Mechanism and data hybrid driven pump set optimization intelligent control method and system

Abstract

The invention discloses a mechanism and data hybrid-driven pump set optimization intelligent control method and system, and relates to the technical field of water supply process control, the system comprises a normal operation model and a special plan model, the normal operation model is used for daily pump set optimization control of a water plant under non-special conditions, and the special plan model is used for daily pump set optimization control of a water plant under non-special conditions. The special plan model is used for predicting special conditions that the flow deviation is too large and the pressure of a water supply pipe orifice is suddenly changed, and the normal operation model comprises the steps of establishing a pump set database, collecting data, updating the database, predicting pipe network demands, preliminarily screening schemes, judging a satisfactory scheme, calculating the fitness of the scheme and outputting an optimal scheme. According to the invention, the security isolation gateway adopts a white list mechanism to filter control instructions, and all optimization instructions need to be verified through an SHA3512 digital signature, so that the system collaboration is improved through OPCUA / Modbus dual channels and RESTfulAPI, and the system stability is improved by carrying out performance attenuation detection through multi-source monitoring and survival analysis model dynamic compensation.

Description

Technical Field

[0001] This invention relates to the field of water supply process control technology, specifically to a mechanism- and data-driven intelligent control method and system for optimizing pump sets. Background Technology

[0002] The water supply pump station is a crucial link in the water supply process, undertaking the important task of pressurizing and transporting water from the clear water tank to the pipeline network system to meet the flow and pressure demands of users. Its core control objective is to achieve economic optimization of water supply energy consumption while ensuring the required flow and head in the pipeline network, and to ensure the safe and stable operation of the pump equipment. Currently, this field faces the following main technical challenges in achieving intelligent optimization configuration: In water plants, how to rationally select operating pumps and their corresponding frequencies through appropriate strategies? Water flow in water plants has significant uncertainty, making accurate prediction and effective utilization difficult. However, existing intelligent control methods and systems for pump optimization have excessively high system computing power requirements and do not fully consider the actual performance degradation of pump units, affecting system efficiency.

[0003] The shortcomings of existing intelligent control methods and systems for pump sets are:

[0004] 1. Patent document CN120891745A discloses a control method for a smart circulating water energy-saving optimization system in thermal power plants. "This method addresses the technical problems of low control accuracy and high energy consumption in traditional thermal power plant cold-end systems by constructing an intelligent control system based on a Transformer deep learning architecture to achieve optimized control. It establishes a smart weather data acquisition network, integrating meteorological data, satellite meteorological data, and a local microclimate sensor network to acquire high-precision temporal and atmospheric meteorological information; it adopts a Transformer encoder-decoder architecture to convert multi-source data into standardized token representations, and achieves complex correlation analysis through a multi-head self-attention mechanism; it constructs a multi-objective collaborative optimization framework with an objective function as its core, establishing a control closed loop of turbine-condenser-cooling tower-circulating water pump to achieve seasonal adaptive control and robust optimization; and it provides transparent decision analysis, explaining the control decision process through attention weight distribution." However, existing intelligent control methods and systems for pump group optimization have excessively high system computing power requirements and do not fully consider the actual performance degradation of the pump groups, affecting system efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a mechanism- and data-driven intelligent control method and system for optimizing pump sets, so as to solve the technical problems of rotational installation and inconvenient disassembly mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a mechanism- and data-driven intelligent control method for pump group optimization, comprising: a normal operation model and a special contingency plan model. The normal operation model is used for daily pump group optimization control of water plants under normal circumstances, and the special contingency plan model is used to predict special situations such as excessive flow deviation and sudden changes in water supply pipe pressure. The normal operation model includes establishing a pump group database, data acquisition and database updating, pipeline demand prediction, initial screening of schemes, determination of satisfactory schemes, calculation of scheme fitness, and output of the optimal scheme. The pump group database needs to contain the flow rate of each pump at its operating frequency. The pump set includes the following formulas: head relationship, flow-axis power relationship, flow-efficiency relationship, shaft power, efficiency, flow rate and high efficiency range at different frequencies. The normal operation model also includes dynamic performance degradation sensing and compensation. The dynamic performance degradation sensing and compensation collects equipment status data through vibration sensors, current harmonic analysis and voiceprint recognition devices, and establishes a pump set life prediction model based on survival analysis. The prediction model is R(t)=exp[∫λ(β·S(u))du], where λ is the failure rate function, S(u) is the multi-source monitoring data fusion feature, outputs the performance degradation compensation factor α, and dynamically updates the mechanism model parameters.

[0007] Preferably, the pump database includes pump data from the water plant, characteristic curves and expressions of each pump under power frequency conditions, relationship formulas of each characteristic curve of the pump under non-power frequency conditions, and a library of pump group schemes for correction and fitting based on real data.

[0008] The water pump data obtained from the water plant includes the actual water pump model, rated flow rate and its corresponding head and rated shaft power, as well as the flow-head characteristic curve, flow-shaft power characteristic curve, and flow-efficiency characteristic curve at the rated frequency.

[0009] Fitting the characteristic curves and expressions of various water pumps under power frequency conditions: Based on the water pump curves provided by the manufacturer under power frequency, the flow-head relationship, flow-axis power relationship, and flow-efficiency relationship of each water pump under power frequency are fitted. When the frequency is constant, the standard form of the flow-head relationship is H = H0 + S·Q 2 Where H0 represents the virtual total head of the pump when the flow rate is 0, and S is the pump internal resistance coefficient. When the frequency is constant, the pump flow-axis power relationship can be approximately expressed as P = a1Q + b1, where a1 and b1 are the power fitting coefficients of the pump. When the frequency is constant, the flow efficiency relationship can be approximately expressed as η = a2Q. 2 +b2Q, where a2 and b2 are the efficiency fitting coefficients of the water pump.

[0010] Preferably, the special contingency plan model includes a trigger condition judgment module, calculates the maximum available flow of the current pump group, selects a pump addition or reduction scheme based on the currently operating pump group, calculates the suitability of the selected scheme, and outputs the optimal scheme. The trigger condition judgment module monitors the deviation between the predicted flow and the actual flow and the pressure value of the water supply pipe inlet in real time according to the system reading value. If the deviation between the predicted flow and the actual flow exceeds the warning value, or the pressure of the water supply pipe inlet drops sharply to the warning value, then data acquisition and database update are initiated; otherwise, the special contingency plan model is not triggered.

[0011] Preferably, the characteristic curves of the water pump under non-power frequency conditions are as follows: when the flow rate is Q, the head H' at frequency f' and the power frequency f' are... N The expression for H' is as follows: H' = H0·(f' / f N ) 2 -S·Q 2 When the frequency is f', the shaft power P' and the power frequency f N Shaft power P at time N The relationship is as follows: P' = (f' / f N ) 2 a1Q+(f' / f N ) 3 b1, When the frequency is f', the expressions for flow rate Q and efficiency η' can be directly derived from the flow-efficiency relationship at power frequency. The specific expressions are as follows: Assuming the efficiency η of the water pump k Not less than ω·η k,max When the pump operates at high efficiency, and the midpoint of the interval is taken as the point of highest efficiency, the high-efficiency operating flow range of the pump at frequency f' can be considered to satisfy the following expression.

[0012] Preferably, the method of correcting based on real data involves using real data obtained from the data acquisition module to calculate the actual shaft power P″ of pump i at frequency f1 and flow rate Q. a,i Actual head H″ i Actual efficiency η′ i The formula for its calculation is as follows: P″ a =P N η m ;

[0013]

[0014] Where P N η is the output power of the motor. m To determine the motor efficiency, the theoretical shaft power P under the current operating condition can be calculated using the QP relationship fitted at frequency f1. a,i and take β P =P″ a,i / Pa,i , as a factor for attenuating the power of the water pump shaft;

[0015] The actual efficiency and actual head can be calculated using the following formulas:

[0016]

[0017] P u =ρgQH;

[0018]

[0019] Where P is u The effective power of the water pump, Δp is the pressure difference between the inlet and outlet of the water pump, and β is taken as... H =H″ i / H i As a correction factor for the pump head; take β η =η' i ' / η i As a correction factor for the actual efficiency of the water pump, each attenuation factor of the water pump is multiplied into the corresponding performance curve expression.

[0020] Preferably, both the normal operation model and the special contingency plan model are equipped with a hybrid drive control platform, a dynamic performance evaluation terminal, a collaborative adapter component, and a containerized deployment architecture. The hybrid drive control platform is equipped with multi-core heterogeneous processors ARM and FPGA, with the FPGA dedicated to parallel computing of the mechanism model. The hybrid drive control platform is configured with a non-volatile memory module and a fingerprint database of the entire life cycle of the storage device. The dynamic performance evaluation terminal integrates a MEMS vibration sensor array, a wideband current analyzer, and an ultrasonic flow meter. The dynamic performance evaluation terminal has a built-in attenuation status visualization dashboard that displays the remaining life distribution cloud map in real time. The collaborative adapter component supports PROFINET and EtherCAT industrial bus protocols and provides a RESTful API interface for third-party systems to call optimization instructions. The containerized deployment architecture encapsulates the optimization algorithm as a Docker microservice and achieves elastic resource scheduling through Kubernetes.

[0021] Preferably, the hybrid drive control platform further includes an energy consumption insight module, which constructs an energy efficiency health index based on the entropy method, EEI=(Σωi·xi) / REF, where ωi is the entropy weight, xi is the sub-item efficiency index, and REF is the benchmark value, and generates a priority ranking list for equipment renovation.

[0022] Preferably, the normal operation model further includes a security isolation gateway, which uses a whitelist mechanism to filter control commands, and all optimization commands must be verified by SHA3512 digital signature.

[0023] Compared with the prior art, the beneficial effects of the present invention are:

[0024] 1. This invention incorporates a hybrid drive control platform, a dynamic performance evaluation terminal, a collaborative adapter component, and a containerized deployment architecture in both the normal operation model and the special contingency plan model. The hybrid drive control platform is equipped with multi-core heterogeneous processors ARM and FPGA, with the FPGA dedicated to parallel computation of the mechanism model. The hybrid drive control platform is configured with a non-volatile memory module and stores a fingerprint database of the entire lifecycle of the storage device. The dynamic performance evaluation terminal integrates a MEMS vibration sensor array, a wideband current analyzer, and an ultrasonic flow meter. The dynamic performance evaluation terminal has a built-in attenuation status visualization dashboard that displays the remaining lifespan distribution cloud map in real time. The collaborative adapter component supports PROFINET and EtherCAT industrial bus protocols and provides a RESTful API interface for third-party systems to call optimization instructions. The containerized deployment architecture encapsulates the optimization algorithm as a Docker microservice and uses Kubernetes to achieve elastic resource scheduling. The normal operation model also includes a security isolation gateway, which uses a whitelist mechanism to filter control instructions. All optimization instructions must be verified by SHA3512 digital signature. This achieves the functions of improving system collaboration through OPCUA / Modbus dual channels and RESTful API, and improving system stability through multi-source monitoring and dynamic compensation of the survival analysis model for performance attenuation detection.

[0025] 2. By adopting a lightweight optimization algorithm model and a distributed computing architecture, this invention greatly reduces the dependence on the computing power of the central server, enabling the system to be easily integrated into the existing intelligent control platform of the water plant and run stably. This effectively saves initial investment and long-term operation and maintenance costs, and is more in line with the actual engineering needs of the gradual transformation of modern water plants.

[0026] 3. By constructing a hybrid decision-making model that integrates pump and pipeline network mechanism models with historical operating data, this invention overcomes the limitations of relying solely on historical experience. Even under conditions of poor historical data quality or minor changes in the water plant's process structure, the system can still make scientific decisions based on the mechanism model, demonstrating excellent adaptability and robustness, and ensuring the continued effectiveness of the optimization results.

[0027] 4. By introducing a pump efficiency model with a performance degradation factor, this invention can identify the performance degradation of pump sets in real time. This allows the optimized configuration strategy to dynamically adapt to the actual operating state of the equipment, always operating within a near-optimal efficiency range, thereby achieving truly refined energy consumption management and energy saving. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the intelligent pump distribution process of the present invention. Detailed Implementation

[0029] 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.

[0030] Example 1: A mechanism- and data-driven intelligent control method and system for pump group optimization. The method includes a normal operation model and a special contingency plan model. The normal operation model is used for routine pump group optimization control in water plants under normal conditions. The special contingency plan model is used to predict special situations such as excessive flow deviation and sudden changes in water supply pipe pressure. The normal operation model includes establishing a pump group database, data acquisition and database updates, pipeline demand prediction, initial screening of solutions, determination of satisfactory solutions, calculation of solution fitness, and output of the optimal solution. The pump group database must contain information on each pump operator. The flow-head relationship, flow-shaft power relationship, and flow-efficiency relationship are defined at different frequencies. The shaft power, efficiency, flow rate, and high-efficiency range at different frequencies are also defined. The normal operation model also includes dynamic performance degradation sensing and compensation. The dynamic performance degradation sensing and compensation collects equipment status data through vibration sensors, current harmonic analysis, and acoustic fingerprint recognition devices to establish a pump group life prediction model based on survival analysis. The prediction model is R(t)=exp[∫λ(β·S(u))du], where λ is the failure rate function, S(u) is the multi-source monitoring data fusion feature, and the performance degradation compensation factor α is output to dynamically update the mechanism model parameters.

[0031] The pump database includes data on water plant pumps, characteristic curves and expressions of each pump under power frequency conditions, relationship formulas of characteristic curves of each pump under non-power frequency conditions, and a library of pump group schemes for correction and fitting based on real data.

[0032] The water pump data obtained from the water plant includes the actual water pump model, rated flow rate and its corresponding head and rated shaft power, as well as the flow-head characteristic curve, flow-shaft power characteristic curve, and flow-efficiency characteristic curve at the rated frequency.

[0033] Fitting the characteristic curves and expressions of various water pumps under power frequency conditions: Based on the water pump curves provided by the manufacturer under power frequency, the flow-head relationship, flow-axis power relationship, and flow-efficiency relationship of each water pump under power frequency are fitted. When the frequency is constant, the standard form of the flow-head relationship is H = H0 + S·Q 2Where H0 represents the virtual total head of the pump when the flow rate is 0, and S is the pump internal resistance coefficient. When the frequency is constant, the pump flow-axis power relationship can be approximately expressed as P = a1Q + b1, where a1 and b1 are the power fitting coefficients of the pump. When the frequency is constant, the flow efficiency relationship can be approximately expressed as η = a2Q. 2 +b2Q, where a2 and b2 are the efficiency fitting coefficients of the water pump;

[0034] The special contingency plan model includes a trigger condition judgment module, calculates the maximum available flow of the current pump group, selects a pump addition or reduction scheme based on the currently operating pump group, calculates the adaptability of the selected scheme, and outputs the optimal scheme. The trigger condition judgment module monitors the deviation between the predicted flow and the actual flow and the pressure value of the water supply pipe inlet in real time according to the system reading value. If the deviation between the predicted flow and the actual flow exceeds the warning value, or the pressure of the water supply pipe inlet drops sharply to the warning value, data acquisition and database update will be initiated; otherwise, the special contingency plan model will not be triggered.

[0035] The relationship between the characteristic curves of the water pump under non-power frequency conditions is as follows: when the flow rate is Q, the head H' at frequency f' and the power frequency f'. N The expression for H' is as follows: H' = H0·(f' / f N ) 2 -S·Q 2 When the frequency is f', the shaft power P' and the power frequency f N Shaft power P at time N The relationship is as follows: P' = (f' / f N ) 2 a1Q+(f' / f N ) 3 b1, When the frequency is f', the expressions for flow rate Q and efficiency η' can be directly derived from the flow-efficiency relationship at power frequency. The specific expressions are as follows: Assuming the efficiency η of the water pump k Not less than ω·η k,max When the pump operates at high efficiency, and the midpoint of the interval is taken as the point of highest efficiency, the high-efficiency operating flow range of the pump at frequency f' can be considered to satisfy the following expression.

[0036] Based on the correction using real data, the actual shaft power P″ of pump i at frequency f1 and flow rate Q is calculated using the real data obtained by the data acquisition module. a,i Actual head H″ i Actual efficiency η′ i The formula for its calculation is as follows: P″ a =P N η m ;

[0037]

[0038] Where P N η is the output power of the motor. m To determine the motor efficiency, the theoretical shaft power P under the current operating condition can be calculated using the QP relationship fitted at frequency f1. a,i and take β P =P″ a,i / P a,i , as a factor for attenuating the power of the water pump shaft;

[0039] The actual efficiency and actual head can be calculated using the following formulas:

[0040]

[0041] P u =ρgQH;

[0042]

[0043] Where P is u The effective power of the water pump, Δp is the pressure difference between the inlet and outlet of the water pump, and β is taken as... H =H″ i / H i As a correction factor for the pump head; take β η =η' i ' / η i As a correction coefficient for the actual efficiency of the water pump, the various attenuation factors of each water pump are multiplied into the corresponding performance curve expression.

[0044] Both the normal operation model and the special contingency plan model are equipped with a hybrid drive control platform, a dynamic performance evaluation terminal, a collaborative adapter component, and a containerized deployment architecture. The hybrid drive control platform is equipped with multi-core heterogeneous processors ARM and FPGA, with the FPGA dedicated to parallel computing of the mechanism model. The hybrid drive control platform is configured with a non-volatile memory module and a fingerprint database of the entire life cycle of the storage device. The dynamic performance evaluation terminal integrates a MEMS vibration sensor array, a wideband current analyzer, and an ultrasonic flow meter. The dynamic performance evaluation terminal has a built-in attenuation status visualization dashboard that displays the remaining life distribution cloud map in real time. The collaborative adapter component supports PROFINET and EtherCAT industrial bus protocols and provides a RESTful API interface for third-party systems to call optimization instructions. The containerized deployment architecture encapsulates the optimization algorithm as a Docker microservice and achieves elastic resource scheduling through Kubernetes.

[0045] The hybrid drive control platform also includes an energy consumption insight module, which constructs an energy efficiency health index based on the entropy method, EEI=(Σωi·xi) / REF, where ωi is the entropy weight, xi is the sub-efficiency index, and REF is the benchmark value, and generates a priority list for equipment modification.

[0046] The normal operation model also includes a security isolation gateway, which uses a whitelist mechanism to filter control commands. All optimization commands must be verified by SHA3512 digital signature. It realizes the functions of improving system coordination through OPCUA / Modbus dual channels and RESTful API, and improving system stability through multi-source monitoring and survival analysis model dynamic compensation for performance degradation detection.

[0047] Example 2: Please refer to Figure 1 A mechanism- and data-driven intelligent control method and system for pump group optimization. The normal operation model is mainly used to meet the daily pump scheduling of water plants. According to the actual situation of the water plant, the operation interval is set for automatic operation. The model is divided into steps 1 to 6.

[0048] Step 1: Establish pump set database

[0049] The water pump database needs to include the flow-head relationship, flow-shaft power relationship, and flow-efficiency relationship for each water pump at its operating frequency; the shaft power, efficiency, flow rate, and high-efficiency range at different frequencies; and the rated power, power factor, and motor efficiency of the motor corresponding to each water pump. Additionally, it needs to include the high-efficiency range for each pump group, with step 1 further divided into steps 101 to 105.

[0050] Step 101: Obtain water pump data from the water plant

[0051] This section of data includes the actual pump model of the water plant, rated flow rate and its corresponding head and rated shaft power, as well as the flow-head characteristic curve, flow-shaft power characteristic curve, and flow-efficiency characteristic curve at the rated frequency.

[0052] Step 102: Fit the characteristic curves and expressions of each water pump under power frequency conditions.

[0053] First, based on the pump curves provided by the manufacturer at the power frequency, we fit the flow-head relationship, flow-axis power relationship, and flow-efficiency relationship of each pump at the power frequency.

[0054] When the frequency is constant, the standard form of the flow-head relationship is:

[0055] H = H0 + S·Q 2

[0056] Where H0 represents the virtual total head of the pump when the flow rate is 0, and S is the pump internal resistance coefficient.

[0057] When the frequency is constant, the relationship between the pump flow rate and shaft power can be approximately expressed as:

[0058] P = a1Q + b1

[0059] Where a1 and b1 are the power fitting coefficients of the water pump.

[0060] When the frequency is constant, the flow efficiency relationship can be approximately expressed as:

[0061] η=a2Q 2 +b2Q

[0062] Where a2 and b2 are the efficiency fitting coefficients of the water pump.

[0063] Step 103: Calculate the relationship between the characteristic curves of the water pump under non-power frequency conditions.

[0064] When the flow rate is Q, the head H′ at frequency f′ and the power frequency f N The expression for time is as follows:

[0065] H′=H0·(f′ / f N ) 2 -S·Q 2

[0066] When the frequency is f′, the shaft power P′ and the power frequency f N Shaft power P at time N The relationship is as follows:

[0067] P'=(f′ / f N ) 2 a1Q+(f′ / f N ) 3 b1

[0068] When the frequency is f′, the expressions for flow rate Q and efficiency η′ can be directly derived from the flow-efficiency relationship at power frequency, as follows:

[0069]

[0070] Assuming the efficiency η of the water pump k Not less than ω·η k,max At frequency f′, the pump operates at high efficiency, and the midpoint of the interval is taken as the point of highest efficiency. The high-efficiency operating flow range of the pump at frequency f′ can be considered to satisfy the following expression:

[0071]

[0072] Step 104: Correction based on real data

[0073] Based on the real data obtained from the data acquisition module, the actual shaft power P″ of pump i at frequency f1 and flow rate Q is calculated. a,i Actual head H″ i Actual efficiency η′ i The formula for its calculation is as follows:

[0074] P″ a =P N η m

[0075]

[0076] Where P N η is the output power of the motor. m For motor efficiency, the theoretical shaft power P under the current operating condition can be calculated using the QP relationship fitted at frequency f1. a,i and take β P =P″ a,i / P a,i , which serves as a factor for attenuating the power of the water pump shaft.

[0077] Similarly, the actual efficiency and actual head can be calculated using the following formulas:

[0078]

[0079] P u =ρgQH

[0080]

[0081] Where P is u The effective power of the water pump, Δp is the pressure difference between the inlet and outlet of the water pump, and β is taken as... H =H″ i / H i As a correction factor for the pump head; take β η =η′ i ' / η i This serves as a correction factor for the actual efficiency of the water pump. Finally, each attenuation factor of the water pump is multiplied into the corresponding performance curve expression.

[0082] Step 105: Fitting the pump set scheme library

[0083] For a pump group consisting of multiple pumps, based on the calculated curves and expressions, the pump operating conditions are divided into different frequency ranges using x Hz as the scale. Different pumps within each frequency range are then grouped together to obtain a scheme library. i Pump i and f at frequency j Pump j at frequency k is a group, and the total flow rate Q of the pumps in the k-th group is... k Satisfy the following formula:

[0084]

[0085] Where δ k Let be the parallel flow loss coefficient of the pump set under scheme k. Therefore, the high-efficiency range of the pump set under scheme k is obtained by adding the endpoints of the high-efficiency range of each pump.

[0086] Step 2: Data Acquisition and Database Update

[0087] The system connects to the water plant's existing data acquisition platform, continuously acquiring multi-dimensional real-time operational data, including flow rate, pressure, pump frequency, energy consumption, and equipment status, and writing it to the core database for dynamic updates. This step ensures that subsequent optimization models always operate in a data-driven mode, providing support for accurate decision-making.

[0088] Step 3: Pipeline Demand Forecasting

[0089] DLinear focuses on the linear portion of deep neural networks and consists of multiple fully connected layers. Its working principle is to first decompose the original data input into trend components using a moving average kernel and residual components. Then, two linear layers are applied to each component, and these two sets of features are summed to obtain the final prediction. By explicitly handling trends, DLinear significantly improves predictive performance when data exhibits a clear trend and demonstrates good performance in time series forecasting. Therefore, DLinear is used to predict pipeline network demand.

[0090] Furthermore, the traditional Dlinear model, which uses two linear layers stacked together, does not perform well in predicting water plant flow rates. Therefore, we considered replacing the linear regression part with multinomial regression to better capture the abrupt changes in water consumption. Testing showed that the improved Dlinear outperformed the traditional Dlinear in all metrics. Moreover, compared to LSTM (Long Short-Term Memory) neural networks, the improved Dlinear requires less computational power, has a faster prediction speed, and higher prediction accuracy.

[0091] Step 4: Initial screening of proposals:

[0092] Step 4 is further divided into steps 401 to 403:

[0093] Step 401: Filtering the Demand Range

[0094] The pipeline demand values ​​from the current control time to the next control time are taken as the two endpoints of the demand range. Before each pump group control, the high-efficiency range of each pump allocation scheme is traversed and compared with the system demand range. The schemes that meet the requirements are initially selected and placed in the reserve pool. Then, the head is determined and calculated.

[0095] Step 402: Head Determination

[0096] The available head range of the optimal solution must correspond to the head range controlled by the water plant. Using the flow-head curve expression for each pump, i.e., the formula, the head range corresponding to the high-efficiency range of each solution is calculated, and it is determined whether the head range intersects with the head range controlled by the water plant. Solutions that do not meet the requirements are eliminated.

[0097] Step 403: Runtime Length Determination

[0098] Water pumps should not run continuously for extended periods; therefore, operating time constraints must be set. Operating time constraints for each water pump:

[0099] T i ≤T max

[0100] The parameters of all candidate solutions need to be compared with the constraints. If the constraints are not met, the solution is eliminated, and finally, n sets of solutions that meet the requirements are selected.

[0101] Step 5: Determine if there is a satisfactory solution.

[0102] Determine whether the initial screening stage effectively selected a solution that meets the criteria. If so, proceed to step 6. If the final number of selected solutions is 0, proceed to step 23 in the emergency plan.

[0103] Step 6: Calculate the fitness of the scheme

[0104] Before each pump unit adjustment, the pipeline flow demand Q during the adjustment period must be considered. j The two endpoints of the pipeline flow demand range are taken as the calculated values, and the pipeline flow demand is evenly distributed to each pump according to the rated flow value of each pump. The pump sets initially selected are then evaluated. The specific evaluation model is as follows:

[0105] Calculate the efficiency of pump i in scheme k using the expression for pump efficiency at f′ in the database:

[0106]

[0107] Q i,j For water pump i in the pipeline flow demand Q j Traffic quota at the location.

[0108] Frequent start-ups and shutdowns of water pumps are detrimental to energy conservation and emission reduction in water plants, therefore a minimum operating time for the pumps needs to be set. Furthermore, prolonged pump operation leads to decreased efficiency, increased operating temperature, accelerated seal aging, and increased cavitation risk; therefore, a maximum operating time for the pumps needs to be set, and an operating time coefficient α for each pump needs to be defined. i :

[0109]

[0110] Where T min T is the minimum operating time for a single water pump. max T is the maximum operating time of a single water pump. S The recommended continuous operating time for a single water pump is determined by the decrease in efficiency during continuous operation.

[0111] While long-term low-frequency operation reduces the risk compared to high-frequency operation, it still carries considerable risk compared to power frequency operation. To maximize the pump's operation at the power frequency, a frequency coefficient β is defined for each pump. i :

[0112]

[0113] The overall energy consumption of the pipeline system is determined by the required flow rate Q and head H. Therefore, when the pump type is fixed, improving the overall pump efficiency can effectively reduce energy loss and achieve energy conservation and carbon reduction. Define the fitness function Fk corresponding to the k-th scheme:

[0114]

[0115] Where F k,i,j Let η be the fitness of pump i at measurement point j under the k-th scheme, J be the number of measurement points, and η be the fitness of pump i at measurement point j. k,i,j Let x be the efficiency of pump i at measurement point j under the k-th scheme. k,i This is a 0-1 variable for the water pump i switch.

[0116] Step 7: Output the optimal solution

[0117] The scheme with the highest fitness obtained in step 6 is selected as the optimal scheme under the current settings and output to the control platform. If multiple schemes have optimal fitness, the pump group scheme with the shortest recent total running time is selected.

[0118] Example 3: Please refer to Figure 1 This paper presents a mechanism- and data-driven intelligent control method and system for pump set optimization. In response to sudden changes in water supply flow demand, it is necessary to comprehensively consider various possible factors and set corresponding contingency plans. The special contingency plan module can be divided into steps 21 to 25. Step 21: Trigger Condition Judgment. First, based on the system readings, the deviation between the predicted flow and the actual flow, as well as the water supply inlet pressure, are monitored in real time. If the deviation between the predicted and actual flow exceeds the warning value, or the water supply inlet pressure suddenly drops to the warning value, then proceed to step 2; otherwise, the special contingency plan module is not triggered. Step 22: Calculate the Maximum Available Flow of the Current Pump Set. Based on the estimated approximate head, calculate the highest operating flow of each scheme under the head requirement.

[0119]

[0120] Where δ k Let be the parallel loss coefficient, if If the current system requirements can be met, then exit the emergency plan; if If the current system requirements cannot be met, proceed to step 23. Step 23: Select a pump addition or reduction scheme based on the currently operating pump group. Since the new pump group scheme needs to meet the system flow and head requirements, and considering that the pumps should not be frequently started and stopped, the new scheme needs to be determined based on the currently operating pump group. If the triggering condition is a sudden drop in the pressure at the water supply pipe or the actual flow is greater than the predicted flow, then the scheme of adding the unit based on the currently operating pump group is selected. If the triggering condition is that the actual flow is less than the predicted flow, then the scheme of removing the unit based on the current scheme and the scheme of keeping the unit unchanged are selected. Step 24: Calculate the fitness of the selected schemes. Based on the pump matching scheme fitness evaluation system in step 6, the fitness of each scheme selected in step 23 can be obtained. Step 25: Output the optimal scheme. Based on the fitness of each alternative scheme calculated in step 24, select the scheme with the highest fitness as the optimal scheme under the current situation, output the optimal scheme, and exit the special contingency plan.

[0121] The technical solutions of Embodiments 3 and 2 achieve the following significant effects and advantages: 1. Low computing power requirement, significantly reduced deployment cost and threshold: By adopting a lightweight optimization algorithm model and distributed computing architecture, this invention greatly reduces the dependence on the computing power of the central server, enabling the system to be easily integrated into the existing intelligent control platform of the water plant and run stably, effectively saving initial investment and long-term operation and maintenance costs, and better meeting the actual engineering needs of the gradual transformation of modern water plants; 2. Strong scientific decision-making, outstanding adaptability and generalization ability: By constructing a hybrid decision-making model that integrates the mechanism model of water pumps and pipelines with historical operating data, this invention overcomes the limitations of simply relying on historical experience. Even under conditions of poor historical data quality or minor changes in the process structure of the water plant, the system can still make scientific decisions based on the mechanism model, demonstrating excellent adaptability and robustness, ensuring the continuous effectiveness of the optimization effect; 3. High operating efficiency, closely matching the actual state of equipment: By introducing a water pump efficiency model with a performance degradation factor, this invention can identify the performance degradation of pump groups in real time. This allows the optimized configuration strategy to dynamically adapt to the actual operating status of the equipment, always operating within a near-optimal efficiency range, thereby achieving truly refined energy consumption management and energy saving; 4. Good system compatibility, significantly reducing transformation costs and time: Through the design of a two-way communication and collaborative control interface with the existing system, this invention can seamlessly integrate with the existing automatic control system of the water plant. This structure maximizes the use and retention of the original control equipment and logic, avoiding "reconstruction from scratch" modifications, significantly shortening the construction period, and reducing the cost and complexity of upgrades; 5. Strong foresight, significantly improving water supply reliability: By introducing a multi-period rolling optimization module based on short-term flow forecasting, this invention achieves a leap from "passive response" to "active prediction". The system can adjust the pump operation strategy in advance based on predicted future flow changes, effectively avoiding unstable water supply pressure or configuration failure caused by changes in operating conditions, and ensuring the continuity and reliability of water supply services; 6. High safety redundancy and excellent system robustness: By constructing an emergency plan library containing various abnormal operating conditions and designing a real-time risk monitoring and self-triggering mechanism, when the system detects an emergency such as a sudden increase in flow or a sudden drop in pressure, it can immediately activate the preset response strategy and automatically and smoothly switch to the special plan mode, which greatly improves the safety assurance capability of the entire water supply system in the face of uncertain interference.

[0122] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A mechanism- and data-driven intelligent control method for pump set optimization, characterized in that, include: The system comprises a normal operation model and a special contingency plan model. The normal operation model is used for the daily optimization and control of pump sets in water plants under normal circumstances. The special contingency plan model is used to predict special situations such as excessive flow deviation and sudden changes in water supply pipe pressure. The normal operation model includes establishing a pump set database, data acquisition and database updates, pipeline demand prediction, initial screening of solutions, determination of satisfactory solutions, calculation of solution adaptability, and output of the optimal solution. The pump set database must contain the flow-head relationship, flow-shaft power relationship, and flow-efficiency relationship of each pump at its operating frequency, as well as the shaft power, efficiency, flow rate, and high-efficiency range at different frequencies. The normal operation model also includes dynamic performance degradation sensing and compensation. The dynamic performance degradation sensing and compensation uses vibration sensors, current harmonic analysis, and acoustic fingerprint recognition devices to collect equipment status data and establish a pump set life prediction model based on survival analysis.

2. The intelligent control method for pump group optimization driven by a hybrid mechanism and data as described in claim 1, characterized in that: The pump database includes data on water plant pumps, characteristic curves and expressions of each pump under power frequency conditions, relationship formulas of characteristic curves of each pump under non-power frequency conditions, and a library of pump group schemes for correction and fitting based on real data. The water pump data obtained from the water plant includes the actual water pump model, rated flow rate and its corresponding head and rated shaft power, as well as the flow-head characteristic curve, flow-shaft power characteristic curve, and flow-efficiency characteristic curve at the rated frequency. Fitting the characteristic curves and expressions of various water pumps under power frequency conditions: Based on the water pump curves provided by the manufacturer under power frequency, the flow-head relationship, flow-axis power relationship, and flow-efficiency relationship of each water pump under power frequency are fitted. When the frequency is constant, the standard form of the flow-head relationship is H = H0 + S·Q 2 Where H0 represents the virtual total head of the pump when the flow rate is 0, and S is the pump internal resistance coefficient. When the frequency is constant, the pump flow-axis power relationship can be approximately expressed as P = a1Q + b1, where a1 and b1 are the power fitting coefficients of the pump. When the frequency is constant, the flow efficiency relationship can be approximately expressed as η = a2Q. 2 +b2Q, where a2 and b2 are the efficiency fitting coefficients of the water pump.

3. The intelligent control method for pump group optimization driven by a hybrid mechanism and data as described in claim 1, characterized in that: The special contingency plan model includes a trigger condition judgment module, calculates the maximum available flow rate of the current pump group, selects a pump addition or reduction scheme based on the currently operating pump group, calculates the suitability of the selected scheme, and outputs the optimal scheme. The trigger condition judgment module monitors the deviation between the predicted flow rate and the actual flow rate, as well as the water supply pipe pressure value in real time based on the system reading value. If the deviation between the predicted flow rate and the actual flow rate exceeds the warning value, or the water supply pipe pressure drops sharply to the warning value, data acquisition and database update will be initiated; otherwise, the special contingency plan model will not be triggered.

4. The mechanism- and data-driven intelligent control method for pump group optimization according to claim 2, characterized in that: The relationship between the characteristic curves of the water pump under non-power frequency conditions is as follows: when the flow rate is Q, the head H' at frequency f' and the power frequency f'. N The expression for H' is as follows: H' = H0·(f' / f N ) 2 -S·Q 2 When the frequency is f', the shaft power P' and the power frequency f N Shaft power P at time N The relationship is as follows: P' = (f' / f N ) 2 a1Q+(f' / f N ) 3 b1, When the frequency is f', the expressions for flow rate Q and efficiency η' can be directly derived from the flow-efficiency relationship at power frequency. The specific expressions are as follows: Assuming the efficiency η of the water pump k Not less than ω·η k,max When the pump operates at high efficiency, and the midpoint of the interval is taken as the point of highest efficiency, the flow range of the pump operating at high efficiency at frequency f′ can be considered to satisfy the following expression.

5. The intelligent control method for pump group optimization driven by a hybrid mechanism and data as described in claim 2, characterized in that: Based on the correction using real data, the actual shaft power P″ of pump i at frequency f1 and flow rate Q is calculated using the real data obtained by the data acquisition module. a,i Actual head H″ i Actual efficiency η′ i The calculation formula is as follows: P″ a =P N η m ; Where P N η is the output power of the motor. m To determine the motor efficiency, the theoretical shaft power P under the current operating condition can be calculated using the QP relationship fitted at frequency f1. a,i and take β P =P″ a,i / P a,i , as a factor for attenuating the power of the water pump shaft; The actual efficiency and actual head can be calculated using the following formulas: P u =ρgQH; Where P is u The effective power of the water pump, Δp is the pressure difference between the inlet and outlet of the water pump, and β is taken as... H =H″ i / H i As a correction factor for the pump head; take β η =η′ i ' / η i As a correction factor for the actual efficiency of the water pump, the various attenuation factors of each water pump are multiplied into the corresponding performance curve expression.

6. A mechanism- and data-driven intelligent control system for pump group optimization, applicable to the mechanism- and data-driven intelligent control method for pump group optimization as described in any one of claims 1-5, characterized in that, Both the normal operation model and the special contingency plan model are equipped with a hybrid drive control platform, a dynamic performance evaluation terminal, a collaborative adapter component, and a containerized deployment architecture. The hybrid drive control platform is equipped with multi-core heterogeneous processors ARM and FPGA, with the FPGA dedicated to parallel computing of the mechanism model. The hybrid drive control platform is configured with a non-volatile memory module and a fingerprint database of the entire life cycle of the storage device. The dynamic performance evaluation terminal integrates a MEMS vibration sensor array, a wideband current analyzer, and an ultrasonic flow meter. The dynamic performance evaluation terminal has a built-in attenuation status visualization dashboard that displays the remaining life distribution cloud map in real time. The collaborative adapter component supports PROFINET and EtherCAT industrial bus protocols and provides a RESTful API interface for third-party systems to call optimization commands. The containerized deployment architecture encapsulates the optimization algorithm as a Docker microservice and achieves elastic resource scheduling through Kubernetes.

7. The mechanism- and data-driven intelligent control system for pump group optimization according to claim 6, characterized in that: The hybrid drive control platform also includes an energy consumption insight module, which constructs an energy efficiency health index based on the entropy method, EEI=(Σωi·xi) / REF, where ωi is the entropy weight, xi is the sub-item efficiency index, and REF is the benchmark value, and generates a priority list for equipment renovation.

8. The mechanism- and data-driven intelligent control system for pump group optimization according to claim 6, characterized in that: The normal operation model also includes a security isolation gateway, which uses a whitelist mechanism to filter control commands, and all optimized commands must be verified by SHA3512 digital signature.

Images