the scheduling model heuristic optimization algorithm module is random, when the long-term, medium-term, and short-term optimization scheduling limits the final water level boundary constraint, the above original values ​​are difficult to accurately deduce the final water level boundary constraint while satisfying the water balance constraint. This means that the original value of the long-term scheduling remaining period reference flow population, the original value of the medium-term scheduling remaining period reference flow population, and the original value of the daily reference flow population given by the scheduling model heuristic optimization algorithm module cannot simultaneously satisfy the water balance constraint and the final water level boundary constraint. Therefore, it is necessary to carry out reference flow correction based on the final water level boundary constraint so that the corrected reference flow can simultaneously satisfy the water balance constraint and the final water level boundary constraint. In addition, in the long-, medium- and short-term optimization scheduling process, the optimal solution of the final generation of population output by the heuristic optimization algorithm module of the scheduling model is also obtained by searching the original values ​​of the reference flow population of each generation. Therefore, it is also necessary to obtain the corrected optimal long-term scheduling remaining period reference flow, the corrected optimal medium-term scheduling remaining period reference flow and the corrected optimal daily reference flow through correction.

[0152] like Figure 3 As shown, the long-term scheduling remaining period reference flow population correction value, the medium-term scheduling remaining period reference flow population correction value and the daily reference flow population correction value, as well as the corrected optimal long-term scheduling remaining period reference flow, the corrected optimal medium-term scheduling remaining period reference flow and the corrected optimal daily reference flow are all obtained through a forward reservoir water level calculation method containing a reverse reference flow correction algorithm and include the following steps:

[0153] S101: Initialize the upper limit V of the reservoir capacity max 、Reservoir capacity lower limit V min 、Citation flow limit Q e_max , the upper limit of the reference flow Q in the first period e1_max and the total number of time periods NT, the upper limit of the reference flow Q e1_max The expression is:

[0154] Q e1_max =[SL L (1)-(V min -V b (1))] / DT(1)-Q s (1) (6)

[0155] In the formula, SL L (1) V b (1), DT(1) and Q s (1) are the reservoir inflow in the first period, the initial storage capacity of the reservoir period, the period duration and the ecological flow;

[0156] S102: Determine the reference flow Q in the first period e (1)>Q e1_max ? Yes, then Q e (1) = Q e1_max , and execute the next step, otherwise, directly execute the next step;

[0157] S103: Initialize t k =1;

[0158] S104: Determine t k <NT-1? If yes, proceed to the next step, otherwise execute the reverse reference flow correction algorithm and go to S108;

[0159] S105: Determine V b (t k )+SL L (t k )-(Q e (t k )+Q s (t k ))*DT(t k )>V max ? Yes, then

[0160]

[0161] Go to S107, if not, go to the next step;

[0162] Where V b (t k ), V e (t k ), V x (t k ), SL L (t k ), Q e (t k ), Q s (t k ) and DT(t k ) are respectively k The reservoir's initial storage capacity for the time period, the reservoir's storage capacity at the end of the time period, the reservoir's discharge, the reservoir's incoming water, the reference flow, the ecological flow and the time period duration;

[0163] S106: Determine V b (t k )+SL L (t k )-(Q e (t k )+Q s (t k ))*DT(t k )<Vmin ? Yes, then V e (t k )=V min , execute the reverse reference flow correction algorithm and proceed to the next step, otherwise,

[0164]

[0165] And execute the next step;

[0166] S107: No. k +1 period reservoir initial water level V b (t k +1)=V e (t k ), t k =t k +1, and transfer to S104;

[0167] S108: The final storage capacity V of the reservoir at all time periods is calculated by the water level-storage capacity relationship curve. e Convert to the final water level of the reservoir in all time periods and end the process.

[0168] like Figure 4 As shown, the reverse reference traffic correction algorithm includes the following steps:

[0169] S111: Initialization t k2 =t k ;

[0170] S112: Determine t k2 ≥1? If yes, proceed to the next step; if no, go to S117.

[0171] S113: Determine V e (t k2 )-V b (t k2 )<SL L (t k2 )-Q s (t k2 )*DT(t k2 )? Yes, then

[0172]

[0173] Go to S117, if not, go to the next step;

[0174] Where V b (t k2 ), V e (t k2 ), V x (t k2 ), SL L (tk2 ), Q e (t k2 ), Q s (t k2 ) and DT(t k2 ) are respectively k2 The reservoir's initial storage capacity for the time period, the reservoir's storage capacity at the end of the time period, the reservoir's discharge, the reservoir's incoming water, the reference flow, the ecological flow and the time period duration;

[0175] S114: Determine t k2 =1? If yes, calculate the initial water level V of the reservoir in the first period according to the following formula: b1

[0176] V b1 =V e (t k2 )-(SL L (t k2 )-Q s (t k2 )*DT(t k2 )) (10)

[0177] And execute the next step, if not, go to S116;

[0178] S115: Judgement | V b (t k2 )-V b1 | / V b (t k2 )<1E-5? Yes, then

[0179]

[0180] Go to S117. If not, it is an error, the reference flow cannot be corrected, and go to S117.

[0181] S116:

[0182] t k2 =t k2 -1, and transfer to S112;

[0183] Where V e (t k2 -1) indicates the tth k2 -1 The reservoir water level at the end of the period;

[0184] S117: The reverse reference flow correction is terminated, and the process ends.

[0185] like Figure 5 As shown in the figure, the operation of the monthly water volume prediction module based on the self-made probability distribution model includes the following steps:

[0186] S201: Determine whether the time is a new year, if yes, proceed to the next step, if no, continue to determine the time;

[0187] S202: Count the monthly rainfall from January to December in the previous years before the current year;

[0188] S203: Calculate the rainfall probability distribution model from January to December respectively, where the rainfall probability distribution model includes the probability density function and probability distribution function of the monthly rainfall;

[0189] S204: Calculate the mathematical expectations of the rainfall probability distribution model from January to December respectively, and use them as the monthly rainfall prediction values ​​of the corresponding months of the current year;

[0190] S205: Calculate the 90% confidence interval of the rainfall probability distribution model from January to December respectively, and use it as the monthly rainfall prediction interval of the corresponding month of the current year;

[0191] S206: Calculate the monthly water inflow forecast value and monthly water inflow forecast interval for the corresponding month of the current year of the reservoir through the monthly water inflow-monthly rainfall relationship curve;

[0192] S207: Determine whether to end, if yes, end the process, if not, go to S201.

[0193] The daily water inflow LSTM prediction module driven by multi-dimensional feature data is executed once at the beginning of each month to predict the daily water inflow from the first day of the month to the last day of the month, and the daily water inflow prediction is carried out using the trained LSTM neural network model. During training, the characteristics of the sample are {daily maximum temperature, daily minimum temperature, daily rainfall, and the solar term to which the day belongs}, the result label of the sample is daily water inflow, the time expansion step is 30 days, and the forecast period is 1 day, that is, the LSTM neural network model uses the {daily maximum temperature, daily minimum temperature, daily rainfall, and the solar term to which the day belongs} from the 29th day in reverse order before the prediction day to the prediction day to fit the daily water inflow on the prediction day. When using the trained LSTM neural network model to carry out daily water inflow prediction, the {daily maximum temperature, daily minimum temperature, daily rainfall, and the solar term to which the day belongs} of the future days are all obtained through weather forecasts.

[0194] The hourly water inflow forecast module is executed once at the beginning of each day to predict the hourly water inflow from the first hour to the last hour of the day, and is obtained by averaging the daily water inflow forecast value over 24 hours.

[0195] Specifically, the above-mentioned long-, medium- and short-term nested cascade hydropower station forecast dispatching system is implemented in the following ways:

[0196] Based on the simulation calculation of Xiecunyuan cascade hydropower station, the hydraulic connection diagram of Xiecunyuan cascade hydropower station is as follows Figure 6 As shown in the figure, the Xiecunyuan Secondary Hydropower Station directly draws water from the Xiecunyuan Secondary Reservoir for power generation, while the tailwater of the Xiecunyuan Secondary Hydropower Station and the discharge flow and ecological flow of the Xiecunyuan Secondary Reservoir are introduced into the Xiecunyuan Thirdary Hydropower Station after being intercepted by the water weir, as the reference flow of the Xiecunyuan Thirdary Hydropower Station. The installed capacity of the Xiecunyuan Secondary Hydropower Station and the Xiecunyuan Thirdary Hydropower Station are 2*8.67MW and 2*2MW respectively, and the rated heads are 219.24m and 48.289m respectively. The power generation price is 0.57 yuan for peak power (peak power period: 7:00 to 11:00 and 13:00 to 23:00), and 0.228 yuan for valley power (valley power period: 11:00 to 13:00, 23:00 to 7:00 the next day).

[0197] The main inequality constraint settings for the Xiecunyuan cascade hydropower station are shown in Table 1.

[0198] Table 1 Parameter setting table of inequality constraints of Xiecunyuan cascade hydropower station

[0199]

[0200] The economic operation module within the cascade hydropower station is executed in a "water-based electricity" manner. The heuristic optimization algorithm module of the scheduling model is based on the iteration and optimization of the adaptive fuzzy particle swarm algorithm (AFPSO algorithm for short). For details of the algorithm, please refer to the document "Accurate Parameter Estimation of a Hydro-Turbine Regulation System UsingAdaptive Fuzzy Particle Swarm Optimization" (author: Dong Liu et al., journal name: Energies, volume number: 2019, issue number: 12, article number: 3903).

[0201] The relationship curve between monthly water inflow and monthly rainfall of Xiecunyuan Secondary Reservoir is as follows: Figure 7 As shown in , the two show a strong linear relationship, that is, monthly water inflow = β0 + β1 * monthly rainfall + ε, where β0 and β1 are the 0th and 1st order coefficients respectively, and ε is the residual. Figure 7 Implementing least squares estimation on the scattered points in , we can obtain β0=-5.457E5, β1=5.138E4, and the determination coefficient is 0.8952.

[0202] Monthly rainfall and monthly water inflow of Xiecunyuan Secondary Reservoir Figure 8-9 As shown in the figure, the trend of the predicted curve is consistent with the actual curve trend, and the 90% confidence interval can basically cover all actual values. At this time, the root mean square error of the monthly water volume forecast is 3.35E6 m 3 .

[0203] The daily water volume of Xiecunyuan Secondary Reservoir Fig.10 As shown in the figure, the trend of the predicted curve and the actual curve are consistent. Except for some peak points, the deviation between the predicted value and the actual value is not large. The root mean square error of the daily water volume prediction is 2.25E5 m 3 .

[0204] The forecast scheduling optimization results of the Xiecunyuan cascade hydropower station based on the embodiment of the present invention are shown in Table 2. It can be seen from the table that after optimization by the embodiment of the present invention, the annual power generation income of the Xiecunyuan cascade hydropower station can be increased by 7.8% in 2020.

[0205] Table 2 Optimization results of the embodiment of the present invention of Xiecunyuan cascade hydropower station

[0206]

[0207] Draw as Figure 11-14 The output curve of Unit 1 of Xiecunyuan Second-Level Hydropower Station before and after optimization from March 1, 2020 to March 31, 2020, the output curve of Unit 2 of Xiecunyuan Second-Level Hydropower Station before and after optimization, the output curve of Unit 1 of Xiecunyuan Third-Level Hydropower Station before and after optimization, and the output curve of Unit 2 of Xiecunyuan Third-Level Hydropower Station before and after optimization are shown to intuitively demonstrate the economic operation effect of the plant of the embodiment of the present invention. As can be seen from the figure, the embodiment of the present invention utilizes the in-plant economic operation module 14 of the cascade hydropower station to reasonably distribute the load, ensuring the maximization of the total output of the Xiecunyuan cascade hydropower station while avoiding the operation of the unit in the vibration zone; when the total output of the power station is greater than the maximum output of one unit, but significantly less than the sum of the maximum outputs of the two units (see the curve from March 13 to March 28), the output of the optimized units is not evenly distributed, but one unit is fully generated and the other unit is partially loaded, so as to avoid the two units operating in the vibration zone at the same time due to the even distribution of loads; and when the total output of the power station has reached the vicinity of the sum of the maximum outputs of the two units (see the curve after March 29), after optimization, the two units evenly distribute the load to ensure that neither operates in the vibration zone; in addition, when the total output of the power station is significantly less than the maximum output of one unit (see the curve from March 6 to March 8), after optimization, one unit operates in a partial load condition in the non-vibration zone, and the other unit is shut down or operates at a low load.

[0208] The above-mentioned specific implementation methods are used to explain the present invention and are only preferred embodiments of the present invention, rather than limiting the present invention. Any modifications, equivalent substitutions, improvements, etc. made to the present invention within the spirit of the present invention and the protection scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A long-, medium- and short-term nested cascade hydropower station forecast dispatching system, characterized by: Including long-, medium- and short-term nested power generation optimization scheduling module and reservoir water inflow forecast module; The long-, medium- and short-term nested power generation optimization scheduling module includes: a long-term optimization scheduling module for cascade hydropower stations, a medium-term optimization scheduling module for cascade hydropower stations, a short-term optimization scheduling module for cascade hydropower stations, an in-plant economic operation module for cascade hydropower stations and a scheduling model heuristic optimization algorithm module; The reservoir water inflow forecast module includes: a monthly water inflow forecast module based on a self-made probability distribution model, a daily water inflow LSTM forecast module driven by multi-dimensional feature data, and an hourly water inflow forecast module.

2. The long-, medium- and short-term nested cascade hydropower station forecast dispatching system according to claim 1 is characterized by: The operation of the long-, medium- and short-term nested power generation optimization scheduling module includes the following steps: S1: Initialize the target final water level of the 12th month of the kth year and use it as the final water level boundary constraint of the long-term optimal scheduling module of the cascade hydropower stations, k = current year, l = 1; S2: Execute the long-term optimization scheduling module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, and obtain the corrected optimal long-term scheduling remaining period reference flow, as well as the end-of-month water levels of all months within the long-term scheduling remaining period, that is, the end-of-month water levels of the reservoir from the first month of the kth year to the 12th month of the kth year. The objective function of the long-term optimization scheduling module of the cascade hydropower station is: Where, T y is the remaining period of long-term scheduling, that is, from the current month to the 12th month, Δt y is the length of the long-term scheduling period, i.e., 1 month, n s and n re are the total number of cascade hydropower stations and the total number of reservoirs, are the total power output and power station reference flow of the i-th power station in the t-th period, For the kth re The discharge flow of a reservoir in period t; S3: The water level at the end of the first month of the long-term scheduling remaining period, that is, the reservoir water level at the end of the month of the kth year, is used as the final water level boundary constraint of the mid-term optimization scheduling module of the cascade hydropower stations; S4: Initialize m=1; S5: Execute the mid-term optimization scheduling module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, and obtain the corrected optimal mid-term scheduling remaining period reference flow, and all end-of-day water levels within the mid-term scheduling remaining period, that is, the end-of-day water levels of the reservoir from the mth day of the lth month of the kth year to the last day of the lth month of the kth year. The objective function of the mid-term optimization scheduling module of the cascade hydropower station is: Where, T m is the remaining period of mid-term scheduling, that is, from the current day to the end of the month, Δt m The duration of the medium-term scheduling period is 1 day; S6: The final water level of the first day of the remaining period of the mid-term optimal dispatch, i.e., the final water level of the reservoir on the mth day of the lth month of the kth year, is used as the final water level boundary constraint of the short-term optimal dispatch module of the cascade hydropower stations; S7: Execute the short-term optimization scheduling module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, and obtain the corrected optimal daily reference flow, daily reference flow plan, daily output plan, and all hourly water levels within the day, that is, the hourly water levels of the reservoir from 1:00 on the mth day of the lth month of the kth year to 24:00 on the mth day of the lth month of the kth year. The objective function of the short-term optimization scheduling module of the cascade hydropower station is: Where, T d is the short-term scheduling period, i.e. 1 day, Δt d is the duration of the short-term scheduling period, i.e. 1 hour, c t is the electricity price for the t-th period; S8: Initialize n=1; S9: Determine whether to adopt the "power-to-water" method. If yes, execute the in-plant economic operation module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, carry out the in-plant economic operation of the cascade hydropower station with "power-to-water" to obtain the optimal unit output. If no, execute the in-plant economic operation module of the cascade hydropower station, and call the scheduling model heuristic optimization algorithm module for optimization, carry out the in-plant economic operation of the cascade hydropower station with "water-to-power" to obtain the optimal unit reference flow. If the "power-to-water" method is adopted, the objective function of the in-plant economic operation module of the cascade hydropower station is: If the "water determines electricity" method is adopted, the objective function of the economic operation module in the cascade hydropower station is: In the formula, q ij , P ij are the unit reference flow and unit output of the jth unit in the i-th power station, Δt is the duration of the current period, i.e., 1 hour, n u is the total number of units in the power station; S10: Check if n≥24? If yes, proceed to the next step; if no, go to S14; S11: Check if m ≥ the total number of days in the first month. If yes, proceed to the next step; if no, go to S15; S12: Check if l≥12? If yes, proceed to the next step; if no, go to S16; S13: Determine whether to end, if yes, end the process, if not, go to S17; S14: Determine whether the time is a new hour. If yes, n=n+1 and go to S9. If no, continue to determine the time. S15: Determine whether the time is a new day, if yes, m=m+1 and go to S5, if no, continue to determine the time; S16: Determine whether the time is a new month, if yes, then l=l+1 and go to S2, if no, continue to determine the time; S17: Determine whether the time is a new year, if yes, go to S1, if no, continue to determine the time.

3. The long-, medium- and short-term nested cascade hydropower station forecast dispatching system according to claim 2 is characterized by: The daily reference flow rate plan = the corrected optimal daily reference flow rate, and the daily output plan = the power station comprehensive output coefficient * the corrected optimal daily reference flow rate * the head.

4. The long-, medium- and short-term nested cascade hydropower station forecast dispatching system according to claim 1 or 2, characterized in that: The long-term scheduling remaining period reference flow population correction value, the medium-term scheduling remaining period reference flow population correction value and the daily reference flow population correction value, as well as the corrected optimal long-term scheduling remaining period reference flow, the corrected optimal medium-term scheduling remaining period reference flow and the corrected optimal daily reference flow are all obtained through a forward reservoir water level calculation method containing a reverse reference flow correction algorithm and include the following steps: S101: Initialize the upper limit V of the reservoir capacity max 、Reservoir capacity lower limit V min 、Citation flow limit Q e_max , the upper limit of the reference flow Q in the first period e1_max and the total number of time periods NT, the upper limit of the reference flow Q e1_max The expression is: Q e1_max =[SL L (1)-(V min -V b (1))] / DT(1)-Q s (1) (6) In the formula, SL L (1) V b (1), DT(1) and Q s (1) are the reservoir inflow in the first period, the initial storage capacity of the reservoir period, the period duration and the ecological flow; S102: Determine the reference flow Q in the first period e (1)>Q e1_max ? Yes, then Q e (1) = Q e1_max , and execute the next step, otherwise, directly execute the next step; S103: Initialize t k =1; S104: Determine t k <NT-1? If yes, proceed to the next step, otherwise execute the reverse reference flow correction algorithm and go to S108; S105: Determine V b (t k )+SL L (t k )-(Q e (t k )+Q s (t k ))*DT(t k )>V max ? Yes, then Go to S107, if not, go to the next step; Where V b (t k ), V e (t k ), V x (t k ), SL L (t k ), Q e (t k ), Q s (t k ) and DT(t k ) are respectively k The reservoir's initial storage capacity for the time period, the reservoir's storage capacity at the end of the time period, the reservoir's discharge, the reservoir's incoming water, the reference flow, the ecological flow and the time period duration; S106: Determine V b (t k )+SL L (t k )-(Q e (t k )+Q s (t k ))*DT(t k )<V min ? Yes, then V e (t k )=V min , execute the reverse reference flow correction algorithm and proceed to the next step, otherwise, And execute the next step; S107: No. k +1 period reservoir initial water level V b (t k +1)=V e (t k ), t k =t k +1, and transfer to S104; S108: The final storage capacity V of the reservoir at all time periods is calculated by the water level-storage capacity relationship curve. e Convert to the final water level of the reservoir in all time periods and end the process.

5. The long-, medium- and short-term nested cascade hydropower station forecast dispatching system according to claim 4 is characterized by: The reverse reference flow correction algorithm includes the following steps: S111: Initialization t k2 =t k ; S112: Determine t k2 ≥1? If yes, proceed to the next step; if no, go to S117; S113: Determine V e (t k2 )-V b (t k2 )<SL L (t k2 )-Q s (t k2 )*DT(t k2 )? Yes, then Go to S117, if not, go to the next step; Where V b (t k2 ), V e (t k2 ), V x (t k2 ), SL L (t k2 ), Q e (t k2 ), Q s (t k2 ) and DT(t k2 ) are respectively k2 The reservoir's initial storage capacity for the time period, the reservoir's storage capacity at the end of the time period, the reservoir's discharge, the reservoir's incoming water, the reference flow, the ecological flow and the time period duration; S114: Determine t k2 =1? If yes, calculate the initial water level V of the reservoir in the first period according to the following formula: b1 V b1 =V e (t k2 )-(SL L (t k2 )-Q s (t k2 )*DT(t k2 )) (10) And execute the next step, if not, go to S116; S115: Judgement | V b (t k2 )-V b1 | / V b (t k2 )<1E-5? Yes, then Go to S117. If not, it is an error, the reference flow cannot be corrected, and go to S117. t k2 =t k2 -1, and transfer to S112; Where V e (t k2 -1) indicates the tth k2 -1 The reservoir water level at the end of the period; S117: The reverse reference flow correction is terminated, and the process ends.

6. The long-, medium- and short-term nested cascade hydropower station forecast dispatching system according to claim 1 is characterized by: The operation of the monthly water volume prediction module based on the self-made probability distribution model includes the following steps: S201: Determine whether the time is a new year, if yes, proceed to the next step, if no, continue to determine the time; S202: Count the monthly rainfall from January to December in the previous years before the current year; S203: Calculate the rainfall probability distribution model from January to December respectively, where the rainfall probability distribution model includes the probability density function and probability distribution function of the monthly rainfall; S204: Calculate the mathematical expectations of the rainfall probability distribution model from January to December respectively, and use them as the monthly rainfall prediction values ​​of the corresponding months of the current year; S205: Calculate the 90% confidence interval of the rainfall probability distribution model from January to December respectively, and use it as the monthly rainfall prediction interval of the corresponding month of the current year; S206: Calculate the monthly water inflow forecast value and monthly water inflow forecast interval for the corresponding month of the current year of the reservoir through the monthly water inflow-monthly rainfall relationship curve; S207: Determine whether to end, if yes, end the process, if not, go to S201.

7. The long-, medium- and short-term nested cascade hydropower station forecast dispatching system according to claim 1 is characterized by: The daily water volume LSTM prediction module driven by multi-dimensional feature data is executed once at the beginning of each month.

8. The long-, medium- and short-term nested cascade hydropower station forecast dispatching system according to claim 1 is characterized by: The hourly water inflow prediction module is executed once at the beginning of each day.

Images