Feed pump energy-saving optimization method and system based on variable frequency speed regulation

Abstract

The invention discloses a feed pump energy-saving optimization method and system based on variable frequency speed regulation, and relates to the technical field of industrial process control and energy-saving optimization, and the method comprises the steps: constructing a causal analysis model by extracting residual statistical characteristics, recognizing the main drive type of carbon emission change, quantifying the causal effect, and estimating the life loss in combination with sensitivity; energy consumption and service life target weights are dynamically adjusted according to causal labels, a dual-objective optimization function is constructed, a multi-objective genetic algorithm is adopted to determine control parameters of frequency adjustment, and frequency self-adaptive adjustment is achieved. The offset abnormal driving reason is effectively identified, and the accuracy of energy-saving control and the equipment service life management and control capability are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial process control and energy saving optimization, in particular to a water supply pump energy saving optimization method and system based on frequency conversion speed regulation. BACKGROUND

[0002] The existing industrial water supply pump system widely adopts frequency conversion speed regulation control strategy to adapt to load fluctuation and realize energy saving operation, but in the actual operation process, the control system usually adjusts only according to the current load or simple energy consumption index, and does not fully consider the dynamic balance between the equipment operation health state and energy efficiency, resulting in lack of pertinence of frequency control decision. In the long-term operation, the water pump equipment faces complex and changeable operation environment, and is easy to appear multi-source disturbance factors such as abnormal current, high vibration, flow fluctuation and the like. These factors not only affect the energy consumption level, but also may cause the equipment wear and tear to be intensified and shorten the service life.

[0003] The traditional state detection method is mostly based on threshold value judgment, which is difficult to accurately identify the potential abnormal driving factors, lacks in-depth modeling of the disturbance generation mechanism, resulting in lag of frequency control response or imbalance of control deviation. In addition, the conventional optimization control strategy often adopts a single objective energy consumption minimization model, ignoring the equipment life loss as a long-term influencing factor, and cannot establish an effective trade-off mechanism between energy efficiency and life. Part of the research introduces a prediction model to estimate the energy consumption trend, but still does not include the causality of disturbance source into the control logic, and cannot realize the precise and interpretable dynamic optimization adjustment.

[0004] With the development of causal analysis and multi-objective optimization technology, the control method integrating equipment operation state, disturbance causal mechanism and energy efficiency evaluation has gradually become a research hotspot, but the existing scheme mostly stays in the theoretical modeling or experimental verification stage, lacking of systematic control strategy design and closed-loop implementation path facing complex industrial environment. SUMMARY

[0005] In view of the above existing problems, the present application is proposed.

[0006] Therefore, the present application provides a water supply pump energy saving optimization method based on frequency conversion speed regulation to solve the problems of existing water supply pump energy saving control method based on frequency conversion speed regulation in aspects of response mechanism lag, lack of causal modeling ability and fixed control strategy in abnormality identification, driving factor discrimination and target optimization strategy dynamic adjustment.

[0007] To solve the above technical problems, the present application provides the following technical scheme: In a first aspect, the present application provides a water supply pump energy saving optimization method based on frequency conversion speed regulation, which comprises collecting multi-dimensional original data in the water pump operation process, predicting the current state, and calculating the residual between the predicted value and the measured value. According to the residual, the dynamic process feature is constructed, and the upper and lower limits of monitoring are calculated in real time to determine whether the current running state exists out-of-limit deviation; When the deviation anomaly is detected, the process feature change of the current period and the running state of the previous period are extracted, the reason for the carbon emission change is analyzed, the causal label reflecting the main driving factor and the corresponding causal effect vector are outputted; Based on the causal effect vector, the equipment life consumption of the current period is estimated, and the life consumption and the energy consumption prediction result are taken as the optimization target to construct a double-target optimization function for adjusting the driving frequency; According to the causal label, the weight ratio between the life target and the energy consumption target is set to balance the control bias in the double-target optimization process; According to the double-target optimization function, the control parameter configuration scheme is determined, and the driving frequency of the water pump is dynamically adjusted accordingly.

[0008] As a preferred scheme of the variable frequency speed regulation energy-saving optimization method of the water pump, wherein: the multi-dimensional original data includes pressure, flow, vibration, current, voltage, frequency, temperature and humidity; The prediction of the current state includes extracting the historical data collected in time sequence, constructing a sliding analysis window covering a preset time length, extracting the statistical characteristics of each monitoring quantity in the target time period from the multi-dimensional original data based on the sliding analysis window, and estimating the prediction value of each monitoring quantity in the current period by using the weighted sliding mean method; the statistical characteristics include mean, slope and trend direction; Wherein, the prediction value of the current and voltage monitoring quantity is used to calculate the energy consumption prediction value of the current period, and the energy consumption prediction value is equal to the product of the current and voltage prediction value and the product of the current period length.

[0009] As a preferred scheme of the variable frequency speed regulation energy-saving optimization method of the water pump, wherein: the judgment of whether the current running state exists out-of-limit deviation includes: based on the residual, the residual statistical characteristics corresponding to the residual are calculated, including sliding mean, sliding standard deviation, range and exponential weighted moving average; according to the historical data, the upper and lower limits of monitoring corresponding to each statistical characteristic are set; the characteristic value in the current period is compared with the upper and lower limits of monitoring corresponding to the statistical characteristics, and when any residual statistical characteristic value exceeds the corresponding monitoring upper limit, it is determined that the current running state exists deviation anomaly.

[0010] As a preferred scheme of the variable frequency speed regulation energy-saving optimization method of the feed water pump, wherein: the analysis of the cause of the carbon emission change comprises, when it is determined that there is an overrun deviation in the current running state, extracting the difference between the residual statistical characteristics of each monitoring quantity in the current period and the corresponding residual statistical characteristics in the previous period as the residual statistical characteristic change value; and the residual statistical characteristic change value and the running state of the previous period are used to construct the input features of the causal analysis model. The causal analysis model is a causal random forest model composed of multiple causal decision trees, and before constructing each causal decision tree, the training samples are divided into structure training samples and effect estimation samples; In each node splitting process, all candidate splitting variables and candidate splitting thresholds are traversed using the structure training samples, for each candidate splitting mode, the conditional average treatment effect between the residual statistical characteristic change value and the carbon emission change is estimated in the effect estimation sample based on the corresponding structure training sample division result, and the corresponding residual sum of squares is calculated; the splitting variable and the splitting threshold corresponding to the minimum residual sum of squares are selected as the optimal splitting scheme of the node; The causal analysis model takes the residual statistical characteristic change value as the treatment variable and the running state as the covariate, and constructs a causal decision path according to the minimum error criterion of the conditional average treatment effect estimation; in the reasoning process of the causal analysis model, the causal label indicating the main driving category of the carbon emission change and the causal effect vector are output, the causal effect vector is the causal influence value of the residual statistical characteristic change value of each monitoring quantity on the carbon emission change; the confidence score corresponding to the causal label is the main driving category probability value output by the causal analysis model in the reasoning process, and the main driving category probability value is the maximum probability value obtained by averaging the category probabilities of the multiple causal decision trees of the causal random forest model.

[0011] As a preferred scheme of the variable frequency speed regulation energy-saving optimization method of the feed water pump, wherein: constructing the double-objective optimization function comprises calculating the life influence factor of each monitoring quantity in the current period based on the residual statistical characteristic change value and the causal effect vector, the life influence factor being the product of the residual statistical characteristic change value of each monitoring quantity and the corresponding causal effect vector element; The life influence factor and the preset monitoring quantity life sensitivity weight are weighted and added to obtain the life loss estimation value of the current period; the life loss estimation value and the energy consumption prediction value of the current period are normalized respectively, and combined with the preset target weight proportion to construct the double-objective optimization function for adjusting the driving frequency.

[0012] As a preferred scheme of the variable frequency speed regulation energy-saving optimization method of the feed water pump, wherein: the weight proportion between the life target and the energy consumption target comprises, according to the causal label and the confidence score corresponding to the causal label, constructing the target weight proportion of the double target optimization function on the basis of the preset energy consumption target basis weight value; the causal label comprises a health driving category, a load driving category, an environment driving category and an unknown category; When the causal label is the health driving category, the energy consumption target basis weight value is negatively adjusted based on the confidence score to obtain the energy consumption target weight value of the current period; When the causal label is the load driving category, the energy consumption target basis weight value is positively adjusted based on the confidence score to obtain the energy consumption target weight value of the current period; when the causal label is the environment driving category or the unknown category, the energy consumption target weight value is equal to the energy consumption target basis weight value; the life target weight value is the difference between the energy consumption target weight value and 1.

[0013] As a preferred scheme of the variable frequency speed regulation energy-saving optimization method of the feed water pump, wherein: the determination of the control parameter configuration scheme comprises, based on the double target optimization function, taking the life loss estimation value and the energy consumption prediction value of the current period as the optimization target, and using a multi-objective genetic algorithm to solve the double target optimization function to obtain the optimal solution satisfying the control target; In the optimization solving process, the control parameter candidate set is taken as the search space, and the target values of each control parameter combination under the double target optimization function are evaluated; according to the solving result of the double target optimization function, the optimal control parameter configuration scheme is selected, the control parameter configuration scheme comprises a target frequency value for adjusting the water pump driving frequency; the target frequency value is taken as the set value of the water pump driving frequency in the current period, the control instruction of the frequency converter or the speed regulation module is updated, and the dynamic adjustment of the water pump driving frequency is realized.

[0014] In the second aspect, the application provides a variable frequency speed regulation energy-saving optimization system of a feed water pump, comprising: a state prediction module for collecting multi-dimensional original data and estimating the prediction value of each monitored quantity; an anomaly detection module for calculating residual error and judging whether there is a deviation anomaly in the running state; a causal analysis module for outputting a causal label and a causal effect vector under the condition of deviation anomaly; a life evaluation module for estimating life loss based on the causal effect vector and constructing a double target optimization function; a weight adjustment module for setting the weight proportion of the life target and the energy consumption target according to the causal label and the confidence score; a control execution module for solving the optimization function and outputting the control parameter of the water pump driving frequency.

[0015] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the variable frequency speed-regulated feed water pump energy-saving optimization method according to the first aspect of the present application.

[0016] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements any step of the variable frequency speed-regulated feed water pump energy-saving optimization method according to the first aspect of the present application.

[0017] The present application has the following beneficial effects: the residual statistical feature is used to construct an offset detection mechanism, which can identify abnormal states in the operation of the water pump in real time; the causal random forest model is used to analyze the driving reasons of carbon emission changes, output causal labels and effect vectors, and accurately identify the reasons for the offset; the effect vectors and statistical feature changes are used to estimate the equipment life loss, and a double-objective optimization function is constructed in combination with the energy consumption prediction results; the life and energy consumption target weight ratio is dynamically set according to the causal label, and the scene adaptability of frequency regulation is enhanced; the optimization algorithm is used to solve the objective function, and the driving frequency set value is determined, so as to realize the comprehensive balance of equipment operation efficiency and life management. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Figure 1 The flowchart of the variable frequency speed-regulated feed water pump energy-saving optimization method. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0021] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0022] Second, the term "one embodiment" or "an embodiment" as used herein means an implementation that can include one or more features, structures, or characteristics, but is not necessarily composed of all of the features. Furthermore, a particular feature, structure, or characteristic can be combined in one or more embodiments with other features, structures, or characteristics. It will be understood that each of the features, structures, or characteristics described herein can be used in any one or more of the embodiments, without limitation.

[0023] Embodiment 1, Reference Figure 1 For one embodiment of the present application, the embodiment provides a variable frequency speed regulation based energy saving optimization method for water supply pumps, including the following steps: S1: Collect multi-dimensional original data during the operation of the water pump, and predict the current state to calculate the residual between the predicted value and the measured value.

[0024] Real-time collection of multi-dimensional parameters during the operation of the water pump through a sensor interface, including pressure , flow , vibration , current , voltage , frequency , temperature and humidity , the sampling frequency is not less than 10Hz, and the data is stored in a ring buffer queue in time series.

[0025] In each control cycle, the latest frame data is extracted from the buffer area to form a sliding analysis window covering a preset time length : Wherein, represents the measured value of any monitored quantity collected at time , such as pressure, flow, etc. represents the number of data points in the sliding window; represents the time interval between adjacent sampling points, in seconds; represents the data sampling frequency, in Hz, which is ; The sliding window constructed at time contains continuous historical data points ; represents the total time length of the sliding window, .

[0026] Three types of statistical features are calculated for each monitored quantity in the window: statistical features include mean, slope and trend direction.

[0027] The weighted moving average prediction applies a weighted moving average method to the feature sequence to estimate the prediction value of the current period : wherein is a smoothing coefficient, controlling the weight of recent data.

[0028] Calculate the residual between the prediction value and the measured value: wherein, represents the monitored quantity in the current period; represents the weight value of the th historical point in the weighted moving average; represents the weight decay coefficient; represents the exponential function, representing the weight factor decaying over time; represents the residual value of the monitored quantity in the current period; represents the prediction value of the monitored quantity at time .

[0029] The residual sequence is used for subsequent dynamic feature analysis and anomaly detection.

[0030] The energy consumption prediction value is calculated using the current and voltage prediction values , to calculate the current period energy consumption prediction value ; the energy consumption prediction value is equal to the product of the current and voltage prediction values and the product of the current period length.

[0031] Compared with the existing method which only relies on static data or a single measuring point to determine the working condition, S1 collects multi-dimensional raw data including pressure, flow, vibration, current, voltage, frequency, temperature and humidity, and combines the sliding window mechanism and time series modeling method to dynamically predict the current state, significantly enhancing the system's ability to describe complex operating states and the ability to anticipate trend changes. By constructing a sliding analysis window covering a preset time length and extracting mean, slope and trend direction from it, the system can accurately capture the state evolution trajectory of multi-dimensional monitored quantities, effectively improving the fitting degree of the prediction value to the true trend, and providing high-quality basic data support for subsequent residual detection and deviation identification.

[0032] The weighted moving average method is used for prediction modeling. Compared with the traditional simple moving average method, the exponential decay weight structure is introduced, so that the recent samples have a higher influence on the current prediction results, thereby enhancing the response sensitivity of the prediction model to short-period disturbances and the adaptability to local mutations. The prediction results of current and voltage are further used for energy consumption estimation, and the residuals are formed by comparing with the measured values, providing real-time difference feedback mechanism for subsequent deviation detection. This mechanism maintains high-precision prediction while having good engineering deployability, which can meet the dual requirements of low delay and high robustness in actual operating environment.

[0033] S2: According to the residual, a dynamic process feature is constructed, and the upper and lower limits of monitoring are calculated in real time to determine whether the current running state is out of limit; when the deviation anomaly is detected, the process feature change of the current period and the running state of the previous period are extracted to analyze the reasons for the change of carbon emissions, and the causal label reflecting the main driving factors and the corresponding causal effect vector are output.

[0034] From the residual value of each monitoring quantity , the following four types of residual statistical features are calculated: The moving average of the residual sequence is calculated using a sliding window with a length of ; in the same sliding window with a length of , the sliding standard deviation of the residual sequence is calculated, and in the window with a length of , the difference between the maximum value and the minimum value of the residual sequence is calculated as the range, and the formula is: The residual value is exponentially weighted with a smoothing coefficient , which is expressed as: The window length above is determined according to the actual control requirements of the system.

[0035] Wherein, represents the residual range of the monitoring quantity at time ; represents the residual value at the previous time points in the sliding window; represents the exponentially weighted moving average of the monitoring quantity ; represents the exponential smoothing coefficient.

[0036] According to the historical data of the feed water pump system, the corresponding monitoring upper and lower limits of each type of residual statistical feature are set, and the 3-sigma rule in statistical process control (SPC) is used for monitoring upper and lower limits: Upper control limit : Lower control limit : wherein, represents the upper control limit; represents the historical mean; represents the historical standard deviation; control factor Commonly takes 3.

[0037] In the real-time control cycle, the residual statistical characteristic values calculated for each monitoring quantity in the current cycle are compared with the corresponding monitoring upper and lower limits one by one: specifically including four types of characteristics of sliding mean, sliding standard deviation, range and exponential weighted moving average. When the current cycle value of any statistical characteristic exceeds its corresponding monitoring upper and lower limit range, it is determined that there is a deviation anomaly in the current running state.

[0038] When a deviation anomaly is detected, the difference between the residual statistical characteristic of each monitoring quantity in the current cycle and the corresponding residual statistical characteristic in the previous cycle is extracted, which is defined as the residual statistical characteristic change value .

[0039] The residual statistical characteristic change values of all monitoring quantities, together with the running state of the previous cycle (i.e. the statistical characteristic values of all monitoring quantities in the previous cycle), jointly constitute the input feature vector of the causal analysis model .

[0040] Causal random forest is used as the causal analysis model, and the specific steps include: Training data sample segmentation: the training data is divided into: Structural training samples: used to determine the split variable and split threshold of the decision tree node.

[0041] Effect estimation samples: used to estimate the conditional average treatment effect (CATE) of each node.

[0042] Decision tree node splitting criterion: use the structural training samples to traverse all possible residual statistical characteristic change values (treatment variables), enumerate candidate split thresholds, and the data division corresponding to each candidate split method is used in the effect estimation sample to calculate the treatment effect and residual sum of squares: the node treatment effect is defined as the conditional average effect of the residual statistical characteristic change value on the carbon emission change; the split variable and split threshold that result in the smallest residual sum of squares are selected as the optimal split strategy for the node.

[0043] When real-time inference is performed, the input features Substitute all the trained causal decision trees to get the main driving category probability value : main driving category probability value The maximum value of the average value of the probability of each category in the inference result of all causal decision trees; according to the probability value, determine the main driving category as the causal label of the current period; meanwhile, calculate and output the corresponding causal effect vector, and each element in the causal effect vector is the causal influence value of the residual statistical feature change value of each monitoring quantity on the carbon emission change.

[0044] Final output: causal label: ; causal effect vector: ; causal label confidence score: .

[0045] The existing water pump control system usually relies on fixed thresholds or simple rules to identify operation abnormalities, and it is difficult to capture potential risks caused by slight fluctuations in a timely manner. By jointly monitoring the sliding mean, sliding standard deviation, range and exponentially weighted moving average of the residual sequence, the changing trend of signals in each dimension can be comprehensively reflected in a short time window. Combined with the automatic setting of upper and lower limits based on historical statistical distribution, real-time comparison algorithm significantly improves the sensitivity and accuracy of abnormality determination, avoiding false negatives and false positives caused by single indicators or empirical thresholds.

[0046] After identifying the deviation anomaly, the change amount of the residual statistical feature is extracted and the running state of the previous period is introduced as a covariate to construct a multi-dimensional causal analysis input. The causal random forest adopts an honest splitting mechanism, which first divides the structural training samples to find the splitting condition, and then estimates the conditional average treatment effect using independent samples, realizing the decoupling of variable splitting and effect estimation. This splitting method fundamentally eliminates the bias of causal estimation and effectively distinguishes the true impact of different disturbance factors on carbon emissions, providing quantifiable driving force indicators for subsequent optimization.

[0047] Closely coupling the quantified causal effect vector with the residual feature change can accurately reflect the contribution of each monitoring quantity to the carbon emission deviation, providing fine-grained input for the wear-out model and double-objective optimization. When dynamically adjusting the target weight ratio, the causal label and confidence score guide the weight allocation, enabling the energy consumption and life target to adaptively balance in various scenarios such as healthy driving, load driving and environmental driving. The overall process breaks through the limitations of traditional single-objective or static multi-objective optimization, realizing a closed-loop collaboration from anomaly detection to causal identification and intelligent regulation.

[0048] S3: estimating the device life consumption in the current period based on the causal effect vector, taking the life consumption and the energy consumption prediction result as optimization objectives, and constructing a double-objective optimization function for adjusting the drive frequency; setting the weight proportion between the life objective and the energy objective according to the causal label, and balancing the control bias in the double-objective optimization process.

[0049] Based on the residual statistical feature change value and the causal effect vector, a life influence factor of each monitoring quantity in the current period is calculated , which is the product result of the residual statistical feature change value of each monitoring quantity and the corresponding causal effect vector element.

[0050] A preset monitoring quantity life sensitivity weight is introduced, which represents the sensitivity degree of the influence of each monitoring quantity on the device life consumption. The life sensitivity weight is determined according to long-term operation experience, satisfying: Based on the life influence factor , an estimated value of the device life consumption in the current period is calculated : Wherein, represents the life influence factor of the monitoring quantity in the current control period; represents the life sensitivity weight of the monitoring quantity ; represents the estimated value of the life consumption in the current control period; represents the monitoring quantity index.

[0051] In order to avoid the dimensional inconsistency between the estimated value of the life consumption and the energy consumption prediction value , the Min-Max normalization method is adopted, and the following definitions are made respectively: Life consumption normalized value: Energy consumption prediction value normalization: Wherein, , represents the minimum value and the maximum value of the life consumption estimated value in all historical control periods; represents the normalized life consumption value; , represents the minimum value and the maximum value of the energy consumption prediction value in all historical control periods; represents the normalized energy consumption prediction value.

[0052] The double-target optimization function for adjusting the driving frequency is constructed by combining the normalized life loss and the energy consumption prediction value, and setting the target weight proportion : wherein the life target weight value and the energy consumption target weight value satisfy: wherein, represents the life target weight value of the current period; represents the energy consumption target weight value of the current period; represents the double-target optimization function value, which is used as the basis for driving frequency decision; represents the energy consumption target basic weight value; represents the weight adjustment coefficient; represents the maximum value of the main driving category probability output by the causal analysis model, i.e., the confidence score.

[0053] According to the causal label and the corresponding confidence score , the energy consumption target basic weight value is dynamically adjusted to determine the target weight proportion of the current period: first, the energy consumption target basic weight value is predefined, and then it is adjusted according to the type of causal label.

[0054] The causal label includes the health driving category, the load driving category, the environment driving category, and the unknown category.

[0055] When the causal label is the health driving category: When the causal label is the load driving category: When the causal label is the environment driving category or the unknown category: wherein is the adjustment coefficient, which is used to control the amplitude of the confidence score adjustment on the weight value, and satisfies to ensure the rationality of the weight adjustment: the final life target weight value is determined as: In the dual-objective optimization phase, the causal effect vector and dynamic statistical characteristic change results output from the causal analysis stage can quantify the specific impact of each monitoring parameter on equipment lifespan in real time. Combined with sensitivity weights, these are accumulated to achieve a precise estimate of lifespan loss. This approach overcomes the shortcomings of traditional methods that rely on coarse weighting using single indicators or empirical coefficients, providing a more accurate and interpretable basis for lifespan loss evaluation.

[0056] The optimized objective function design employs a dual-objective normalized modeling approach, encompassing both lifetime loss and energy consumption prediction. By introducing historical extreme value intervals, it achieves direct fusion of parameters with different dimensions, effectively improving the convergence and physical rationality of the objective function. Simultaneously, the normalization process eliminates the bias risk caused by differences in monitoring data distribution, enhancing the optimization calculation's adaptability to changes in actual operating conditions.

[0057] In the dynamic target weight setting mechanism, the weights of energy consumption and lifespan targets are adjusted based on causal label categories and confidence scores, achieving intelligent adaptive adjustment of frequency regulation bias. Especially under different operating scenarios such as health-driven, load-driven, and environmental disturbances, the control system can automatically identify the dominant factors and optimize the target weight ratio accordingly, improving the dynamic response capability of multi-objective balance and the actual energy-saving lifespan consideration effect.

[0058] S4: Determine the control parameter configuration scheme based on the bi-objective optimization function, and dynamically adjust the pump drive frequency accordingly.

[0059] Within each control cycle, the bi-objective optimization function constructed in S3 is used as the objective function to optimize the control parameters. The goal is to minimize the weighted lifetime and normalized energy loss value within the current cycle. The control parameter optimization employs a non-dominated sorting genetic algorithm (NSGA-II), with its search space defined as the candidate set of frequency control parameters. This represents the set of adjustable target frequency values. Each candidate frequency value... For each optimization scheme, its fitness is determined by substituting it into the objective function. The final function value determines the result. The genetic algorithm initializes the population size to 100 and uses real-number encoding for chromosome expression; the crossover operation uses simulated binary crossover (SBX) with a crossover probability of 0.9; the mutation operation uses polynomial mutation with a mutation probability of [missing value]. ,in To control the dimensionality of parameters, in each iteration, individuals are selected based on Pareto ranking and crowding comparison, and selection, crossover, and mutation operations are performed based on the objective function value.

[0060] During the optimization process, each candidate frequency value All of them were substituted into the normalized value of the lifetime loss for the current cycle. With energy consumption normalized value Through the objective function Evaluate its fitness. After iteration, select the optimal frequency value that best matches the current state from the Pareto optimal solution set. This serves as the target value for the pump drive frequency in the current cycle. Target frequency value The control command input is set as the inverter's control command input and sent to the actuator via bus or control interface to achieve frequency update control. The frequency update operation is limited by physical constraints, such as the upper limit of frequency variation. and equipment safety zone This ensures the stability and engineering feasibility of frequency change operations. The process achieves adaptive frequency adjustment under dual-objective constraints, providing a closed-loop feedback basis for state prediction and residual construction in subsequent cycles.

[0061] in, Indicates the first of the candidate frequency control values One option; Indicates the current period The optimal frequency control value obtained from the solution is used as the drive frequency setpoint. This represents the upper limit of the frequency adjustment range, indicating the maximum allowable frequency variation within a single cycle. This indicates the permissible driving frequency range for device operation, ensuring that frequency adjustments do not exceed the safe operating range of the device.

[0062] Compared to conventional control methods based on fixed empirical parameters or simple weight combinations, this approach introduces a multi-objective control parameter solution mechanism based on a non-dominated sorting genetic algorithm, establishing an optimization relationship between execution efficiency and control accuracy. By using lifetime loss estimates and energy consumption predictions as objective function inputs, a dual-objective optimization structure is actively constructed, ensuring that the output of control parameter solutions that meet equipment sustainability and energy efficiency constraints can still be achieved even in the presence of offset anomalies. Compared to traditional control methods that use single-objective minimization as the criterion, this optimization mechanism supports a dynamic balance strategy that simultaneously considers lifetime and energy consumption, improving the system's ability to cope with multi-objective conflict scenarios.

[0063] In the specific parameter solution process, the candidate set of control parameters used covers multiple alternative solutions for the driving frequency. The optimization algorithm performs multiple rounds of individual selection, crossover, and mutation around the objective function value, outputting the optimal frequency setting value for the current operating state. Simultaneously, by introducing state data-driven weight ratio settings in real time during the optimization process, dynamic coupling between individual fitness and control bias is achieved, making the generated control parameter configuration more timely and state-adaptive. This approach effectively avoids the limitation of fixed-threshold control logic in adapting to complex operating states.

[0064] The final frequency updating operation introduces a physical constraint mechanism to limit the frequency adjustment range and optional interval, so as to ensure that the frequency adjustment is operated within the physical bearing range of the device and improve the stability and safety of the control system. By taking the target frequency value in the optimization result as the water pump driving frequency set value of the current period and outputting it to the frequency converter or speed regulation module to update the execution instruction, the closed-loop landing of the adjustment path is realized, and the implementability, constraint compliance and logical integrity of the control action are ensured.

[0065] The embodiment also provides a water supply pump energy-saving optimization system based on frequency conversion and speed regulation, which comprises: A state prediction module is configured to collect multi-dimensional original data and estimate the predicted values of the monitored quantities; an anomaly detection module is configured to calculate the residual and determine whether there is a deviation anomaly in the running state; a causal analysis module is configured to output a causal label and a causal effect vector under the condition of a deviation anomaly; a life assessment module is configured to estimate the life loss based on the causal effect vector and construct a double-objective optimization function; a weight adjustment module is configured to set the weight proportion of the life target and the energy consumption target according to the causal label and the confidence score; and a control execution module is configured to solve the optimization function and output the control parameters of the water pump driving frequency.

[0066] The embodiment also provides a computer device suitable for the water supply pump energy-saving optimization method based on frequency conversion and speed regulation, which comprises a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the water supply pump energy-saving optimization method based on frequency conversion and speed regulation proposed in the above embodiment.

[0067] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is configured to perform wired or wireless communication with external terminals. The wireless communication can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0068] The embodiment also provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the energy-saving optimization method for the variable frequency speed regulation of the feed water pump according to the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0069] In summary, the application achieves the synergy of life protection and energy-saving control by: constructing a causal model through residual analysis, identifying the main driving factors of carbon emission offset, estimating life loss and constructing a double-target optimization function combined with energy consumption prediction value, and dynamically adjusting driving frequency.

[0070] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application but not limit the application, and although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the application, and all of them should be covered in the scope of the claims of the application.

Claims

1. A variable frequency speed regulation based energy saving optimization method for feed water pumps, characterized in that: The method comprises collecting multi-dimensional original data of a water pump during operation, predicting a current state, calculating a residual between a predicted value and a measured value, constructing a dynamic process feature according to the residual, and calculating upper and lower limits of monitoring in real time to determine whether the current operation state has an out-of-limit deviation. When an out-of-limit deviation is detected, the process feature change of the current period and the operation state of the previous period are extracted to analyze the cause of the carbon emission change, and a causal label reflecting the main driving factor and a corresponding causal effect vector are output. Based on the causal effect vector, the device life loss of the current period is estimated, and the life loss and the energy consumption prediction result are taken as optimization objectives to construct a double-objective optimization function for adjusting the driving frequency. According to the double-objective optimization function, a control parameter configuration scheme is determined, and the driving frequency of the water pump is dynamically adjusted accordingly. The multi-dimensional original data includes pressure, flow, vibration, current, voltage, frequency, temperature, and humidity. The prediction of the current state includes extracting historical data collected in chronological order, constructing a sliding analysis window covering a preset time length, extracting statistical features of each monitoring quantity in the target time period from the multi-dimensional original data based on the sliding analysis window, and estimating the predicted value of each monitoring quantity in the current period by using a weighted sliding mean method; the statistical features include mean, slope, and trend direction.

2. The variable frequency speed-adjustable feed water pump energy-saving optimization method of claim 1, wherein: The predicted values of the current and voltage monitoring quantities are used to calculate the energy consumption prediction value of the current period, which is equal to the product of the predicted values of the current and voltage and the product of the time length of the current period. The determination of whether the current operation state has an out-of-limit deviation includes calculating residual statistical features corresponding to the residual based on the residual, including sliding mean, sliding standard deviation, range, and exponentially weighted moving average; setting the upper and lower limits of monitoring corresponding to each statistical feature according to historical data; comparing the feature values in the current period with the upper and lower limits of monitoring corresponding to the statistical features; and determining that the current operation state has an out-of-limit deviation when any residual statistical feature value exceeds the corresponding upper limit of monitoring. The analysis of the cause of the carbon emission change includes extracting the difference between the residual statistical features of each monitoring quantity in the current period and the corresponding residual statistical features in the previous period as residual statistical feature change values when it is determined that the current operation state has an out-of-limit deviation; and constructing input features of a causal analysis model together with the residual statistical feature change values and the operation state of the previous period.

3. The variable frequency speed-adjustable feed water pump energy-saving optimization method of claim 2, wherein: The causal analysis model is a causal random forest model composed of multiple causal decision trees, and before constructing each causal decision tree, the training samples are divided into structure training samples and effect estimation samples.

4. The variable frequency speed-adjustable feed water pump energy-saving optimization method of claim 3, wherein: ​ ​ In each node splitting process, all candidate splitting variables and candidate splitting thresholds are traversed by using the structure training sample, for each candidate splitting mode, based on the corresponding structure training sample division result, the conditional average treatment effect between the residual statistical characteristic change value and the carbon emission change in the effect estimation sample is estimated, and the corresponding residual sum of squares is calculated; the splitting variable and the splitting threshold corresponding to the minimum residual sum of squares are selected as the optimal splitting scheme of the node; The causal analysis model takes the residual statistical characteristic change value as a treatment variable, takes the operating state as a covariate, and constructs a causal decision path according to a minimum error criterion of conditional average treatment effect estimation; in the reasoning process of the causal analysis model, a causal label indicating a main driving category of carbon emission change and a causal effect vector are output, the causal effect vector being a causal influence value of the residual statistical characteristic change value of each monitoring quantity on the carbon emission change; the confidence score corresponding to the causal label is a main driving category probability value output in the reasoning process of the causal analysis model, and the main driving category probability value is a maximum probability value obtained by averaging the category probabilities of multiple causal decision trees by the causal random forest model.

5. The variable frequency speed-adjustable feed water pump energy-saving optimization method of claim 4, wherein: The construction of the double-target optimization function includes calculating a life impact factor of each monitoring quantity in the current period based on the residual statistical characteristic change value and the causal effect vector, the life impact factor being a product result of the residual statistical characteristic change value of each monitoring quantity and a corresponding causal effect vector element; The life impact factor and a preset monitoring quantity life sensitivity weight are weighted and accumulated to obtain a life loss estimation value of the current period; The life loss estimation value and an energy consumption prediction value of the current period are normalized respectively, and a target weight proportion is combined to construct a double-target optimization function for adjusting the driving frequency.

6. The variable frequency speed-adjustable feed water pump energy-saving optimization method of claim 5, characterized in that: The setting of the weight proportion between the life target and the energy consumption target includes, on the basis of the causal label and the confidence score corresponding to the causal label, constructing a target weight proportion for the double-target optimization function on the basis of a preset energy consumption target basic weight value; The causal label includes a health driving category, a load driving category, an environment driving category and an unknown category; When the causal label is the health driving category, the energy consumption target basic weight value is adjusted in a negative direction based on the confidence score to obtain an energy consumption target weight value of the current period; When the causal label is the load driving category, the energy consumption target basic weight value is adjusted in a positive direction based on the confidence score to obtain an energy consumption target weight value of the current period; When the causal label is the environment driving category or the unknown category, the energy consumption target weight value is equal to the energy consumption target basic weight value; and the life target weight value is a difference between the energy consumption target weight value and 1.

7. The variable frequency speed-adjustable feed water pump energy-saving optimization method of claim 6, wherein: The determination of the control parameter configuration scheme includes, based on the double-target optimization function, taking the life loss estimation value and the energy consumption prediction value of the current period as optimization targets, solving the double-target optimization function by using a multi-objective genetic algorithm to obtain an optimal solution satisfying the control target; In the optimization solving process, a candidate set of control parameters is taken as a search space, and objective values of each control parameter combination under the double-objective optimization function are evaluated; According to the solving result of the double-objective optimization function, an optimal control parameter configuration scheme is selected, the control parameter configuration scheme including a target frequency value for adjusting the water pump driving frequency; the target frequency value is taken as a set value of the current cycle water pump driving frequency, a control instruction of the frequency converter or the speed regulation module is updated, and dynamic adjustment of the water pump driving frequency is realized.

8. The energy-saving optimization system of the water supply pump based on the variable frequency speed regulation, the energy-saving optimization method of the water supply pump based on any one of claims 1-7, characterized in that: The method comprises: a state prediction module for collecting multi-dimensional original data and estimating predicted values of each monitoring quantity; an abnormality detection module for calculating a residual error and judging whether there is a deviation abnormality in the operating state; a cause-effect analysis module for outputting a cause-effect label and a cause-effect effect vector under the condition of the deviation abnormality; a life assessment module for estimating life loss based on the cause-effect effect vector and constructing a double-objective optimization function; a weight adjustment module for setting a weight ratio of a life objective and an energy consumption objective according to the cause-effect label and a confidence score; and a control execution module for solving the optimization function and outputting a control parameter of the water pump driving frequency. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the variable frequency speed regulation energy-saving optimization method of the feed water pump according to any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the variable frequency speed regulation energy-saving optimization method of the feed water pump according to any one of claims 1-7.

Images