Water pump system energy consumption abnormity detection method and system

Abstract

The invention relates to the technical field of water pump energy consumption detection, in particular to a water pump system energy consumption anomaly detection method and system. 2, data preprocessing; 3, setting a data range; step 4, abnormity judgment; and step 5, performing abnormity early warning and processing, performing dimension reduction processing on multiple parameters through principal component analysis and other technologies, extracting main features, constructing a multi-index comprehensive analysis model, and meanwhile, accurately judging an energy consumption abnormity condition in combination with single-index abnormity judgment, multi-index comprehensive abnormity judgment and energy consumption deviation rate calculation. According to the multi-parameter fusion and intelligent judgment mode, the accuracy and reliability of energy consumption anomaly detection are greatly improved, the misjudgment and missed judgment conditions are effectively reduced, and more powerful guarantee is provided for stable operation of a water pump system.

Description

Technical Field

[0001] The present invention relates to the technical field of pump energy consumption detection, and particularly to a method and system for detecting abnormal energy consumption of a pump system. Background Art

[0002] In modern industrial production and various infrastructure construction, as a key power equipment, the pump system is widely used in many fields such as water supply and drainage, heating, ventilation and air conditioning, and industrial circulating water. With the increasingly prominent energy problem, the energy consumption management of the pump system has become the focus of attention. During the long-term operation of the pump, due to various factors such as equipment aging, working condition changes, and improper operation, the energy consumption situation is complex and changeable. The traditional pump energy consumption monitoring means are relatively single, mostly relying only on regular manual inspections or simple instrument measurements, unable to comprehensively and real-time grasp the energy consumption status of the pump system, and difficult to detect in time at the initial stage of abnormal energy consumption. This not only causes energy waste, but also may lead to equipment damage due to long-term abnormal operation, increasing maintenance costs and downtime, and having an adverse impact on the continuity and stability of production activities.

[0003] At present, although some technologies attempt to monitor the energy consumption of pumps, there are generally problems such as inaccurate detection methods and imperfect system functions. Some methods only judge energy consumption anomalies based on a single index, ignoring the correlation between multiple operating parameters, and are prone to misjudgment or missed judgment; there are also deficiencies in data processing of existing detection systems, lacking effective means for cleaning, normalizing, and in-depth analyzing a large amount of complex data collected, and unable to fully exploit the data value. In addition, the abnormal warning and processing mechanism is not intelligent enough to take corresponding measures in a timely and accurate manner according to the type and severity of the anomaly. Therefore, we propose a method and system for detecting abnormal energy consumption of a pump system, which can achieve efficient detection and processing of abnormal energy consumption of the pump system through comprehensive analysis of multiple key operating parameters, advanced data processing technologies, and intelligent abnormal determination and processing mechanisms. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a method and system for detecting abnormal energy consumption of a pump system, thereby solving the technical problems mentioned in the background art.

[0005] To achieve the above object, the present invention is realized through the following technical solutions:

[0006] A method for detecting abnormal energy consumption of a pump system includes the following steps:

[0007] Step 1: Data collection. Install a Hall effect current sensor on the power supply line of the water pump motor to measure the working current, install a resistive voltage divider type voltage sensor at the motor power input terminal to measure the voltage, install strain type pressure sensors on the inlet and outlet pipes of the water pump to measure the inlet and outlet pressures respectively, install a magnetoelectric speed sensor near the water pump shaft to measure the speed, and install an electromagnetic flow sensor at the water pump outlet to measure the flow rate. According to the scale of the water pump system, collect data at a frequency of once per minute for large systems and once every 10 seconds for small systems, and store the data in a MySQL relational database. The database table contains fields such as timestamp, sensor type, and measurement value, and indexes are established for the timestamp and sensor type fields, and the data is backed up regularly;

[0008] Step 2: Data preprocessing. Based on historical normal operation data and equipment technical parameters, set a reasonable value range for each sensor data to judge outliers, and process outliers using deletion, interpolation, and smoothing methods; use the min-max normalization method to normalize the cleaned data according to the minimum and maximum values of the historical normal operation data of each sensor, eliminating the differences in dimension and value range;

[0009] Step 3: Set data range. Collect a large amount of historical data of the water pump system under different working conditions and time periods during normal operation, and remove outliers and missing values; calculate the mean and standard deviation of each sensor data. According to statistical principles, set the normal range as [μ - kσ, μ + kσ], where k is determined according to the actual situation and the requirements for anomaly detection accuracy; use principal component analysis technology to centralize the normalized multi-sensor data matrix, calculate the covariance matrix, perform eigenvalue decomposition, select the eigenvectors corresponding to the first m largest eigenvalues to form a projection matrix, project the data into the principal component space to obtain the principal component score matrix, and set the normal range of the principal component scores;

[0010] Step 4: Anomaly determination. Compare each sensor data collected in real time with the set single-index normal range. If it exceeds, it is determined that the index is abnormal, and mild and severe anomaly levels are set according to the degree of exceeding; project the normalized multi-sensor data collected in real time into the principal component space, compare the obtained real-time principal component scores with the set normal range of the principal component scores. If it exceeds, it is determined that there is a multi-index comprehensive anomaly, and analyze the anomaly cause by calculating the principal component contribution rate; calculate the actual energy consumption of the water pump system according to the current, voltage, and power factor, calculate the theoretical energy consumption according to the flow rate, inlet and outlet pressure difference, and water pump efficiency, and then calculate the energy consumption deviation rate. When at least one single-index anomaly, multi-index comprehensive anomaly, and the energy consumption deviation rate exceeds 15%, it is determined that there is an energy consumption anomaly in the water pump system;

[0011] Step 5: Abnormal warning and handling. When it is determined that there is abnormal energy consumption, relevant personnel are notified by combining acoustic-optical alarm, SMS alarm, and email alarm. The content of the SMS and email includes the abnormal occurrence time, abnormal indicators, abnormal types, and severity information. The email also attaches a detailed analysis report of the abnormal data and a comparison chart of historical data. After receiving the alarm, relevant personnel check the sensor connection status, the wear condition of the pump mechanical components, and the electrical system faults. After processing, the system is retested, and the abnormal handling process is recorded.

[0012] In a possible implementation, in the abnormal value processing of the data preprocessing, the linear interpolation method calculates the values of the missing and abnormal points based on the values of two adjacent normal data points through a linear relationship. The formula is where I1 and I3 are the current values of two adjacent normal data points, and t1, t2, and t3 are the corresponding time points.

[0013] In a possible implementation, in the step of setting the data range, the calculation formula of the covariance matrix in the principal component analysis is where n is the number of data samples, and X c is the matrix after data centralization.

[0014] In a possible implementation, in the abnormal determination step, the formula for calculating the energy consumption deviation rate is where P real is the actual energy consumption, and P theo is the theoretical energy consumption.

[0015] In a possible implementation, a pump system energy consumption abnormal detection system includes the following modules:

[0016] Intelligent data acquisition module: The acquisition frequency adjustment unit, based on the fuzzy control model, takes the rotational speed, flow rate, and pressure change rate of the pump as input variables to automatically adjust the data acquisition frequency. The data transmission and caching sub-module transmits data through a high-speed industrial Ethernet, uses the cyclic redundancy check technology to prevent data transmission errors, and stores data in the local large-capacity cache using the first-in-first-out strategy to ensure data order and real-time performance.

[0017] Data preprocessing module: The noise filtering sub-module combines the moving average filtering, exponential smoothing filtering, and wavelet transform filtering algorithms to remove the noise in the collected data. The data missing and abnormal value repair sub-module uses linear interpolation, spline interpolation, and filling algorithms based on machine learning to handle data missing problems, and analyzes the causes of abnormal values by establishing an abnormal value discrimination model. The data normalization sub-module uses a method combining principal component analysis and min-max normalization to reduce the dimension of the data and eliminate the differences in dimension and scale.

[0018] Data range setting module: The indicator update submodule adjusts the normal range of a single indicator in real time by establishing a dynamic model based on the statistical analysis results of historical data, combined with the real-time operating status of the water pump system and environmental changes; the comprehensive analysis submodule uses principal component analysis and cluster analysis techniques to explore the potential correlation patterns between multiple indicators. The cluster analysis uses the K-Means algorithm to continuously calculate the Euclidean distance between the data point and the cluster center and adjust the cluster center until the algorithm converges;

[0019] Abnormal data judgment module: The multi-model fusion abnormal judgment submodule combines the judgment model based on threshold, the judgment model based on statistical analysis and the judgment model based on machine learning, performs abnormality detection by weighted fusion of the judgment results of each model, and dynamically adjusts the weight of each model according to different working conditions; the identification and classification submodule analyzes the change characteristics and mutual relationship of each indicator when an abnormality occurs, and combines the historical abnormal case library to judge the abnormality type and classify it into mild, moderate and severe abnormalities;

[0020] Abnormal warning and processing module: The abnormal warning submodule provides multiple methods such as sound and light alarm, SMS alarm, email alarm and mobile phone APP push alarm. Users can customize the warning method and level. The warning information includes basic information of the abnormality, detailed analysis report and possible cause suggestions; the abnormal processing decision submodule automatically generates processing decisions based on the abnormality type and level, based on historical abnormality processing experience and expert knowledge base. It automatically adjusts the operating parameters for minor abnormalities, provides processing suggestions for moderate abnormalities, activates emergency plans for severe abnormalities and notifies technical personnel for emergency repairs, and the processing decisions can be optimized in real time according to the actual processing results.

[0021] In a possible implementation, in the wavelet transform filtering algorithm of the data preprocessing module, the continuous wavelet transform formula is: in yes The conjugate function of is the wavelet mother function, a is the scale parameter, and b is the translation parameter.

[0022] In a possible implementation, in the data range setting module, the formula for calculating the Euclidean distance between a data point and a cluster center by the K-Means algorithm is: Where n is the dimension of data points, x ik and C jk The data points x are i and cluster center C j The kth eigenvalue of .

[0023] Beneficial effects compared with the prior art:

[0024] 1. In this solution, techniques such as principal component analysis (PCA) are used to reduce the dimensionality of multi-parameters, extract the main features, and construct a comprehensive multi-index analysis model. At the same time, by combining single-index anomaly determination, multi-index comprehensive anomaly determination, and energy consumption deviation rate calculation, the energy consumption anomaly situation can be accurately judged. This multi-parameter fusion and intelligent determination method greatly improves the accuracy and reliability of energy consumption anomaly detection, effectively reduces misjudgment and missed judgment situations, and provides a more powerful guarantee for the stable operation of the water pump system;

[0025] 2. In this solution, a combination of multiple algorithms is adopted in the data preprocessing stage, and the single-index normal range can be dynamically updated by the data range setting module according to the real-time operation status of the system and environmental changes. Techniques such as cluster analysis are used to deeply explore the multi-index correlation patterns, better adapt to the complex and changeable operating conditions of the water pump system, discover potential anomalies in a timely manner, provide more reliable data support for energy consumption anomaly detection, and ensure the timeliness and accuracy of the detection results;

[0026] 3. In this solution, through various methods such as sound and light alarm, SMS alarm, email alarm, and mobile APP push alarm, the abnormal information is notified to relevant personnel in a timely manner. The alarm information not only includes the basic abnormal information, but also provides a detailed analysis report and suggestions on possible reasons. Moreover, the abnormal handling decision sub-module will automatically generate a handling decision according to the abnormal type and level. This intelligent and accurate early warning and handling mechanism can quickly respond to abnormal situations, reduce energy waste and equipment damage caused by abnormal operation, reduce maintenance costs and downtime, and ensure the efficient and stable operation of the water pump system. Brief Description of the Drawings

[0027] The above description is only an overview of the technical solution of the present invention. In order to understand the technical means of the present invention more clearly and implement it according to the content of the specification, the following describes the preferred embodiments of the present invention in detail in conjunction with the drawings.

[0028] Figure 1 is the flowchart of the method for detecting energy consumption anomalies in the water pump system of the present invention;

[0029] Figure 2 is the framework diagram of the system for detecting energy consumption anomalies in the water pump system of the present invention. Detailed Embodiments

[0030] The preferred embodiments of the present invention will be described in detail with reference to the drawings. However, the present invention can be implemented in various different forms. Therefore, the present invention is not limited to the embodiments described below. In addition, in order to describe the present invention more clearly, components not connected to the invention will be omitted from the drawings;

[0031] The technical solutions in the embodiments of the present application are to solve the problems in the above-mentioned background technology. The general idea is as follows:

[0032] Example 1:

[0033] This example introduces a method for detecting abnormal energy consumption in a water pump system. By comprehensively analyzing multiple key operating parameters, the accuracy and reliability of abnormal energy consumption detection are improved, including data acquisition, data preprocessing, setting the normal range of data, determining data anomalies, and abnormal warning and handling.

[0034] Step 1: Data acquisition

[0035] 1. Sensor deployment

[0036] Current sensor: To accurately measure the working current I of the water pump motor, a Hall effect current sensor is selected. This sensor has the advantages of high precision, wide frequency band, good linearity, etc., and can effectively avoid the influence of electromagnetic interference on the measurement results. It is installed on the power supply line of the motor to ensure that the sensor is closely attached to the line to obtain accurate current data. At the same time, in order to prevent the sensor from being damaged by external environmental factors, a protective shell is used for protection.

[0037] Voltage sensor: The voltage sensor uses a resistive voltage divider type sensor, which can convert high voltage into low voltage signals for measurement. It is installed at the power input end of the motor, and through an accurate voltage division ratio, it ensures that the measured voltage U is accurate and reliable. During the installation process, attention should be paid to the insulation performance of the sensor to avoid electric leakage accidents.

[0038] Pressure sensor: Pressure sensors are installed on the inlet and outlet pipes of the water pump respectively to measure the inlet pressure P in and the outlet pressure P out . A strain type pressure sensor is selected, which has the characteristics of high sensitivity and fast response speed. During installation, it is necessary to ensure that the sensor is tightly connected to the pipe to avoid leakage, and at the same time, attention should be paid to the installation position of the sensor to avoid being affected by factors such as water flow impact.

[0039] Speed sensor: A magnetoelectric speed sensor is used to measure the speed n of the water pump. This sensor calculates the speed by detecting the pulse signal generated by the magnetic element on the rotating shaft, and has the advantages of non-contact measurement and strong anti-interference ability. It is installed near the rotating shaft of the water pump, and the distance between the sensor and the rotating shaft is adjusted to ensure that the pulse signal can be accurately detected.

[0040] Flow sensor: An electromagnetic flow sensor is installed at the outlet of the water pump to measure the flow Q of the water pump. The electromagnetic flow sensor works based on Faraday's law of electromagnetic induction, has high measurement accuracy for the flow of conductive liquids, and is not affected by factors such as fluid density and viscosity. During installation, it is necessary to ensure that the fluid in the pipe fills the measuring tube of the sensor to avoid factors such as bubbles affecting the measurement results.

[0041] 2. Data Acquisition and Storage

[0042] For large pump systems, since their operation is relatively stable and the operating conditions change slowly, after a large number of experiments and data analyses, the data acquisition frequency is set to once per minute. This can not only ensure that the collected data can reflect the operating state of the system but also avoid generating excessive data redundancy, reducing the burden of subsequent data processing.

[0043] For small pump systems, their operating conditions change relatively frequently. To capture the dynamic changes of the system in a timely manner, the data acquisition frequency is increased to once every 10 seconds. Through high-frequency data acquisition, the operating state of the system can be monitored more accurately, and potential abnormal situations can be detected in a timely manner.

[0044] Regarding data storage, a MySQL relational database is selected to store the collected data. An independent table is created for the data of each sensor, and the table structure is designed as follows:

[0045] Timestamp field: A high-precision timestamp data type is used to accurately record the data acquisition time, ensuring the chronological order and accuracy of the data.

[0046] Sensor type field: Used to identify which sensor the data comes from, facilitating subsequent data query and analysis.

[0047] Measured value field: Stores the actual measured data values of the sensor.

[0048] To improve the data retrieval efficiency, indexes are established for the timestamp field and the sensor type field. At the same time, the database is backed up regularly to prevent data loss. The backup data is stored in an external storage device to ensure data security and recoverability.

[0049] Step 2: Data Preprocessing

[0050] 1. Data Cleaning

[0051] Outlier judgment criteria: For the data of each sensor, a reasonable value range is set according to its historical normal operating data and the technical parameters of the device. Specifically, for the current I, through statistical analysis of a large amount of historical data, its normal range is determined to be [I min , I max . When the collected current value exceeds this range, it is regarded as an outlier. The reasons for the appearance of outliers may be due to sensor failures, electromagnetic interference, poor line contacts, etc.

[0052] Outlier handling methods: For outliers, the following several handling methods are adopted:

[0053] Deletion method: When the outlier significantly deviates from the normal range and cannot be corrected by other methods, it is directly deleted from the dataset. Specifically, if the collected current value is several times that of the normal range and it is confirmed through inspection that the sensor is fault-free, then this data is considered an outlier and is deleted.

[0054] Interpolation method: For missing or abnormal data points, methods such as linear interpolation or spline interpolation can be used for supplementation. Linear interpolation calculates the value of the missing or abnormal point based on the values of two adjacent normal data points through a linear relationship. Specifically, if the current values collected at times t1 and t3 are I1 and I3 respectively, and the current value collected at time t2 is an outlier, then the estimated value of I2 can be calculated through the linear interpolation formula to calculate the estimated value of I2.

[0055] Smoothing method: For data with noise interference, methods such as moving average or exponential smoothing can be used for smoothing. The moving average method averages the data within a certain time window to eliminate data fluctuations. For example, using the 3-point moving average method, for the i-th data point xi, its smoothed value

[0056] 2. Data normalization

[0057] To eliminate the differences in dimension and value range between data from different sensors, the min-max normalization method is used to normalize the cleaned data. For any sensor data x, its normalized value xnorm is calculated using the formula: where xmin and xmax are the minimum and maximum values of the sensor data in the historical normal operation data respectively.

[0058] In actual calculations, first, statistical analysis needs to be performed on the historical normal operation data to determine the xmin and xmax values of each sensor data. Then, the real-time collected data is substituted into the above formula for calculation to obtain the normalized data. The normalized data will be used for subsequent normal range setting and anomaly determination.

[0059] Step three: Set the data range

[0060] 1. Statistical analysis of historical data

[0061] Data collection and collation: Collect a large amount of historical data of the water pump system under normal operating conditions, and these data should cover different working conditions and operating periods. Collate the collected data, remove the outliers and missing values, and ensure the data quality.

[0062] Mean and standard deviation calculation: Statistical analysis is performed on the data of each sensor to calculate its mean μ and standard deviation σ. The mean μ reflects the average level of the data, and the standard deviation σ reflects the degree of dispersion of the data. The calculation formulas are as follows: where n is the number of data samples, and xi is the value of the i-th data point.

[0063] Normal range determination: According to statistical principles, the normal range is set as [μ - kσ, μ + kσ], where k is an empirical coefficient, usually taking values of 2 or 3. When k = 2, the normal range covers approximately 95% of the data; when k = 3, the normal range covers approximately 99.7% of the data. In this embodiment, according to the actual situation and the requirements for the accuracy of anomaly detection, k = 2.5 is taken. For example, for the current I, its normal range is [μ I - 2.5σ I , μ I + 2.5σ I .

[0064] 2. Comprehensive analysis of multiple indicators

[0065] Principle and application of principal component analysis (PCA): Principal component analysis is a commonly used data dimensionality reduction method that transforms the original multiple correlated variables into a set of uncorrelated principal components through linear transformation. In this embodiment, the principal component analysis method is used to perform dimensionality reduction processing on the data of multiple sensors and extract the main features of the data.

[0066] Specific steps of PCA:

[0067] Data centering: The normalized data matrix X of multiple sensors is centered, that is, each data point is subtracted by its corresponding mean to obtain the matrix X c .

[0068] Covariance matrix calculation: Calculate the covariance matrix S of the matrix X c , and the formula is: where n is the number of data samples.

[0069] Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix S to obtain the eigenvalues λ i and the corresponding eigenvectors e i . The eigenvalues reflect the variance sizes of each principal component, and the eigenvectors represent the directions of the principal components.

[0070] Projection matrix selection: Select the eigenvectors corresponding to the first m largest eigenvalues to form the projection matrix P, and project the data matrix X c onto the principal component space to obtain the principal component score matrix Y: Y = X c P

[0071] Determination of the normal range of principal component scores: Statistical analysis is performed on each principal component of the principal component score matrix Y to calculate its mean and standard deviation, and the normal range of the principal component scores is set. The method of [μ - kσ, μ + kσ] is also used, where the value of k is determined according to the actual situation.

[0072] Step 4: Abnormality determination

[0073] 1. Single-index abnormality determination

[0074] For the data of each sensor, the real-time collected data is compared with the set single-index normal range. If the data exceeds the normal range, it is determined that the index is abnormal. Specifically, if the real-time collected current I real satisfies I real <μ I -2.5σ I or I real >μ I +2.5σ I , it is determined that the current index is abnormal.

[0075] To improve the accuracy of abnormality determination, different abnormality levels can also be set. When the degree to which the data exceeds the normal range is small, it is determined as a mild abnormality; when the degree to which the data exceeds the normal range is large, it is determined as a severe abnormality. Specifically, when I real exceeds the normal range but does not exceed 1.5 times the upper or lower limit of the normal range, it is determined as a mild abnormality; when I real exceeds 1.5 times the upper or lower limit of the normal range, it is determined as a severe abnormality.

[0076] 2. Multi-index comprehensive abnormality determination

[0077] After normalizing the real-time collected data of multiple sensors and projecting it into the principal component space, the real-time principal component scores are obtained. The real-time principal component scores are compared with the set normal range of the principal component scores. If the real-time principal component scores exceed the normal range, it is determined that there is a multi-index comprehensive abnormality.

[0078] To further analyze the reasons for the multi-index comprehensive abnormality, the contribution rate of each principal component can be calculated. The contribution rate reflects the degree of explanation of the variation of the original data by each principal component. By analyzing the original variables corresponding to the principal components with larger contribution rates, it can be determined which sensor data has a greater impact on the abnormality.

[0079] 3. Energy consumption abnormality determination

[0080] Calculation of actual energy consumption: The actual energy consumption P of the water pump system real can be calculated from the current I and voltage U, taking into account the power factor The formula is: Power factor It can be measured by a power factor meter or estimated according to the characteristic curve of the motor.

[0081] Theoretical energy consumption calculation: According to the flow rate Q of the water pump, the pressure difference ΔP = P out -P in and the efficiency η of the water pump, the theoretical energy consumption P of the water pump system can be calculated theo , and the formula is: The efficiency η of the water pump can be obtained through the performance curve of the water pump or actual tests.

[0082] Calculation and determination of energy consumption deviation rate: Calculate the energy consumption deviation rate δ, and the formula is:

[0083]

[0084] When the single - index anomaly, multi - index comprehensive anomaly, and energy consumption deviation rate δ all meet certain conditions, it is determined that there is an energy consumption anomaly in the water pump system. Specifically, when at least one single - index anomaly exists, and there is a multi - index comprehensive anomaly, and at the same time the energy consumption deviation rate δ exceeds 15%, it is determined as an energy consumption anomaly.

[0085] Step Five: Abnormality warning and handling

[0086] 1. Abnormality warning

[0087] When it is determined that there is an energy consumption anomaly in the water pump system, the system immediately triggers an alarm mechanism. The alarm method combines audible and visual alarms, SMS alarms, and email alarms to ensure that relevant personnel can receive alarm information in a timely manner.

[0088] Audible and visual alarm: Install an audible and visual alarm in the control room of the water pump system. When an anomaly is detected, the alarm emits a loud sound and flashing lights to alert on - site staff.

[0089] SMS alarm: Send an alarm SMS to a pre - set mobile phone number through the SMS gateway. The SMS content includes details such as the time of anomaly occurrence, anomaly indicators, anomaly types (single - index anomaly, multi - index comprehensive anomaly, or energy consumption anomaly), and the severity of the anomaly.

[0090] Email alarm: Send an alarm email to the email of relevant management personnel. In addition to the content of the SMS alarm, the email can also attach a detailed analysis report of the anomaly data and a historical data comparison chart to facilitate in - depth analysis and decision - making by management personnel.

[0091] 2. Abnormality handling

[0092] After receiving the alarm information, the relevant personnel should promptly inspect and maintain the water pump system. First, determine the possible causes of the abnormality based on the alarm information. Specifically, check whether the sensors are working properly, whether there are mechanical failures in the water pump, and whether there are short circuits in the electrical system, etc.

[0093] Sensor inspection: Inspect the sensors involved in the alarm to check whether the connections of the sensors are loose and whether there are signs of damage. The sensors can be calibrated by replacing the spare sensors or using calibration equipment to determine whether the sensors are working properly.

[0094] Water pump mechanical inspection: Check whether there are wear, damage or jamming in the mechanical components such as the impeller, shaft and bearings of the water pump. The water pump can be disassembled for detailed inspection, and the damaged components can be replaced or repaired.

[0095] Electrical system inspection: Check whether there are faults such as short circuits and open circuits in the electrical components such as the motor windings, contactors and relays. Electrical detection equipment such as multimeters and insulation resistance meters can be used for detection, and the faulty components can be replaced or repaired.

[0096] After the processing is completed, retest the water pump system to ensure that the system resumes normal operation. At the same time, record the abnormal handling process for subsequent data analysis and experience summary.

[0097] Embodiment 2:

[0098] This embodiment introduces an abnormal energy consumption detection system for a water pump system. This system is composed of multiple cooperating modules and can realize the full-process automatic operation from data collection to abnormal handling, greatly improving the detection efficiency and accuracy. It includes an intelligent data collection module, a data preprocessing module, a data range setting module, an abnormal data determination module, and an abnormal warning and handling module.

[0099] I. Intelligent data collection module

[0100] 1. Acquisition frequency adjustment unit

[0101] This unit can automatically adjust the data acquisition frequency according to the operating state of the water pump system. When the system is running stably, the acquisition frequency is reduced to reduce the data storage and processing pressure; when the system shows fluctuations or abnormal signs, the acquisition frequency is automatically increased to capture data changes in a timely manner.

[0102] Specifically, by analyzing historical data and the dynamic characteristics of the system, an acquisition frequency adjustment model based on fuzzy control is established. This model takes the rotational speed, flow rate and pressure change rate of the water pump as input variables and outputs the appropriate acquisition frequency according to the preset fuzzy rules.

[0103] 2. Data transmission and caching sub-module

[0104] The collected data is transmitted through a high-speed industrial Ethernet to ensure the real-time and stability of the data. At the same time, to prevent data loss or errors during the data transmission process, the cyclic redundancy check (CRC) technology is used to verify the data.

[0105] A large-capacity data cache is set up locally. When the network fails or there is a transmission delay, the data can be temporarily stored in the cache and then transmitted after the network returns to normal. The cache adopts the first-in, first-out (FIFO) storage strategy to ensure the sequentiality of the data.

[0106] II. Data Preprocessing Module

[0107] 1. Noise Filtering Sub-module

[0108] For the complex noise that may exist in the collected data, a combination of multiple filtering algorithms is used for processing. In addition to the moving average filtering and exponential smoothing filtering in Embodiment 1, the wavelet transform filtering algorithm is also introduced.

[0109] The wavelet transform filtering technology effectively removes high-frequency noise and pulse interference through time-frequency analysis of the collected data, while retaining the important features of the signal. Based on the scaling and translation characteristics of the wavelet function, the signal is decomposed into different frequency channels. Let the wavelet mother function be which needs to satisfy The wavelet function family generated by it is where a is the scale parameter, controlling the scaling degree of the wavelet function; b is the translation parameter, controlling its position on the time axis. The continuous wavelet transform (CWT) of the water pump system data x(t) is calculated by the formula Here is 's conjugate function. The discrete wavelet transform (DWT) is the discretization process of the scale parameter a and the translation parameter b. The common discretization form is a = 2 j , b = k2 j (j, k ∈ Z). By selecting appropriate wavelet basis functions (such as the Daubechies wavelet series) and the decomposition level, and through multiple experimental optimizations, the best filtering effect can be achieved, improving the data quality and providing a more reliable data basis for subsequent data processing.

[0110] 2. Data Missing and Outlier Repair Sub-module

[0111] For the data missing problem, in addition to the linear interpolation and spline interpolation methods, a filling algorithm based on machine learning is also adopted. A neural network model is trained using historical data to predict the missing values according to the data at adjacent time points and other relevant parameters.

[0112] For outliers, in addition to simple deletion and replacement, in-depth cause analysis was also carried out. By establishing an outlier discrimination model, it was determined whether the outliers were caused by sensor failures, system interference, or real abnormal operating conditions, and corresponding treatment measures were taken.

[0113] 3. Data normalization sub-module

[0114] Considering the complex correlations between different sensor data, a method combining principal component analysis (PCA) and min-max normalization was adopted for data normalization. First, PCA was used to reduce the dimensionality of the data and extract the main characteristic components of the data, and then min-max normalization was performed on these characteristic components to eliminate the dimensional and scale differences of the data.

[0115] III. Data range setting module

[0116] 1. Index update sub-module

[0117] This sub-module not only determines the normal range of single indicators based on historical data statistical analysis, but also can be dynamically updated according to the real-time operating state of the system and environmental changes. Specifically, when the load of the water pump system changes or factors such as environmental temperature and humidity change, the normal range of each indicator is adjusted in real time by establishing a dynamic model.

[0118] 2. Comprehensive analysis sub-module

[0119] Data mining and machine learning algorithms were used to deeply analyze the correlation relationships between multiple indicators. In addition to the principal component analysis method in Embodiment 1, clustering analysis technology was also used to find potential correlation patterns between different indicators.

[0120] Clustering analysis technology is used to mine potential correlation patterns between multiple indicators. Specifically, the K-Means algorithm, which aims to divide the data set into K clusters, so that the data points within the clusters have high similarity and the similarity between clusters is low.

[0121] In its implementation steps, first randomly select K data points as the initial clustering centers C1, C2,..., C K . For each data point x in the data set i , by calculating the Euclidean distance from each clustering center (n is the dimension of the data point, x ik and C jk are the k-th eigenvalue of the data point x i and the clustering center C j respectively), x iAllocate it to the cluster to which the nearest cluster center belongs. Then recalculate the cluster center of each cluster (the average value of the eigenvalues of all data points in the cluster), and continuously repeat this process until the cluster center no longer changes or changes very little, and the algorithm converges. When processing the data of the water pump system, use parameters such as rotational speed, flow rate, and pressure as the characteristics of data points for clustering. If it is found that the cluster to which some data points belong is different from most other data points and corresponds to the abnormal time of the system, it can assist in judging that the system has an abnormal operating state and help determine the multi-index comprehensive normal range model.

[0122] IV. Abnormal Data Judgment Module

[0123] 1. Multi-Model Fusion Abnormal Judgment Sub-Module

[0124] Adopt a method of combining multiple abnormal judgment models for abnormal detection, including a judgment model based on thresholds, a judgment model based on statistical analysis, and a judgment model based on machine learning.

[0125] The judgment model based on thresholds makes a preliminary judgment according to the single-index and multi-index comprehensive normal ranges; the judgment model based on statistical analysis identifies abnormalities by calculating the statistical characteristics (such as mean, variance, etc.) and distribution characteristics of the data; the judgment model based on machine learning uses trained models such as neural networks and support vector machines to classify and predict the data.

[0126] Comprehensively integrate the judgment results of multiple models through weighted fusion to improve the accuracy and reliability of abnormal judgment. Dynamically adjust the weights of each model according to the performance of different models under different working conditions.

[0127] 2. Identification and Classification Sub-Module

[0128] This sub-module can identify and classify the detected abnormalities. By analyzing the change characteristics and mutual relationships of each index when the abnormality occurs and combining with the historical abnormal case library, judge the type of the abnormality, specifically sensor failure abnormality, water pump mechanical failure abnormality, electrical system failure abnormality, etc.

[0129] According to the severity of the abnormality and the degree of influence on the system operation, classify the abnormalities into different levels, such as mild abnormality, moderate abnormality, and severe abnormality. Different levels of abnormalities correspond to different warning and processing strategies.

[0130] V. Abnormal Warning and Processing Module

[0131] 1. Abnormal Warning Sub-Module

[0132] This sub-module provides multiple warning methods, including sound and light alarm, SMS alarm, email alarm, and mobile APP push alarm. Users can set different warning methods and warning levels according to their own needs and preferences.

[0133] The early warning information not only includes the basic information of the anomaly (specifically the time of anomaly occurrence, anomaly indicators, anomaly types, etc.), but also provides a detailed anomaly analysis report and suggestions for possible causes. At the same time, according to the level and type of the anomaly, different early warning tones and reminder frequencies are adopted.

[0134] 2. Anomaly handling decision sub-module

[0135] When an anomaly is detected, this sub-module automatically generates corresponding handling decisions according to the type and level of the anomaly. For minor anomalies, the system can automatically adjust the operating parameters of the water pump for self-repair; for moderate anomalies, the system issues an early warning and provides detailed handling suggestions to guide the operator to handle them; for severe anomalies, the system immediately activates the emergency plan, such as automatic shutdown protection, and at the same time notifies relevant technical personnel for emergency repair.

[0136] The handling decisions are generated based on historical anomaly handling experience and the expert knowledge base, and can be optimized and adjusted in real time according to the actual handling effects.

[0137] Finally, it should be noted that: Obviously, the above embodiments are only examples for clearly illustrating the present invention, rather than limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation manners here. And the obvious changes or modifications derived therefrom are still within the protection scope of the present invention.

Claims

1. A method for detecting abnormal energy consumption of a water pump system, characterized in that: The following steps are involved: Step 1: Data collection: install a Hall effect current sensor on the water pump motor power supply line to measure the working current, install a resistor divider voltage sensor on the motor power input end to measure the voltage, install a strain gauge pressure sensor on the water pump inlet and outlet pipes to measure the inlet and outlet pressures, install a magnetoelectric speed sensor near the water pump shaft to measure the speed, and install an electromagnetic flow sensor at the water pump outlet to measure the flow rate; according to the scale of the water pump system, collect data once a minute for large systems and once every 10 seconds for small systems, and store the data in a MySQL relational database. The database table contains timestamp, sensor type, and measurement value fields, and indexes are established for the timestamp and sensor type fields, and data is backed up regularly; Step 2: Data preprocessing: According to historical normal operation data and equipment technical parameters, a reasonable value range is set for each sensor data to determine abnormal values, and abnormal values ​​are processed by deletion, interpolation and smoothing methods; the minimum-maximum normalization method is used to normalize the cleaned data according to the minimum and maximum values ​​of the historical normal operation data of each sensor to eliminate the differences in dimension and value range; Step 3: Set the data range, collect a large amount of historical data under different working conditions and time periods when the water pump system is operating normally, and remove outliers and missing values; calculate the mean and standard deviation of each sensor data, and set the normal range to [μ-kσ,μ+kσ] according to statistical principles, where k is determined according to the actual situation and the accuracy requirements of anomaly detection; use principal component analysis technology to centralize the normalized multiple sensor data matrices, calculate the covariance matrix, perform eigenvalue decomposition, select the eigenvectors corresponding to the first m largest eigenvalues ​​to form a projection matrix, project the data into the principal component space to obtain the principal component score matrix, and set the normal range of the principal component score; Step 4: Abnormal determination: compare each sensor data collected in real time with the normal range of the set single indicator. If it exceeds the normal range, the indicator is determined to be abnormal, and the mild and severe abnormal levels are set according to the degree of excess. After normalizing the multiple sensor data collected in real time, project them into the principal component space, and compare the obtained real-time principal component score with the set normal range of the principal component score. If it exceeds the normal range, it is determined that the multi-indicator comprehensive abnormality exists, and the cause of the abnormality is analyzed by calculating the principal component contribution rate. The actual energy consumption of the water pump system is calculated based on the current, voltage and power factor, and the theoretical energy consumption is calculated based on the flow rate, inlet and outlet pressure difference and water pump efficiency, and then the energy consumption deviation rate is calculated. When at least one single indicator is abnormal, multiple indicators are comprehensively abnormal, and the energy consumption deviation rate exceeds 15%, it is determined that the water pump system has energy consumption abnormality. Step 5: Abnormal warning and processing. When it is determined that there is abnormal energy consumption, the relevant personnel are notified by combining sound and light alarms, SMS alarms and email alarms. The content of the SMS and email contains the time of the abnormality, abnormal indicators, abnormality type and severity information. The email also attaches a detailed analysis report of the abnormal data and a historical data comparison chart. After receiving the alarm, the relevant personnel check the sensor connection status, the wear of the mechanical parts of the water pump, and the electrical system failure. After processing, retest the system and record the abnormality handling process.

2. A method for detecting abnormal energy consumption of a water pump system according to claim 1, characterized in that: In the outlier processing of the data preprocessing, the linear interpolation method is to calculate the values ​​of missing and outlier points through a linear relationship based on the values ​​of two adjacent normal data points. The formula is: Among them, I1 and I3 are the current values ​​of adjacent normal data points, and t1, t2, and t3 are the corresponding time points.

3. A method for detecting abnormal energy consumption of a water pump system according to claim 1, characterized in that: In the step of setting the data range, the calculation formula of the covariance matrix in the principal component analysis is: Where n is the number of data samples, X c is the matrix after data centering.

4. A method for detecting abnormal energy consumption of a water pump system according to claim 1, characterized in that: In the abnormality determination step, the formula for calculating the energy consumption deviation rate is: Where P real is the actual energy consumption, P theo The theoretical energy consumption.

5. A water pump system energy consumption abnormality detection system, characterized in that: Includes the following modules: Intelligent data acquisition module: The acquisition frequency adjustment unit is based on the fuzzy control model, taking the speed, flow rate and pressure change rate of the water pump as input variables to automatically adjust the data acquisition frequency; The data transmission and cache submodule transmits data through high-speed industrial Ethernet, uses cyclic redundancy check technology to prevent data transmission errors, and uses a local large-capacity cache with a first-in-first-out strategy to store data to ensure data sequence and real-time performance; Data preprocessing module: The noise filtering submodule combines moving average filtering, exponential smoothing filtering and wavelet transform filtering algorithms to remove noise from the collected data; The missing data and outlier repair submodule uses linear interpolation, spline interpolation, and filling algorithms based on machine learning to handle missing data problems, and analyzes the causes of outliers by establishing an outlier discrimination model; the data normalization submodule uses a combination of principal component analysis and minimum-maximum normalization to reduce the dimension of the data and eliminate dimension and scale differences; Data range setting module: The indicator update submodule adjusts the normal range of a single indicator in real time by establishing a dynamic model based on the statistical analysis results of historical data, combined with the real-time operating status of the water pump system and environmental changes; the comprehensive analysis submodule uses principal component analysis and cluster analysis techniques to explore the potential correlation patterns between multiple indicators. The cluster analysis uses the K-Means algorithm to continuously calculate the Euclidean distance between the data point and the cluster center and adjust the cluster center until the algorithm converges; Abnormal data judgment module: The multi-model fusion abnormal judgment submodule combines the judgment model based on threshold, the judgment model based on statistical analysis and the judgment model based on machine learning, performs abnormality detection by weighted fusion of the judgment results of each model, and dynamically adjusts the weight of each model according to different working conditions; the identification and classification submodule analyzes the change characteristics and mutual relationship of each indicator when an abnormality occurs, and combines the historical abnormal case library to judge the abnormality type and classify it into mild, moderate and severe abnormalities; Abnormal warning and processing module: The abnormal warning submodule provides multiple ways of sound and light alarm, SMS alarm, email alarm and mobile phone APP push alarm. Users can customize the warning method and level. The warning information includes basic abnormal information, detailed analysis report and possible cause suggestions; The exception handling decision submodule automatically generates handling decisions based on the exception type and level, historical exception handling experience and expert knowledge base. It automatically adjusts operating parameters to repair minor exceptions, provides handling suggestions for moderate exceptions, and activates emergency plans for severe exceptions and notifies technicians for emergency repairs. The handling decisions can be optimized in real time based on the actual handling results.

6. A water pump system energy consumption abnormality detection system as claimed in claim 5, characterized in that: In the wavelet transform filtering algorithm of the data preprocessing module, the continuous wavelet transform formula is: in yes The conjugate function of is the wavelet mother function, a is the scale parameter, and b is the translation parameter.

7. A water pump system energy consumption abnormality detection system as claimed in claim 5, characterized in that: In the data range setting module, the formula for calculating the Euclidean distance between a data point and a cluster center by the K-Means algorithm is: Where n is the dimension of the data point, x ik and C jk The data points x are i and cluster center C j The kth eigenvalue of .

Images