Hydropower station spare part inventory intelligent prediction method and system based on digital twinning

Abstract

The invention belongs to the technical field of hydropower station operation and maintenance management, and relates to a hydropower station spare part inventory intelligent prediction method and system based on digital twinning. According to the method, the residual life of the key component of the equipment is predicted according to the historical fault data and the real-time monitoring data of the hydropower station equipment, and the operation condition and the health degree of the key component of the equipment can be known more comprehensively and accurately by fusing the historical fault data and the real-time monitoring data. According to the residual life of the key parts of the equipment, in combination with the equipment maintenance plan and the supply chain period, the expected demand of each spare part in the future time period is output, multiple factors such as the residual life of the key parts of the equipment, the equipment maintenance plan and the supply chain period are comprehensively considered, and prediction of the expected demand of the spare part is more comprehensive and reasonable. The safety inventory threshold value and the purchase priority are dynamically adjusted according to the expected demand of each spare part, matching of equipment loss prediction and the spare part demand is realized, the inventory structure is optimized, the spare part fund occupation is reduced, and the inventory management is more flexible and efficient.

Description

Technical Field

[0001] The present invention belongs to the technical field of operation and maintenance management of hydropower stations, and relates to an intelligent prediction method and system for spare parts inventory of hydropower stations based on digital twin. Background Technique

[0002] As a core element in the equipment maintenance and operation guarantee system of hydropower stations, spare parts refer to various spare parts that are pre-purchased and stocked to shorten the downtime caused by lack of parts during the maintenance of hydropower station equipment failures or planned maintenance, and to ensure normal power supply. They are used to replace faulty parts or deteriorated parts. Its essence is to achieve the dual goals of rapid repair of equipment failures and guarantee of power supply continuity through material reserve strategies. The management of spare parts inventory in hydropower stations is an important link to ensure the safe and stable operation of the power station, and involves transactions such as classification, procurement, storage, requisition and renewal of spare parts.

[0003] Currently, the traditional management of spare parts inventory in hydropower stations mainly has the following deficiencies: The demand prediction is extensive, relying on historical experience or fixed-cycle replacement, and cannot match the actual wear state of the equipment in real time, which is likely to lead to inventory backlogs or shortages. The backlog of high-value spare parts inventory occupies a large amount of funds, and in case of sudden failures, downtime losses are easily caused due to shortages of some spare parts inventory. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent prediction method and system for spare parts inventory of hydropower stations based on digital twin to solve the technical problem of extensive spare parts demand prediction.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions:[[]]END]] In a second aspect, the present invention discloses an intelligent prediction method for spare parts inventory of hydropower stations based on digital twin, including the following steps: Predict the remaining life of the key components of the equipment based on the historical failure data and real-time monitoring data of the hydropower station equipment; Based on the remaining life of the key components of the equipment, combined with the equipment maintenance plan and the supply chain cycle, output the expected demand for each spare part within the future time period; Dynamically adjust the safety inventory threshold and procurement priority according to the expected demand for each spare part.

[0006] Furthermore, it also includes the following steps: Establish a digital twin model based on the physical equipment of the hydropower station, collect the real-time data of the equipment operation, and dynamically associate the real-time data of the equipment operation with the digital twin model, specifically as follows: Generate a parametric three-dimensional model based on the CAD drawings of the physical equipment of the hydropower station; Assign physical property parameters to the parametric 3D model to obtain the parametric 3D model; the physical property parameters include mechanical parameters, electrical parameters, and thermodynamic parameters; Obtain the real-time data of the operation of the physical equipment in the hydropower station in real time; the real-time data of the equipment operation includes vibration, temperature, pressure, and lubrication status; Use the Kalman filtering algorithm to perform time series alignment on the operation data of all equipment, and remove outliers through data cleaning to obtain preprocessed data; Use a hierarchical coding system to dynamically associate the preprocessed data with the parametric 3D model to achieve automatic mapping of sensor data to corresponding components in the 3D scene.

[0007] Further, predicting the remaining life of the key components of the equipment based on the historical failure data and real-time monitoring data of the hydropower station equipment is as follows: Obtain the historical failure data and real-time monitoring data of the hydropower station equipment; the historical failure data includes maintenance records, sensor historical data, and environmental logs; the real-time monitoring data includes vibration, temperature, lubrication status, and equipment operation parameters; Preprocess the historical failure data and real-time monitoring data; the preprocessing process includes multi-source data time alignment, outlier removal, and missing value filling; Extract health features from the preprocessed data; the health features include vibration spectrum abnormality degree, temperature gradient change rate, and lubricating medium deterioration index; Construct a Weibull distribution survival analysis model for remaining life probability analysis based on the health features; Train the Weibull distribution survival analysis model based on the loss function.

[0008] Further, the calculation method of the vibration spectrum abnormality degree is as follows:

[0009] In the formula, is the vibration spectrum abnormality degree, is the current vibration spectrum, is the reference spectrum in the healthy state of the equipment, is the frequency point, is the number of frequency points; The calculation method of the temperature gradient change rate is as follows:

[0010] In the formula, is the temperature gradient change rate, is the temperature value at the current time point of, and are the temperature values at adjacent time points, is the sampling interval time; The calculation method of the lubricant medium deterioration index is as follows:

[0011] In the formula, is the lubricant medium deterioration index, is the concentration of ferromagnetic particles in the oil, is the particle concentration threshold, is the current oil viscosity, is the nominal viscosity of the new oil, is the water content, is the maximum allowable water content, , and are the weight coefficients, and + + = 1.

[0012] Furthermore, the Weibull distribution survival analysis model is as follows:

[0013] In the formula, is the remaining life of the key components of the equipment, is the shape parameter, is the scale parameter, is the preset failure probability threshold.

[0014] Furthermore, the expected demand for each spare part includes predictive replacement demand and replacement demand quantity; The triggering condition for the predictive replacement demand is: ≤ T1 + T2, where is the remaining life of the key components of the equipment, T1 represents the planned maintenance time, and T2 represents the supplier delivery cycle; The calculation method of the replacement demand quantity is as follows:

[0015] In the formula, is the replacement demand quantity, represents the mean time between failures statistically based on historical data, represents the total number of similar components, represents the exponential decay function.

[0016] Furthermore, it also includes the following steps: Adjust the replacement demand quantity according to the minimum order quantity of the supply chain to obtain the actual demand quantity. The adjustment formula is as follows:

[0017] In the formula, represents the adjusted actual demand quantity, is the replacement demand quantity, is the minimum order quantity of the supply chain.

[0018] Furthermore, the safety stock threshold is dynamically adjusted according to the expected demand of each spare part, specifically as follows: Obtain the safety stock threshold according to the standard normal distribution quantile corresponding to the target service level, the average time from placing an order to delivery by the supplier, the standard deviation of the spare part demand quantity, the average monthly demand quantity of the spare part, the standard deviation of the supplier delivery time, and the proportion of on-time delivery by the supplier within the agreed time. The acquisition formula is as follows:

[0019] In the formula, is the safety stock threshold, represents the standard normal distribution quantile corresponding to the target service level, represents the average time from placing an order to delivery by the supplier, represents the standard deviation of the spare part demand quantity, represents the average monthly demand quantity of the spare part, represents the standard deviation of the supplier delivery time, represents the proportion of on-time delivery by the supplier within the agreed time.

[0020] Furthermore, the method for obtaining the procurement priority is as follows: Rank the procurement priority according to the urgency score. The urgency score formula is specifically as follows:

[0021] In the formula, is the urgency score, and are the weight coefficients, and + = 1, represents the failure probability of the equipment within the remaining life, represents the inventory gap, represents the safety stock threshold.

[0022] In the second aspect, the present invention also discloses an intelligent prediction system for hydropower station spare part inventory based on digital twin, including: A remaining life prediction module, configured to predict the remaining life of the key components of the equipment according to the historical failure data and real-time monitoring data of the hydropower station equipment; An expected demand module, configured to output the expected demands of each spare part within a future time period according to the remaining life of the key components of the device, in combination with the device maintenance plan and the supply chain cycle; A dynamic update module, configured to dynamically adjust the safety inventory threshold and the procurement priority according to the expected demands of each spare part.

[0023] Compared with the prior art, the present invention has the following beneficial effects: The method of the present invention predicts the remaining life of the key components of the device based on the historical failure data and real-time monitoring data of the hydropower station equipment. By fusing the historical failure data and real-time monitoring data, it can more comprehensively and accurately understand the operating conditions and health status of the key components of the device. According to the remaining life of the key components of the device, in combination with the device maintenance plan and the supply chain cycle, it outputs the expected demands of each spare part within a future time period, comprehensively considering multiple factors such as the remaining life of the key components of the device, the device maintenance plan, and the supply chain cycle, making the prediction of the expected demands of the spare parts more comprehensive and reasonable. Dynamically adjusting the safety inventory threshold and the procurement priority according to the expected demands of each spare part avoids the problem that high-value spare parts occupy a large amount of funds, and it is easy to cause downtime losses due to shortage of spare parts during sudden failures, realizes the matching of equipment wear prediction and spare part demands, optimizes the inventory structure, reduces the occupation of spare part funds, and makes the inventory management more flexible and efficient.

[0024] The present invention realizes the deep integration of the device state and the three-dimensional scene through a digital twin model. At the same time, by combining the device health prediction and supply chain parameters, it realizes the high matching of equipment wear prediction and spare part demands, avoiding the problems of inventory backlog or shortage existing in traditional inventory management that relies on historical experience or fixed-cycle replacement.

[0025] The system of the present invention includes: a life estimation module, an expected demand module, and a dynamic update module; the life estimation module is configured to predict the remaining life of the key components of the device based on the historical failure data and real-time monitoring data of the hydropower station equipment; the expected demand module is configured to output the expected demands of each spare part within a future time period according to the remaining life of the key components of the device, in combination with the device maintenance plan and the supply chain cycle; the dynamic update module is configured to dynamically adjust the safety inventory threshold and the procurement priority according to the expected demands of each spare part. Each module cooperates with each other, and can realize the matching of equipment wear prediction and spare part demands, optimize the inventory structure, reduce the occupation of spare part funds, and make the inventory management more flexible and efficient. Description of the Drawings

[0026] Figure 1 It is the flowchart of the method of the embodiment of the present invention; Figure 2 It is the system module diagram of the embodiment of the present invention; Figure 3 It is the flowchart of the method of another embodiment of the present invention. Detailed implementation manners

[0027] To enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0028] It should be noted that the terms "first", "second", etc. in the specification of the present invention and the above-mentioned accompanying drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data used in appropriate cases can be interchanged so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0029] The present invention will be further described in detail below in conjunction with the accompanying drawings: See Figure 1 , the present invention discloses an intelligent prediction method for spare parts inventory of a hydropower station based on digital twin, including the following steps: S1. Predict the remaining life of the key components of the equipment according to the historical failure data and real-time monitoring data of the hydropower station equipment. By fusing the historical failure data and real-time monitoring data, the operating conditions and health levels of the key components of the equipment can be understood more comprehensively and accurately. Specifically as follows: Obtain the historical failure data and real-time monitoring data of the hydropower station equipment; the historical failure data includes maintenance records, sensor historical data and environmental logs; the real-time monitoring data includes vibration, temperature, lubrication status and equipment operating parameters; Preprocess the historical failure data and real-time monitoring data; the preprocessing process includes multi-source data time alignment, outlier removal and missing value filling; Extract health features from the preprocessed data; the health features include vibration spectrum abnormality degree, temperature gradient change rate and lubricating medium deterioration index; According to the health features, construct a Weibull distribution survival analysis model for remaining life probability analysis; Train the Weibull distribution survival analysis model based on the loss function.

[0030] In this embodiment, the calculation method of the vibration spectrum abnormality is as follows:

[0031] In the formula, is the vibration spectrum abnormality, is the current vibration spectrum, is the reference spectrum under the healthy state of the device, is the frequency point, is the number of frequency points; In this embodiment, the calculation method of the temperature gradient change rate is as follows:

[0032] In the formula, is the temperature gradient change rate, is the current time point of the temperature value, and are the temperature values of adjacent time points respectively, is the sampling interval time; In this embodiment, the calculation method of the lubricant medium deterioration index is as follows:

[0033] In the formula, is the lubricant medium deterioration index, is the concentration of ferromagnetic particles in the oil, is the particle concentration threshold, is the current oil viscosity, is the nominal viscosity of the new oil, is the water content, is the maximum allowable water content, , and are the weight coefficients, and + + = 1.

[0034] In this embodiment, the Weibull distribution survival analysis model is as follows:

[0035] In the formula, is the remaining life of the key components of the device, is the shape parameter, is the scale parameter, is the preset failure probability threshold.

[0036] In this embodiment, the following steps are further included: Build a digital twin model based on the physical equipment of the hydropower station, collect the real-time data of the equipment operation, and dynamically associate the real-time data of the equipment operation with the digital twin model, specifically as follows: Generate a parametric 3D model based on the CAD drawings of the physical equipment of the hydropower station; Assign physical property parameters to the parametric 3D model to obtain the parametric 3D model; the physical property parameters include mechanical parameters, electrical parameters, and thermodynamic parameters; Obtain the real-time data of the operation of the physical equipment of the hydropower station in real time; the real-time data of the equipment operation includes vibration, temperature, pressure, and lubrication status; Use the Kalman filter algorithm to perform time series alignment on the operation data of all equipment, and eliminate outliers through data cleaning to obtain preprocessed data; Use a hierarchical coding system to dynamically associate the preprocessed data with the parametric 3D model, realize the automatic mapping of sensor data to the corresponding components in the 3D scene, and realize the matching of sensor data and components.

[0037] S2. According to the remaining life of the key components of the equipment, combined with the equipment maintenance plan and the supply chain cycle, output the expected demand for each spare part in the future time period; comprehensively consider multiple factors such as the remaining life of the key components of the equipment, the equipment maintenance plan, and the supply chain cycle to make the prediction of the expected demand for spare parts more comprehensive and reasonable.

[0038] In this embodiment, the expected demand for each spare part includes predictive replacement demand and replacement demand quantity; The triggering condition for the predictive replacement demand is: ≤T1+T2, where is the remaining life of the key components of the equipment, T1 represents the planned maintenance time, and T2 represents the supplier delivery cycle; The calculation method for the replacement demand quantity is as follows:

[0039] In the formula, is the replacement demand quantity, represents the mean time between failures statistically based on historical data, represents the total number of similar components, represents the exponential decay function.

[0040] In this embodiment, the following steps are further included: Adjust the replacement demand quantity according to the minimum order quantity of the supply chain to obtain the actual demand quantity. The adjustment formula is as follows:

[0041] In the formula, represents the adjusted actual demand quantity, For the replacement demand quantity is the minimum order quantity of the supply chain

[0042] S3. Dynamically adjust the safety inventory threshold and procurement priority according to the expected demand of each spare part. On the one hand, it avoids the problem that high-value spare parts occupy a large amount of funds and it is easy to cause downtime losses due to spare part shortages during sudden failures. On the other hand, after optimizing the inventory structure, it is estimated that the capital occupation of spare parts is reduced by 20%-40%, and at the same time, the stock-out risk is reduced by 50%, making inventory management more flexible and efficient. Specifically as follows: Obtain the safety inventory threshold according to the standard normal distribution quantile corresponding to the target service level, the average time from placing an order to delivery by the supplier, the standard deviation of the spare part demand quantity, the average monthly demand quantity of the spare part, the standard deviation of the supplier's delivery time, and the proportion of the supplier delivering on time within the agreed time. The acquisition formula is as follows:

[0043] In the formula is the safety inventory threshold represents the standard normal distribution quantile corresponding to the target service level represents the average time from placing an order to delivery by the supplier represents the standard deviation of the spare part demand quantity represents the average monthly demand quantity of the spare part represents the standard deviation of the supplier's delivery time represents the proportion of the supplier delivering on time within the agreed time

[0044] In this embodiment, the method for obtaining the procurement priority is as follows: Rank the procurement priority according to the urgency score. The urgency score formula is specifically as follows:

[0045] In the formula is the urgency score and are the weight coefficients, and + = 1 represents the failure probability of the equipment within the remaining life represents the inventory gap represents the safety inventory threshold

[0046] Based on the above method, the present invention also discloses an intelligent prediction system for hydropower station spare part inventory based on digital twin. See Figure 2 , including: A remaining life estimation module, which is used to predict the remaining life of the key components of the equipment according to the historical failure data and real-time monitoring data of the hydropower station equipment An expected demand module, which is used to output the expected demand of each spare part in the future time period according to the remaining life of the key components of the equipment, combined with the equipment maintenance plan and the supply chain cycle; A dynamic update module, which is used to dynamically adjust the safety inventory threshold and procurement priority according to the expected demand of each spare part.

[0047] Each module of the system of the present invention cooperates with each other, which can reduce the occupation of spare part funds and make the inventory management more flexible and efficient.

[0048] Embodiment 2: Refer to Figure 3 , this embodiment discloses an intelligent prediction method for the inventory of spare parts in a hydropower station based on digital twin, including the following steps: Step S1: Establish a digital twin model based on the physical equipment of the hydropower station, collect the real-time data of the equipment operation, and dynamically associate the real-time data with the twin model by using a coding system to realize the automatic mapping of the real-time data and the three-dimensional scene. The real-time data includes equipment operation parameters and environmental data; Step S2: Construct an equipment health status evaluation model, analyze the historical failure data and real-time monitoring data, and predict the remaining life of the key components of the equipment; Step S3: According to the prediction result of Step S2, combined with the equipment maintenance plan and the supply chain cycle, output the expected demand of each spare part in the future time period; Step S4: Dynamically adjust the safety inventory threshold and procurement priority according to the expected demand of each spare part.

[0049] In an implementation scheme, the specific method of Step S1 is as follows: Step S1.1: Generate a parametric three-dimensional model based on the CAD drawings of the physical equipment of the hydropower station; Step S1.2: Assign physical property parameters to the physical equipment of the hydropower station. The physical property parameters include mechanical parameters, electrical parameters and thermodynamic parameters; Step S1.3: Real-time collect the operation data of the physical equipment of the hydropower station through the sensor network deployed on the physical equipment of the hydropower station. The operation data includes vibration, temperature, pressure and lubrication status; Step S1.4: Use the Kalman filter algorithm to align the time series data of all sensors, and eliminate outliers through data cleaning; Step S1.5: Dynamically associate the sensor data with the parametric three-dimensional model by using a hierarchical coding system to realize the automatic mapping of the sensor data and the corresponding components in the three-dimensional scene.

[0050] In an implementation scheme, the specific method of Step S2 is as follows: Step S2.1: Collect historical fault data and real-time monitoring data of hydropower station equipment. The historical fault data includes maintenance records, sensor historical data, and environmental logs. The real-time monitoring data includes vibration, temperature, lubrication status, and equipment operation parameters. Step S2.2: Preprocess the historical fault data and real-time monitoring data, including multi-source data time alignment, outlier removal, and missing value filling. Step S2.3: Extract health features from the preprocessed data. The health features include vibration spectrum abnormality degree, temperature gradient change rate, and lubricant medium deterioration index. Step S2.4: Based on the health features, construct a Weibull distribution survival analysis model for remaining life probability analysis. Step S2.5: Train the Weibull distribution survival analysis model based on the loss function.

[0051] In one implementation, the step S2.3: The calculation method of the vibration spectrum abnormality degree is as follows:

[0052] In the formula, is the current vibration spectrum, is the reference spectrum in the healthy state of the equipment; The calculation method of the temperature gradient change rate is as follows:

[0053] In the formula, is the temperature value at the current time point t, are the temperature values at adjacent time points respectively, is the sampling interval time; The calculation method of the lubricant medium deterioration index is as follows:

[0054] In the formula, is the lubricant medium deterioration index, is the concentration of ferromagnetic particles in the oil, is the particle concentration threshold, is the current oil viscosity, is the nominal viscosity of the new oil, is the water content, is the maximum allowable water content, 、 and are weight coefficients, and + + = 1.

[0055] In one embodiment, in step S2.4, the remaining life is calculated by the Weibull distribution survival analysis model using the following formula:

[0056] where, is the remaining life of the critical component of the equipment, is the shape parameter, is the scale parameter, is the preset failure probability threshold.

[0057] In one embodiment, the expected demand in step S3 includes predictive replacement demand and replacement quantity demand. The triggering condition for the predictive replacement demand is: ≤ T1 + T2, where T1 represents the planned maintenance time and T2 represents the supplier delivery cycle; the replacement quantity demand is calculated as follows:

[0058] where, represents the mean time between failures statistically based on historical data, represents the total number of similar components, represents the exponential decay function.

[0059] In one embodiment, step S3 further includes adjusting the demand according to the minimum order quantity MOQ of the supply chain: · where, represents the actual demand.

[0060] In one embodiment, the calculation method for dynamically adjusting the safety stock threshold in step S4 is as follows:

[0061] where, represents the standard normal distribution quantile corresponding to the target service level, represents the average time from when the supplier places an order to delivery, represents the standard deviation of the spare part demand, represents the average monthly demand for spare parts, represents the standard deviation of the supplier delivery time, represents the proportion of the supplier delivering on time within the agreed time.

[0062] In one embodiment, the procurement priority ranking in step S4 is based on the urgency score E:

[0063] In the formula, and are weight coefficients, and + = 1, represents the failure probability of the device within the remaining life, represents the inventory gap, represents the safety inventory threshold.

[0064] To achieve the above object, the present invention also provides an intelligent prediction system for hydropower station spare parts inventory based on digital twin, including: Digital twin model module: Establish a digital twin model based on the physical equipment of the hydropower station, collect real-time data of equipment operation, and dynamically associate the real-time data with the twin model using a coding system to achieve automatic mapping of real-time data and three-dimensional scenes. The real-time data includes equipment operation parameters and environmental data; Prediction module: Construct an equipment health status evaluation model, analyze historical failure data and real-time monitoring data, and predict the remaining life of key components of the equipment; Expected demand module: According to the prediction results, combined with the equipment maintenance plan and supply chain cycle, output the expected demand for each spare part within the future time period; Dynamic update module: Dynamically adjust the safety inventory threshold and procurement priority.

[0065] Compared with the prior art, the present invention has the following beneficial effects: The present invention realizes the deep integration of equipment status and three-dimensional scene through the digital twin model. At the same time, by combining equipment health prediction and supply chain parameters, it realizes a high degree of matching between equipment wear prediction and spare part demand, avoiding the problems of inventory backlog or shortage existing in traditional inventory management that relies on historical experience or fixed-cycle replacement.

[0066] The present invention can dynamically adjust the safety inventory threshold and procurement priority. On the one hand, it avoids the problem that high-value spare parts occupy a large amount of funds and are prone to downtime losses due to spare part shortages during sudden failures. On the other hand, after optimizing the inventory structure, it is calculated that the capital occupation of spare parts is reduced by 20% - 40%, and the shortage risk drops by 50% at the same time.

[0067] Embodiment 3: As Figure 3 shown, this embodiment provides an intelligent prediction method for hydropower station spare parts inventory based on digital twin. This method uses the digital twin model of the physical equipment of the hydropower station as a carrier, predicts the remaining life and failure probability of key components of the equipment, and combines the equipment maintenance plan and supply chain cycle to output the expected demand for each spare part within the future time period, so as to achieve high-precision prediction of hydropower station spare part demand and inventory optimization.

[0068] Step S1: Establish a digital twin model based on the physical equipment of the hydropower station, collect the real-time data of the equipment operation, and dynamically associate the real-time data with the twin model using a coding system.

[0069] The physical equipment of the hydropower station should include core components such as turbines, generators, transformers, and gates. The specific construction method of the digital twin model is as follows: Step S1.1: Generate a parametric 3D model based on the CAD engineering drawings of the physical equipment of the hydropower station; Model accuracy requirements: The geometric error of key components (such as bearings, windings) ≤ 0.1 mm, and the error of non-key components ≤ 1 mm.

[0070] Step S1.2: Assign physical property parameters to the physical equipment of the hydropower station. The physical property parameters include mechanical parameters, electrical parameters, and thermodynamic parameters; The detailed descriptions of each parameter are as follows: Mechanical parameters: material density, elastic modulus, fatigue strength; Electrical parameters: winding resistance, insulation class, electromagnetic characteristics; Thermodynamic parameters: thermal conductivity, coefficient of thermal expansion.

[0071] Step S1.3: Real-time collect the operation data of the physical equipment of the hydropower station through the sensor network deployed on the physical equipment of the hydropower station. The operation data includes vibration, temperature, pressure, and lubrication status; The sensor network includes: Vibration monitoring: Piezoelectric acceleration sensor (such as PCB 352C33), sampling frequency ≥ 10 kHz; Temperature monitoring: Infrared thermal imager (FLIRA65) and PT100 platinum resistance; Lubrication status monitoring: Online oil sensor (such as Spectro ScientificFerroCheck) to detect the ferromagnetic particle concentration; Environmental monitoring: Ultrasonic water level gauge, temperature and humidity sensor (HoneywellHIH8000 series).

[0072] Step S1.4: Align all sensor time-series data using the Kalman filter algorithm and remove outliers through data cleaning; The Kalman filter algorithm is a mature existing algorithm. The Kalman Filter is an efficient recursive filtering algorithm used to estimate the state of a dynamic system from observation data containing noise and can also be indirectly used for multi-sensor time-series alignment. Its core is to fuse sensor information at different time points through a dynamic model and observation data. If there are timestamp deviations in the data of each sensor, the out-of-sync measurement values can be unified to the same time point through the state prediction and update steps. For example, the data of sensor A at time t1 and sensor B at time t2 can be predicted to the same reference time t and then fused; The delayed update mechanism of the Kalman filter (such as "lag filtering") can handle the time-series inconsistency problem caused by communication delays. For example, recalculate the Kalman gain based on the current state with the old measurement value and update; The Kalman filter usually requires the known time deviation of the sensor (such as a fixed clock offset). If the time deviation is unknown, the time difference needs to be estimated first through other methods (such as cross-correlation analysis, timestamp calibration) and then compensated in the filtering.

[0073] Step S1.5: Dynamically associate the sensor data with the parametric 3D model using a hierarchical coding system to achieve automatic mapping of the sensor data to the corresponding components in the 3D scene; The coding structure of the hierarchical coding system: power plant code (4 digits) - equipment type code (3 digits) - component ID (6 digits) - sensor ID (5 digits), and a conventional coding system can also be used; Mapping rule: Automatically bind the sensor data to the corresponding components in the 3D model according to the coding, and support reverse query of associated data by clicking on the components.

[0074] After the construction of the digital twin model is completed, the digital twin model needs to be verified. In an offline environment, verify the model accuracy through fault injection testing.

[0075] Step S2: Construct a device health status assessment model, analyze historical fault data and real-time monitoring data, and predict the remaining life of the key components of the device.

[0076] The specific implementation method of this step is as follows: Step S2.1: Collect historical fault data and real-time monitoring data of hydropower station equipment. The historical fault data includes maintenance records, sensor historical data, and environmental logs. The real-time monitoring data includes vibration, temperature, lubrication status, and equipment operation parameters. The maintenance records are sourced from the maintenance work order database, which records the equipment fault time, faulty components, fault modes (such as wear, corrosion, fracture), maintenance measures, and information on replaced spare parts. The sensor historical data is sourced from the sensor historical repository, which stores the equipment operation data (such as time series data of vibration, temperature, pressure, current, etc.) for 30 days to 1 year before the fault occurs. The environmental logs are associated with the environmental parameters (such as humidity, water level, load fluctuation) at the time of the fault.

[0077] Step S2.2: Preprocess the historical fault data and real-time monitoring data, including multi-source data time alignment, outlier removal, and missing value filling. The purpose of multi-source data time alignment is to unify the timestamps of multi-source data and ensure the precise synchronization of fault events and sensor data. Outlier removal can use the Isolation Forest algorithm to identify and remove abnormal data points. Missing value filling: Linear interpolation is used to fill the sensor data.

[0078] Step S2.3: Extract health features from the preprocessed data. The health features include vibration spectrum abnormality degree, temperature gradient change rate, and lubricant medium deterioration index. The calculation methods for each health feature are as follows: The calculation method for the vibration spectrum abnormality degree is as follows: , where is the current vibration spectrum, is the reference spectrum under the healthy state of the equipment; The calculation method for the temperature gradient change rate is as follows: TGCR = , where is the temperature value at the current time point t, are the temperature values at adjacent time points respectively, is the sampling interval time; The calculation method for the lubricant medium deterioration index is as follows:

[0079] where is the concentration of ferromagnetic particles in the oil (ppm), is the particle concentration threshold (such as the corresponding value for level 18 / 16 / 13 in the ISO 4406 standard), is the current oil viscosity (cSt), is the nominal viscosity of the new oil, is the water content (ppm), is the maximum allowable water content (e.g., 500 ppm), and α, β, γ are weight coefficients, and α + β + γ = 1.

[0080] Step S2.4: Construct a Weibull distribution survival analysis model for remaining life probability analysis according to the health characteristics; the remaining life is calculated by the following formula in the Weibull distribution survival analysis model:

[0081] In the formula, is the shape parameter, is the scale parameter, is the preset failure probability threshold (e.g., = 0.9).

[0082] In a further preferred solution, a hybrid prediction model can be constructed, including an LSTM neural network for time series feature learning and a Weibull distribution survival analysis model for remaining life probability analysis. Among them, the LSTM neural network includes: Input layer: a multi-dimensional feature sequence with a time window length of T (e.g., T = 30 days) (dimension = 15, including features such as vibration, temperature, lubrication, etc.); Hidden layer: 2 layers of LSTM, with 128 units in each layer, and Dropout = 0.2 to prevent overfitting; Output layer: predict the health status score (HHI, range 0 - 100) within the future Δt time (e.g., 7 days). Based on the HHI trend output by the LSTM and the Weibull distribution survival analysis model, the predicted value of the remaining life of the device can be dynamically adjusted.

[0083] Step S2.5: Optimize the Weibull distribution survival analysis model based on the loss function; the Weibull distribution survival analysis model is one of the common parametric models, and its loss function is usually constructed based on the negative log-likelihood function, as shown below: , optimizing this loss function can estimate the shape parameter k and the scale parameter λ, and then predict the survival probability or risk; k > 0 is the shape parameter, λ > 0 is the scale parameter, t ≥ 0 is the time, and μ is the logarithmic transformation of the scale parameter λ.

[0084] If a hybrid prediction model is adopted, a composite loss function needs to be constructed for model training, and the mathematical expression of the composite function is as follows: , in the formula, is the weight coefficient (e.g., 0.6 / 0.7), represents the mean square error of the health score, represents the negative log-likelihood loss of the Weibull distribution.

[0085] The mathematical expression of The mathematical expression of , the description of each parameter: N is the number of training samples, is the predicted health score and the true score of the i-th sample, δi is the event indicator variable (δi = 1 indicates component failure, δi = 0 indicates component normality), f(ti) is the probability density function of the Weibull distribution (a known function), and S(ti) is the survival function (a known function).

[0086] Step S3: According to the prediction results, combined with the equipment maintenance plan and the supply chain cycle, output the expected demand for each spare part in the future time period.

[0087] The expected demand in this step (the expected demand for a certain component) includes predictive replacement demand and replacement demand quantity. The trigger condition for the predictive replacement demand is: ≤T1 + T2, where T1 represents the planned maintenance time and T2 represents the supplier delivery cycle; the calculation method of the replacement demand quantity is as follows (rounding up):

[0088] In the formula, represents the mean time between failures statistically based on historical data, represents the total number of similar components, represents the exponential decay function, which is used to describe the probability of no failure of the equipment within the remaining life.

[0089] Example: If = 0 (the remaining life is exhausted): e0 = 1, so 1 - 1 = 0, indicating that replacement is necessary (directly set R = N in the formula); If ≫ (the remaining life is much greater than the mean time between failures, then L≈0, so R≈N, that is, almost all components need to be replaced.

[0090] The calculated value is the theoretical value, and there is still a deviation from the actual demand. To avoid the difference between theory and practice, this step takes into account the supply chain issue and also includes adjusting the demand according to the minimum order quantity MOQ of the supply chain: · , in the formula, represents the actual demand.

[0091] Step S4: Dynamically adjust the safety stock threshold and procurement priority.

[0092] This step is used for optimizing the inventory structure, where the safety stock threshold is dynamically adjusted The calculation method is as follows:

[0093] In the formula, represents the standard normal distribution quantile corresponding to the target service level (data source: normal distribution table, set according to requirements (such as 95%, 99%)), represents the average time from the supplier's order placement to delivery (statistically based on the supplier's historical delivery data), d represents the standard deviation of the spare part demand (reflecting demand volatility, calculated based on historical demand data), d represents the average monthly demand for spare parts (using the mean of historical demand data), represents the standard deviation of the supplier's delivery time (reflecting delivery delay volatility, statistically based on the supplier's historical delivery time), represents the proportion of the supplier's on-time delivery within the agreed time (0 ≤ ρ ≤ 1, recorded by the procurement system).

[0094] In the formula, reflects the risk caused by spare part demand fluctuations during the delivery cycle. The longer the delivery cycle or the greater the demand volatility, the higher the safety stock required; reflects the risk caused by the supplier's delivery delay fluctuations. The greater the demand or the more unstable the delivery time, the higher the safety stock required; the square root term jointly models the uncertainties of demand and supply and integrates them into the total volatility through the square root; is the supplier reliability correction. If the supplier's on-time rate is low ( small), the safety stock is amplified to compensate for the risk; otherwise, the inventory can be reduced.

[0095] The basis for the procurement priority ranking is the urgency score E:

[0096] In the formula, and are the weight coefficients (weight distribution coefficients for failure risk and inventory gap), and + = 1. The weights can be set according to empirical values and adjusted according to the equipment risk level. For example, the weights of spare parts for high-risk equipment can be increased by 30% - 50% based on the benchmark weights. Generally, > , thus emphasizing the priority of failure risk. The weight coefficients can also be adjusted dynamically. For example, is increased under high-load operating conditions, and is increased when the supply chain is unstable; represents the failure probability of the equipment within the remaining service life ( ), reflecting the potential impact of component failures on production. The higher the failure probability, the higher the urgency score. represents the inventory gap (the difference between the current inventory level and the demand). represents the safety stock threshold.

[0097] Through the above method, the matching of equipment wear prediction and spare part requirements can be achieved. At the same time, the inventory structure is optimized.

[0098] Based on the above method, this embodiment provides a digital-twin-based intelligent prediction system for hydropower station spare part inventory, including: Digital twin model module: Establish a digital twin model based on the physical equipment of the hydropower station, collect real-time data of equipment operation, and dynamically associate the real-time data with the twin model using a coding system to achieve automatic mapping of real-time data and three-dimensional scenes. The real-time data includes equipment operation parameters and environmental data; Prediction module: Construct an equipment health status assessment model, analyze historical failure data and real-time monitoring data, and predict the remaining life of key components of the equipment; Expected demand module: According to the prediction results, combined with the equipment maintenance plan and the supply chain cycle, output the expected demand for each spare part in the future time period; Dynamic update module: Dynamically adjust the safety stock threshold and procurement priority.

[0099] It should be noted that the above functional modules correspond one by one to the steps of the digital-twin-based intelligent prediction method for hydropower station spare part inventory provided in Embodiment 1. The specific functions implemented are the same as those of the digital-twin-based intelligent prediction method for hydropower station spare part inventory provided in Embodiment 1, and the beneficial effects achieved are also the same as those of the digital-twin-based intelligent prediction method for hydropower station spare part inventory provided in Embodiment 1.

[0100] An electronic device, including: a processor; a memory for storing computer program instructions; for implementing the steps of the digital-twin-based intelligent prediction method for hydropower station spare part inventory when executing the computer program.

[0101] A storage medium stores computer program instructions. When the computer program instructions are loaded and run by a processor, the processor executes the digital-twin-based intelligent prediction method for hydropower station spare part inventory.

[0102] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0103] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0104] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means realizes the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0105] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0106] The above content is only to illustrate the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. Any modification made on the basis of the technical solution according to the technical idea proposed by the present invention falls within the protection scope of the present invention.

Claims

1. A method for intelligent prediction of spare parts inventory of hydropower stations based on digital twins, characterized in that: The following steps are involved: Predict the remaining life of key components of hydropower station equipment based on historical failure data and real-time monitoring data; Output the expected demand for each spare part in the future period based on the remaining life of key equipment components, combined with equipment maintenance plan and supply chain cycle; Dynamically adjust safety stock thresholds and procurement priorities based on expected demand for each spare part.

2. According to claim 1, a method for intelligent prediction of spare parts inventory of hydropower stations based on digital twins is characterized in that: The following steps are also included: A digital twin model is established based on the physical equipment of the hydropower station, and the real-time data of the equipment operation is collected. The real-time data of the equipment operation is dynamically associated with the digital twin model, as follows: Generate parametric 3D models based on CAD drawings of hydropower station physical equipment; Assigning physical property parameters to the parameterized three-dimensional model to obtain the parameterized three-dimensional model; the physical property parameters include mechanical parameters, electrical parameters and thermodynamic parameters; Real-time acquisition of real-time data of the operation of the physical equipment of the hydropower station; the real-time data of the operation of the equipment includes vibration, temperature, pressure and lubrication status; The Kalman filter algorithm is used to align the time series of the operating data of all devices, and outliers are removed through data cleaning to obtain pre-processed data; A hierarchical coding system is used to dynamically associate the preprocessed data with the parameterized three-dimensional model, thereby achieving automatic mapping of sensor data with corresponding components in the three-dimensional scene.

3. According to claim 1, a method for intelligent prediction of spare parts inventory of hydropower stations based on digital twins is characterized in that: The remaining life of key components of the equipment is predicted based on the historical failure data and real-time monitoring data of the hydropower station equipment, as follows: Obtain historical fault data and real-time monitoring data of hydropower station equipment; the historical fault data includes maintenance records, sensor historical data and environmental logs; the real-time monitoring data includes vibration, temperature, lubrication status and equipment operating parameters; Preprocessing the historical fault data and real-time monitoring data; the row preprocessing process includes multi-source data time alignment, outlier removal and missing value filling; Extracting health features from the preprocessed data; the health features include vibration spectrum abnormality, temperature gradient change rate and lubricating medium degradation index; According to the health characteristics, a Weibull distribution survival analysis model for remaining life probability analysis is constructed; The Weibull distribution survival analysis model is trained based on the loss function.

4. According to claim 3, a method for intelligent prediction of spare parts inventory of hydropower stations based on digital twins is characterized in that: The calculation method of the vibration spectrum abnormality is as follows: In the formula, is the abnormality of the vibration spectrum, is the current vibration spectrum, is the reference spectrum when the device is in a healthy state. is the frequency point, is the frequency point number; The calculation method of the temperature gradient change rate is as follows: In the formula, is the temperature gradient change rate, For the current time point The temperature value, and are the temperature values ​​at adjacent time points, is the sampling interval; The calculation method of the lubricating medium degradation index is as follows: In the formula, is the lubricating medium degradation index, is the concentration of ferromagnetic particles in the oil, is the particle concentration threshold, is the current oil viscosity, is the nominal viscosity of new oil, is the water content, is the maximum allowable moisture content, , and is the weight coefficient, and + + =1.

5. According to claim 3, a method for intelligent prediction of spare parts inventory of hydropower stations based on digital twins is characterized in that: The Weibull distribution survival analysis model is as follows: In the formula, The remaining life of key components of the equipment. is the shape parameter, is the scale parameter, is the preset failure probability threshold.

6. According to claim 1, a method for intelligent prediction of spare parts inventory of hydropower stations based on digital twins is characterized in that: The expected demand for each spare part includes a predicted replacement demand and a replacement demand quantity; The triggering conditions for the predictive replacement demand are: ≤T1+T2, where is the remaining life of the key components of the equipment, T1 represents the planned maintenance time, and T2 represents the supplier's delivery cycle; The calculation method of the replacement demand is as follows: In the formula, To replace the demand, It represents the mean time between failures based on historical data statistics. Indicates the total number of similar parts. represents an exponential decay function.

7. The method for intelligent prediction of spare parts inventory of hydropower station based on digital twin according to claim 6 is characterized in that: The following steps are also included: Adjust the replacement demand according to the minimum order quantity of the supply chain to obtain the actual demand. The adjustment formula is as follows: In the formula, represents the actual demand after adjustment, To replace the demand, It is the minimum order quantity for the supply chain.

8. The method for intelligent prediction of spare parts inventory of hydropower station based on digital twin according to claim 1 is characterized in that: The safety stock threshold is dynamically adjusted according to the expected demand for each spare part, as follows: The safety stock threshold is obtained based on the standard normal distribution quantile corresponding to the target service level, the average time from order placement to delivery by suppliers, the standard deviation of spare parts demand, the average monthly demand for spare parts, the standard deviation of supplier delivery time, and the proportion of suppliers delivering on time within the agreed time. The formula is as follows: In the formula, is the safety stock threshold, represents the quantile of the standard normal distribution corresponding to the target service level, It represents the average time from order placement to delivery by suppliers. represents the standard deviation of spare parts demand, represents the average monthly demand for spare parts, represents the standard deviation of supplier delivery time, Indicates the proportion of suppliers delivering goods on time within the agreed time.

9. The method for intelligent prediction of spare parts inventory of hydropower station based on digital twin according to claim 1 is characterized in that: The method for obtaining the procurement priority is as follows: The procurement priority is sorted according to the urgency score. The urgency score formula is as follows: In the formula, Rate the urgency, and is the weight coefficient, and + =1, represents the probability of failure of the equipment within its remaining life, Indicates inventory gap, Indicates the safety stock threshold.

10. A hydropower station spare parts inventory intelligent prediction system based on digital twins, characterized in that: include: The life estimation module is used to predict the remaining life of key components of the equipment based on the historical failure data and real-time monitoring data of the hydropower station equipment; The expected demand module is used to output the expected demand for each spare part in the future time period based on the remaining life of key equipment components, combined with equipment maintenance plans and supply chain cycles; Dynamic update module, used to dynamically adjust safety stock thresholds and procurement priorities based on expected demand for each spare part.

Images