New energy equipment part aging prediction method and system based on digital twin and multi-source data fusion

Abstract

The invention discloses a new energy equipment part aging prediction method and system based on digital twinning and multi-source data fusion, and relates to the technical field of digital twinning and intelligent prediction of new energy equipment, and the method comprises the steps: building a digital twinning model, and carrying out the working condition label labeling of multi-source monitoring data according to the current operation working condition of the equipment; extracting working condition aging characteristic parameters, calculating change gradients and working condition switching stress factors, and storing the change gradients and the working condition switching stress factors into a working condition associated aging characteristic library; inputting the working condition aging characteristic parameters and the working condition switching stress factors into a cross-working condition aging accumulation model, and outputting an aging accumulation index; and taking the aging accumulation index as a simulation initial state, simulating an attenuation trajectory of a component performance parameter under a preset future working condition sequence, when the performance parameter reaches a preset failure threshold value for the first time, extracting a running time length, and taking a difference value between the running time length and a current actual running time length as a component aging prediction remaining life. According to the invention, prospective residual life prediction based on the current aging state is realized.

Description

Technical Field

[0001] This invention relates to the field of digital twin and intelligent prediction technology for new energy equipment, and in particular to a method and system for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion. Background Technology

[0002] With the continuous expansion of the installed capacity of new energy equipment, the operating environment of key equipment such as wind power, photovoltaics, and energy storage is becoming increasingly complex. Their core components exhibit significant nonlinear degradation characteristics under the long-term influence of multiple coupled factors. Therefore, how to characterize the aging mechanism and predict the lifespan of equipment components based on multi-source monitoring data has gradually become an important development direction for intelligent operation and maintenance in the new energy industry. Existing research attempts to use digital twin technology to construct virtual equipment, achieving dynamic mapping between physical operating states and virtual models, and using sensor data for condition assessment and fault prediction. However, these methods typically focus only on a single operating condition or fixed equipment characteristic parameters, making it difficult to accurately reflect the cumulative degradation process of component performance in complex operating condition switching scenarios. Furthermore, the heterogeneity between multi-source data, the differences in stress load under different operating conditions, and the additional aging effects brought about by transient operating condition switching are not fully described in existing models. In addition, traditional lifespan prediction often relies on fixed characteristic thresholds or single statistical indicators, lacking the ability to fuse aging trajectories across operating conditions, resulting in insensitivity of remaining lifespan estimation to changes in the operating environment and insufficient prediction accuracy under dynamic load scenarios. As digital twin models become more real-time and high-fidelity, how to build a component aging prediction system that can span operating conditions, accumulate data, and simulate future operating scenarios has become a key challenge that current intelligent operation and maintenance technologies urgently need to overcome.

[0003] CN115438726A discloses a method and system for predicting equipment lifespan and failure types based on digital twin technology. It proposes to map operating states by constructing a physical equipment model and a digital twin virtual model, and extract comprehensive indicators from the operating twin data for determining lifespan end-of-life parameters. This method uses Kalman filtering or calculates remaining lifespan based on specific performance parameters reaching critical values, and can adapt well to the lifespan-determining characteristics of different equipment. However, this technology mainly relies on a single dominant lifespan parameter for prediction, failing to identify the segmented degradation behavior of equipment under multiple operating conditions, and lacking the ability to model stress mutations and cumulative effects caused by operating condition switching. It is therefore difficult to address the complex aging mechanisms of new energy equipment operating in environments with frequent wind speed fluctuations, light variations, and load disturbances.

[0004] CN120180869A discloses a device lifetime prediction method based on a multi-source fusion model. This method introduces feature normalization, RUL label generation, and sliding window sample construction, and trains the model using a CNN-Informer to ultimately construct a DCNN-Informer multi-source fusion model, improving lifetime prediction accuracy under complex degradation features. This method fully utilizes multi-source monitoring data, addressing the shortcomings of traditional methods in characterizing degradation patterns. However, the proposed model still relies on a pre-defined sample window and fixed data structure, making it difficult to synchronously update with real-time operating condition changes. Furthermore, it lacks a dynamic twin model corresponding to the virtual and real equipment, fails to correlate degradation rate differences under different operating conditions, and lacks the ability to visualize and simulate performance evolution under future operating condition sequences. Therefore, its prediction results are insufficient in responding to accelerated aging phenomena caused by sudden operating conditions and operating condition switching. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0006] Given that existing digital twin models are unable to accurately characterize the degradation mechanism of components under multiple coupled operating conditions, lack an effective fusion mechanism between multi-source monitoring data, and that life prediction models cannot quantify the additional aging effects caused by operating condition switching, nor can they simulate the performance degradation trajectory under future operating condition sequences in a virtual environment, resulting in insufficient accuracy in predicting the remaining life of components, this invention is proposed.

[0007] Therefore, the problem to be solved by this invention is how to achieve high-precision and interpretable prediction of the remaining life of components under complex operating conditions.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion, comprising, A digital twin model of a new energy equipment component is constructed. Virtual sensor nodes that are synchronized with the physical component in real time are set in the digital twin model to collect multi-source monitoring data of the physical component. The multi-source monitoring data is labeled with operating condition tags according to the current operating condition of the equipment. Based on the multi-source monitoring data, the aging characteristic parameters under different working conditions are extracted, the change gradient and the stress factor for working condition switching are calculated, and stored in the working condition-related aging characteristic library. The cross-condition aging accumulation model is constructed by inputting the aging characteristic parameters of the operating conditions and the stress factor of the operating condition switching, and outputs the aging accumulation index that characterizes the current aging degree of the component. Using the aging accumulation index as the initial state of the simulation, the decay trajectory of the component performance parameters under a preset future operating condition sequence is simulated in the digital twin model. When the performance parameter first reaches the preset failure threshold, the corresponding running time is extracted, and the difference between the running time and the current actual running time is used as the predicted remaining life of the component aging.

[0009] Secondly, embodiments of the present invention provide a new energy equipment component aging prediction system based on digital twin and multi-source data fusion, comprising: The digital twin model and data management module is used to construct digital twin models of new energy equipment components. Virtual sensor nodes that are synchronized with physical components in real time are set in the digital twin model to collect multi-source monitoring data of physical components and label the multi-source monitoring data with operating condition tags according to the current operating conditions of the equipment. The feature and stress management module extracts aging feature parameters under different working conditions based on the multi-source monitoring data, calculates the change gradient and working condition switching stress factor, and stores them in the working condition associated aging feature library. The health status assessment module is used to input the working condition aging characteristic parameters and the working condition switching stress factor into the constructed cross-working condition aging accumulation model, and output the aging accumulation index characterizing the current aging degree of the component. The life prediction simulation module is used to take the aging accumulation index as the initial state of the simulation, simulate the decay trajectory of component performance parameters under a preset future working condition sequence in the digital twin model, and extract the corresponding running time when the performance parameter first reaches the preset failure threshold, and use the difference between the running time and the current actual running time as the remaining life predicted by component aging.

[0010] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the steps of the above-described method for predicting the aging of new energy equipment components based on digital twins and multi-source data fusion.

[0011] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the above-described method for predicting the aging of new energy equipment components based on digital twins and multi-source data fusion.

[0012] Compared with existing technologies, the advantages of this invention are as follows: By constructing a digital twin model with virtual sensor nodes and labeling it with operating conditions, real-time mirror mapping of the physical component's operating status and accurate correlation of operating conditions with multi-source monitoring data are achieved, solving the problem that traditional methods cannot distinguish the differences in the aging effects of different operating conditions; by extracting operating condition aging characteristic parameters and calculating the operating condition switching stress factor, the steady-state aging process and the transient impact of operating condition switching are separated and quantitatively characterized, revealing the mechanism of the accelerated aging of components due to stress mutation during operating condition transition, thus overcoming the shortcomings of existing technologies that only focus on cumulative damage under a single operating condition while ignoring the influence of stress during operating condition switching; and through a cross-operating condition aging accumulation model... By integrating steady-state aging accumulation branches and dynamic stress accumulation branches, and designing an adaptive weight allocation mechanism, we have achieved collaborative modeling and adaptive weighted fusion of steady-state aging and dynamic stress. This enables the aging accumulation index to accurately reflect the true aging degree under complex operating condition sequences, overcoming the limitation of traditional linear accumulation models in handling operating condition coupling effects. By using the aging accumulation index as the initial state for simulation in a digital twin model to simulate the performance degradation trajectory under future operating condition sequences, we have achieved forward-looking remaining life prediction based on the current aging state. Compared with traditional methods based on historical data extrapolation, this method can consider the impact of future actual operating conditions on the aging process, significantly improving prediction accuracy and engineering practicality. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 A flowchart of a method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion; Figure 2 This is a structural diagram of a new energy equipment component aging prediction system based on digital twin and multi-source data fusion. Detailed Implementation

[0014] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0015] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0017] As mentioned in the background section, existing digital twin models struggle to accurately depict component degradation mechanisms under coupled multi-operating-condition conditions. Furthermore, they lack effective fusion mechanisms between multi-source monitoring data, and their lifespan prediction models cannot quantify the additional aging effects caused by operating condition switching. They also cannot simulate performance degradation trajectories under future operating condition sequences in a virtual environment, resulting in insufficient accuracy in predicting the remaining lifespan of components. To address these issues, this invention provides a method for predicting the aging of new energy equipment components based on digital twins and multi-source data fusion.

[0018] Reference Figure 1 , Figure 1 This is a flowchart illustrating a method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion, according to an embodiment of the present invention. Figure 1 As shown, a method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion includes: S1: Construct a digital twin model of new energy equipment components, set up virtual sensor nodes that are synchronized with the physical components in real time in the digital twin model, collect multi-source monitoring data of the physical components, and label the multi-source monitoring data with operating condition labels according to the current operating conditions of the equipment. Specifically, the method for establishing a digital twin model is as follows: obtain the three-dimensional geometric design drawings and material property parameter tables of the target components of new energy equipment, and establish a geometric mapping model; assign the parameters in the material property parameter tables to the corresponding regions of the geometric mapping model to construct an initial digital twin model with material properties; embed a stress calculation module based on finite element discretization and a temperature field distribution module based on the heat transfer equation into the initial digital twin model to form a digital twin model capable of physical field simulation.

[0019] It should be noted that the geometric mapping model includes the component's external outline, internal structure, and connection interfaces; the parameters in the material property parameter table include elastic modulus, coefficient of thermal expansion, fatigue limit, and density parameters.

[0020] Preferably, a virtual component degradation simulation module is set in the digital twin model, wherein the virtual component degradation simulation module includes a performance degradation calculation engine, a working condition sequence parser, and a failure diagnosticator; and a performance parameter decay model describing the evolution law of component performance parameters with running time and cumulative damage is established in the performance degradation calculation engine.

[0021] Furthermore, based on the spatial coordinates of the sensors already deployed on the physical components, virtual sensor nodes are established at corresponding positions in the geometric mapping model of the digital twin model; a bidirectional data channel is established between the physical components and the digital twin model through the Industrial Internet communication protocol, and the data receiving port of the virtual sensor node is configured so that the virtual sensor node and the corresponding physical sensor form a data mapping relationship with time-stamp synchronization.

[0022] It should be noted that the sensors include temperature sensors, vibration sensors, strain sensors, and current sensors; the deviation between the coordinate position of each virtual sensor node and the corresponding physical sensor position is less than five-thousandths of the component feature size.

[0023] Preferably, a unified timestamp server is set up to timestamp the raw data collected by all physical sensors, and the timestamped raw data is sent to the data receiving module of the digital twin model through the data transmission interface; the data receiving module of the digital twin model distributes the received raw data to the corresponding virtual sensor nodes according to the data receiving channel identifier; the virtual sensor nodes store the received data values ​​in the time order of the timestamps, forming a virtual measurement data stream that is synchronized with the physical components in real time.

[0024] Furthermore, temperature monitoring data, vibration acceleration monitoring data, strain monitoring data, and current monitoring data are extracted from virtual sensor nodes and aligned along the time axis according to timestamps to form a multi-source monitoring data matrix. Based on the multi-source monitoring data matrix, a subset of data within a preset time window is extracted, and the mean and variance of the current monitoring data, the rate of change of the temperature monitoring data, and the peak frequency of the vibration acceleration monitoring data in the subset of data are calculated to generate an operating condition feature vector.

[0025] Specifically, a benchmark library of operating conditions is pre-established, including rated operating conditions, light-load operating conditions, heavy-load operating conditions, start-stop operating conditions, and variable-load operating conditions. The Euclidean distance between the operating condition feature vector and each benchmark feature vector in the benchmark library is calculated. The benchmark feature vector with the smallest Euclidean distance is selected as the primary candidate operating condition, and the benchmark feature vector with the second smallest Euclidean distance is selected as the secondary candidate operating condition. The difference in similarity scores between the primary candidate operating condition and the secondary candidate operating condition is calculated as the confidence index for operating condition identification.

[0026] Preferably, the specific formula for the confidence index of working condition identification is as follows: ; in, Confidence indicators for identifying operating conditions The similarity score is calculated between the feature vector of the operating condition and the baseline feature vector of the main candidate operating condition. The similarity score is calculated between the feature vector of the operating condition and the baseline feature vector of the next candidate operating condition.

[0027] Furthermore, when the confidence index is greater than a preset first threshold, it is determined whether each feature component of the operating condition feature vector falls within the allowable fluctuation range corresponding to the main candidate operating condition; if all fall within the allowable fluctuation range, the operating condition type corresponding to the main candidate operating condition is taken as the current operating condition identification result and labeled with an operating condition tag; if any feature component exceeds the allowable fluctuation range, it is marked as a boundary abnormal operating condition; when the confidence index is between the preset first threshold and the preset second threshold, the operating condition identification result of the previous time window is extracted, and the operating condition transition probability matrix is ​​queried; if the transition probability between the main candidate operating condition and the previous operating condition is greater than the preset transition threshold... If the probability of transition between the primary candidate condition and the previous condition is less than the preset transition threshold, it is marked as a mixed condition. If the confidence index is less than the preset second threshold, the condition identification sequence of the most recent consecutive time windows is extracted, and the time series fluctuation index is calculated. If the time series fluctuation index is less than the preset stability threshold, the primary candidate condition is accepted as the identification result. If the time series fluctuation index is greater than the preset stability threshold, it is marked as an oscillation abnormal condition. If the current variance characteristic exceeds the preset multiple of the allowable range of all benchmark conditions or the absolute value of the temperature change rate exceeds the preset rapid change threshold, it is directly determined as a transient impact condition.

[0028] Furthermore, the data marked as abnormal operating conditions are classified according to the degree of abnormality. When the deviation of the feature components of the boundary abnormal operating condition is less than the preset mild threshold, it is judged as mildly abnormal, the abnormal feature vector is recorded and the operating condition label of the previous moment is maintained; when the oscillation abnormal operating condition or the deviation of the feature components exceeds the preset moderate threshold, it is judged as moderately abnormal and manual review is triggered; when the transient impact operating condition or the deviation of the feature components exceeds the preset severe threshold, it is judged as severely abnormal and equipment safety warning is triggered; the final determined operating condition identification results, together with the confidence index and the abnormality level, are labeled to the multi-source monitoring data of the corresponding time period to form a monitoring dataset with operating condition labels and stored in the operating condition association database.

[0029] It should be noted that the operating condition characteristic benchmark library stores the benchmark feature vectors, allowable fluctuation ranges, and operating condition transition probability matrices corresponding to each operating condition type; the preset first threshold is determined based on the distribution dispersion of each benchmark feature vector in the operating condition characteristic benchmark library; the preset second threshold is determined based on the confidence distribution characteristics of historical operating condition identification results. The preset transition threshold is determined based on the allowable operating condition switching frequency in the equipment operating procedures; the preset stability threshold is determined based on the time-series stationarity requirements of the operating condition feature vectors under normal operating conditions. The preset mild threshold is determined based on the tolerance for the expansion of the allowable fluctuation range of the benchmark operating condition; the preset moderate threshold is determined based on the equipment performance degradation warning limit; the preset severe threshold is determined based on the equipment safe operation protection limit. The preset multiple threshold is determined based on the critical conditions of the impact of transient processes on equipment lifespan; the preset rapid change threshold is determined based on the material's thermal shock resistance capacity.

[0030] Preferably, a working condition label field is added to the data rows in the multi-source monitoring data matrix corresponding to the time period of the operating condition feature vector. The current working condition identification result is filled in the working condition label field to establish the association between each row of data in the multi-source monitoring data matrix and the corresponding working condition type. The multi-source monitoring data matrix with completed working condition labeling is stored in the working condition data storage area. The working condition data storage area is divided into multiple data partitions according to the working condition type. Each data partition stores the monitoring data with the same working condition label.

[0031] Specifically, the system continuously monitors the operating condition labels corresponding to two adjacent preset time windows. When the operating condition label of the current time window is detected to be different from that of the previous time window, the system records the timestamp of the operating condition switch. It extracts the jump amplitude of strain monitoring data and temperature monitoring data in the multi-source monitoring data matrix within each time window before and after the operating condition switch. The system combines the timestamp of the operating condition switch, the operating condition label before the switch, the operating condition label after the switch, and the jump amplitude into an operating condition switch record. The operating condition switch record is then passed to the subsequent operating condition switch stress factor calculation module.

[0032] S2: Based on multi-source monitoring data, extract the aging characteristic parameters of different working conditions, calculate the change gradient and working condition switching stress factor, and store them in the working condition associated aging characteristic library. Furthermore, monitoring data with operating condition labels stored in each data partition are read from the sub-operating condition data storage area, and the multi-source monitoring data matrix is ​​divided into rated operating condition dataset, light load operating condition dataset, heavy load operating condition dataset, start-stop operating condition dataset, and variable load operating condition dataset according to the operating condition labels.

[0033] Preferably, each working condition dataset is arranged in chronological order, and the number of data rows and time span of each working condition dataset are counted. The working condition datasets with a time span greater than the preset minimum duration are selected as valid sub-working condition datasets. Temperature monitoring data, vibration acceleration monitoring data, strain monitoring data, and current monitoring data are extracted from the valid sub-working condition datasets, and the monotonicity coefficient and fluctuation dispersion of each type of monitoring data in the time dimension are calculated. The monitoring data with a monotonicity coefficient greater than the preset monotonicity threshold are selected as aging trend sensitive data, and the monitoring data with fluctuation dispersion greater than the preset dispersion threshold are selected as aging fluctuation sensitive data. The aging trend sensitive data and the aging fluctuation sensitive data are merged into an aging sensitive monitoring dataset.

[0034] Furthermore, for each valid sub-condition dataset, time series of temperature-related aging-sensitive data are extracted from the corresponding aging-sensitive monitoring dataset. Linear fitting is performed on the time series of temperature-related aging-sensitive data to obtain the temperature degradation slope parameter and temperature reference offset parameter. Time series of vibration-related aging-sensitive data are extracted, and spectral analysis is performed on the time series of vibration-related aging-sensitive data to obtain the dominant frequency drift parameter and high-frequency energy proportion parameter. Time series of strain-related aging-sensitive data are extracted, and the cumulative strain variable parameter and strain cycle count parameter of the time series of strain-related aging-sensitive data are calculated. Time series of current-related aging-sensitive data are extracted, and the average current drift parameter and current fluctuation amplitude parameter of the time series of current-related aging-sensitive data are calculated. All the above parameters are combined into a working condition aging characteristic parameter vector.

[0035] Specifically, the effective working condition dataset is divided into multiple time-segment subsets according to time sequence. The difference between each feature parameter in the working condition aging feature parameter vector corresponding to adjacent time-segment subsets is calculated to obtain the gradient of change within the working condition. Statistical analysis is performed on the gradient of change within the working condition, and the mean and standard deviation of the gradient of change within the working condition are calculated, which are used as the average degradation rate parameter and degradation fluctuation parameter of the component aging under this working condition, respectively. Based on the working condition switching record, the working condition label before switching and the working condition label after switching are extracted. The working condition aging feature parameter vector of the last data window corresponding to the working condition label before switching and the first working condition label after switching are read from the working condition data storage area, respectively. The aging characteristic parameter vectors of each data window are denoted as the feature vector before switching and the feature vector after switching. The difference vector between the feature vector after switching and the feature vector before switching is calculated, and the Euclidean norm of this difference vector is obtained. The Euclidean norm is then divided by the switching time interval to obtain the instantaneous gradient of the switching condition. Based on the average degradation rate parameter, the baseline degradation rate is extracted, and the ratio of the instantaneous gradient of the switching condition to the baseline degradation rate is calculated. This ratio is then used as the stress amplification factor of the switching condition. The stress amplification factor of the switching condition is weighted and summed with the jump amplitude to obtain the stress factor of the switching condition, which characterizes the degree of acceleration of component aging caused by the switching condition.

[0036] Preferably, the specific formula for the stress factor during operating condition switching is as follows: ; in, For stress factor switching of operating conditions, The gradient weight coefficients are for instantaneous transitions between operating conditions. The jump amplitude weighting coefficient, This is the difference vector between the feature vector after the switch and the feature vector before the switch. Let be the Euclidean norm of the difference vector. This is the time interval for switching operating conditions. The average degradation rate parameter corresponding to the previous operating condition label is used for switching. To monitor the number of parameter types, and These are the measured values ​​of the k-th type of monitoring parameter before and after the operating condition switch, respectively.

[0037] It should be noted that the gradient weight coefficient during the instantaneous switching of operating conditions... and jump amplitude weighting coefficient The preset working condition switching coefficient table is consulted based on the working condition combination type of the working condition labels before and after the switch, and the conditions are satisfied. .

[0038] Furthermore, a data table structure for the working condition-related aging feature library is created. The working condition aging feature parameter vectors corresponding to each working condition type are stored in the feature parameter field, the corresponding average degradation rate parameters and degradation fluctuation parameters are stored in the gradient parameter field, and the working condition switching stress factor is stored in the stress factor field according to the working condition combination index of the working condition label before switching and the working condition label after switching. This establishes an associated storage structure between working condition type and aging feature parameters, change gradient, and switching stress factor.

[0039] It should be noted that the data table structure includes a working condition type field, a characteristic parameter field, a gradient parameter field, and a stress factor field.

[0040] Furthermore, a feature library update trigger condition is set. When the accumulated data volume of a certain working condition type in the working condition data storage area increases beyond a preset incremental threshold, the recalculation of the working condition aging feature parameter vector corresponding to that working condition type is triggered. The new working condition aging feature parameter vector calculated using the newly added data is exponentially smoothed and fused with the historical working condition aging feature parameter vector stored in the working condition-related aging feature library. The fused working condition aging feature parameter vector is then updated to the corresponding feature parameter field in the working condition-related aging feature library. The corresponding gradient parameter field and stress factor field are updated synchronously to maintain the consistency between the aging feature information stored in the working condition-related aging feature library and the actual operating state of the component. The updated working condition-related aging feature library is then passed to the subsequent cross-working condition aging accumulation model.

[0041] S3: The cross-condition aging accumulation model is constructed by inputting the aging characteristic parameters of the working condition and the stress factor of the working condition switching, and outputs the aging accumulation index that characterizes the current aging degree of the component. Specifically, the aging characteristic parameters under operating conditions are divided into steady-state aging characteristic sequences, while the stress factors for switching operating conditions are divided into dynamic stress characteristic sequences. The cross-operating-condition aging accumulation model includes a steady-state aging accumulation branch and a dynamic stress accumulation branch. The steady-state aging accumulation branch adopts a long short-term memory neural network structure, while the dynamic stress accumulation branch adopts a convolutional neural network structure. The two branches handle the aging evolution characteristics at different time scales, respectively.

[0042] It should be noted that the steady-state aging characteristic sequence reflects the gradual degradation process of the component under various working conditions; the dynamic stress characteristic sequence reflects the impact damage at the moment of working condition switching.

[0043] Furthermore, the steady-state aging feature sequence is input into the steady-state aging accumulation branch, and the temporal dependency of aging features under each working condition is extracted through the forget gate and input gate mechanism of the long short-term memory neural network, outputting the steady-state accumulated aging vector; the dynamic stress feature sequence is input into the dynamic stress accumulation branch, and the local stress pattern of the working condition switching event is extracted by sliding the convolution kernel of the convolutional neural network in the time dimension, outputting the dynamic stress accumulation vector; an adaptive weight allocation mechanism is designed to calculate the dynamic weight coefficient and steady-state weight coefficient according to the working condition switching frequency within the historical time window.

[0044] It should be noted that the components of the steady-state cumulative aging vector correspond to the aging contribution under different working conditions; the dynamic stress cumulative vector reflects the cumulative effect of working condition switching frequency and intensity on component fatigue damage; the dynamic weight coefficient is positively correlated with the working condition switching frequency, the steady-state weight coefficient is negatively correlated with the working condition switching frequency, and the sum of the dynamic weight coefficient and the steady-state weight coefficient is normalized to a unit value.

[0045] Furthermore, the steady-state cumulative aging vector is multiplied by the steady-state weight coefficients to obtain the weighted steady-state aging component. Simultaneously, the dynamic stress accumulation vector is multiplied by the dynamic weight coefficients to obtain the weighted dynamic stress component. The weighted steady-state aging component and the weighted dynamic stress component are then concatenated to obtain the fused aging feature vector. This fused aging feature vector is input into the fully connected output layer of the cross-condition aging accumulation model. The sigmoid activation function maps the fused aging feature vector to a preset numerical range, outputting the aging accumulation index. The aging accumulation index is then associated and stored with its corresponding timestamp and current condition label to establish a time-series database of the aging accumulation index.

[0046] The preferred formula for the aging accumulation index is as follows: ; in, As an indicator of aging accumulation, It is the Sigmoid activation function. The weight vector of the fully connected output layer. This is the transpose of the fully connected layer weight vector. For the bias term of the fully connected output layer, For the weighted steady-state aging component, For weighted dynamic stress components, This is the vector concatenation operator. These are the steady-state weighting coefficients. These are dynamic weighting coefficients.

[0047] It should be noted that the closer the value of the aging accumulation index is to the upper limit of the range, the more severe the current aging of the component; the aging accumulation index time series database supports subsequent steps to trace the aging evolution trend of the component.

[0048] S4: Using the aging accumulation index as the initial state of the simulation, the decay trajectory of the component performance parameters under the preset future working condition sequence is simulated in the digital twin model. When the performance parameter reaches the preset failure threshold for the first time, the corresponding running time is extracted, and the difference between the running time and the current actual running time is used as the remaining life of the component aging prediction. Specifically, the virtual component degradation simulation module is called from the digital twin model, the initial performance parameter table of the component is loaded, and the performance parameter values ​​in the initial performance parameter table are set as the initial performance parameter values ​​of the virtual component degradation simulation module; the average degradation rate parameter and degradation fluctuation parameter corresponding to each working condition type are read from the working condition associated aging feature library, a correlation lookup table between working condition type and performance degradation rate is established, and the correlation lookup table is loaded into the working condition-performance mapping module of the virtual component degradation simulation module.

[0049] It should be noted that the initial performance parameter table includes the rated output power, operating efficiency baseline value, vibration intensity baseline value, and temperature rise baseline value of the component at the time of manufacture.

[0050] Furthermore, the aging accumulation index is converted into a performance degradation correction coefficient and applied to the initial value of the performance parameter to obtain a corrected performance parameter value that takes into account the current degree of aging. The corrected performance parameter value is set as the simulation starting performance state for the virtual component degradation simulation module to perform future working condition simulation calculations, and the aging accumulation index is set as the simulation starting damage state for the virtual component degradation simulation module, thus completing the dual configuration of the simulation initial state.

[0051] It should be noted that the performance degradation correction coefficient is calculated based on the mapping relationship between the aging accumulation index and the fatigue degradation curve of the component material; the corrected performance parameter values ​​include the corrected output power, the corrected operating efficiency, the corrected vibration intensity, and the corrected temperature rise.

[0052] Furthermore, historical operating records are extracted from the sub-operating condition data storage area, and the frequency of occurrence, average duration, and operating condition transition probability matrix of each operating condition type within the historical time period are statistically analyzed. Based on the operating condition transition probability matrix, a Markov operating condition prediction model is constructed. The current operating condition type and the preset future simulation duration are input to generate a preset future operating condition sequence that includes the operating condition type time series and the duration series of each operating condition.

[0053] It should be noted that the working condition transition probability matrix records the probability distribution of switching from any working condition type to other working condition types; the Markov working condition prediction model is constructed based on the statistical analysis of the transition probabilities between various working conditions according to historical working condition switching records. By statistically analyzing the conversion frequency and conversion patterns between different working condition types, a working condition state transition probability matrix is ​​established to predict future working condition evolution trends.

[0054] In an optional embodiment, the time schedules of planned maintenance conditions, seasonal load adjustment conditions, and special operating mode conditions are read from the operation and maintenance plan database of new energy equipment. The condition information in the time schedule is inserted into the corresponding time nodes of the preset future condition sequence to obtain a corrected future condition sequence that integrates historical statistical patterns and planned operation arrangements. The corrected future condition sequence is then passed to the condition sequence parser of the virtual component degradation simulation module. In the virtual component degradation simulation module, a multi-physics field coupled performance parameter attenuation model is established, and output power attenuation equation, operating efficiency attenuation equation, vibration intensity growth equation, and temperature rise growth equation are defined. Preferably, the performance parameter decay model describes the evolution of component performance parameters with operating time and cumulative damage; the output power decay equation expresses the current output power as a function of the initial output power, cumulative operating time, current damage state, and operating load coefficient; the operating efficiency decay equation considers the combined effects of increased friction loss, decreased heat transfer efficiency, and increased electrical loss on efficiency; the vibration intensity growth equation relates to the vibration amplitude changes caused by component fatigue crack propagation and increased gaps; the temperature rise growth equation relates to the temperature rise caused by deterioration of heat dissipation performance and increased internal losses. The above decay equations and growth equations are integrated into the calculation kernel of the performance parameter decay model.

[0055] Specifically, the corrected future operating condition sequence is divided into multiple simulation time steps, and the average degradation rate parameter is multiplied by the duration of the simulation time step to obtain the damage increment generated in that time step; the damage increment is accumulated to the cumulative damage state of the previous time step to obtain the updated cumulative damage state of the current time step.

[0056] It should be noted that each simulation time step corresponds to a working condition type and duration, and iterative calculations are performed starting from the first simulation time step.

[0057] Furthermore, the load parameters of the updated cumulative damage state and the current operating condition type are input into the performance parameter decay model to calculate the predicted value of the component performance parameters at the end of the current time step. The predicted value of the component performance parameters is stored in the simulation trajectory database as the input performance state for the next simulation time step. The calculation is repeated for all simulation time steps to generate the component performance parameter decay trajectory for correcting future operating condition sequences.

[0058] Furthermore, a preset failure threshold is set for each performance parameter in the failure diagnostic unit. During the simulation iterative calculation process, after each simulation time step is completed, the predicted value of the component performance parameter at the current time step is compared with the corresponding preset failure threshold to determine whether any performance parameter exceeds its corresponding preset failure threshold.

[0059] It should be noted that the preset failure thresholds include the minimum allowable output power threshold, the minimum allowable operating efficiency threshold, the maximum allowable vibration intensity threshold, and the maximum allowable temperature rise threshold; the preset failure thresholds are determined according to the component technical specifications and safe operation requirements.

[0060] Specifically, when the output power in the predicted value of the component performance parameters is lower than the minimum allowable output power threshold, or the operating efficiency is lower than the minimum allowable operating efficiency threshold, or the vibration intensity exceeds the maximum allowable vibration intensity threshold, or the temperature rise exceeds the maximum allowable temperature rise threshold, the component is determined to have reached a failure state at that simulation time step, the simulation iteration calculation is terminated, the cumulative duration from the simulation start time to the current failure time is extracted, and the cumulative duration is recorded as the simulation runtime.

[0061] Furthermore, the cumulative operating time of the components from their first use to the current moment is read from the operation monitoring system of the new energy equipment, and the cumulative operating time is recorded as the current actual running time.

[0062] Furthermore, the remaining lifespan of the component aging prediction, the type of performance parameter that first reaches failure, the corresponding preset failure threshold, and the component performance parameter decay trajectory are combined into a remaining lifespan prediction report, which is output to the operation and maintenance management platform of new energy equipment. This allows operation and maintenance personnel to formulate component replacement plans and maintenance strategies. At the same time, the remaining lifespan prediction report is stored in the prediction history database of the digital twin model to establish a comparison and verification archive between the prediction results and the actual lifespan.

[0063] It should be noted that the difference represents the expected operational time of a component from its current aging state to its failure state.

[0064] Specifically, multiple operating condition fluctuation scenarios are generated based on the corrected future operating condition sequence, and the simulation iterative calculation and failure determination process is repeatedly executed on the operating condition fluctuation scenarios to obtain the remaining lifetime of the optimistic scenario, the remaining lifetime of the baseline scenario, and the remaining lifetime of the pessimistic scenario, respectively. The mean and standard deviation of the remaining lifetime of the three scenarios are calculated. The mean is used as the comprehensive predicted remaining lifetime, and the standard deviation is used as the prediction uncertainty, which are added to the remaining lifetime prediction report.

[0065] It should be noted that the operating condition fluctuation scenario is obtained by adjusting and correcting the duration and frequency of each operating condition type in the future operating condition sequence, including optimistic scenario, baseline scenario and pessimistic scenario.

[0066] In summary, this invention, by constructing a digital twin model with virtual sensor nodes and labeling it with operating condition tags, achieves real-time mirror mapping of the physical component's operating status and accurate correlation of operating conditions with multi-source monitoring data, solving the problem that traditional methods cannot distinguish the differences in the aging effects of different operating conditions. By extracting characteristic parameters of operating condition aging and calculating the stress factor during operating condition switching, it achieves the separate quantitative characterization of the steady-state aging process and the transient impact during operating condition switching, revealing the mechanism of the accelerated aging effect of stress mutation during operating condition transition, thus overcoming the shortcomings of existing technologies that only focus on cumulative damage under a single operating condition while ignoring the impact of stress during operating condition switching. Furthermore, by fusing steady-state aging with a cross-operating-condition aging accumulation model, this invention addresses the issue of traditional methods failing to distinguish the differences in the aging effects of different operating conditions. By incorporating cumulative and dynamic stress cumulative branches and designing an adaptive weighting mechanism, we have achieved collaborative modeling and adaptive weighted fusion of steady-state aging and dynamic stress. This enables the aging accumulation index to accurately reflect the true aging degree under complex operating condition sequences, overcoming the limitation of traditional linear cumulative models in handling operating condition coupling effects. By using the aging accumulation index as the initial state for simulation in a digital twin model to simulate the performance degradation trajectory under future operating condition sequences, we have achieved forward-looking remaining life prediction based on the current aging state. Compared with traditional methods based on historical data extrapolation, this method can consider the impact of future actual operating conditions on the aging process, significantly improving prediction accuracy and engineering practicality.

[0067] Following the teachings of the above embodiments, refer to Figure 2 Other aspects disclosed in the embodiments of the present invention also propose a new energy equipment component aging prediction system based on digital twin and multi-source data fusion, comprising: The digital twin model and data management module is used to construct digital twin models of new energy equipment components. Virtual sensor nodes that are synchronized with physical components in real time are set in the digital twin model to collect multi-source monitoring data of physical components and label the multi-source monitoring data with operating condition tags according to the current operating conditions of the equipment. The feature and stress management module extracts aging feature parameters under different working conditions based on the multi-source monitoring data, calculates the change gradient and working condition switching stress factor, and stores them in the working condition associated aging feature library. The health status assessment module is used to input the working condition aging characteristic parameters and the working condition switching stress factor into the constructed cross-working condition aging accumulation model, and output the aging accumulation index characterizing the current aging degree of the component. The life prediction simulation module is used to take the aging accumulation index as the initial state of the simulation, simulate the decay trajectory of component performance parameters under a preset future working condition sequence in the digital twin model, and extract the corresponding running time when the performance parameter first reaches the preset failure threshold, and use the difference between the running time and the current actual running time as the remaining life predicted by component aging.

[0068] This embodiment also provides a computer device applicable to the aging prediction method for new energy equipment components based on digital twin and multi-source data fusion, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the aging prediction method for new energy equipment components based on digital twin and multi-source data fusion as proposed in the above embodiment.

[0069] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0070] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for predicting the aging of new energy equipment components based on digital twins and multi-source data fusion as proposed in the above embodiments.

[0071] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0072] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion, characterized in that: include, A digital twin model of a new energy equipment component is constructed. Virtual sensor nodes that are synchronized with the physical component in real time are set in the digital twin model to collect multi-source monitoring data of the physical component. The multi-source monitoring data is labeled with operating condition tags according to the current operating condition of the equipment. Based on the multi-source monitoring data, the aging characteristic parameters under different working conditions are extracted, the change gradient and the stress factor for working condition switching are calculated, and stored in the working condition-related aging characteristic library. The cross-condition aging accumulation model is constructed by inputting the aging characteristic parameters of the operating conditions and the stress factor of the operating condition switching, and outputs the aging accumulation index that characterizes the current aging degree of the component. Using the aging accumulation index as the initial state of the simulation, the decay trajectory of the component performance parameters under a preset future operating condition sequence is simulated in the digital twin model. When the performance parameter first reaches the preset failure threshold, the corresponding running time is extracted, and the difference between the running time and the current actual running time is used as the predicted remaining life of the component aging.

2. The method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion as described in claim 1, characterized in that: The multi-source monitoring data is labeled with operating condition tags based on the current operating conditions of the equipment, including: Based on the multi-source monitoring data matrix, a subset of data within a preset time window is extracted, and the mean and variance of the current monitoring data, the rate of change of the temperature monitoring data, and the peak frequency of the vibration acceleration monitoring data in the subset of data are calculated to generate an operating condition feature vector. A baseline library of operating condition characteristics, including rated operating condition, light load operating condition, heavy load operating condition, start-stop operating condition, and variable load operating condition, is established in advance. Calculate the Euclidean distance between the operating condition feature vector and each reference feature vector in the operating condition feature benchmark library, select the reference feature vector with the smallest Euclidean distance as the primary candidate operating condition, and the reference feature vector with the second smallest Euclidean distance as the secondary candidate operating condition. The difference in similarity scores between the primary candidate working condition and the secondary candidate working condition is calculated as a working condition identification confidence index. When the confidence index is greater than the preset first threshold, it is determined whether each feature component of the operating condition feature vector falls within the allowable fluctuation range corresponding to the main candidate operating condition; if all fall within the allowable fluctuation range, the operating condition type corresponding to the main candidate operating condition is taken as the current operating condition identification result and the operating condition label is marked; if there is a feature component that exceeds the allowable fluctuation range, it is marked as a boundary abnormal operating condition. When the confidence index is between the preset first threshold and the preset second threshold, the working condition identification result of the previous time window is extracted, and the working condition transition probability matrix is ​​queried. If the transition probability between the main candidate working condition and the previous working condition is greater than the preset transition threshold, it is marked as a transition working condition type. If the transition probability between the main candidate working condition and the previous working condition is less than the preset transition threshold, it is marked as a mixed working condition type. When the confidence index is less than the preset second threshold, the working condition identification sequence of the most recent consecutive time windows is extracted and the time series fluctuation index is calculated; if the time series fluctuation index is less than the preset stability threshold, the main candidate working condition is accepted as the identification result; if the time series fluctuation index is greater than the preset stability threshold, it is marked as an oscillation abnormal working condition. When the current variance characteristic exceeds a preset multiple of the allowable range of all reference operating conditions or the absolute value of the temperature change rate exceeds the preset rapid change threshold, it is directly determined as a transient impact condition.

3. The method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion as described in claim 1, characterized in that: The method for obtaining the aging accumulation index is as follows: The aging characteristic parameters under working conditions are divided into a steady-state aging characteristic sequence, and the stress factor of working condition switching is divided into a dynamic stress characteristic sequence. The steady-state aging feature sequence is input into the steady-state aging accumulation branch, and the temporal dependency of aging features under each working condition is extracted through the forget gate and input gate mechanism of the long short-term memory neural network, and the steady-state accumulation aging vector is output. The dynamic stress feature sequence is input into the dynamic stress accumulation branch, and the local stress pattern of the working condition switching event is extracted by sliding the convolution kernel of the convolutional neural network in the time dimension, and the dynamic stress accumulation vector is output. Design an adaptive weight allocation mechanism to calculate dynamic weight coefficients and steady-state weight coefficients based on the operating condition switching frequency within the historical time window. Multiply the steady-state cumulative aging vector by the steady-state weighting coefficient to obtain the weighted steady-state aging component; at the same time, multiply the dynamic stress cumulative vector by the dynamic weighting coefficient to obtain the weighted dynamic stress component. A vector concatenation operation is performed on the weighted steady-state aging component and the weighted dynamic stress component to obtain a fused aging feature vector. The fused aging feature vector is input into the fully connected output layer of the cross-condition aging accumulation model, and the fused aging feature vector is mapped to a preset numerical range through the sigmoid activation function to output the aging accumulation index.

4. The method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion as described in claim 3, characterized in that: The cross-condition aging accumulation model includes a steady-state aging accumulation branch and a dynamic stress accumulation branch; the steady-state aging accumulation branch adopts a long short-term memory neural network structure; and the dynamic stress accumulation branch adopts a convolutional neural network structure.

5. The method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion as described in claim 3, characterized in that: The method for obtaining the stress factor for the working condition switching is as follows: The effective working condition dataset is divided into multiple time period subsets in chronological order. The difference between each feature parameter in the working condition aging feature parameter vector corresponding to adjacent time period subsets is calculated to obtain the change gradient within the working condition. Statistical analysis is performed on the gradient variation within the operating condition to calculate the mean and standard deviation of the gradient variation within the operating condition, which are then used as the average degradation rate parameter and degradation fluctuation parameter of the component aging under this operating condition, respectively. Based on the working condition switching record, the working condition label before switching and the working condition label after switching are extracted. The working condition aging feature parameter vector of the last data window and the working condition aging feature parameter vector of the first data window corresponding to the working condition label before switching are read from the working condition data storage area respectively, and are denoted as the feature vector before switching and the feature vector after switching. Calculate the difference vector between the feature vector after the switch and the feature vector before the switch, find the Euclidean norm of this difference vector, and divide the Euclidean norm by the working condition switching time interval to obtain the instantaneous gradient of the working condition switching. Based on the average degradation rate parameter, the baseline degradation rate is extracted, the ratio of the instantaneous gradient of the working condition switching to the baseline degradation rate is calculated, and this ratio is used as the stress amplification factor for the working condition switching. The stress amplification factor of the operating condition switching and the jump amplitude are weighted and summed to obtain the operating condition switching stress factor, which characterizes the degree of acceleration of component aging caused by the operating condition switching.

6. The method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion as described in claim 5, characterized in that: The specific formula for the stress factor for switching operating conditions is as follows: ； in, For stress factor switching of operating conditions, The gradient weight coefficients are for instantaneous transitions between operating conditions. The jump amplitude weighting coefficient, This is the difference vector between the feature vector after the switch and the feature vector before the switch. Let be the Euclidean norm of the difference vector. This is the time interval for switching operating conditions. The average degradation rate parameter corresponding to the previous operating condition label is used for switching. To monitor the number of parameter types, and These are the measured values ​​of the k-th type of monitoring parameter before and after the operating condition switch, respectively.

7. The method for predicting the aging of new energy equipment components based on digital twin and multi-source data fusion as described in claim 1, characterized in that: The method for establishing the digital twin model is as follows: Obtain the three-dimensional geometric design drawings and material property parameter tables of the target components of new energy equipment, and establish a geometric mapping model; Assign the parameters from the material property parameter table to the corresponding regions of the geometric mapping model to construct an initial digital twin model with material properties; A stress calculation module based on finite element discretization and a temperature field distribution module based on heat transfer equations are embedded in the initial digital twin model to form a digital twin model capable of physical field simulation. A virtual component degradation simulation module is set in the digital twin model, wherein the virtual component degradation simulation module includes a performance degradation calculation engine, a working condition sequence parser, and a failure diagnosticator; In the performance degradation calculation engine, a performance parameter decay model is established to describe the evolution of component performance parameters with running time and cumulative damage.

8. A new energy equipment component aging prediction system based on digital twin and multi-source data fusion, based on the new energy equipment component aging prediction method based on digital twin and multi-source data fusion as described in any one of claims 1 to 7, characterized in that: include, The digital twin model and data management module is used to construct digital twin models of new energy equipment components. Virtual sensor nodes that are synchronized with physical components in real time are set in the digital twin model to collect multi-source monitoring data of physical components and label the multi-source monitoring data with operating condition tags according to the current operating conditions of the equipment. The feature and stress management module extracts aging feature parameters under different working conditions based on the multi-source monitoring data, calculates the change gradient and working condition switching stress factor, and stores them in the working condition associated aging feature library. The health status assessment module is used to input the working condition aging characteristic parameters and the working condition switching stress factor into the constructed cross-working condition aging accumulation model, and output the aging accumulation index characterizing the current aging degree of the component. The life prediction simulation module is used to take the aging accumulation index as the initial state of the simulation, simulate the decay trajectory of component performance parameters under a preset future working condition sequence in the digital twin model, and extract the corresponding running time when the performance parameter first reaches the preset failure threshold, and use the difference between the running time and the current actual running time as the remaining life predicted by component aging.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the new energy equipment component aging prediction method based on digital twin and multi-source data fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the new energy equipment component aging prediction method based on digital twin and multi-source data fusion as described in any one of claims 1 to 7.

Images