Intelligent collaborative control method, device and product of a hydraulic residual pressure power generation system

Abstract

The application provides an intelligent collaborative control method, device and product of a hydraulic residual pressure power generation system, comprising the following steps: acquiring system data of a pipe network system, a time-of-use electricity price signal and a demand response instruction; based on historical water consumption data, weather information and calendar information, a time series prediction model is used to rollingly predict a water consumption load curve of a future set period; based on the water consumption load curve, the electricity price signal, the demand response instruction and the system data, a candidate strategy set containing multiple operation modes is pre-generated in a rolling time window; in each control period, the candidate strategy set is real-time optimized and adjusted according to the latest system data, and a target control instruction set of the current control period is output. The problems of single traditional control strategy and slow response are effectively overcome, the maximization of power generation income, water supply reliability and safe operation of equipment can be considered under a complex dynamic environment, so that the collaborative optimization of multiple targets is realized.

Description

Technical Field

[0001] This invention relates to the field of energy recovery and water and energy conservation technology, and in particular to an intelligent collaborative control method, equipment and product for a hydropower generation system with residual pressure. Background Technology

[0002] With the acceleration of urbanization and increasingly severe energy resource constraints, energy conservation and consumption reduction in water supply systems have become a key aspect of the green transformation of urban infrastructure. In traditional municipal water supply, industrial circulating water, and building secondary pressurization systems, a large amount of residual water pressure often exists in the pipeline network to ensure the flow and pressure at the end. This portion of energy, which could be recovered, is currently mostly dissipated as heat energy through pressure reducing valves, throttling devices, etc., which not only causes continuous energy waste but may also lead to pipeline vibration and noise. There is significant room for improvement in system energy efficiency. Therefore, efficiently recovering and utilizing hydraulic residual pressure and converting it into electrical energy is an important technical direction for achieving energy recycling in water systems and reducing operating costs.

[0003] Currently, hydropower generation mainly relies on two technological approaches: one is to use dedicated turbines for energy conversion, and the other is to generate electricity by reversing the operation of existing pumps (pump-driven turbine mode). The former is technologically mature but involves high equipment investment and complex installation, and is particularly uneconomical in low-head, low-flow scenarios; the latter has low modification costs and good adaptability, but its control strategy is relatively simple. Regardless of the technological approach, existing systems generally suffer from insufficient intelligence in control: the operating modes are singular and rigidly switched, unable to dynamically optimize based on diverse information such as water load, grid electricity price, and equipment status; hydraulic and electrical shocks are easily generated during the transition between pumping and power generation modes, threatening the safety of the pipeline network and equipment; at the same time, there is a lack of consideration for the long-term operational health of the equipment, making it difficult to achieve optimal synergy between pursuing power generation revenue and ensuring water supply security and extending equipment life. Therefore, developing a method that can adapt to different forms of hydropower generation and achieve multi-objective intelligent collaborative control has become crucial for promoting the large-scale application of hydropower recovery technology. Summary of the Invention

[0004] This invention provides an intelligent collaborative control method, equipment, and product for a hydropower residual pressure generation system, which solves the technical problem that existing hydropower residual pressure generation systems generally suffer from single control strategies and slow response, making it difficult to achieve multi-objective collaborative optimization among power generation revenue, water supply security, and equipment lifespan under dynamically changing water consumption, electricity prices, and equipment operating conditions.

[0005] This invention provides an intelligent collaborative control method for a hydroelectric residual pressure power generation system, the method comprising the following steps: Acquire system data from the pipeline network system, time-of-use electricity price signals from the power grid, and demand response commands; Based on historical water usage data, weather information, and calendar information, a time-series forecasting model is used to predict the water load curve for a future set period. Based on the water load curve, electricity price signal, demand response command and system data, with the objective function of minimizing operating cost and the constraints of ensuring water supply security and equipment operation security, a candidate strategy set containing multiple operating modes is pre-generated within a rolling time window. In each control cycle, the candidate strategy set is optimized and adjusted in real time based on the latest system data, and the target control instruction set for the current control cycle is output.

[0006] According to the present invention, an intelligent collaborative control method for a hydroelectric power generation system with residual pressure is provided, wherein the time-series prediction model is a long short-term memory network model or a Transformer model; and the future set time period is 6 to 24 hours.

[0007] According to the present invention, an intelligent collaborative control method for a hydropower generation system with residual pressure is provided, wherein the system data includes pipeline pressure data and pipeline flow data; The method further includes a sudden water use event detection step, which includes: Based on the real-time monitored pipeline pressure and flow data, the pressure change rate, flow change rate, and deviation of the pressure and flow data from the baseline operating range are calculated within a set time window and used as event characteristic parameters. When any of the event feature parameters exceeds its corresponding adaptive dynamic threshold, and the change pattern of the event feature parameter matches any type of event pattern stored in the preset sudden water use event feature library, a sudden water use event is determined to have occurred, and the corresponding event type identifier is output.

[0008] According to the present invention, an intelligent collaborative control method for a hydropower residual pressure generation system, after determining that a sudden water use event has occurred, executes a predictive dynamic fusion step, the predictive dynamic fusion step including: Based on the event type identifier and the event characteristic parameters, retrieve the historical water load impact template with the highest matching degree with the current event from the historical event case library, or call the preset impact function corresponding to the event type identifier; The load change curve represented by the historical water load impact template or the preset impact function is superimposed on the basic water load curve predicted by the time series prediction model to generate the fused predicted load curve during the event impact period. The fused predicted load curve replaces the original water load curve and is used in subsequent optimization decision-making steps based on the water load curve.

[0009] According to the present invention, an intelligent collaborative control method for a hydroelectric residual pressure power generation system, wherein ensuring equipment operation safety is a constraint, specifically includes constructing and applying a dynamic safety constraint model for the equipment: The system data also includes water temperature data and water quality parameters; Based on the water temperature data, water quality parameters and current operating conditions of the hydropower unit in the real-time acquired system data, the safety thresholds for the operation of key equipment are dynamically calculated and updated. The safety thresholds include at least the maximum allowable speed, the maximum allowable pressure fluctuation rate and the minimum required net positive suction head (NPSH). During the pre-generated candidate strategy set and real-time optimization adjustment process, based on the water load curve and system topology model, the rotational speed, pressure fluctuations and net positive suction head (NPSH) of the hydropower unit in future time periods are predicted, and the predicted values ​​are compared with the corresponding dynamically updated safety thresholds as hard constraints.

[0010] According to the intelligent collaborative control method for a hydroelectric residual pressure power generation system provided by the present invention, the dynamic safety constraint model of the equipment is constructed and updated in the following manner: A mapping model is established with water temperature data, water quality parameters, and historical load rate of hydropower units as input, and the operating safety threshold of the key equipment as output. The mapping model can be a lookup table, a multinomial regression model, or a machine learning model. During system operation, the actual monitored safety-related parameters are compared with the safety thresholds output by the mapping model. If the deviation between the two exceeds the preset tolerance for multiple consecutive control cycles, the newly collected operating data is used to perform incremental training or parameter calibration on the mapping model.

[0011] According to the present invention, an intelligent collaborative control method for a hydropower generation system with residual pressure is provided, wherein the system data further includes the rotational speed of the hydropower generation unit, bearing vibration data, and water level data of the regulating pool; The method also includes a stand-alone safety monitoring step, which includes: The real-time acquired speed, bearing vibration data, and water level data are continuously compared with the corresponding preset safety thresholds. When any parameter exceeds its corresponding safety threshold, the protection logic is triggered, generating a protective control command and replacing the current target control command set; The protective control commands include reducing the load, switching to a preset safe operating mode, or emergency shutdown.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the above-described intelligent collaborative control methods.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent cooperative control method as described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the intelligent collaborative control methods described above.

[0015] This invention provides an intelligent collaborative control method for a hydropower generation system with residual pressure. Based on a time-series prediction model, it anticipates changes in water load in advance and combines time-of-use pricing and demand response commands to generate a multi-mode candidate strategy set while ensuring water supply security and equipment operation safety. Furthermore, within each control cycle, the strategy set is adjusted in real time based on the latest system data to output the optimal control command. This effectively overcomes the problems of traditional control strategies being singular and having a delayed response. It can balance maximizing power generation revenue, water supply reliability, and equipment safety in complex dynamic environments, thereby achieving multi-objective collaborative optimization. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is one of the flowcharts illustrating the intelligent collaborative control method for a hydroelectric residual pressure power generation system provided by the present invention; Figure 2 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0019] To address the aforementioned technical problems, this invention provides an intelligent collaborative control method and system for a hydropower residual pressure generation system. The aim is to achieve synergistic optimization among power generation revenue, water supply security, and equipment lifespan in complex dynamic environments through a solution that deeply integrates software and hardware.

[0020] Specifically, minimizing operating costs mainly involves deducting the power generation revenue and equipment start-up and shutdown losses from the grid purchase cost. This is calculated based on the time-of-use electricity price, the difference between the grid purchase cost and the hydropower residual pressure generation revenue, and then adding the loss costs due to frequent or unreasonable equipment start-ups and shutdowns and over-limit operation. Minimizing this cost is the core economic objective of optimized control.

[0021] The "multiple operating modes" involved in this invention refer to system operating states that are dynamically switched according to different optimization objectives and technical constraints. Their essential characteristics include at least: a first mode aimed at maximizing power generation revenue; a second mode aimed at ensuring the safety of the minimum water supply pressure in the pipeline network; and a third mode aimed at rapidly responding to demand-side commands from the power grid (such as peak shaving and valley filling). These modes are achieved by adjusting the setpoint combinations of actuators such as valves, frequency converters, and circuit breakers.

[0022] The "rolling time window" and the "control cycle" have a clear temporal relationship: the rolling time window is used for forward-looking optimization planning, and its length (e.g., 6-24 hours) is much longer than a single control cycle (e.g., 15 minutes). In each control cycle, the system makes short-term decisions and outputs control commands based on the candidate strategy set generated within the rolling window and the latest data, thus realizing a closed loop of "long-term predictive planning" and "short-term feedback correction".

[0023] The following is combined with Figure 1 The intelligent cooperative control method for the hydroelectric residual pressure power generation system of the present invention includes the following steps: Acquire system data from the pipeline network system, time-of-use electricity price signals from the power grid, and demand response commands; Based on historical water usage data, weather information, and calendar information, a time-series forecasting model is used to predict the water load curve for a future set period. Based on the water load curve, electricity price signal, demand response command and system data, with the objective function of minimizing operating cost and the constraints of ensuring water supply security and equipment operation security, a candidate strategy set containing multiple operating modes is pre-generated within a rolling time window. In each control cycle, the candidate strategy set is optimized and adjusted in real time based on the latest system data, and the target control instruction set for the current control cycle is output.

[0024] Based on the time-series prediction model, water load changes are predicted in advance. Combined with time-of-use pricing and demand response commands, a multi-mode candidate strategy set is generated under the premise of ensuring water supply security and equipment operation safety. Furthermore, the strategy set is adjusted in real time based on the latest system data in each control cycle to output the optimal control command. This effectively overcomes the problems of traditional single control strategies and delayed response. It can balance maximizing power generation revenue, water supply reliability and equipment operation safety in complex dynamic environments, thereby achieving multi-objective synergistic optimization.

[0025] Specifically, the time-series prediction model is a long short-term memory network model or a Transformer model; the future time period is set to 6 to 24 hours.

[0026] In one specific embodiment, the hydroelectric pressure relief power generation system is deployed at the pressure regulating node of the city's main water supply network. The system includes a pressure relief generator set, pressure / flow sensors, an intelligent controller, and a communication module. The intelligent controller embeds a Long Short-Term Memory (LSTM) network model as a time-series prediction model. Every day at 00:00, the system automatically acquires historical hourly water consumption data from the past 30 days, weather forecasts for the next 24 hours (including temperature and probability of rainfall), and daily calendar information (such as whether it is a weekday or holiday). Using this as input, the system uses the LSTM model to continuously predict the water load curve for the next 24 hours, with a time resolution of 15 minutes.

[0027] Simultaneously, the system receives real-time time-of-use electricity price signals (such as peak, flat, and valley periods and corresponding prices) from the power grid and demand response instructions issued by the superior dispatcher (such as reducing or increasing electricity load during specific periods) through the communication module. Based on the above prediction results and external signals, combined with system data such as current pipeline pressure, flow rate, and generator unit status, the intelligent controller constructs a candidate strategy set at the beginning of each hour with the objective function of minimizing the operating cost for the next 6 hours. This set includes multiple operating modes such as "priority power generation," "pressure protection and power restriction," and "response peak shaving," while strictly meeting safety constraints such as minimum water supply pressure ≥ 0.28 MPa and generator unit start-stop frequency ≤ 3 times / day.

[0028] At the start of each 15-minute control cycle, the controller reads the latest pipeline pressure, actual water consumption, and electricity price status. It then re-evaluates the expected costs and risks of each strategy in the original candidate strategy set online, selects the optimal strategy using model predictive control (MPC), and outputs the corresponding power generation setpoint, valve opening commands, and unit start / stop signals. For example, when the predicted off-peak electricity price at night and stable water demand, the system prioritizes generating electricity under residual pressure. However, before the morning peak and when the predicted water supply pressure approaches its lower limit, it switches to a "pressure-maintaining and power-limiting" mode, appropriately reducing power generation to maintain pipeline pressure safety. Through this implementation, the system significantly improves its economic operation and response flexibility in dynamic environments while ensuring water supply safety and equipment lifespan.

[0029] Preferably, the system data includes pipeline pressure data and pipeline flow data; The method further includes a sudden water use event detection step, which includes: Based on the real-time monitored pipeline pressure and flow data, the pressure change rate, flow change rate, and deviation of the pressure and flow data from the baseline operating range are calculated within a set time window and used as event characteristic parameters. When any of the event feature parameters exceeds its corresponding adaptive dynamic threshold, and the change pattern of the event feature parameter matches any type of event pattern stored in the preset sudden water use event feature library, a sudden water use event is determined to have occurred, and the corresponding event type identifier is output.

[0030] In the specific implementation of the aforementioned sudden water use event sensing steps, it is first necessary to collect pipeline pressure data P(t) and pipeline flow data Q in real time through pressure sensors and flow meters deployed at key nodes of the pipeline network. (t) Where t represents the current time. To characterize the transient changes in the system state, two core dynamic indicators need to be calculated within a set time window ΔT (e.g., 30 seconds to 5 minutes): the pressure change rate R... P (t) and the rate of change of flow R Q (t). The calculation formulas are as follows:

[0031] In the formula, P(t−ΔT) and Q(t−ΔT) are the pressure and flow rate values ​​before time ΔT, respectively, reflecting the instantaneous intensity of the hydraulic state of the pipeline network.

[0032] At the same time, in order to measure the degree of deviation between the current operating state and the normal state, it is necessary to calculate the deviation index D. P (t) and D Q (t), the calculation of which depends on the baseline operating range obtained from prior statistical analysis.

[0033] For example, a pressure baseline interval can be defined as the range formed by adding or subtracting a certain number of standard deviations from the mean pressure values ​​of the same historical period (such as the same time period on similar working days), where k is a coefficient set according to system characteristics (usually 2 or 3). The deviation of the current pressure value from this baseline interval can be calculated using standardized distance:

[0035] Flow deviation D Q The calculation of (t) is similar. Therefore, the event feature parameter vector can be expressed as:

[0037] The adaptive dynamic threshold is initialized based on historical system operating data and continuously calibrated during operation. Its general setting principle is: take the moving average of each characteristic parameter sequence under historical normal operating conditions, plus N times its moving standard deviation (N is set according to the tolerance for false alarm rate, usually 2-3). In the initial stage of system deployment, a conservative threshold (a larger N) can be used. After stable operation, the N value can be automatically fine-tuned based on the statistics of false alarms and missed alarms over a period of time, achieving threshold self-adaptation. For example, the pressure change rate threshold can be set to |R| within the recent period (e.g., the past hour). P | The moving average plus n times its moving standard deviation:

[0039] Other parameters The method for determining the threshold is similar, which allows the threshold to adapt to normal fluctuations in different load stages of the system and reduces the probability of misjudgment.

[0040] A pre-built database of sudden water use event characteristics stores feature pattern templates for various typical events. For example, the "pipe burst event" pattern might manifest as a sudden drop in pressure (R... P Significantly negative overthreshold) accompanied by a sudden increase in flow (R) Q (significantly positive overthreshold), and the deviation between the two is D. P D Q Rapidly increasing. The "large user concentrated start-stop" mode may manifest as a step change in traffic (R). Q (Exceeding the threshold) while pressure changes are relatively gradual. The feature library can be constructed by analyzing historical event data or based on hydraulic model simulations.

[0041] During actual operation, the system continuously calculates real-time event characteristic parameters. F(t) The decision logic is: when F(t) Any parameter (such as) R P (t) ) exceeds its corresponding current adaptive dynamic threshold (e.g. ), and at the same time, F(t) When the matching degree between the trajectory of change in multidimensional space (i.e., the combination of multiple parameters changing over time) and a preset event pattern in the feature library exceeds a set similarity threshold, the system determines that a sudden water use event has occurred. The matching degree can be calculated using Euclidean distance, cosine similarity, or more complex sequence alignment methods. Once a match is successful, the system outputs the event type identifier corresponding to the pattern, such as "Type_A: Major pipeline leak".

[0042] For illustration, assume that on a certain weekday afternoon, the system monitors a drop in regional pipeline pressure from 0.35 MPa to 0.28 MPa within 2 minutes, while the flow rate increases from 200 m³ / h to 280 m³ / h. The calculations yield... R P It is -0.058 MPa / min. R Q The flow rate was 40 m³ / h / min, far exceeding its current dynamic threshold; meanwhile, the deviation between pressure and flow rate was... D P and D Q All values ​​rapidly increased to over 3.5. The time series of this parameter combination and its rapid evolution matched the "pipe burst event" pattern in the feature library with a 90% accuracy. Therefore, the system determined that a sudden water use event had occurred and immediately output the event type identifier "pipe burst," thus providing clear anomaly category information for subsequent prediction fusion and control adjustments.

[0043] Through the specific implementation methods described above, this sudden water use event perception step achieves a quantitative, adaptive, and patterned processing flow from raw data to event judgment. It does not rely solely on limit-breaking alarms for a single parameter, but integrates multi-dimensional characteristics such as rate of change and deviation, along with their temporal patterns. Furthermore, it employs dynamic thresholds to distinguish between normal fluctuations and genuine anomalies, thereby significantly improving the accuracy, timeliness, and reliability of event perception. This lays a crucial information perception foundation for the entire intelligent collaborative control system to respond to sudden operating conditions.

[0044] Preferably, after determining that a sudden water use event has occurred, a predictive dynamic fusion step is performed, the predictive dynamic fusion step including: Based on the event type identifier and the event characteristic parameters, retrieve the historical water load impact template with the highest matching degree with the current event from the historical event case library, or call the preset impact function corresponding to the event type identifier; The load change curve represented by the historical water load impact template or the preset impact function is superimposed on the basic water load curve predicted by the time series prediction model to generate the fused predicted load curve during the event impact period. The fused predicted load curve replaces the original water load curve and is used in subsequent optimization decision-making steps based on the water load curve.

[0045] Specifically, after a sudden water usage event is determined to have occurred, a dynamic prediction fusion step needs to be executed to ensure that the water load forecast on which the system's optimization decisions rely reflects the true impact of the event. The core of this step is to use the specific information output from the event identification phase to correct the basic water load curve obtained from historical rolling forecasts in real time, thereby generating a fused prediction result that more accurately reflects load changes in future periods.

[0046] First, the system uses the event type identifier (e.g., "pipe burst", "large-scale centralized water use") and the quantified event feature parameter vector obtained from the preceding steps. F(t) The system initiates the fusion prediction process. It pre-stores a historical event case library, where each case records the start time, event characteristic parameters, and the measured water load change curve relative to the pre-event prediction baseline during the event's duration. This curve is stored as a "historical water load impact template." To retrieve the template that best matches the current event from the case library, the matching degree S between the current event characteristics and the historical case characteristics needs to be calculated. A feasible formula for calculating the matching degree is the reciprocal of the weighted Euclidean distance:

[0048] in, This represents the matching degree with the i-th historical case. This represents the j-th characteristic parameter of the current event (such as the rate of pressure change). R P Flow rate of change R Q wait), Indicates the first i The corresponding historical case is the j Each feature parameter value, where is the weight of the j-th feature parameter, used to characterize the difference in importance of different features in the event influence pattern, and m is the total number of feature parameters. The system retrieves the matching degree. The highest historical case, and its impact template stored. Extracted, among which This is the time offset with the time of the event as the origin.

[0049] For certain regular event types with distinct characteristics and clear mechanisms, the system can also pre-set standardized influence functions. For example, a planned large-scale water outage followed by water supply restoration might trigger a short-term water consumption peak that then declines to normal levels. Its impact function can be modeled as an exponential decay function:

[0051] Where A represents the initial load surge amplitude, which can be calibrated based on event characteristic parameters (such as the estimated user scale of the recovery area); This is the attenuation coefficient, which is related to the characteristics of the pipe network and the user's water usage habits. When the event type identifier corresponds to this type of preset function, the system directly calls the corresponding function. It also instantiates the key parameters of the event using the current event characteristic parameters.

[0052] Subsequently, the system dynamically fuses the predicted curves. Let the basic water load curve, predicted by the time-series forecasting model during the period of future event impact, be... L base (t) Where t is absolute time. The time when the event occurs is denoted as . t 0 The fusion process involves superimposing the load changes represented by the retrieved impact templates or instantiated impact functions onto the base forecast curve. Taking the use of historical templates as an example, the fused forecast load curve... L fused (t) Calculated by the following formula:

[0054] in, This means aligning the template on the timeline to the moment the event occurred. t 0 , is an adaptive scaling factor ( This is used to fine-tune the magnitude of the impact based on subtle differences between the current event's characteristic parameters and the characteristics of selected historical cases (for example, if the current rate of change in traffic is 1.2 times that of historical cases, then it can be set...). ) The typical duration of the impact of this type of event is determined by the template or function domain. Outside the event's impact period, the fused prediction curve remains equal to the base prediction curve. If a preset influence function is called, then the above formula will be... Replace with That's all.

[0055] To illustrate this further, let's continue with the example of the pipe burst incident mentioned earlier. After the system determines the event type as "pipe burst," it retrieves the most matching case from the historical case database based on its characteristic parameters (the magnitude and rate of pressure drop and flow surge). The historical template for this case shows that within approximately 30 minutes of the pipe burst, the water load in the affected area would abnormally increase by about 25% due to the pipe network leak and uncontrolled water flow (equivalent to a pulse with an amplitude of 0.25 times the base load). Subsequently, as valves were closed or the leak was controlled, the load gradually returned to normal levels over the next hour. The system uses this template... Extracted. Meanwhile, the rolling forecasting model has already provided the base load curve for the region for the next two hours. L base (t) The system aligns the template on the timeline to the moment of pipe burst detection. t 0 Considering that the traffic change rate for this event is 10% higher than historical cases, a scaling factor is set. Superposition via formula Generate the forecast for the next 1.5 hours (i.e. The combined predicted load curve is shown. t 0 A significant abnormal load peak occurs after time point [time], followed by a slow decline. This fusion prediction curve [is shown in the original text]. It will immediately replace the original This serves as a key input for subsequent optimization decision-making modules.

[0056] Through the specific implementation methods described above, the predictive dynamic fusion step organically combines long-term prediction based on statistical regularities with short-term correction based on event characteristics when facing sudden water usage events. It not only relies on historical templates but also supports mechanism-based modeling, and achieves precise quantitative adaptation to specific events through parameter matching and adaptive scaling. Therefore, the entire intelligent collaborative control system can anticipate load disturbances caused by sudden events, providing reliable forward-looking information for formulating optimized control strategies that both ensure water supply security and effectively utilize residual pressure for power generation, significantly improving the system's robustness under unsteady conditions.

[0057] Preferably, the constraint of ensuring equipment operation safety specifically includes constructing and applying a dynamic safety constraint model for the equipment: The system data also includes water temperature data and water quality parameters; Based on the water temperature data, water quality parameters and current operating conditions of the hydropower unit in the real-time acquired system data, the safety thresholds for the operation of key equipment are dynamically calculated and updated. The safety thresholds include at least the maximum allowable speed, the maximum allowable pressure fluctuation rate and the minimum required net positive suction head (NPSH). During the pre-generated candidate strategy set and real-time optimization adjustment process, based on the water load curve and system topology model, the rotational speed, pressure fluctuations and net positive suction head (NPSH) of the hydropower unit in future time periods are predicted, and the predicted values ​​are compared with the corresponding dynamically updated safety thresholds as hard constraints.

[0058] To achieve refined safety assurance of the hydropower generation unit's operating status, the construction and application of the equipment dynamic safety constraint model follows a dynamic and adaptive process. Traditional fixed safety thresholds cannot reflect the impact of water temperature, water quality changes, and cumulative effects on the equipment's actual withstand capacity. This method establishes a real-time data-driven mapping model, enabling the safety thresholds to be dynamically adjusted according to the operating environment and equipment status.

[0059] Specifically, constructing the mapping model first requires defining input and output variables. The input variable X mainly includes: real-time monitored water temperature data T (unit: °C), and water quality parameter vector W (e.g., may include pH value, NTU turbidity, and chloride ion concentration Cl). - (etc.), and historical load rate reflecting the recent workload of the equipment. Historical load factor can be calculated as the ratio of the average output power of the generating unit to its rated power over a past period (e.g., 24 hours). The output variable Y represents the critical equipment operating safety threshold to be determined, and at least includes: the maximum permissible rotational speed. N max (Unit: rpm), Maximum permissible pressure fluctuation rate ΔP max (Unit: MPa / s), and minimum required net positive suction head (NPSH). NPSH r,min (Unit: m).

[0060] The mapping model is constructed based on historical equipment operating data and safety test data. A simplified multinomial regression model is used as an example to illustrate its construction process: A large amount of operating data covering different water temperatures, water qualities, and load rates is collected, along with safety threshold boundaries determined through experiments or long-term monitoring under these conditions, such as rotational speed and pressure fluctuation rates during slight vibrations. Using this data as a sample, the least squares method is used to fit the coefficients of the regression equation. For example, for the maximum permissible rotational speed... N max The regression equation is as follows:

[0062] in, The regression coefficients are obtained by fitting historical data, and the error term is denoted as . This model can capture the influence of water temperature, water quality and their interaction on material strength and bearing lubrication conditions, and thus affect the maximum allowable speed.

[0063] The device dynamic safety constraint model is constructed and updated in the following ways: A mapping model is established with water temperature data, water quality parameters, and historical load rate of hydropower units as input, and the operating safety threshold of the key equipment as output. The mapping model can be a lookup table, a multinomial regression model, or a machine learning model. During system operation, the actual monitored safety-related parameters are compared with the safety thresholds output by the mapping model. If the deviation between the two exceeds the preset tolerance for multiple consecutive control cycles, the newly collected operating data is used to perform incremental training or parameter calibration on the mapping model.

[0064] During system operation, in each control cycle, the system uses the real-time collected water temperature T as the basis for... real Water quality Wreal and calculated recent load rate The dynamic safety threshold under the current operating conditions is calculated using a mapping model. N max,dyn , ΔP max,dyn , NPSH r,min,dyn Meanwhile, when making optimization decisions based on water load curves and network hydraulic models (either by pre-generating a set of candidate strategies or by adjusting in real time), the system predicts the possible changes in rotational speed of the hydroelectric generating units in future time periods caused by implementing each candidate strategy. N pred (t) Pressure volatility ΔP pred (t) and cavitation margin NPSH a,pred (t) The constraint is stated as follows: For all candidate strategies and their corresponding prediction timelines, the following condition must be met.

[0065]

[0066] Any candidate strategy that violates any of the above hard constraints will be directly eliminated, thus ensuring that the optimization search is always conducted within the device's dynamic safety boundaries.

[0067] To ensure the long-term accuracy and reliability of the mapping model, an online update mechanism was designed for the system. Model updates are controlled by a trigger mechanism based on deviation monitoring. The specific process is as follows: During operation, the system continuously monitors the actual safety-related parameters of the hydroelectric power generation unit, such as obtaining the actual rotational speed through a speed sensor. N act The actual pressure fluctuation rate is calculated using a high-frequency pressure sensor. ΔP act And to estimate the actual net positive suction head (NPSH) using formulas. NPSH a,act In each control cycle, these measured values ​​are compared with the current mapping model based on the real-time input. X real The output dynamic safety thresholds are compared, and the relative deviation is calculated: The system sets a preset tolerance limit. The deviation is 5%. When the measured value of a certain safety parameter deviates from its corresponding dynamic threshold... For more than K consecutive control cycles If all values ​​are greater than 6, for example, K=6, then it is determined that the current mapping model may have a significant deviation in this working condition region, triggering the model update process.

[0068] The update process uses newly acquired runtime data to incrementally train or calibrate the original model. For lookup table models, the update operation is performed on the current input parameters. X real The corresponding data region is then smoothed by applying new measured safety threshold data (or threshold boundaries inferred from measured safety conditions) to the existing table values. For multinomial regression or machine learning models, the update involves updating the latest input-output data pairs (…). X real , Y adjusted Add to the training dataset, where Y adjusted This involves adjusting the original threshold output based on the actual safety status of the measured equipment, such as whether abnormal vibrations have occurred. Subsequently, the model is incrementally trained or rapidly fine-tuned using the augmented dataset in the background. Once completed, the old model is seamlessly replaced with the new model, achieving online evolution and optimization of the safety constraint threshold.

[0069] For example, consider a specific application scenario: a hydraulic residual pressure power generation unit operates using a pump-turbine mode. During the high-temperature period in summer, the circulating water temperature rises to 35℃ (normally 25℃), and the pH value of the water quality slightly decreases due to seasonal changes in the water source. The fixed threshold model still uses the upper limit of the safe operating speed designed based on normal temperature clean water conditions. However, the dynamic model of this method, based on real-time inputs T=35℃ and pH=6.5, The current dynamic maximum permissible speed is calculated using the trained SVR model. N max,dyn =1420 rpm, which is more conservative than the fixed threshold of 1500 rpm, because the model learned that high temperature and slightly acidic water quality will accelerate bearing wear and material fatigue. In the optimization control, a candidate strategy predicts that the speed will reach 1480 rpm during peak electricity consumption periods, but this strategy violates... N pred The system was rejected due to the constraint of ≤1420 rpm. A few days later, the system was monitored at T=34℃ and pH=6.6. Under similar operating conditions, the actual speed was stable at 1450 rpm and the equipment vibration was normal, but the dynamic threshold output by the model was still 1415 rpm, resulting in a deviation. The tolerance was exceeded for several consecutive cycles. The system then triggered an update, adding new data points to the training set. The incrementally trained SVR model output under these conditions... N max,dyn The speed has been adjusted to approximately 1460 rpm, which better reflects the actual safe operating limits of the equipment.

[0070] Through the above specific implementation methods, the equipment dynamic safety constraint model transforms static and conservative safety rules into dynamic and precise safety boundaries. It can not only set safety limits that are more matched according to the real-time environment and equipment status, thereby releasing the greater operating potential of the equipment within the safety range, but also has the ability to continuously learn and self-calibrate, enabling the safety model to follow the slow changes in equipment performance or the migration of environmental characteristics and maintain high reliability in the long term.

[0071] The system data also includes the rotational speed of the hydroelectric power generation unit, bearing vibration data, and water level data of the regulating pool; The method also includes a stand-alone safety monitoring step, which includes: The real-time acquired speed, bearing vibration data, and water level data are continuously compared with the corresponding preset safety thresholds. When any parameter exceeds its corresponding safety threshold, the protection logic is triggered, generating a protective control command and replacing the current target control command set; The protective control commands include reducing the load, switching to a preset safe operating mode, or emergency shutdown.

[0072] Protection commands are divided into three levels based on the severity of the anomaly: Reduce load: For minor over-limit situations, quickly reduce the output of the power generation unit to bring it back to a safe range.

[0073] Switch to Safe Mode: In case of obvious anomalies, the system will be forcibly switched to a pre-set simplified operating mode with safety as the sole objective.

[0074] Emergency shutdown: For severe or complex anomalies, immediately execute the sequence of disconnecting the network, shutting off valves, and shutting down the system. This is the highest level of protection.

[0075] For example, if the system encounters a sudden and severe fluctuation in upstream pressure during optimized power generation, causing the rotational speed to surge within seconds and exceed the safety limit, the independent safety monitoring module will first perform a brief delay to confirm the situation, then quickly determine that the exceedance is valid, and immediately issue an "emergency shutdown" command to forcibly interrupt the current operation, thereby avoiding potential equipment damage during the main control system's response delay.

[0076] The present invention also provides a hardware system for implementing the above method, comprising: Hydropower generation unit, which is a dedicated water turbine or reversible water pump unit, is used to convert the residual pressure of the pipeline network into electrical energy; The actuator assembly includes an electrically operated regulating valve installed in the pipeline, a frequency converter for controlling the speed of the generator unit, and a grid-connected circuit breaker; The data acquisition module includes sensors for acquiring pipeline pressure, flow rate, water temperature, and water quality; sensors for monitoring the rotational speed of the hydropower generation unit and bearing vibration; and a water level gauge for monitoring the water level in the regulating pool. The intelligent controller is an embedded industrial computer or programmable logic controller. It stores a program that implements the above-mentioned intelligent collaborative control method and is connected to the sensor network and the actuator group. It is used to output instructions according to the algorithm to physically drive changes in valve opening, frequency adjustment of frequency converter and circuit breaker opening and closing, thereby realizing precise control of the hydraulic state of the pipeline network and the power generation process.

[0077] The intelligent collaborative control method, by driving the aforementioned hardware system, specifically addresses the following technical issues: by predicting and adjusting valve opening and unit speed in real time, it smooths pressure fluctuations, suppresses water hammer effects, and reduces the pressure fluctuation rate of key pipeline nodes by more than 30%; by using dynamic safety constraints to prevent equipment from operating beyond limits under unfavorable water temperature and quality conditions, it reduces the bearing abnormal vibration alarm rate by 40%; and by using rapid safety monitoring and protection commands, it triggers protection actions within 200 milliseconds after an abnormal speed spike, preventing mechanical overspeed damage.

[0078] Figure 2 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 2 As shown, the electronic device may include a processor 810, a communication interface 820, a memory 830, and a communication bus 840. The processor 810, communication interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logic instructions from the memory 830 to execute an intelligent collaborative control method for the hydroelectric residual pressure power generation system.

[0079] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0080] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the intelligent collaborative control method for the hydropower generation system provided by the above methods.

[0081] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent collaborative control method for the hydroelectric residual pressure power generation system provided by the above methods.

[0082] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0083] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart collaborative control method for a hydroelectric residual pressure power generation system, characterized in that, Includes the following steps: Acquire system data from the pipeline network system, time-of-use electricity price signals from the power grid, and demand response commands; Based on historical water usage data, weather information, and calendar information, a time-series forecasting model is used to predict the water load curve for a future set period. Based on the water load curve, electricity price signal, demand response command and system data, with the objective function of minimizing operating cost and the constraints of ensuring water supply security and equipment operation security, a candidate strategy set containing multiple operating modes is pre-generated within a rolling time window. In each control cycle, the candidate strategy set is optimized and adjusted in real time based on the latest system data, and the target control instruction set for the current control cycle is output.

2. The intelligent collaborative control method for a hydroelectric residual pressure power generation system according to claim 1, characterized in that, The time-series prediction model is a long short-term memory network model or a Transformer model; the future time period is set to 6 to 24 hours.

3. The intelligent collaborative control method for a hydroelectric residual pressure power generation system according to claim 2, characterized in that, The system data includes pipeline pressure data and pipeline flow data; The method further includes a sudden water use event detection step, which includes: Based on the real-time monitored pipeline pressure and flow data, the pressure change rate, flow change rate, and deviation of the pressure and flow data from the baseline operating range are calculated within a set time window and used as event characteristic parameters. When any of the event feature parameters exceeds its corresponding adaptive dynamic threshold, and the change pattern of the event feature parameter matches any type of event pattern stored in the preset sudden water use event feature library, a sudden water use event is determined to have occurred, and the corresponding event type identifier is output.

4. The intelligent collaborative control method for a hydroelectric residual pressure power generation system according to claim 3, characterized in that, After determining that a sudden water use event has occurred, a predictive dynamic fusion step is performed, which includes: Based on the event type identifier and the event feature parameters, retrieve the historical water load impact template with the highest matching degree with the current event from the historical event case library, or call the preset impact function corresponding to the event type identifier; The load change curve represented by the historical water load impact template or the preset impact function is superimposed on the basic water load curve predicted by the time series prediction model to generate the fused predicted load curve during the event impact period. The fused predicted load curve replaces the original water load curve and is used in subsequent optimization decision-making steps based on the water load curve.

5. The intelligent collaborative control method for a hydroelectric residual pressure power generation system according to claim 1, characterized in that, The constraint of ensuring equipment operation safety specifically includes constructing and applying a dynamic safety constraint model for the equipment: The system data also includes water temperature data and water quality parameters; Based on the water temperature data, water quality parameters and current operating conditions of the hydropower unit in the real-time acquired system data, the safety thresholds for key equipment operation are dynamically calculated and updated. The safety thresholds include at least the maximum allowable speed, the maximum allowable pressure fluctuation rate and the minimum required net positive suction head (NPSH). During the pre-generated candidate strategy set and real-time optimization adjustment process, based on the water load curve and system topology model, the rotational speed, pressure fluctuations and net positive suction head (NPSH) of the hydropower unit in future time periods are predicted, and the predicted values ​​are compared with the corresponding dynamically updated safety thresholds as hard constraints.

6. The intelligent collaborative control method for a hydroelectric residual pressure power generation system according to claim 5, characterized in that, The device dynamic safety constraint model is constructed and updated in the following ways: A mapping model is established with water temperature data, water quality parameters, and historical load rate of hydropower units as input, and the operating safety threshold of the key equipment as output. The mapping model can be a lookup table, a multinomial regression model, or a machine learning model. During system operation, the actual monitored safety-related parameters are compared with the safety thresholds output by the mapping model. If the deviation between the two exceeds the preset tolerance for multiple consecutive control cycles, the newly collected operating data is used to perform incremental training or parameter calibration on the mapping model.

7. The intelligent collaborative control method for a hydroelectric residual pressure power generation system according to claim 1, characterized in that, The system data also includes the rotational speed of the hydroelectric power generation unit, bearing vibration data, and water level data of the regulating pool; The method also includes a stand-alone safety monitoring step, which includes: The real-time acquired speed, bearing vibration data, and water level data are continuously compared with the corresponding preset safety thresholds. When any parameter exceeds its corresponding safety threshold, the protection logic is triggered, generating a protective control command and replacing the current target control command set; The protective control commands include reducing the load, switching to a preset safe operating mode, or emergency shutdown.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the control method as described in any one of claims 1 to 7.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the control method as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the control method as described in any one of claims 1 to 7.

Images