JP7467315B2 - Water system operation planning method and water system operation planning system - Google Patents

Description

Embodiments of the present invention relate to a water system operation planning method and a water system operation planning system.

The amount of water flowing into a dam water system generally varies depending on factors such as the amount of rainfall upstream as well as snowmelt and spring water. When operating the water system, it is necessary to plan and operate the water intake for power generation in accordance with the amount of water flowing into the dam water system in order to obtain the amount of power generation that corresponds to the amount of water flowing in.

Patent Document 1 shows a flow prediction device that uses rainfall information from multiple locations to generate predicted flow data with a flow prediction model based on correlation. Patent Document 2 also shows a method for planning the amount of water used for power generation by each generator, taking into account operational constraints such as the amount of water stored in the dam, so as to maximize the amount of power generated by an interconnected water system. A dam operation method is known that uses a combination of these methods to improve the amount of power generated in a water system.

Japanese Patent No. 4807565 JP 2005-285032 A

In general, it is possible to predict dam inflows at a future time of about several hours with high accuracy by making full use of recent technologies including machine learning, but when the prediction is for a future time of about several days, the difficulty of predicting inflows increases and the error in inflow predictions (prediction error) tends to increase. Even when dam inflows suddenly increase due to heavy rain or snowmelt, it is generally often difficult to predict inflows with high accuracy.

On the other hand, when operating a dam, it is desirable to maintain the dam's standard high water level and operate it continuously and efficiently to generate electricity. However, for example, if the dam's inflow increases more than predicted and the dam's water level exceeds the upper operational limit, water will be released, resulting in a reduction in the amount of electricity generated relative to the actual inflow.

The present invention has been made in consideration of the above-mentioned circumstances, and aims to provide a water system operation planning method and a water system operation planning system that can minimize the impact of errors in inflow prediction.

According to an embodiment, a water system operation planning method is provided which includes an inflow prediction step of predicting a time series inflow of water flowing into a water system, a prediction error estimation step of estimating a time series prediction error for the inflow predicted in the inflow prediction step based on past inflow predictions and past inflow actual results, and a water system operation planning step of creating information for a water system operation plan based on the inflow predicted in the inflow prediction step and the prediction error estimated in the prediction error estimation step, in which the water system operation planning step creates information for the water system operation plan by performing an optimization calculation aimed at any one of maximizing the amount of generated power, maximizing the value of generated power, minimizing the dam discharge amount, or maximizing the amount of water used for power generation, and creates information indicating the water system operation plan by performing the optimization calculation using an objective function to which an influence when the inflow increases or decreases corresponding to the prediction error estimated in the prediction error estimation step is added.

The present invention makes it possible to minimize the impact of errors in inflow prediction.

FIG. 1 is a diagram showing an example of the functional configuration of a river system operation planning system for implementing a river system operation planning method according to an embodiment. FIG. 2 is a diagram showing an example of a model of a river system that is the subject of a river system management plan in the embodiment. FIG. 13 is a diagram showing an example of a processing procedure for a water system operation plan. FIG. 13 is a diagram showing an example of a calculation result of an inflow prediction Qin _i (t), which is a predicted value of an inflow, and an estimated value ΔQin _i (t) of a corresponding prediction error, using a regression equation. A diagram showing the relationship between various decision variables at each time step. FIG. 13 is a diagram showing an example of correcting the upper limit of the planned water level to be lower than the reference water level Hnmax _i . FIG. 13 is a diagram showing an example of transfer operation in which, when a period with a large prediction error is expected to be included, the power generation output P _i (t) is increased or decreased so as to increase or decrease the amount of water intake for power generation before that period.

The following describes the embodiment with reference to the drawings.

Figure 1 is a diagram showing an example of the functional configuration of a water system operation planning system for implementing a water system operation planning method according to one embodiment. Figure 2 is a diagram showing an example of a model of a water system that is the subject of a water system operation plan in the embodiment.

(composition)
1 is realized, for example, by using one or more information processing devices (computers), and includes, as various processing functions, an inflow prediction unit 10, a prediction error estimation unit 20, and a river system operation planning unit 30. Also, as various storage functions, the system includes a first storage unit DB1, a second storage unit DB2, and a third storage unit DB3.

The inflow prediction unit 10, the prediction error estimation unit 20, and the water system operation planning unit 30 may be configured as functions of a program executed by a computer processor. The first memory unit DB1, the second memory unit DB2, and the third memory unit DB3 may be realized using a recording medium that can be read and written under the control of the computer processor. These various processing functions and storage functions may be integrated into one computer, or may be distributed across several computers and installed.

Function of the inflow prediction unit 10 The inflow prediction unit 10 has a function of predicting the time-series inflow of water flowing into a water system.

This inflow prediction unit 10 inputs weather forecasts, rainfall information, and snowfall information provided from the outside, references past weather forecasts and past rainfall information stored in memory unit DB1, and references past actual data on inflows of each dam stored in memory unit DB2, and outputs a predicted value QF of the inflow of water flowing into each dam in chronological order. Here, instead of the actual data on inflows of each dam, actual data on the discharge amount, water intake amount for power generation, and water level change of each dam can be input, and the inflow of each dam calculated can be used. The prediction of the inflow of each dam may also be a prediction of the flow rate of the river flowing into each dam.

Functions of the Prediction Error Estimation Unit 20 The prediction error estimation unit 20 is a function that estimates a time-series prediction error for the inflow predicted by the inflow prediction unit 10, based on the actual data of the past inflow of each dam stored in the memory unit DB2 and the predicted data of the past inflow of each dam stored in the memory unit DB3. The prediction error is estimated based on the inflow predicted at each time for each predetermined time step within a certain period of time, for example.

This prediction error estimation unit 20 inputs a time series of the predicted inflow values QF output by the inflow prediction unit 10, and by referring to the past predicted inflow data for each dam and the past actual inflow data for each dam, outputs a time series of the prediction error estimate value EE corresponding to the predicted inflow value QF output by the inflow prediction unit 10.

Functions of the River System Operation Planning Unit 30 The river system operation planning unit 30 is a function that creates information on a river system operation plan based on the inflow predicted by the inflow prediction unit 10 and the prediction error estimated by the prediction error estimation unit 20. The information on the river system operation plan may be created, for example, by performing an optimization calculation using an objective function that adds the influence of an increase or decrease in the inflow corresponding to the prediction error estimated by the prediction error estimation unit 20. Alternatively, the information on the river system operation plan may be created by adding a water level constraint or a water storage volume constraint that is influenced by a change in the dam water level corresponding to the prediction error estimated by the prediction error estimation unit 20, or by changing the water level constraint or the water storage volume constraint to be narrowed, and then performing an optimization calculation.

This water system operation planning unit 30 inputs a time series of the inflow prediction values QF output by the inflow prediction unit 10 and a time series of the prediction error estimate values EE output by the prediction error estimation unit 20, and outputs planning information for the target water system operation including the power plant's water intake volume QP and the water discharge volume QR from the dam gate calculated by optimization calculations aimed at maximizing the amount of generated power, maximizing the value of generated power, minimizing the dam discharge volume, or maximizing the amount of water used for power generation.

The inflow prediction unit 10, prediction error estimation unit 20, and water system operation planning unit 30 described above may be provided near the target water system. In addition, a device that collects and records data such as rainfall information via a mobile phone communication network or the Internet and calculates a water system operation plan may be provided in a server or the like in a remote location geographically distant from the water system, and information on the water system operation plan may be transmitted to a device used by the water system operator.

Water system model The model shown in Figure 2 is an example of a model for a dam water system in which multiple dams 1, 2, 3, ... are connected to each other. Dams 1, 2, 3, ... are each equipped with generators 11, 12, 13, ... that generate electricity by taking in water from each dam. After being used by the generators of each dam, the water is sent to the next dam together with the water discharged from each dam.

The amounts of water flowing into dams 1, 2, 3, . . . are denoted as Qin ₁ , Qin ₂ , Qin ₃ , .

The water levels of dams 1, 2, 3, . . . are assumed to be H ₁ , H ₂ , H ₃ , .

The water volumes of dams 1, 2, 3, . . . are denoted as V ₁ , V ₂ , V ₃ , .

The intake amounts of water (intake amounts for power generation) supplied from the dams ₁ , ₂ , ₃ , . . . to the generators 11, 12, 13, .

The amounts of water discharged from the dam gates of dams 1, 2, 3, . . . are denoted as Qr ₁ , Qr ₂ , Qr ₃ , .

The amounts of power generated by the generators 11, 12, 13, . . . are denoted as P ₁ , P ₂ , P ₃ , .

(Processing Procedure)
Next, an example of a processing procedure for a water system operation plan will be described with reference to FIG.

Here, the predicted inflow value QF as explained in Figure 1 is used to calculate the planned information for the power generation intake QP and discharge QR based on the evaluation function for the power generation volume of the target water system and the constraints on the dam water system operation. Also, for the water system model explained in Figure 2, the number of dams is M, and at a fixed time every day (t = 0), inflow prediction, prediction error estimation, and water system operation planning are performed for each time t = 1, ..., N up to the time N steps ahead at time intervals Δt. The time interval Δt for the water system operation plan can be selected to be 30 minutes, 1 hour, etc.

In the inflow prediction in step S1, the inflow prediction unit 10 calculates a predicted value QF of the inflow for each time interval Δt of the water system operation plan based on weather forecasts and rainfall information at multiple points, and by referring to past weather forecasts, past rainfall information, and past actual data on inflows of each dam, using machine learning or the like. Here, the predicted value QF of the inflow is data on the inflow prediction Qin _i (t) (where i = 1, ..., M, t = 1, ..., N) of dam i.

A model may be constructed that inputs weather forecasts and precipitation information to calculate the predicted inflow value QF by referring to past weather forecasts, past precipitation information, and past actual inflow data for each dam, and by applying techniques such as neural networks and deep learning as machine learning, etc. Furthermore, a model may be constructed that calculates the predicted inflow value QF so as to also refer to actual snowfall information, and the accuracy of the inflow prediction may be improved by adding snowfall information as an input.

In the prediction error estimation in step S2, the prediction error estimation unit 20 constructs an equation for estimating the expected value of the prediction error estimate EE corresponding to the predicted inflow QF as a function of the predicted inflow QF and time t from the past inflow prediction data and past actual inflow data of each dam. The prediction error estimate EE is an estimate ΔQin _i (t) of the difference between the predicted inflow and actual inflow of each dam, and can be expressed by the following regression equation.

ΔQin _i (t) = β1 _i · Qin _i (t) + β2 _i · t + β3 _i ... (1)
Here, t is the time for each predetermined time step, and _β1i , β2i _{, and β3i} are the coefficients of the regression equation for each dam. The coefficients of the regression equation can be selected by multiple regression analysis from the predicted data of the past inflow of each dam and the actual data of the past inflow of each dam.

In addition to the above, the method of calculating the prediction error estimate EE may also involve expressing the prediction error as a probability distribution and taking the expected value of the increasing distribution of the prediction error as the prediction error estimate EE.

4 is a graph showing an example of the calculation results of the inflow prediction Qin _i (t), which is the predicted value of the inflow, and the corresponding estimated value ΔQin _i (t) of the prediction error by the regression equation. In this example graph, ΔQin _i (t) becomes large in the time period when Qin _i (t) becomes large, but when Qin _i (t) becomes small after the time period has passed, it does not follow suit and become small in the same way, but tends to remain at a relatively large value according to the elapsed time t.

In the water system operation plan of step S3, the water system operation planning unit 30 creates information for a water system operation plan that uses the dam water volume V _i (t), dam water level H _i (t), power generation water intake Qp _i (t), discharge Qr _i (t), and power generation P _i (t) as decision variables and maximizes the total power generation as an objective function under the constraints of water system operation.

Here, Fig. 5 shows the relationship between various decision variables for each time step. In Fig. 5, the amount of water intake for power generation Qp _i (t) and the amount of discharge Qr _i (t) are collectively represented as Q _i (t). As shown in Fig. 5, V _i (t) and H _i (t) indicate values corresponding to each time from time t = 0 to time t = N + 1 at each time interval Δt. On the other hand, Q _i (t) and P _i (t) indicate values corresponding to the time interval Δt between each time.

The objective function, along with the equality and inequality constraints, is set as follows:

Objective function:
max Σ _i Σ _t {C(t) · P _i (t)} ... (2)
Equality constraints:
V _i (t+1) - V _i (t)
= {Qin _i (t) + Qp _i-1 (t) - Qp _i (t) + Qr _i-1 (t) - Qr _i (t)} Δt ... (3)
H _i (t) = _ahi · V _i (t) + _bhi ... (4)
P _i (t) = a p _i · Q p _i (t) + b _{p i} · { H _i (t) - H _i0 } + c _{p i} ... (5)
Inequality constraints:
Hnmin _i ≦ H _i (t) ≦ Hnmax _i ... (6)
Pmin _i ≦ P _i (t) ≦ Pmax _i or P _i (t) = 0 ... (7)
0≦ _Qpi (t)≦ _Qpmaxi ... (8)
Qrmin _i ≦ Qr _i (t) ≦ Qrmax _i ... (9)
Here, C(t) is the hourly electricity value (unit price of electricity sales), _ahi and _bhi are coefficients representing the characteristics of the dam's water storage volume and water level, _api , _bpi and _cpi are coefficients representing the power generation output characteristics of the power plant, and _Qrmini and _Qrmaxi are the minimum maintained discharge and maximum discharge of each dam, respectively.

The objective function shown in equation (2) may be either maximization of the sum of the generated power amounts P _i (t) or maximization of the sum of the generated power values C(t) x P _i (t), or minimization of the sum of the discharge amounts Qr _i (t) or maximization of the sum of the water intake amounts for power generation Qp _i (t).

Here, in the planning period of the dam where the estimated value of the prediction error ΔQin _i (t) may exceed the water storage volume between the upper water level limit Hmax _i for dam operation and the reference water level Hnmax _i , the following correction of the inequality constraint in Method 1 or the correction of the objective function in Method 2 can be applied.

Method 1 (correction of constraint equations)
As illustrated in FIG. 6, at the planning time t when the estimated value ΔQin _i (t) of the prediction error is greater than the threshold value, equation (6) is corrected by ΔH _i (t) so that the upper limit of the planned water level in the dam water constraint is lowered below the reference water level Hnmax _i .

Hnmin _i ≦ Hnmin _i ≦ H _i (t) ≦ Hnmax _{i −} ΔH _i (t) … (6A)
In addition, an inequality constraint regarding the equivalent water storage volume V _i (t) may be used. ΔH _i (t) can be estimated as follows using the coefficients of equation (4).

ΔH _i (t) = _ahi · ΔQin _i (t) Δt + bhi _- (Hmax _{i -} Hnmax _i ) ... (10)
In addition, by performing a similar correction on the operational water level lower limit Hnmin _i , the lower limit of the planned water level can be prevented from falling below the water level lower limit.

Method 2 (correction of objective function)
The objective function of equation (2) is corrected so as to reduce the expected value ΔL _i (t) of the decrease in the amount of generated power P _i (t) accompanying an increase in the discharge amount.

max Σ _i Σ _t {C(t) · (P _i (t) - ΔL _i (t) )} ... (2A)
Here, ΔL _i (t) can be approximately estimated using the coefficients in equation (5).

ΔL _i (t) = a p _i (t) · ΔQ _{in i} (t) + c p _i ... (11)
In this case, the same effect as in method 1 can be obtained.

In addition, it is also possible to use each formula in combination, applying method 1 to dams with large dam capacity and method 2 to dams with small dam capacity. Also, if a solution to the optimization calculation cannot be obtained using method 1, the process can be switched to method 2 and recalculated.

In the water system operation plan of step S3, a linear programming or mixed integer programming solver may be used to process the plan to obtain a more optimal solution.

In order to reduce the impact of prediction errors, the latest inflow forecasts can be obtained every few hours, and replanning can be performed by creating prediction error estimates and water system operation plan information.

In addition, a water system operation plan may be implemented by adding the processing of the prediction error estimation unit 20 to an existing computer system that performs water system operation planning and has processing units equivalent to the inflow prediction unit 10 and the water system operation planning unit 30, and modifying the water system operation planning unit so that it is the same as the processing of the above-mentioned embodiment.

(effect)
By predicting the inflow into the water system at a future time, estimating the prediction error, and correcting the water level constraints and objective function of the optimal planning problem to minimize the impact of the prediction error on power generation and discharge volumes, and calculating information for the optimal plan for water system operation, including water intake and discharge volumes for power generation, based on this, it is possible to realize a water system operation plan that minimizes the decrease in optimality of the water system operation plan even when there is an error in the inflow prediction.

Furthermore, when a river system operation plan is expected to include a period with large prediction errors, it is effective to carry out transfer operation to increase or decrease the power generation output P _i (t) within a range that does not exceed the power generation output upper limit so as to increase or decrease the power generation water intake before that period, as shown in Figure 7, for example. By applying this to optimization calculations that include the effects of prediction errors, it is possible to realize a more optimal river system operation plan that includes information on the transfer period and transfer power generation amount for a dam river system in which multiple dams are connected.

Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the scope of the invention and its equivalents described in the claims.

10...inflow prediction unit, 20...prediction error estimation unit, 30...water system operation planning unit, 100...water system operation planning system.

Claims (6)

an inflow prediction step of predicting a time series inflow amount of water flowing into the water system;
a prediction error estimating step of estimating a time series prediction error for the inflow predicted in the inflow prediction step based on past inflow predictions and past inflow results;
a river system operation planning step of creating information on an operation plan of the river system based on the inflow predicted in the inflow prediction step and the prediction error estimated in the prediction error estimation step;
Including,
In the water system operation planning step,
Create information for the operation plan of the water system by performing optimization calculations aimed at maximizing the amount of generated power, maximizing the value of generated power, minimizing dam discharge, or maximizing the amount of water used for power generation.
generating information showing an operation plan for the water system by performing the optimization calculation using an objective function that includes an influence of an increase or decrease in the inflow amount corresponding to the prediction error estimated in the prediction error estimation step;
A computer-implemented method for planning water system operations. an inflow prediction step of predicting a time series inflow amount of water flowing into the water system;
a prediction error estimating step of estimating a time series prediction error for the inflow predicted in the inflow prediction step based on past inflow predictions and past inflow results;
a river system operation planning step of creating information on an operation plan of the river system based on the inflow predicted in the inflow prediction step and the prediction error estimated in the prediction error estimation step;
Including,
In the water system operation planning step,
Create information for the operation plan of the water system by performing optimization calculations aimed at maximizing the amount of generated power, maximizing the value of generated power, minimizing dam discharge, or maximizing the amount of water used for power generation.
a water level constraint or a water storage volume constraint corresponding to the influence of the dam water level change corresponding to the prediction error estimated in the prediction error estimation step is added, or the water level constraint or the water storage volume constraint is changed to be narrowed, and then the optimization calculation is performed to create an operation plan for the water system.
A computer-implemented method for planning water system operations. The prediction error estimation step includes:
estimating a prediction error based on a predicted inflow volume at each time for each predetermined time step within a certain period of time;
The water system operation planning method according to claim 1 or 2 . This is for dam water systems with multiple dams connected to each other.
A method for planning a water system operation according to any one of claims 1 to 3 . An inflow prediction means for predicting a time series inflow amount of water flowing into the water system;
a prediction error estimation means for estimating a time series prediction error for the inflow predicted by the inflow prediction means based on past inflow predictions and past inflow results;
a river system operation planning means for creating information on an operation plan of the river system based on the inflow predicted by the inflow prediction means and the prediction error estimated by the prediction error estimation means;
Equipped with
The water system operation planning means includes:
Create information for the operation plan of the water system by performing optimization calculations aimed at maximizing the amount of generated power, maximizing the value of generated power, minimizing dam discharge, or maximizing the amount of water used for power generation.
generating information showing an operation plan for the water system by performing the optimization calculation using an objective function obtained by adding an influence of an increase or decrease in the inflow amount corresponding to the prediction error estimated by the prediction error estimation means;
Water system operational planning system. An inflow prediction means for predicting a time series inflow amount of water flowing into the water system;
a prediction error estimation means for estimating a time series prediction error for the inflow predicted by the inflow prediction means based on past inflow predictions and past inflow results;
a river system operation planning means for creating information on an operation plan of the river system based on the inflow predicted by the inflow prediction means and the prediction error estimated by the prediction error estimation means;
Equipped with
The water system operation planning means includes:
Create information for the operation plan of the water system by performing optimization calculations aimed at maximizing the amount of generated power, maximizing the value of generated power, minimizing dam discharge, or maximizing the amount of water used for power generation.
adding a water level constraint or a water storage volume constraint for the influence of the dam water level change corresponding to the prediction error estimated by the prediction error estimation means, or modifying the water level constraint or the water storage volume constraint to be narrower, and then performing the optimization calculation to create an operation plan for the water system.
Water system operational planning system.