Self-adaptive PID (Proportion Integration Differentiation) control system for water conservancy reservoir environment

Abstract

The invention relates to the technical field of water conservancy automation control and edge computing, in particular to a self-adaptive PID control system for a water conservancy reservoir environment, which comprises an edge computing gateway, a water level sensor, a flow sensor and a servo execution mechanism, fluid static geometric parameters are obtained through a topological mapping module, an initial state matrix of a fluid momentum state observer is established, a momentum decoupling module inputs a dam front water level, upstream incoming water flow and opening data to update the state matrix, and transient fluctuation energy characteristics and real reserve residual characteristics are generated through decoupling. The priori fitting module calculates proportion, integral and differential coefficients according to the parameters, and the closed-loop evolution module generates a control instruction and corrects bottom resistance parameters; physical separation of real reservoir capacity change and surface wave disturbance is achieved, invalid oscillation of the gate is reduced, control stability under the flood peak working condition is improved, and the service life of an executing mechanism is prolonged.

Description

Technical Field

[0001] This invention relates to the field of water conservancy automation control and edge computing technology, specifically to an adaptive PID control system for water conservancy reservoir environments. Background Technology

[0002] In the current water conservancy reservoir operation and maintenance environment, large and medium-sized reservoirs, especially valley-type regulating reservoirs, will experience complex surface wave propagation and reflection phenomena on the free liquid surface during the passage of flood peaks or rapid opening and closing of gates. Existing technologies generally adopt traditional PID control schemes, which directly use the single-point water level data collected by sensors in front of the dam as feedback for closed-loop regulation. Although this scheme can maintain the basic water level, it cannot distinguish between the slow-changing component in the water level gauge reading that represents the increase or decrease in reservoir capacity and the fast-changing disturbance component formed by water surface fluctuations. As a result, the controller is very likely to misjudge false water level fluctuations as regulation deviations.

[0003] This not only induces frequent reciprocating movements of the gate actuator, resulting in severe mechanical shocks and control oscillations and shortening the service life of the servo equipment, but also makes it difficult for traditional solutions to cope with the long-term evolution of boundary conditions caused by sediment deposition and scouring in the bedrock during the adjustment process with large inertia and long time. Therefore, how to effectively decouple the real mass residual from the transient fluctuation energy in the complex dynamic environment of multiple physical quantities superimposed, and how to achieve adaptive optimization of control parameters according to the topological evolution law of fluid space, so as to improve the control stability under flood peak conditions and the system's adaptability to the evolution of the bedrock environment, has become an urgent technical problem to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive PID control system for water conservancy and reservoir environments, solving the following technical problems:

[0005] It achieves physical separation between actual reservoir capacity changes and surface wave disturbances, avoiding misjudging false fluctuations caused by waves as water level errors, thereby reducing ineffective oscillations and mechanical impacts of the gate, improving control stability and actuator lifespan under flood peak conditions, and can adapt to the long-term evolution of the reservoir bed boundary, preventing control performance degradation caused by model aging.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] An adaptive PID control system for a water conservancy reservoir environment includes an edge computing gateway, a water level sensor, a flow sensor, and a servo actuator for driving the reservoir gate.

[0008] The water level sensor, flow sensor, and servo actuator are used to collect real-time reservoir water level data, real-time inflow data, and current actuator opening data, respectively, and send the collected data to the edge computing gateway.

[0009] The edge computing gateway, which operates according to a preset control cycle, is equipped with a topology mapping module, which is used to acquire preset fluid static geometric parameters including bottom resistance parameters, construct a fluid space topology model, and establish an initial state matrix of the fluid momentum state observer based on the fluid space topology model.

[0010] The momentum decoupling module is used to input the received real-time reservoir water level data, real-time inflow data, and current actuator opening data into the fluid momentum state observer, drive the initial state matrix to update the state to obtain the current state matrix, and perform frequency domain decomposition on the current state matrix to separate the fast-changing component representing surface waves and the slow-changing component representing reservoir capacity changes, thereby decoupling and generating transient fluctuation energy characteristics and true storage residual characteristics.

[0011] The prior fitting module is used to calculate the proportional coefficient based on the transient fluctuation energy characteristics, calculate the integral coefficient based on the real storage residual characteristics, and calculate the differential coefficient by extracting the time change rate of the real-time inflow data.

[0012] The closed-loop evolution module is used to generate control commands based on the proportional coefficient, the integral coefficient, and the derivative coefficient and send them to the servo actuator, and extract the steady-state deviation of the transient fluctuation energy characteristics to correct the bottom resistance parameter in the fluid static geometry parameters;

[0013] The servo actuator is also used to receive the control command and drive the reservoir gate to move.

[0014] Furthermore, the topology mapping module is specifically used for constructing fluid space topology models:

[0015] The reservoir fluid space is divided into a preset number of virtual nodes;

[0016] The fluid static geometry parameters are assigned to each of the virtual nodes;

[0017] Establish the physical connection relationships between each of the virtual nodes to generate the fluid space topology model;

[0018] The fluid static geometric parameters also include the reservoir capacity curve parameters.

[0019] Furthermore, the momentum decoupling module, when decoupling the generation of transient fluctuation energy characteristics and actual reserve residual characteristics, is specifically used for:

[0020] The current state matrix of the fluid momentum state observer is decomposed in the frequency domain using an orthogonal decomposition algorithm.

[0021] High-frequency rapidly changing components are extracted as the transient fluctuation energy characteristics.

[0022] Low-frequency, slowly varying components are extracted as the residual features of the actual reserves;

[0023] The transient fluctuation energy characteristic represents the kinetic and potential energy of the waves on the fluid surface, and the real storage residual characteristic represents the difference between the current total fluid volume and the target volume preset based on scheduling requirements.

[0024] Furthermore, when calculating the proportionality coefficient based on the transient fluctuation energy characteristics, the prior fitting module is specifically used for:

[0025] Obtain the preset baseline scaling factor and the preset attenuation factor;

[0026] Calculate the product of the attenuation factor and the transient fluctuation energy characteristic;

[0027] Using the natural constant as the base and the negative of the product as the exponent, the negative exponential decay weight is calculated.

[0028] The benchmark scaling factor is multiplied by the negative exponential decay weight to obtain the scaling factor.

[0029] Furthermore, when calculating the integral coefficients based on the actual reserve residual characteristics, the prior fitting module is specifically used for:

[0030] Obtain the preset baseline integral coefficients;

[0031] The actual reserve residual features are input into a preset Sigmoid nonlinear driving function to generate integral driving weights;

[0032] The integral coefficient is obtained by multiplying the baseline integral coefficient by the integral driving weight.

[0033] Furthermore, when calculating the differential coefficients of the time change rate of the real-time inbound flow data, the prior fitting module is specifically used for:

[0034] Calculate the time derivative of the real-time inbound flow data within the current control cycle to obtain the time change rate;

[0035] The time change rate is used as a feedforward suppression compensation term and added to a preset benchmark differential coefficient to calculate the differential coefficient.

[0036] Furthermore, when the closed-loop evolution module extracts the steady-state deviation of the transient fluctuation energy characteristics and uses the steady-state deviation to correct the bottom drag parameter in the fluid static geometry parameters, it is specifically used for:

[0037] Obtain the fluctuation frequency characteristics of the transient fluctuation energy characteristics within a preset time window;

[0038] The difference between the fluctuation frequency characteristic and the preset reference frequency characteristic is calculated as the steady-state deviation;

[0039] The bottom resistance parameters of the corresponding virtual nodes in the fluid space topology model are updated based on the steady-state deviation to adaptively compensate for the boundary morphology evolution trend of the fluid bottom.

[0040] Furthermore, the system is applied to water conservancy and reservoir environments;

[0041] Among them, the real-time reservoir water level data is the water level data in front of the dam, the real-time inflow data is the upstream inflow data, and the current actuator opening data is the reservoir gate opening data;

[0042] The servo actuator is either a hydraulically driven gate actuator or a motor-driven gate actuator.

[0043] Furthermore, the boundary morphology evolution trend is the evolution trend of the sediment deposition thickness at the bottom of the reservoir.

[0044] The beneficial effects of this invention are:

[0045] 1. This invention addresses the problem of frequent reciprocating gate actions caused by surface waves in the mixed water level in front of the dam. This system uses a momentum decoupling module to perform frequency domain decomposition on the state matrix, accurately separating the transient wave energy characteristics representing surface waves from the actual storage residual characteristics representing the deviation of the actual reservoir capacity. This effectively avoids misjudging short-term pseudo fluctuations as water level errors that must be corrected immediately, eliminates ineffective gate oscillations, greatly reduces mechanical impact, and improves the overall lifespan of the actuator.

[0046] 2. To address the problem of insufficient adaptability of fixed parameter adjustment, the system dynamically calculates the control coefficient through a priori fitting module; the system drives integral adjustment based on the actual quality residual according to the wave energy suppression proportional response, and extracts the time change rate of the inflow as feedforward suppression compensation; this mechanism can automatically adjust the sensitivity according to the inflow and disturbance, enhance the damping capacity in advance, and perfectly avoid sudden overshoot and response lag.

[0047] 3. This invention addresses the problem that empirical regulation is difficult to adapt to environmental evolution. The closed-loop evolution module continuously extracts the steady-state deviation of the fluctuation frequency characteristics and adaptively updates the bottom resistance parameters in the fluid space topology model. This design enables the control system to keenly perceive and automatically compensate for the long-term boundary morphological evolution trend such as sediment accumulation at the bottom of the reservoir, ensuring that subsequent control is always based on the updated physical bed state, thus guaranteeing the long-term stability of control performance. Attached Figure Description

[0048] The invention will now be further described with reference to the accompanying drawings.

[0049] Figure 1 This is a schematic diagram of a module of an adaptive PID control system for a water conservancy reservoir environment provided in an embodiment of this application. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Please see Figure 1 An adaptive PID control system for a water conservancy reservoir environment includes an edge computing gateway, a water level sensor, a flow sensor, and a servo actuator for driving the reservoir gate.

[0052] The water level sensor, flow sensor, and servo actuator are used to collect real-time reservoir water level data, real-time inflow data, and current actuator opening data, respectively, and send the collected data to the edge computing gateway.

[0053] The edge computing gateway, which operates according to a preset control cycle, is equipped with a topology mapping module, which is used to obtain preset fluid static geometric parameters including bottom resistance parameters, construct a fluid space topology model, and establish the initial state matrix of the fluid momentum state observer based on the fluid space topology model.

[0054] The momentum decoupling module is used to input the received real-time reservoir water level data, real-time inflow data, and current actuator opening data into the fluid momentum state observer, drive the initial state matrix to update the state to obtain the current state matrix, and perform frequency domain decomposition on the current state matrix to separate the fast-changing component representing surface waves and the slow-changing component representing reservoir capacity changes, thereby decoupling and generating transient wave energy characteristics and true storage residual characteristics.

[0055] The prior fitting module is used to calculate the proportional coefficient based on the transient fluctuation energy characteristics, calculate the integral coefficient based on the real storage residual characteristics, and extract the time change rate of the real-time inflow data to calculate the differential coefficient.

[0056] The closed-loop evolution module is used to generate control commands based on the proportional coefficient, integral coefficient, and derivative coefficient and send them to the servo actuator, and extract the steady-state deviation of transient fluctuation energy characteristics to correct the bottom resistance parameter in the fluid static geometry parameters.

[0057] The servo actuator is also used to receive the control commands and drive the reservoir gate to move.

[0058] This embodiment provides an adaptive PID control mechanism for the passage of flood peaks in a valley-type multi-year regulating reservoir. Specifically, water level sensors are deployed in front of the main dam, flow sensors are deployed at the upstream inlet section, and opening degree acquisition units are deployed at the gate hoist or hydraulic cylinder. Each sensor uploads data to the edge computing gateway at a fixed control cycle.

[0059] The edge computing gateway is installed in the dam crest control room or the gate local control cabinet. It has floating-point computing capabilities and industrial communication interfaces, and is used to complete fluid state identification and control command generation before the gate is executed.

[0060] Specifically, this embodiment does not directly equate the single-point water level in front of the dam with the actual change in reservoir capacity. The reason is that when the flood peak arrives or the gate is opened and closed rapidly, surface waves will appear on the free liquid surface in the reservoir area, which will propagate and be reflected along the way. The water level read by the water level gauge in front of the dam includes both the slow-changing component that truly represents the increase or decrease in reservoir capacity and the fast-changing component formed by the fluctuation of the water surface.

[0061] If the controller confuses these two types of changes, it will misjudge the false fluctuations caused by waves as water level errors that must be corrected immediately, thereby repeatedly pushing the gate to open and close, forming mechanical shocks and control oscillations. To this end, this embodiment sets up a topology mapping module in the edge computing gateway to pre-import the fluid static geometric parameters corresponding to the current reservoir, including at least the bottom resistance parameters, local shoal and channel changes, and parameter information corresponding to the overall geometric distribution of the reservoir area.

[0062] The topology mapping module abstracts the actual reservoir area into a set of virtual fluid units with spatial sequence, enabling the edge gateway to perceive at the control level that the reservoir area is not a point, but a hydrodynamic space with propagation, dissipation and reflection behaviors.

[0063] Based on this, a fluid momentum state observer is established according to the fluid space topology model. This observer is not used for full three-dimensional fluid simulation, but for tracking the momentum propagation trend, fluctuation accumulation trend and overall storage change trend of the reservoir water body within the control period.

[0064] The momentum decoupling module inputs the real-time water level in front of the dam, the real-time inflow from upstream, and the current actuator opening data into the fluid momentum state observer to update the current state matrix. The state update here corresponds to a physically meaningful process: the upstream inflow determines the new water volume and the inflow impact, the gate opening determines the discharge boundary conditions, and the water level in front of the dam reflects the comprehensive response of the reservoir area to these two boundary effects.

[0065] Specifically, the state matrix of the fluid momentum state observer is defined as follows:

[0066]

[0067] in, Indicates the current control cycle. Let be the discrete water level vector for each virtual node. This provides real-time inbound traffic data from upstream sources. The discharge flow boundary is calculated based on the current operating data of the implementing agency. This is a vector of equivalent water storage at each node, calculated by combining the parameters of the reservoir capacity curves at each node. Through this matrix dimension design, the system can completely track the evolution of hydrodynamics and water quality within a single framework.

[0068] After the state update, the system decouples from the current state to obtain two types of features: one is the transient fluctuation energy feature, which is used to characterize whether there is strong wave reflection and free surface disturbance on the current water surface; the other is the real storage residual feature, which is used to characterize the degree of deviation of the current total water volume in the reservoir area from the scheduling target.

[0069] Furthermore, the prior fitting module does not rely on data-driven closed models trained on historical samples, but directly generates adaptive PID parameters based on physical priors; among them, the proportional coefficient is used to determine the sensitivity of the controller to the current deviation, so it should be suppressed by the transient fluctuation energy characteristics.

[0070] When the wave surface in the reservoir is chaotic, it indicates that there are many false fluctuations in the sensor readings. At this time, the proportional response should be conservative to prevent the gate from being induced by the waves to operate at high frequency. When the wave surface is stable, the proportional term can recover higher sensitivity, thus responding to the real water level changes more quickly. The integral coefficient is related to the characteristics of the real storage residual, because the integral term is inherently suitable for dealing with long-term accumulated storage deviations, but not suitable for chasing short-term waves.

[0071] By allowing the integral term to follow only the changes in the actual quality residual, the system can achieve stable convergence around the actual reservoir capacity target. The differential coefficients extract the time rate of change of real-time inflow data to form feedforward suppression. This means that when the upstream water flow shows a rapid upward trend, even if the water level in front of the dam has not yet risen significantly, the damping capacity against the flood peak can be increased in advance to suppress subsequent sudden overrush. The mathematical model for updating the differential coefficients of the above feedforward compensation mechanism is as follows:

[0072]

[0073] in, The differential coefficients generated for the current control cycle. As the reference differential coefficient, and These are the real-time inbound flow data for the current and previous control cycles, respectively. To control the cycle duration, For feedforward compensation gain ( ); the formula utilizes The function ensures that the feedforward enhancement suppression of the differential coefficients is triggered only when the upstream water flow shows an upward trend;

[0074] The closed-loop evolution module generates control commands based on the aforementioned proportional coefficient, integral coefficient, and derivative coefficient, and sends them to the servo actuator. The servo actuator can be a hydraulic gate hoist, an electric screw gate hoist, or a winch-type gate hoist. After receiving the control commands, it drives the gate opening to change, thereby forming new outflow boundary conditions. At the same time, the closed-loop evolution module will also continuously extract the steady-state deviation of transient fluctuation energy characteristics over a longer time scale to identify whether the bottom resistance parameter has undergone a continuous shift.

[0075] The mechanism is that if the propagation and attenuation of water surface waves deviate from the original model for a long time under similar inflow and gate conditions, it often means that the siltation at the bottom of the reservoir, local scouring, or changes in the roughness of the bottom bed have altered the wave energy dissipation characteristics. Therefore, the system uses steady-state deviation to correct the bottom resistance parameters, so that subsequent control is always based on the updated fluid static geometric parameters.

[0076] To improve the robustness of the system, if the upstream water level data is missing, the flow data is interrupted, or the current actuator opening data is abnormal during a certain control cycle, the edge computing gateway will prioritize maintaining the previously verified state observer parameters and the PID parameters that have been limited, and will suspend the adaptive correction of the bottom resistance parameters to prevent the topology model from being mistakenly modified due to the lack of key inputs.

[0077] If the sensor malfunction persists for more than the preset time, the system switches to conservative operation mode, retaining only the basic PID parameters and the gate safety opening change rate that have been manually confirmed, until the sensor recovers; if the transient fluctuation energy characteristics are abnormally high and do not match the opening change and the inflow water change, the system can determine that it is a non-process disturbance caused by wind and waves, floating object impact, or local contamination of the sensor. At this time, the proportional term response is limited and the bottom resistance self-learning is prohibited to avoid erroneous adaptation.

[0078] For example, during an upstream flood peak during the flood season, the daytime inflow to a valley-type reservoir continues to rise, and the scheduling objective requires the gates to be gradually opened to release water in order to maintain a safe control water level in front of the dam;

[0079] When the gate is first opened wide, the water level gauge in front of the dam will briefly show obvious fluctuations, which may be mistaken for a drastic change in water level and cause repeated gate corrections.

[0080] In this embodiment, the edge computing gateway identifies that the current fluctuations are mainly manifested as the enhancement of transient fluctuation energy, while the real quality residual still shows that the reservoir capacity is continuously accumulating. Therefore, the system actively reduces the proportional response to prevent the gate from swinging back and forth, while retaining the integral regulation driven by the real quality residual, and increasing the differential feedforward suppression in combination with the upward trend of upstream water.

[0081] As a result, the gate operation is continuous, smooth and responsive, the water level in front of the dam is stably controlled, and the mechanical impact of the gate is significantly reduced. After the flood peak, long-term system records show that the fluctuation decay is slower than in previous years under the same conditions. Therefore, the bottom resistance parameters are gradually corrected to provide a model basis that is closer to the actual bed conditions for subsequent control.

[0082] The purpose of this step is to upgrade reservoir control from a single-point error feedback to a control process oriented towards fluid dynamics characteristics, thereby achieving physical separation between actual reservoir capacity changes and surface wave disturbances, reducing ineffective gate oscillations, and improving control stability and actuator lifespan under flood peak conditions.

[0083] In a preferred embodiment of the present invention, the topology mapping module is specifically used to: divide the reservoir fluid space into a preset number of virtual nodes; allocate the fluid static geometric parameters to each virtual node; establish the physical connection relationship between each virtual node, and generate the fluid space topology model; wherein the fluid static geometric parameters also include reservoir capacity curve parameters.

[0084] This embodiment provides a refined construction mechanism for the topology mapping module; specifically, when only the overall reservoir capacity parameter is used to describe the entire reservoir, although it can initially reflect the total water storage capacity, it cannot express the differences in propagation speed, energy dissipation and water level response between river-type reservoir bays, deep channels, near-dam areas and backwater areas.

[0085] Especially in long and narrow reservoirs or reservoir areas with obvious tributaries, surface waves and inflow momentum do not act synchronously on the entire area. If the entire reservoir is still treated as a whole, the state observer's identification of wave propagation paths will be too coarse, thereby weakening the subsequent decoupling effect. Therefore, this embodiment introduces a virtual node-based fluid space topology model.

[0086] Specifically, the topology mapping module divides the reservoir fluid space into a preset number of virtual nodes; each virtual node is not arbitrarily divided, but is divided according to the actual topography of the reservoir, water depth changes, typical backwater length, tributary confluence location, and gate influence range;

[0087] For example, the upstream main channel inflow area, the reservoir diffusion area, and the near-dam rectification area can be regarded as different node segments. The purpose of this division is to make each node represent a water body area with relatively consistent hydrodynamic behavior. The fluid static geometric parameters are allocated to each virtual node, including the reservoir capacity curve parameters corresponding to the storage and discharge relationship of each node, in addition to the bottom resistance parameters.

[0088] The reservoir capacity curve parameter is used to represent the sensitivity of the corresponding water volume change when the water level rises or falls at the node; the bottom resistance parameter is used to represent the dissipation capacity of the bottom bed at the node to flow and fluctuations; furthermore, the physical connection relationship between each node is established to express the energy transfer from upstream to downstream, echo reflection, and storage-dissipation coupling between adjacent nodes; the topology structure formed in this way not only preserves the spatial propagation characteristics, but also avoids high-complexity fluid solutions.

[0089] In the real physical world, the introduction of reservoir capacity curve parameters is particularly important; because the increase in water level is the same for a narrow, deep channel and a wide, shallow shoal.

[0090] Without reservoir capacity curve parameters, the system may misinterpret water level changes of the same magnitude as storage changes of the same magnitude, and incorrectly estimate the true quality residual. After incorporating reservoir capacity curve parameters into each virtual node, the state observer can distinguish between areas where water level changes seem obvious but the actual increase is small and areas where water level changes are small but volume changes are significant, thus providing a more reliable basis for the true storage of the integral term.

[0091] To improve the robustness of the system, if some remote nodes lack the latest measurement data, the topology mapping module allows the use of historical terrain measurement results as initial geometric parameters, but will mark the node as a low-confidence node and reduce its weight in bottom resistance correction during closed-loop evolution.

[0092] If a sudden event causes the normal connection between nodes to be temporarily distorted, such as a local landslide, accumulation of floating debris, or unconventional diversion, the system can retain the topological skeleton unchanged and only freeze the online correction of the corresponding nodes to avoid frequent changes in the model structure that could cause the observer to become unstable. If the reservoir is small or has a simple geometry, fewer virtual nodes can be used to achieve control. However, in this case, manual configuration should be used to ensure that the area near the dam is at least a separate section to retain the ability to identify gate wave reflections.

[0093] For example, in the aforementioned flood peak dispatching scenario, the valley-type reservoir is long and narrow, with a deep main channel upstream, an expanding bay in the middle reaches, and deep-hole gates and spillway facilities near the dam. The system divides the reservoir area into upstream inflow nodes, midstream energy storage nodes, and near-dam response nodes. The upstream nodes mainly receive the impact of upstream water inflow, the midstream nodes reflect the buffering effect of the reservoir capacity, and the near-dam nodes mainly reflect the local fluctuations and reflections caused by the opening and closing of the gates.

[0094] Because the reservoir capacity curve is steeper in the area near the dam, the same water level fluctuation may not correspond to a large amount of actual water storage change in this area; while a small rise in water level in the bay in the middle reaches may correspond to a large volume change; through this zonal modeling, the system can more accurately identify which changes belong to near-dam waves and which changes belong to the actual water storage of the entire reservoir after the gate is activated.

[0095] The purpose of this step is to provide a basic model with spatial resolution for the fluid momentum state observer, thereby enabling differentiated representation of the hydrodynamic behavior of different reservoir areas and improving the accuracy of identifying true mass residuals and transient fluctuation energy characteristics.

[0096] In a preferred embodiment of the present invention, the momentum decoupling module, when decoupling and generating transient fluctuation energy characteristics and real storage residual characteristics, is specifically used to: perform frequency domain decomposition on the current state matrix of the fluid momentum state observer using an orthogonal decomposition algorithm; and extract high-frequency fast-changing components as transient fluctuation energy characteristics.

[0097] Low-frequency slow-varying components are extracted as the true reserve residual features; among them, transient fluctuation energy features characterize the kinetic and potential energy of the waves on the fluid surface, and true reserve residual features characterize the difference between the current total fluid volume and the target volume preset based on scheduling requirements.

[0098] This embodiment provides a refined mechanism for the momentum decoupling module; specifically, directly driving control based solely on the comprehensive state updated by the state observer may still leave behind the problem of fast fluctuations masking slow storage.

[0099] Especially during the control cycles following a large gate movement, the state in front of the dam is most sensitive to waves. If the rapid and slow changes are not further distinguished, the system may still over-respond to surface waves. Therefore, this embodiment uses orthogonal decomposition to physically separate the current state in the frequency domain. Specifically, the adaptive update mathematical model for the proportional coefficient is as follows:

[0100]

[0101] in, The proportional coefficient generated for the current control cycle. This is the preset baseline ratio coefficient. This represents the current amplitude of the transient fluctuation energy characteristic. The preset attenuation factor ( When wave energy During a surge, It converges rapidly in an exponential manner, thereby suppressing the controller's erroneous response to rapidly changing surface water levels through deterministic mathematical relationships;

[0102] Here, the orthogonal decomposition algorithm specifically employs the discrete wavelet transform algorithm; the edge computing gateway uses a preset wavelet basis function; for example, the Daubechies4 wavelet basis function is preferred, and the decomposition level is set to 3 levels, decomposing the time-series state matrix sequence composed of multiple consecutive control cycles into low-frequency approximation coefficients and high-frequency detail coefficients. Subsequently, the approximation coefficients representing the low-frequency trend are reconstructed as low-frequency slow-varying components, representing the true residual characteristics of the reserves, and the detail coefficients representing high-frequency fluctuations are reconstructed as high-frequency fast-varying components, representing the transient fluctuation energy characteristics, thereby achieving decoupling of the physical signal in the control frequency band;

[0103] Specifically, the role of the orthogonal decomposition algorithm in this embodiment is not to pursue complex mathematical forms, but to separate different physical processes; the high-frequency fast-changing component corresponds to the rapid fluctuations of the water surface in a short period of time, which usually comes from the opening and closing of gates, water hammer effect, local reflection and local disturbance.

[0104] These components are characterized by rapid changes, short duration, and strong impact on the free surface, but do not necessarily represent changes in total water volume. Therefore, they are extracted as transient fluctuation energy characteristics. This characteristic can be understood as how turbulent the water surface in the reservoir is at present, and its strength directly reflects whether the controller is in a stage that is easily misled by waves.

[0105] In contrast, the low-frequency slow-varying component reflects the changing trend of inflow, outflow, water storage and discharge balance over a longer period of time. This type of component is closer to the real change of the total volume of the reservoir area, and is therefore extracted as the real storage residual feature. This feature does not chase instantaneous peaks, but reflects how much the current total fluid volume deviates from the target volume.

[0106] Furthermore, to avoid misinterpreting the current state matrix as containing only static quantities at a single moment, in this embodiment, the current state matrix is ​​preferably a temporal state expression formed by the edge computing gateway combining the observer states of the most recent consecutive control cycles in chronological order after the current control cycle is updated.

[0107] In other words, frequency domain decomposition targets a short-term state sequence ending with the current control cycle, rather than isolated single-frame data. The reason for this is that fast-changing waves and slow-changing reserves are essentially reflected in the laws of time evolution. Only by observing state changes within a preset sliding time window can high-frequency and low-frequency components be stably distinguished.

[0108] When implemented in engineering, the sliding time window can be preset to several continuous sampling points according to the control cycle, the echo propagation time in the reservoir area, and the sensor refresh frequency. When the window moves forward with the arrival of a new cycle, the system always performs an orthogonal decomposition once with the state sequence in the latest window, thereby keeping the decoupling result synchronized with the current control action.

[0109] Furthermore, in this embodiment, the transient wave energy characteristics are not directly equivalent to the water level amplitude at a single point, but are preferably obtained by comprehensively considering the amplitude, duration, and spatial distribution of the high-frequency fast-changing components at each virtual node; this can avoid individual measurement point spike noise being misjudged as full-reservoir wave enhancement.

[0110] The actual storage residual characteristics are preferably calculated based on the low-frequency slow-varying components combined with the reservoir capacity curve parameters corresponding to each virtual node, so that they represent the equivalent volume deviation relative to the target volume, rather than simply the low-frequency average water level change. In this way, the two types of characteristics output by the momentum decoupling module correspond to the strength of short-term surface disturbance and the degree of long-term overall storage and discharge deviation, respectively, which are consistent with the objects of action of the subsequent proportional and integral terms.

[0111] From an engineering perspective, this high- and low-frequency separation conforms to the actual response pattern of a reservoir; the waves on the free surface can swing up and down significantly in a short period of time, but if the total inflow and total outflow do not change accordingly, the total water storage does not change significantly in sync.

[0112] Conversely, when the water inflow increases continuously but the surface remains relatively calm, the actual reservoir capacity may continue to deviate from the target value. By separating these two aspects, the controller can assign the suppression of short-term illusions and the correction of long-term real deviations to different PID components.

[0113] To improve the robustness of the system, if the sampling density is insufficient in the current control cycle, resulting in unclear boundaries between fast and slow components, the momentum decoupling module will prioritize maintaining the previous verified decomposition scale to avoid control jitter caused by frequent switching of the decoupling scale. If a large-scale wind and wave caused by extreme strong winds significantly enhances the high-frequency fast component, but the upstream water flow and gate boundary changes are relatively stable, the system will mark this situation as dominated by environmental disturbances and allow the proportional term to be further restricted, while not including this abnormal fluctuation in the long-term bottom resistance correction.

[0114] If the low-frequency slow-varying component is significantly contradictory to the reservoir capacity scheduling boundary, for example, if it shows continuous water storage but the gate discharge and upstream water inflow do not support this conclusion, the system can trigger a data consistency check and prioritize checking for zero drift of the water level gauge or failure of the flow meter.

[0115] For example, in the aforementioned flood peak operation, the gate is continuously opened to receive the upstream flood peak, and a visible reflected wave train appears on the water surface near the dam area; at this time, the updated comprehensive state of the state observer includes both wave fluctuations and actual changes in the inflow balance.

[0116] The momentum decoupling module identifies short-cycle repeated rises and falls as high-frequency fast-changing components and submits them as transient fluctuation energy characteristics to the proportional term suppression mechanism; it identifies the overall storage increase trend caused by the continuous inflow of flood peaks as low-frequency slow-changing components and submits them as true storage residual characteristics to the integral term.

[0117] In this way, the controller will not close the gate in the opposite direction due to temporary fluctuations in the water level near the dam, but will be able to continuously track the actual water storage pressure brought by the flood peak.

[0118] To further clarify, to maintain consistency with the detailed description in this embodiment, the current state matrix in this embodiment still corresponds to the current state matrix obtained after the state update. However, when performing frequency domain decomposition, the current state matrix is ​​organized using a short-time sliding window method that includes the current period to form a time-series matrix expression that can separate fast and slow components.

[0119] Therefore, the temporal state representation here is to further define the data organization form of the current state matrix, rather than introducing a new object independent of the current state matrix; accordingly, the high-frequency fast-changing component and the low-frequency slow-changing component mentioned in the text are only used as sources of description of transient fluctuation energy characteristics and real reserve residual characteristics, respectively, and are not confused with the proportional and integral terms in the subsequent PID.

[0120] The purpose of this step is to separate the different physical processes that are mixed in the same set of sensor data, so as to enable the controller to be immune to short-term wave disturbances and to accurately correct long-term reservoir capacity deviations.

[0121] In a preferred embodiment of the present invention, the prior fitting module, when calculating the proportional coefficient based on the transient fluctuation energy characteristics, is specifically used to: obtain a preset benchmark proportional coefficient and a preset attenuation factor; calculate the product of the attenuation factor and the transient fluctuation energy characteristics; calculate the negative exponential attenuation weight with the natural constant as the base and the negative number of the product as the exponent; and multiply the benchmark proportional coefficient by the negative exponential attenuation weight to obtain the proportional coefficient.

[0122] This embodiment provides a refined scheme for the adaptive generation mechanism of the proportional coefficient. Specifically, if the proportional coefficient is always kept at a fixed value, the response may be appropriate when the water surface is stable, but in the stage of strong waves and heavy reflection, the fixed proportional gain will be too sensitive and will easily take the short-term surface fluctuations as real deviations that must be corrected immediately, thereby pushing the gate into high-frequency reciprocating motion. Based on this, this embodiment introduces a negative exponential decay mechanism driven by the transient wave energy characteristics.

[0123] Specifically, the baseline proportional coefficient represents the basic response capability that the control system is expected to have when the wave surface in the reservoir area is relatively stable and the reliability of the measurement value is high; the attenuation factor is used to characterize the system's conscientiousness to wave disturbances; when the transient wave energy characteristics increase, the negative exponential attenuation weight will cause the proportional coefficient to decrease rapidly but continuously, rather than abruptly cut off.

[0124] The significance is that the more violent the waves, the more the controller should reduce its direct amplification of instantaneous deviations, making the gate action more restrained, so as to wait for the real water volume changes to appear in subsequent cycles. The reason for using negative exponential decay is that in engineering control, when the waves change from weak to strong, the system's need to reduce proportional sensitivity does not increase linearly, but should become conservative quickly after reaching a certain disturbance level to prevent the controller from being pulled by high-frequency illusions.

[0125] This mechanism does not negate the proportional term, but rather endows it with the ability to adapt to changes in the environment. When the wave surface is gentle, the proportional coefficient is close to the reference value, and the system can still correct itself quickly. When the wave surface is violent, the proportional coefficient converges significantly, and the system prioritizes maintaining the stability of the gate and mechanical safety. This change conforms to the physical laws of free liquid flow and can also directly protect the actuator.

[0126] To improve system robustness, if the transient wave energy characteristics are abnormally high and the duration is extremely short, the system can first adopt a short-term confirmation mechanism to avoid mistaking a single impact noise for a continuous wave. If the wave energy characteristics remain high for a long time, it indicates that the reservoir area is always under intense disturbance. At this time, although the proportional coefficient is suppressed, the closed-loop evolution module still needs to retain the minimum safe proportional response to prevent the controller from completely losing its basic perception of the real rapid deviation. If the transient wave energy characteristics fail to be acquired, the proportional coefficient will fall back to the conservative benchmark range set manually and a maintenance prompt will be triggered instead of continuing to execute adaptive decay.

[0127] For example, in the aforementioned reservoir flood peak scheduling scenario, after the gate is opened at a higher degree for a short period of time to facilitate flood discharge, the surface wave near the dam is significantly enhanced;

[0128] The system recognizes that such fluctuations are mainly caused by free surface disturbances due to gate operation, rather than by a sudden increase in upstream water flow leading to an instantaneous change in the actual reservoir capacity. Therefore, it rapidly reduces the proportional coefficient according to the negative exponential decay law, so that the controller no longer immediately reverses the gate for every peak and trough of the water level in front of the dam. As the wave surface gradually calms down, the proportional coefficient is then smoothly restored, thereby maintaining the agility of subsequent control.

[0129] The purpose of this mechanism is to adaptively adjust the proportional response intensity according to the strength of the waves in the reservoir area, thereby achieving rapid suppression of surface wave-induced errors and avoiding high-frequency reciprocating oscillations of the gate.

[0130] In a preferred embodiment of the present invention, the prior fitting module, when calculating the integral coefficient based on the actual reserve residual characteristics, is specifically used to: obtain a preset benchmark integral coefficient; input the actual reserve residual characteristics into a preset Sigmoid nonlinear driving function to generate integral driving weights; and multiply the benchmark integral coefficients by the integral driving weights to obtain the integral coefficients.

[0131] This embodiment provides a refined scheme for the adaptive generation mechanism of integral coefficients; specifically, when working only based on the accumulation of instantaneous errors, the traditional integral term is prone to integral accumulation distortion in the stage of frequent wave reflection and obvious false fluctuations in water level, and then concentrated release after the waves disappear, causing the gate to continue to over-operate.

[0132] Even if the basis of the integral term has been shifted to the actual quality residual, if the integral coefficient is still a fixed value, it may react too strongly in the small deviation stage and increase too sharply in the large deviation stage, which is not conducive to balancing fine adjustment and rapid handling of peaks; therefore, this embodiment uses the Sigmoid nonlinear driving function to form the integral driving weight.

[0133] Specifically, the actual reserve residual characteristic represents the degree of deviation between the current total volume and the target volume; the integral term's role is to slowly and steadily correct this persistent imbalance between storage and release; the physical meaning of the Sigmoid nonlinear drive is that when the actual quality residual is small and the storage capacity is close to the target, the integral drive weight remains at a relatively mild level, avoiding excessive accumulation of small residuals by the integral term; the specific expression of the preset Sigmoid nonlinear drive function is as follows:

[0134]

[0135] in, As integral-driven weights, This represents the absolute value of the residual characteristics of the actual reserves. For nonlinear driving slope parameters, This is the preset residual tolerance threshold;

[0136] As the actual quality residual continues to increase, the integral-driven weight gradually strengthens, making the controller more determined to push the gate to continue moving in a direction that is conducive to restoring the target storage capacity.

[0137] When the deviation is already large, the driving weight tends to be smooth and will not increase indefinitely, thus avoiding the oversaturation of the integral term and causing long-term lag. This characteristic of being sensitive in the middle and smooth at both ends is in line with the reservoir scheduling law, because when the reservoir capacity deviation is close to the target, stability should be emphasized, and when the deviation is widening, correction should be emphasized, but the correction force still needs to be constrained by engineering safety.

[0138] Through this mechanism, the integral term truly becomes a regulating component oriented towards real changes in reserves, rather than a slow follower of surface waves; this can significantly reduce the risk of integral saturation and reverse overshoot, and is particularly suitable for reservoirs, which are large inertia and long-term regulating objects.

[0139] To improve the robustness of the system, if the real reserve residual characteristics change sign frequently in a short period of time, it indicates that the reservoir area is in a small-range fluctuation state near the target. At this time, the system can limit the speed of change of the integral-driven weight to avoid damage to the stability of the integral term due to switching back and forth.

[0140] If a large real quality residual exists for a long period of time but the gate is close to the mechanical or scheduling limit, the system should retain the upper limit constraint of the integral coefficient and output a warning of insufficient control margin to the upper-level scheduling system, instead of simply continuing to amplify the integral effect; if the real reserve residual characteristics cannot be obtained stably for the time being, the integral coefficient should be returned to the conservative benchmark level to prevent integral drift caused by incorrect reserve judgment.

[0141] For example, during the same flood peak scheduling process, although the gates had been gradually increased in the early stages, the system identified that the actual storage capacity of the entire reservoir was still slowly deviating from the target due to the continuous increase in upstream water flow.

[0142] At this point, although the proportional response has become restrained due to the instability of the wave surface near the dam, the integral term will not become invalid. Instead, it will gradually increase the correction intention based on the actual quality residual, so that the gate maintains a stable trend in the direction of increasing discharge in multiple control cycles. When the flood peak recedes and the actual quality residual decreases, the integral driving weight will smoothly decrease again to prevent the gate from continuing to be over-opened when the reservoir capacity is close to the target.

[0143] The purpose of this mechanism is to ensure that the integral term works only around the deviation of the actual storage, and to achieve robust regulation of the reservoir's long-term storage and release balance by taking into account both the stability of small deviations and the ability to correct large deviations through nonlinear driving.

[0144] In a preferred embodiment of the present invention, when the prior fitting module extracts the time change rate of the real-time inbound flow data to calculate the differential coefficient, it is specifically used to: calculate the time derivative of the real-time inbound flow data within the current control cycle to obtain the time change rate; and add the time change rate as a feedforward suppression compensation term to a preset benchmark differential coefficient to calculate the differential coefficient.

[0145] This embodiment provides a refined scheme for the adaptive generation mechanism of differential coefficients; specifically, if the differential term works only based on the rate of change of the water level in front of the dam, although it can suppress rapid changes to a certain extent, its observation object is still mainly located near the dam measuring point, which is easily affected by wave fluctuations.

[0146] For situations like flood peak inflow with obvious upstream precursors, relying solely on changes in upstream measuring points to adjust the differential term is often too late. To compensate for this deficiency, this embodiment introduces the time variation rate of real-time inflow into the differential coefficient calculation, giving the differential term a feedforward suppression function.

[0147] Specifically, the time derivative of the real-time inflow within the current control cycle reflects the rate of change in upstream water flow. If the inflow is rising rapidly, it indicates that the flood peak is approaching, and the water and momentum input that the reservoir will bear in the next few cycles will increase significantly.

[0148] By incorporating the rate of change of time as a feedforward suppression compensation term into the baseline differential coefficient, the controller is essentially made to have a damping awareness of the impending shock in advance, rather than passively responding only after the water level in front of the dam changes significantly.

[0149] This differs from the general differential term which only suppresses changes in error that have already occurred. This embodiment is more physically consistent with the propagation law of floodwater in a reservoir. Since changes in upstream cross-section flow usually precede changes in water level in front of the dam, the differential term can adjust the changing trend of gate action in advance, reducing the risk of over-rush when the flood peak reaches the vicinity of the dam.

[0150] Meanwhile, the time rate of change, as a feedforward compensation term, can also reduce the dependence of the differential term on local water level noise. Its core is not to amplify any short-term changes, but to identify whether the inflow is becoming rapid or slow. When the inflow increases significantly, the differential term enhances damping and pre-control. When the inflow changes slowly, the differential term returns to the basic level, avoiding unnecessary increase in control rigidity.

[0151] To improve the robustness of the system, if upstream flow data experiences short-term packet loss, discrete jumps, or interference from floating objects, the system should first perform a continuity check on the flow change trend. The feedforward compensation should only be activated if the change trend is higher than the measurement noise level and consistent for several consecutive cycles. If the predicted rate of change is lower than the noise background, the system should pause the feedforward enhancement and retain only the reference differential coefficient to avoid mistaking the measurement noise as a precursor to the flood peak.

[0152] If the upstream section sensor fails completely, the differential term reverts to the conservative mode dominated by the upstream feedback and sends a data degradation prompt to the dispatch terminal.

[0153] For example, in the aforementioned main scenario, rainfall in the upstream tributary caused a significant steep rise in the inflow curve, but the water level in front of the dam had not yet fully responded;

[0154] The system detects from the upstream flow sensor that the momentum of the incoming water is rapidly increasing. Therefore, it incorporates this trend as a feedforward compensation into the differential coefficient, so that the gate control has a stronger overrush suppression capability before the water level in front of the dam rises significantly. In this way, when the flood peak actually arrives near the dam area, the gate will not be forced to suddenly and drastically move due to insufficient preparation.

[0155] The purpose of this mechanism is to introduce the leading information of upstream water flow into the differential regulation process, thereby achieving early suppression of flood peak impact and prevention of excessive water level sluice near the dam.

[0156] In a preferred embodiment of the present invention, when the closed-loop evolution module extracts the steady-state deviation of transient fluctuation energy characteristics and uses the steady-state deviation to correct the bottom resistance parameter in the fluid static geometric parameters, it is specifically used to: obtain the fluctuation frequency characteristics of transient fluctuation energy characteristics within a preset time window; calculate the difference between the fluctuation frequency characteristics and the preset reference frequency characteristics as the steady-state deviation.

[0157] The bottom resistance parameters of the corresponding virtual nodes in the fluid space topology model are updated based on the steady-state deviation to adaptively compensate for the boundary morphology evolution trend at the bottom of the fluid.

[0158] This embodiment provides a long-term adaptive correction mechanism for the closed-loop evolution module; specifically, if the system only adjusts the PID parameters according to the instantaneous state in each control cycle without updating the static geometric parameters of the fluid, then as the reservoir's operating years increase, the sedimentation, scouring, and local roughness changes will gradually alter the propagation and attenuation patterns of waves.

[0159] At this point, although the original topology model is effective in the early stages of operation, it will become disconnected from the actual reservoir area after long-term operation, resulting in a decrease in decoupling accuracy and a weakening of control effect; therefore, this embodiment introduces a bottom resistance parameter correction mechanism based on the steady-state deviation of transient fluctuation energy characteristics.

[0160] Specifically, the system continuously acquires the frequency characteristics of transient fluctuation energy characteristics within a preset time window; the frequency characteristics here reflect: under similar inflow and gate operation conditions, how long does the free liquid surface fluctuation in the reservoir area decay, reflect, and redistribute?

[0161] If the bed is rougher, the sediment is thicker, or the local topography changes, the wave propagation resistance and reflection path will change accordingly, and its frequency performance in the time domain will deviate from the historical benchmark. Therefore, the system compares the wave frequency characteristics in the current window with the preset benchmark frequency characteristics to obtain the steady-state deviation.

[0162] This steady-state deviation is not used to directly drive the gate, but rather serves as a basis for reflecting the long-term differences between the model and reality. The bottom resistance parameters of the corresponding virtual nodes in the fluid space topology model are updated based on this steady-state deviation, so that the model can once again closely match the current bed boundary conditions.

[0163] Furthermore, the update in this embodiment is not fixed as a one-way increase or one-way decrease, but the correction direction is determined based on the direction of propagation and dissipation change reflected by the steady-state deviation. If, under similar boundary conditions, the wave attenuation of the target node is faster than the reference and the reflection duration is shorter, it indicates that the equivalent dissipation capability of the node is enhanced, and the corresponding bottom resistance parameter can be adjusted upward in small steps.

[0164] If, under similar boundary conditions, the wave attenuation at the target node is slower and the reflection duration is longer than the baseline, it indicates that the equivalent dissipation capacity of the node is weakened. In this case, the corresponding bottom resistance parameter can be adjusted downwards in small steps. The system updates the parameters based on the following first-order linear adaptive correction law:

[0165]

[0166] in, and The first The bottom resistance parameters after and before the correction of each virtual node. This represents the steady-state deviation corresponding to the wave dissipation index within this node, with the sign of the value consistent with the direction of change in dissipation capacity. Update the step size for a preset minimum constant, such as By introducing a minimal update step size This limits the magnitude of a single evolution, thereby ensuring the mathematical convergence and stability of the long-term bed evolution process of the topological model;

[0167] If only the frequency structure changes but the total attenuation level does not change significantly, the system will prioritize reducing the correction magnitude or only record the trend without rewriting the parameters, so as to avoid simply misprojecting the change in propagation path as a change in drag strength. Through this correction rule that is consistent with the deviation direction and parameter direction, the bottom drag parameter and the wave energy dissipation capability it represents will remain consistent in meaning.

[0168] This correction mechanism has obvious engineering rationale; because the bottom resistance parameter essentially corresponds to the bed's ability to dissipate water flow and fluctuations, it does not change abruptly like the water level, but evolves slowly; by basing its correction on a longer time window and steady-state deviation, it can avoid mistaking a single wave or a one-time operational shock as a change in the bed; in this way, the control system can perform closed-loop gate control in the short term, and dynamically update the model parameters of the bed state in the reservoir area in the long term, achieving coordinated adaptation between control and environmental evolution;

[0169] To improve the robustness of the system, if a significant abnormal event occurs within the preset time window, such as continuous strong winds, earthquake disturbances, large-scale accumulation of floating objects, or unconventional scheduling tests, the fluctuation frequency characteristics within that window should not be used for bottom resistance correction, so as to avoid mistakenly solidifying the sudden environmental event as a change in the bottom parameters.

[0170] If the steady-state deviation of a certain virtual node persists for a long time but differs too much from that of adjacent nodes, the system should first prompt manual verification of whether the node has sensor deviation or local engineering changes, rather than immediately making a large correction; if the current window data is insufficient to form reliable frequency characteristics, the original bottom resistance parameters should be kept unchanged, and only recorded for subsequent accumulation;

[0171] For example, after the aforementioned reservoir has been operating continuously for a flood season, the system found that the rate of decrease in water surface fluctuations at the near-dam node and the midstream node under similar inflow and gate opening conditions was significantly slower than the historical benchmark, and this change was not caused by monsoon waves.

[0172] Based on this, the closed-loop evolution module judges that the boundary conditions at the bottom of the reservoir have changed over a long period of time. Therefore, it slowly lowers the bottom resistance parameters of the corresponding virtual nodes or redistributes the equivalent resistance levels between adjacent nodes so that the dissipation capacity represented in the model is consistent with the actual observation. After the update, the state observer is more accurate in separating the subsequent wave energy and the true mass residual, and the gate control is restored to a more stable state.

[0173] Furthermore, in this embodiment, the virtual node division of the corresponding virtual nodes maintains a one-to-one correspondence. That is, when the closed-loop evolution module corrects the bottom resistance parameters, it uses the node segment to which the fluctuation frequency characteristics belong as the basic correction unit and does not arbitrarily merge or replace parameters across nodes that have not established physical connections. The target nodes, corresponding virtual nodes, and adjacent nodes mentioned in the text are all referred to based on the same set of topology mapping results.

[0174] Accordingly, the evolution trend of boundary morphology, changes in subsurface boundary conditions, and long-term changes in bottom sediment or roughness all point to the same long-term physical causes used to explain the slow correction of bottom resistance parameters in this embodiment, without introducing any new correction target parameters.

[0175] The purpose of this mechanism is to enable the control model to adapt to the long-term evolution of the reservoir bed boundary, thereby achieving continuous tracking of long-term operating condition changes and preventing the control performance from deteriorating due to model aging.

[0176] In a preferred embodiment of the present invention, the system is applied to a water conservancy reservoir environment; wherein, the real-time reservoir water level data is the water level data in front of the dam, the real-time inflow data is the upstream inflow data, and the current actuator opening data is the reservoir gate opening data; wherein, the servo actuator is a hydraulically driven gate actuator or a motor-driven gate actuator.

[0177] This embodiment provides a mechanism for defining application scenarios and hardware boundaries. Specifically, this invention is not intended for closed full-pipe flow systems, nor is it applicable to pressure vessel regulation scenarios without free liquid surfaces. Instead, it is specifically applied to water conservancy reservoir environments with free water surfaces, wave propagation, and echo reflection. In this environment, the upstream water level, upstream inflow, and gate opening together constitute the core physical quantities required for control.

[0178] Specifically, the real-time reservoir water level data is preferably the water level data in front of the dam; because the area in front of the dam is the area most directly affected by the gate boundary and is also a key observation position for dispatching safety control, which can reflect the response of the water surface near the dam in a timely manner; the real-time inflow data is preferably the upstream inflow data, which serves to provide leading information on the flood and the boundary conditions of the total inflow.

[0179] The preferred data for the current actuator opening is the reservoir gate opening, which represents the outflow regulation action that the system has already applied. Only when all three are input together can a complete closed-loop chain of inflow-reservoir response-outflow execution be formed in the state observer. If any one of these quantities is missing, the system will find it difficult to accurately determine whether the current water level change is due to increased inflow, gate action, or surface fluctuations.

[0180] Regarding servo actuators, hydraulically driven gate actuators are suitable for scenarios with large loads and rapid opening and closing, especially for deep-hole gates and floodgates requiring large opening and closing forces; motor-driven gate actuators are suitable for small and medium-sized gates or scenarios with high requirements for smooth opening and closing speeds; regardless of the type of actuator used, the control logic of this embodiment requires the actuator to be able to transmit the actual opening degree or equivalent displacement so that the edge computing gateway can incorporate it as a boundary condition into the state update.

[0181] To improve the robustness of the system, if the accuracy of the water level gauge in front of the dam decreases due to fog, floating objects or echo obstruction, the system can temporarily use redundant water level gauge data as a substitute, but it must not be separated from the measuring point in front of the dam for a long time and rely solely on other cross sections for estimation.

[0182] If the upstream inflow cannot be directly measured, the upstream cross-section estimate can be used when hydrological station data and a reliable conversion model are available. However, the confidence level must be marked and the differential feedforward compensation strength must be limited accordingly. If the actuator can only receive control in one direction and lacks effective opening feedback, the system can only operate in a degraded manner and it is not advisable to enable the complete momentum observation and adaptive model correction functions.

[0183] For example, in the aforementioned valley-type reservoir, radar level gauges are installed in front of the dam for continuous observation of water surface elevation, ultrasonic flow monitoring devices are installed at the upstream inlet section or the flow is measured by a hydrological station, and the gate actuator is equipped with a hydraulic hoist and an opening encoder.

[0184] The edge computing gateway periodically collects these three types of data through the industrial bus to form a complete input chain for the control of the free liquid surface in the reservoir. If the system is transplanted to a closed pressure pipeline, it will not fall within the application boundary of this embodiment because there is no decoupling logic between surface wave energy and the actual reservoir capacity.

[0185] The purpose of this step is to clarify the applicable objects, core measuring points and actuator types of the present invention, so as to ensure that the control logic is consistent with the physical mechanism of the free liquid surface in actual water conservancy reservoirs.

[0186] In a preferred embodiment of the present invention, the boundary morphology evolution trend is the evolution trend of the sediment deposition thickness at the bottom of the reservoir.

[0187] This embodiment provides a specific mechanism for the evolution trend of boundary morphology; specifically, the long-term correction of the bottom resistance parameter can refer to the change of the bottom boundary conditions, but if its main source is not further clarified, it is difficult to reflect the typical engineering significance of this correction mechanism in the reservoir environment.

[0188] Considering that the main boundary morphology evolution of most reservoirs during long-term operation comes from sediment deposition and local scouring, this embodiment specifically defines the boundary morphology evolution trend as the evolution trend of sediment deposition thickness at the bottom of the reservoir.

[0189] Specifically, siltation can alter the elevation, roughness, and local cross-sectional shape of the reservoir bottom, changing the propagation and reflection paths of surface waves and the efficiency of momentum transfer between different nodes. As siltation gradually raises the bottom, the shallow water area expands, and the wave propagation speed and attenuation characteristics change. When siltation embankments or scour channels are formed locally, certain frequencies of waves are more easily reflected or retained.

[0190] Therefore, the long-term frequency shift of transient fluctuation energy characteristics is not only an anomaly at the control level, but can also serve as indirect observation information on the evolution of bottom sediment morphology. By accumulating such shift patterns over a long period of time, the system corrects the bottom resistance parameters of virtual nodes, which essentially transforms control operation data into a dynamic perception of the trend of sediment thickness change.

[0191] Furthermore, it should be understood that the relationship between the increase in sediment deposition thickness and the change in bottom resistance parameters is not a simple unidirectional one at all nodes; for wide and shallow shoal areas, the increased roughness and shallow water dissipation after siltation may result in an upward adjustment of the equivalent bottom resistance parameter; for areas where the main channel is filled with silt, the reflection path is extended, or the local rectification capacity is reduced, it may also result in a downward adjustment of the equivalent bottom resistance parameter, or a change in the resistance distribution relationship between adjacent nodes.

[0192] Therefore, in this embodiment, the evolution trend of sediment deposition thickness is understood as the physical source driving the change of bed boundary conditions, while the specific parameter correction direction is still based on the actual propagation and attenuation results corresponding to the steady-state deviation, thereby avoiding mechanically equating all deposition situations to the same direction of correction.

[0193] It should be understood that the trend of sediment thickness evolution here emphasizes trend identification rather than providing absolute topographic measurement values ​​as a replacement for dedicated depth sounding equipment; without the need for additional interruption of operation for high-frequency underwater measurement, the control system can continuously sense which reservoir nodes are experiencing changes in bottom dissipation characteristics and optimize the control model accordingly; for the scheduling and maintenance department, this trend can also serve as a basis for arranging dredging, re-measurement, or local engineering inspections.

[0194] To improve the robustness of the system, factors such as continuous strong winds, ship disturbances, floating debris, or temporary construction cofferdams in the reservoir area during a certain period may also cause changes in the frequency characteristics of the fluctuations.

[0195] Therefore, the system only attributes similar offsets to sedimentation trends when they occur repeatedly across multiple time windows and operating conditions, and when they are consistent with the logic of changes in inflow and gate conditions. If the offset is sudden and unreproducible, it is considered an external event and is not used to determine the sedimentation thickness trend. If there is a significant difference between the subsequent manual sounding results and the system's inference, the corresponding virtual nodes are corrected based on the manual sounding results, and the reference frequency characteristics are updated.

[0196] For example, after the aforementioned reservoir operated for two consecutive flood seasons, the system found that the reflection lag time of surface waves at the midstream bay node was stably longer than that in the initial operation stage under the condition of receding flood peaks, and this phenomenon was repeated under multiple similar inflow processes.

[0197] Based on the historical situation of large sediment-laden water inflow in the upstream of the reservoir area, the system interprets this long-term offset as a trend of increasing sediment deposition thickness in the midstream reservoir bay, and makes small-step corrections to the bottom resistance parameters of the node according to the actual dissipation changes corresponding to the steady-state deviation.

[0198] The maintenance department can then arrange key depth measurements and verifications based on this information, thereby transforming the long-term operational information of the control system into an auxiliary understanding of the evolution of sedimentation.

[0199] The purpose of this mechanism is to link the adaptive correction of the bottom resistance parameters with the real engineering evolution process of reservoir sediment deposition, thereby enabling the long-term self-updating of the control model and providing usable information on the changing trends of the bottom bed for reservoir operation and maintenance.

[0200] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. An adaptive PID control system for a water conservancy reservoir environment, characterized in that, This includes edge computing gateways, water level sensors, flow sensors, and servo actuators for driving the reservoir gates. The water level sensor, flow sensor, and servo actuator are used to collect real-time reservoir water level data, real-time inflow data, and current actuator opening data, respectively, and send the collected data to the edge computing gateway. The edge computing gateway, which operates according to a preset control cycle, is equipped with a topology mapping module, which is used to acquire preset fluid static geometric parameters including bottom resistance parameters, construct a fluid space topology model, and establish an initial state matrix of the fluid momentum state observer based on the fluid space topology model. The momentum decoupling module is used to input the received real-time reservoir water level data, real-time inflow data, and current actuator opening data into the fluid momentum state observer, drive the initial state matrix to update the state to obtain the current state matrix, and perform frequency domain decomposition on the current state matrix to separate the fast-changing component representing surface waves and the slow-changing component representing reservoir capacity changes, thereby decoupling and generating transient fluctuation energy characteristics and true storage residual characteristics. The prior fitting module is used to calculate the proportional coefficient based on the transient fluctuation energy characteristics, calculate the integral coefficient based on the real storage residual characteristics, and calculate the differential coefficient by extracting the time change rate of the real-time inflow data. The closed-loop evolution module is used to generate control commands based on the proportional coefficient, the integral coefficient, and the derivative coefficient and send them to the servo actuator, and extract the steady-state deviation of the transient fluctuation energy characteristics to correct the bottom resistance parameter in the fluid static geometry parameters; The servo actuator is also used to receive the control command and drive the reservoir gate to move.

2. The adaptive PID control system for a water conservancy reservoir environment according to claim 1, characterized in that, The topology mapping module is specifically used for constructing the fluid space topology model: The reservoir fluid space is divided into a preset number of virtual nodes; The fluid static geometry parameters are assigned to each of the virtual nodes; Establish the physical connection relationships between each of the virtual nodes to generate the fluid space topology model; The fluid static geometric parameters also include the reservoir capacity curve parameters.

3. The adaptive PID control system for a water conservancy reservoir environment according to claim 1, characterized in that, The momentum decoupling module, when decoupling the transient fluctuation energy characteristics and the actual storage residual characteristics, is specifically used for: The current state matrix of the fluid momentum state observer is decomposed in the frequency domain using an orthogonal decomposition algorithm. High-frequency rapidly changing components are extracted as the transient fluctuation energy characteristics. Low-frequency, slowly varying components are extracted as the residual features of the actual reserves; The transient fluctuation energy characteristic represents the kinetic and potential energy of the waves on the fluid surface, and the real storage residual characteristic represents the difference between the current total fluid volume and the target volume preset based on scheduling requirements.

4. The adaptive PID control system for a water conservancy reservoir environment according to claim 1, characterized in that, When calculating the proportionality coefficient based on the transient fluctuation energy characteristics, the prior fitting module is specifically used for: Obtain the preset baseline scaling factor and the preset attenuation factor; Calculate the product of the attenuation factor and the transient fluctuation energy characteristic; Using the natural constant as the base and the negative of the product as the exponent, the negative exponential decay weight is calculated. The benchmark scaling factor is multiplied by the negative exponential decay weight to obtain the scaling factor.

5. The adaptive PID control system for a water conservancy reservoir environment according to claim 1, characterized in that, When calculating the integral coefficients based on the actual reserve residual characteristics, the prior fitting module is specifically used for: Obtain the preset baseline integral coefficients; The actual reserve residual features are input into a preset Sigmoid nonlinear driving function to generate integral driving weights; The integral coefficient is obtained by multiplying the baseline integral coefficient by the integral driving weight.

6. The adaptive PID control system for a water conservancy reservoir environment according to claim 1, characterized in that, When the prior fitting module extracts the time change rate of the real-time inbound flow data to calculate the differential coefficient, it is specifically used for: Calculate the time derivative of the real-time inbound flow data within the current control cycle to obtain the time change rate; The time change rate is used as a feedforward suppression compensation term and added to a preset benchmark differential coefficient to calculate the differential coefficient.

7. An adaptive PID control system for a water conservancy reservoir environment according to claim 2, characterized in that, When the closed-loop evolution module extracts the steady-state deviation of the transient fluctuation energy characteristics and uses the steady-state deviation to correct the bottom drag parameter in the fluid static geometry parameters, it is specifically used for: Obtain the fluctuation frequency characteristics of the transient fluctuation energy characteristics within a preset time window; The difference between the fluctuation frequency characteristic and the preset reference frequency characteristic is calculated as the steady-state deviation; The bottom resistance parameters of the corresponding virtual nodes in the fluid space topology model are updated based on the steady-state deviation to adaptively compensate for the boundary morphology evolution trend of the fluid bottom.

8. The adaptive PID control system for a water conservancy reservoir environment according to claim 1, characterized in that, The system is applied to water conservancy and reservoir environments; Among them, the real-time reservoir water level data is the water level data in front of the dam, the real-time inflow data is the upstream inflow data, and the current actuator opening data is the reservoir gate opening data; The servo actuator is either a hydraulically driven gate actuator or a motor-driven gate actuator.

9. An adaptive PID control system for a water conservancy reservoir environment according to claim 7, characterized in that, The boundary morphology evolution trend is the evolution trend of the sediment deposition thickness at the bottom of the reservoir.

Images