Model-free adaptive control reservoir scheduling method and terminal equipment

Abstract

The invention belongs to the technical field of water conservancy scheduling, and particularly discloses a model-free adaptive control reservoir scheduling method and terminal equipment, and the method comprises the steps: obtaining an actual observation water level at a moment in a scheduling period; determining a disturbance estimation value at the moment in the scheduling period according to the actual observation water level, the predicted water level and the disturbance estimation value at the moment in the scheduling period; determining a water level error at the moment in the scheduling period according to the actual observation water level and the predicted water level at the moment in the scheduling period; and determining the control quantity of the moment in the scheduling period according to the disturbance estimated value of the moment in the scheduling period, the water level error of the moment in the scheduling period and the control quantity of the moment in the scheduling period. The method does not depend on a traditional hydrological physical model, and the difficulty of modeling and parameter identification can be remarkably reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of water conservancy dispatching, and specifically discloses a reservoir dispatching method and terminal equipment with model-free adaptive control. Background Art

[0002] As water resource systems become increasingly complex and dynamic, reservoirs, as core infrastructure for flood control, water supply security, and ecological regulation, face unprecedented challenges in their scheduling and management. Currently, mainstream reservoir scheduling methods rely on optimization control strategies based on physical models. These methods describe the hydrological processes in the reservoir area by establishing hydraulic and hydrological models, and use mathematical optimization algorithms to solve scheduling instructions. However, actual hydrological systems often have strong nonlinearity, strong uncertainty, and significant external disturbance influences. This leads to the following typical problems with these model-based scheduling methods:

[0003] It is highly dependent on model accuracy. When faced with external disturbances such as sudden changes in rainfall, sudden water releases upstream and downstream (such as local heavy rainfall or abnormal drainage upstream), the prediction deviation of traditional models may be significantly amplified, thereby affecting the safety of scheduling. The modeling process is complex and the update frequency is low. For systems with non-stationary input characteristics, model reconstruction and parameter correction are often time-consuming and labor-intensive, making it difficult to meet the high-frequency scheduling needs in rapidly changing environments. It lacks a data-driven mechanism and is not intelligent enough to fully explore the potential laws in historical and real-time hydrological information for dynamic scheduling. Summary of the Invention

[0004] The purpose of the present invention is to provide a reservoir scheduling method and terminal equipment with model-free adaptive control to solve the technical problems of existing reservoir scheduling methods, such as strong model dependence and lack of data-driven mechanism, which affect scheduling safety.

[0005] A first aspect of the present invention provides a reservoir operation method with model-free adaptive control, comprising:

[0006] Step 1. Get During the scheduling cycle The actual observed water level at the moment;

[0007] Step 2: According to During the scheduling cycle The actual observed water level, predicted water level and disturbance estimate at the moment are determined During the scheduling cycle The disturbance estimate at time t;

[0008] Step 3: According to During the scheduling cycle The actual observed water level and predicted water level at the moment are determined During the scheduling cycle Water level error at each moment;

[0009] Step 4: According to During the scheduling cycle The disturbance estimate at time During the scheduling cycle The water level error at the time and During the scheduling cycle Determination of control quantity at each moment During the scheduling cycle The amount of control at any moment.

[0010] Preferably, step 2 is specifically:

[0011] Step 2.1, confirm During the scheduling cycle The difference between the actual observed water level and the predicted water level at the moment;

[0012] Step 2.2, determining a water level variable according to the difference and a preset gain value;

[0013] Step 2.3: Based on the water level variable and During the scheduling cycle The sum of the disturbance estimates at time During the scheduling cycle The disturbance estimate at time .

[0014] Preferably, after step 2.3, the method further includes:

[0015] when During the scheduling cycle When the disturbance estimate at the moment is greater than the preset saturation value constant, During the scheduling cycle The disturbance estimate at time is a saturation constant;

[0016] when During the scheduling cycle When the disturbance estimate at time t is less than the negative saturation constant, During the scheduling cycle The disturbance estimate at time t is a negative saturation constant.

[0017] Preferably, step 3 is specifically:

[0018] according to During the scheduling cycle The difference between the actual observed water level and the predicted water level at the time is determined During the scheduling cycle Water level error at any moment.

[0019] Preferably, step 4 is specifically:

[0020] Step 4.1, confirm During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment;

[0021] Step 4.2, according to During the scheduling cycle The estimated value of the dynamic adjustment factor at each moment and During the scheduling cycle The control amount at a moment determines the control inheritance item;

[0022] Step 4.3, according to During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment, During the scheduling cycle The disturbance estimate at time During the scheduling cycle The water level error at the moment determines the disturbance correction term;

[0023] Step 4.4: Determine based on the sum of the control inheritance term and the disturbance correction term During the scheduling cycle The amount of control at any moment.

[0024] Preferably, the control inheritance item in step 4.2 is determined according to a first formula, which is:

[0025]

[0026] Where, is the energy consumption penalty factor; is the weight; for During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment; for During the scheduling cycle The amount of control at any moment.

[0027] Preferably, the disturbance correction term in step 4.3 is determined according to a second formula, which is:

[0028]

[0029] Where, for During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment; is the energy consumption penalty factor; is the weight; for During the scheduling cycle The water level error at the moment, for During the scheduling cycle The actual observed water level at the moment, for During the scheduling cycle Predicted water level at the moment; for During the scheduling cycle The disturbance estimate at time .

[0030] Preferably, step 4.1 is specifically as follows:

[0031] Sure During the scheduling cycle Time-of-day water level error and rate of change of controlled quantity;

[0032] according to During the scheduling cycle Determination of the estimated value of the momentary water level error, the rate of change of the control quantity and the dynamic adjustment factor During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment.

[0033] Preferably, before step 1, the method further includes:

[0034] The reservoir in The historical control quantity before the scheduling period is used as input and the historical water level value is used as output to train the machine learning model. The initial control quantity of the scheduling period.

[0035] The second aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned model-free adaptive control reservoir scheduling method when executing the computer program.

[0036] Compared with the prior art, the reservoir scheduling method and terminal device of the present invention with model-free adaptive control have the following beneficial effects:

[0037] The present invention is based on a model-free adaptive control reservoir scheduling method based on a machine learning model and an extended state observer. It fully integrates the predictive ability of machine learning, the disturbance estimation compensation mechanism and the data-driven control algorithm, breaking through the bottleneck problems of traditional scheduling methods such as strong dependence on system modeling, delayed response, and unpredictable disturbances. The present invention uses a machine learning model to achieve efficient prediction of the initial control quantity, uses an extended state observer to dynamically estimate external disturbances such as sudden inflow and rainfall, and uses a model-free adaptive controller to iteratively optimize the scheduling instructions based on error feedback, ultimately forming a set of closed-loop control mechanisms with adaptability, robustness and real-time performance. The algorithm of the present invention runs stably and can be embedded in the existing reservoir scheduling automation platform. It is suitable for a variety of reservoir scheduling scenarios with different scales and different operating objectives, and has significant engineering promotion value and practical application significance.

[0038] The present invention can achieve optimal control of reservoir scheduling strategies without relying on traditional hydrological physical models, significantly reducing the difficulty of modeling and parameter identification, and is suitable for complex, nonlinear, and time-varying hydrological system environments.

[0039] The present invention processes historical hydrological data and real-time hydrological data through a machine learning model, and outputs an initial control quantity for controller initialization, which can effectively reduce the initial control error, improve the control accuracy and accelerate the convergence speed of the control strategy.

[0040] By extending the state observer, the present invention can estimate external disturbance factors such as sudden rainfall and upstream water changes in real time, and use them for control compensation, thereby improving the operating stability and scheduling robustness of the reservoir under non-ideal or extreme working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The figure is a flow chart of a reservoir scheduling method based on model-free adaptive control according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In the following description, specific details such as particular system structures and techniques are provided for purposes of illustration, not limitation, to facilitate a thorough understanding of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.

[0043] The first aspect of the embodiment of the present invention provides a reservoir scheduling method without model adaptive control, which can solve the problems of traditional scheduling methods such as time-consuming and labor-intensive establishment of accurate models, external interference, and inability to respond quickly to sudden hydrological changes. The process is as follows: Figure 1 shown.

[0044] The embodiment of the present invention firstly places the reservoir in The historical control quantity before the scheduling period is used as input and the historical water level value is used as output to train the machine learning model. Initial control quantity of the scheduling period .

[0045] The above-mentioned historical control quantities and initial control quantities are the reservoir's inflow and outflow. In the embodiment of the present invention, the inflow is the inflow transmitted by the water level stations on the tributaries and main stream, the outflow is measured by flow meters at the tributaries and main stream gates, and the water level is the reservoir water level measured by a radar level gauge.

[0046] The machine learning models described above can include random forest models, support vector machines, and decision trees. Due to the advantages of the random forest model, such as its ability to automatically analyze the impact of each feature on the prediction results, high accuracy and robustness, its ability to process high-dimensional data, its strong noise immunity, and its support for parallel computing, the random forest model is preferably used in embodiments of the present invention. The random forest model outputs the initial control variable used to initialize the model-free adaptive control. As the initial input value of the iterative controller, it can effectively reduce initial scheduling errors, improve control efficiency, and reduce the number of iterations.

[0047] Determine the initial control quantity using random forest model The specific process is as follows:

[0048] Train a regression random forest model consisting of multiple decision trees, and then During the scheduling period, the vector consisting of the water level values ​​of different main streams and tributaries at the initial time Input into the regression random forest model, and send the water level value into each regression tree to predict Control initial value of the scheduling cycle , For the A regression tree in The control initial value of the scheduling cycle. The prediction results of all regression trees are averaged and integrated to generate the final output, as follows:

[0049] (1)

[0050] in, for The initial control quantity of the scheduling cycle; is the total number of regression trees in the random forest; Indicates the A regression tree in The initial control value of the scheduling cycle.

[0051] As the starting input of the subsequent model-free controller, it can effectively reduce the initial control error and improve the convergence speed and scheduling stability of iterative control.

[0052] In determining Initial control quantity of the scheduling period After that, subsequent model-free adaptive control of reservoir operation can be carried out, specifically:

[0053] Step 1. Get During the scheduling cycle Actual observed water level at the moment .

[0054] Step 2: According to During the scheduling cycle Actual observed water level at the moment , predict water level and the perturbation estimate Sure During the scheduling cycle The disturbance estimate at time , specifically:

[0055] Step 2.1, confirm During the scheduling cycle Actual observed water level at the moment and predicted water level The difference .

[0056] Step 2.2: Based on the difference and preset gain values Determine water level variables .

[0057] Step 2.3: Based on the water level variable and During the scheduling cycle The disturbance estimate at time The sum is determined During the scheduling cycle The disturbance estimate at time , as shown in formula (2):

[0058] (2)

[0059] Disturbance terms in this embodiment of the present invention include, but are not limited to, external disturbance sources such as localized heavy rainfall, sudden upstream inflow changes, and operational fluctuations in dispatching facilities. Step 2 is implemented using an extended state observer, which estimates external disturbance terms in the system during the dispatch process, facilitating subsequent disturbance compensation for control commands. The extended state observer iteratively determines disturbance estimates, eliminating reliance on modeling of disturbance sources and improving the control system's adaptability to sudden changes.

[0060] To prevent the disturbance estimate from being too large and causing oscillation of the control variable, the embodiment of the present invention further includes, after step 2.3:

[0061] when During the scheduling cycle The disturbance estimate at time Greater than the preset saturation constant hour, During the scheduling cycle The disturbance estimate at time is the saturation constant ;when During the scheduling cycle The disturbance estimate at time Less than the negative saturation constant hour, During the scheduling cycle The disturbance estimate at time is the negative saturation constant , as follows:

[0062] (3)

[0063] Perturbation estimate is passed to the controller to compensate for the error term, thereby improving the stability and robustness of the controller under non-ideal conditions.

[0064] Step 3: According to During the scheduling cycle Actual observed water level at the moment and predicted water levels Determine the difference During the scheduling cycle Water level error at the moment , as shown in formula (4):

[0065] (4)

[0066] Step 4: According to During the scheduling cycle The disturbance estimate at time 、 During the scheduling cycle Water level error at the moment as well as During the scheduling cycle The amount of control at any moment ,Sure During the scheduling cycle The amount of control at any moment.

[0067] In the embodiment of the present invention, a controller is used to implement step 4. The controller implements control update based on error drive, does not require system modeling, has a fast convergence speed, and is suitable for complex and time-varying hydrological scheduling scenarios.

[0068] The above step 4 is specifically as follows:

[0069] Step 4.1, confirm During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment .

[0070] The above step 4.1 is specifically as follows:

[0071] Sure During the scheduling cycle Time disturbance estimate and the rate of change of the controlled variable ;

[0072] according to During the scheduling cycle Time disturbance estimate , the rate of change of the controlled quantity and the estimated value of the dynamic adjustment factor Sure During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment , as shown in formula (5):

[0073] (5)

[0074] Where, To adjust the step length; is the updated smoothing factor.

[0075] Dynamic regulatory factors It is used to balance the control speed, which is unpredictable. In the embodiment of the present invention, the cost function of formula (6) is minimized to estimate it online, thereby obtaining :

[0076] (6)

[0077] Where, is the cost function; for During the scheduling cycle Dynamic adjustment factor of the moment; for During the scheduling cycle The estimated value of the dynamic adjustment factor at time .

[0078] Step 4.2, according to During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment and During the scheduling cycle The amount of control at any moment Determine the control inheritance term, as shown in formula (7):

[0079] (7)

[0080] Where, is the energy consumption penalty factor, used to limit the scheduling intensity; is the weight; for During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment; for During the scheduling cycle The amount of control at any moment.

[0081] Step 4.3, according to During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment 、 During the scheduling cycle The disturbance estimate at time and During the scheduling cycle Water level error at the moment Determine the disturbance correction term, as shown in formula (8):

[0082] (8)

[0083] Where, for During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment; is the energy consumption penalty factor; is the weight; for During the scheduling cycle The water level error at the moment, for During the scheduling cycle The actual observed water level at the moment, for During the scheduling cycle Predicted water level at the moment; for During the scheduling cycle The disturbance estimate at time .

[0084] Step 4.4: According to the control inheritance item and the disturbance correction term The sum is determined During the scheduling cycle The control quantity at the moment is shown in formula (9):

[0085] (9)

[0086] The control inheritance item of the embodiment of the present invention controls the degree to which the current scheduling instruction inherits the scheduling instruction at the previous moment, and the disturbance correction item corrects the scheduling instruction based on the water level error and the disturbance estimation value.

[0087] The controller used in the scheduling process of the present invention uses the objective function as the optimization criterion, and its objective function is as follows:

[0088] (10)

[0089] Where, for During the scheduling cycle The amount of control at any moment The objective function of for During the scheduling cycle Water level error at each moment; for During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment; for During the scheduling cycle The rate of change of the controlled variable at each moment, and ; for During the scheduling cycle The disturbance estimate at time t; is the energy consumption penalty factor; is the weight; for During the scheduling cycle The amount of control at any moment.

[0090] By taking the partial derivative of the above formula (10) and setting it to zero, we can obtain the optimal control law shown in formula (9).

[0091] The control law of formula (9) can be simplified as follows:

[0092] (11)

[0093] Where, is the historical retention factor, is the error amplification factor. The above two parameters can adjust the iteration speed and sensitivity of the controller.

[0094] The embodiment of the present invention determines the dispatching instructions according to the above-mentioned reservoir dispatching method, including adjusting the opening of the reservoir gate to achieve water inlet flow and water outlet flow regulation to ensure that the reservoir water level is within the set range, thereby achieving precise control of the reservoir water level and efficient utilization and intelligent management of water resources.

[0095] This embodiment of the present invention regularly backs up and stores dispatch operation records and hydrological data, ensuring data recovery in the event of a failure or emergency, and ensuring the long-term stable operation of the system. It also continuously optimizes dispatch strategies by analyzing historical hydrological data and uses this data to adjust controller parameters, further improving dispatch efficiency.

[0096] The method of the present invention controls the adaptive adjustment of input through real-time output data, thus getting rid of the dependence on precise mathematical models and greatly improving the flexibility and adaptability of scheduling.

[0097] The method of the present invention can achieve real-time regulation of reservoir water level, quickly respond to sudden hydrological changes, and ensure the safety and efficiency of reservoir operation.

[0098] The method of the present invention performs self-learning and optimized scheduling based on historical data and real-time data, which significantly improves the intelligent management level of the reservoir.

[0099] The present invention combines meteorological forecast data and real-time hydrological data to make scheduling decisions, achieve early response to extreme weather events, and further improve the accuracy of scheduling.

[0100] The second aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned model-free adaptive control reservoir scheduling method when executing the computer program.

[0101] The essence of the present invention is a reservoir scheduling method based on model-free adaptive iterative learning control of random forest models and extended state observers. It fully integrates the prediction capabilities of machine learning, disturbance estimation and compensation mechanisms, and data-driven control algorithms, breaking through the bottleneck problems of traditional scheduling methods such as strong dependence on system modeling, delayed response, and unpredictable disturbances. The present invention uses a random forest model to achieve efficient prediction of the initial control quantity, uses an extended state observer to dynamically estimate external disturbances such as sudden inflows and rainfall, and uses a model-free adaptive controller to iteratively optimize scheduling instructions based on error feedback, ultimately forming a closed-loop control mechanism with adaptability, robustness, and real-time performance. The algorithm of the present invention is stable in operation and can be widely used in various types of reservoir scheduling scenarios. It has significant engineering promotion value and practical application significance.

[0102] The above descriptions are merely several embodiments of the present invention and do not constitute any form of limitation to the present invention. Although the present invention is disclosed as above in terms of preferred embodiments, they are not intended to limit the present invention. Any technician familiar with the present profession, without departing from the scope of the technical solution of the present invention, who makes slight changes or modifications using the technical contents disclosed above, is equivalent to an equivalent implementation case and falls within the scope of the technical solution.

Claims

1. A reservoir operation method based on model-free adaptive control, characterized in that: include: Step 1. Get During the scheduling cycle The actual observed water level at the moment; Step 2: According to During the scheduling cycle The actual observed water level, predicted water level and disturbance estimate at the moment are determined During the scheduling cycle The disturbance estimate at time t; Step 3: According to During the scheduling cycle The actual observed water level and predicted water level at the moment are determined During the scheduling cycle Water level error at each moment; Step 4: According to During the scheduling cycle The disturbance estimate at time During the scheduling cycle The water level error at the time and During the scheduling cycle Determination of control quantity at each moment During the scheduling cycle The amount of control at any moment.

2. The reservoir operation method of model-free adaptive control according to claim 1, characterized in that: Step 2 is as follows: Step 2.1, confirm During the scheduling cycle The difference between the actual observed water level and the predicted water level at the moment; Step 2.2, determining a water level variable according to the difference and a preset gain value; Step 2.3: Based on the water level variable and During the scheduling cycle The sum of the disturbance estimates at time During the scheduling cycle The disturbance estimate at time .

3. The reservoir operation method of model-free adaptive control according to claim 2, characterized in that: After step 2.3 also include: when During the scheduling cycle When the disturbance estimate at the moment is greater than the preset saturation value constant, During the scheduling cycle The disturbance estimate at time is a saturation constant; when During the scheduling cycle When the disturbance estimate at time t is less than the negative saturation constant, During the scheduling cycle The disturbance estimate at time t is a negative saturation constant.

4. The reservoir operation method of model-free adaptive control according to claim 1, characterized in that: Step 3 is as follows: according to During the scheduling cycle The difference between the actual observed water level and the predicted water level at the time is determined During the scheduling cycle Water level error at any moment.

5. The reservoir operation method of model-free adaptive control according to claim 1, characterized in that: Step 4 is as follows: Step 4.1, confirm During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment; Step 4.2, according to During the scheduling cycle The estimated value of the dynamic adjustment factor at each moment and During the scheduling cycle The control amount at a moment determines the control inheritance item; Step 4.3, according to During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment, During the scheduling cycle The disturbance estimate at time During the scheduling cycle The water level error at the moment determines the disturbance correction term; Step 4.4: Determine based on the sum of the control inheritance term and the disturbance correction term During the scheduling cycle The amount of control at any moment.

6. The reservoir operation method of model-free adaptive control according to claim 5, characterized in that: In step 4.2, the control inheritance item is determined according to the first formula, which is: ； Where, is the energy consumption penalty factor; is the weight; for During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment; for During the scheduling cycle The amount of control at any moment.

7. The reservoir operation method of model-free adaptive control according to claim 6, characterized in that: The disturbance correction term in step 4.3 is determined according to the second formula, which is: ； Where, for During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment; is the energy consumption penalty factor; is the weight; for During the scheduling cycle The water level error at the moment, for During the scheduling cycle The actual observed water level at the moment, for During the scheduling cycle Predicted water level at the moment; for During the scheduling cycle The disturbance estimate at time .

8. The reservoir operation method according to any one of claims 5 to 7, characterized in that: Step 4.1 is as follows: Sure During the scheduling cycle Time-of-day water level error and rate of change of controlled quantity; according to During the scheduling cycle Determination of the estimated value of the momentary water level error, the rate of change of the control quantity and the dynamic adjustment factor During the scheduling cycle Estimated value of the dynamic adjustment factor at each moment.

9. The reservoir operation method of model-free adaptive control according to claim 1, characterized in that: Also include before step 1: The reservoir in The historical control quantity before the scheduling period is used as input and the historical water level value is used as output to train the machine learning model. The initial control quantity of the scheduling period.

10. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the reservoir scheduling method with model-free adaptive control are implemented as described in any one of claims 1 to 9.

Images