Fault diagnosis method, system and equipment for power equipment and storage medium

Abstract

The invention discloses a fault diagnosis method, system and device for power equipment and a storage medium. The method comprises the steps that all original power signals in the operation process of a driving motor are collected; performing data preprocessing on the plurality of original power signals to obtain time domain feature information, frequency domain feature information and time-frequency feature information corresponding to each normalized power signal; selecting a plurality of excellent power signals from the plurality of normalized power signals through an improved genetic algorithm internal fitness evaluation formula according to the time domain feature information, the frequency domain feature information and the time frequency feature information; determining an optimized feature set according to the plurality of excellent power signals; a plurality of fault classification models are trained based on the optimized feature set; and carrying out fault diagnosis on the power equipment through a dynamic weighted integrated fault judgment algorithm according to the plurality of trained fault classification models. According to the invention, by introducing adaptive parameter adjustment, improved selection and variation strategies, high efficiency and accuracy of fault diagnosis are realized.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a method, system, device and storage medium for fault diagnosis of power equipment. Background Technology

[0002] Disconnect switches play a crucial role in the protection and isolation of power systems; their failure can lead to serious power system accidents. Traditional fault diagnosis methods mainly rely on fixed threshold settings and expert experience. While these methods are effective for simple faults, they have significant limitations when dealing with complex and diverse faults. Therefore, achieving both high efficiency and accuracy in fault diagnosis has become an urgent problem to be solved.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide a method, system, device, and storage medium for fault diagnosis of power equipment, aiming to solve the technical problem of how to achieve high efficiency and accuracy in fault diagnosis.

[0005] To achieve the above objectives, the present invention provides a fault diagnosis method for power equipment, the fault diagnosis method for power equipment comprising:

[0006] Collect all raw power signals during the operation of the drive motor;

[0007] Multiple raw power signals are preprocessed to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information for each normalized power signal.

[0008] Based on the time-domain feature information, the frequency-domain feature information, and the time-frequency feature information, multiple excellent power signals are selected from multiple normalized power signals using an improved genetic algorithm internal fitness evaluation formula.

[0009] Determine the optimal feature set based on multiple excellent power signals;

[0010] Multiple fault classification models are trained based on the optimized feature set;

[0011] Fault diagnosis of power equipment is performed by using a dynamic weighted ensemble fault judgment algorithm based on multiple trained fault classification models.

[0012] Optionally, the step of performing data preprocessing on multiple raw power signals to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal includes:

[0013] Denoising is performed on multiple original power signals to obtain the denoised power signals corresponding to each original power signal.

[0014] Multiple denoised power signals are normalized to obtain multiple normalized power signals;

[0015] Feature extraction is performed on multiple normalized power signals to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal.

[0016] Optionally, the step of selecting multiple excellent power signals from multiple normalized power signals based on the time-domain feature information, the frequency-domain feature information, and the time-frequency feature information using an improved genetic algorithm internal fitness evaluation formula further includes:

[0017] Based on the time-domain feature information, the frequency-domain feature information, and the time-frequency feature information, the individual fitness corresponding to each normalized power signal is calculated using the improved genetic algorithm's fitness evaluation formula.

[0018] Multiple excellent power signals are selected from multiple normalized power signals based on the individual fitness. Optionally, the fitness evaluation formula is:

[0019] F(P i ) = F(features(P i ))

[0020]

[0021] In the formula, F is the fitness function, F(P) i P represents the individual fitness corresponding to the i-th normalized power signal. i For the i-th normalized power signal, features(P) i Let f(peak) be the individual feature value corresponding to the i-th normalized power signal, μ be the mean of the power signal, σ be the standard deviation of the power signal, p be the peak value of the power signal, E be the spectral energy, f(peak) be the main frequency component, and W be the peak value of the power signal. mean W is the mean of the time-frequency characteristics. var denoted as the variance of the time-frequency features, ∈ is a constant, and w1, w2, w3, w4, w5, w6, w7, and w8 are weighting coefficients.

[0022] Optionally, the step of selecting multiple excellent power signals from multiple normalized power signals based on the individual fitness includes:

[0023] Multiple normalized power signals are sorted according to their individual fitness scores from highest to lowest to obtain a fitness ranking result. A preset number of excellent power signals are then selected sequentially based on this fitness ranking result; or

[0024] Select multiple excellent power signals from multiple normalized power signals whose individual fitness is greater than a preset fitness threshold.

[0025] Optionally, the step of determining the optimized feature set based on multiple excellent power signals includes:

[0026] By performing cross-operation on multiple excellent power signals using a cross-function, a set of cross-individuals is obtained, which contains multiple cross-individuals.

[0027] By performing mutation operations on multiple crossover individuals using a mutation function, a set of mutated individuals is obtained;

[0028] Obtain the feature information corresponding to each mutated individual within the set of mutated individuals;

[0029] An optimized feature set is constructed based on the characteristic information corresponding to each mutated individual.

[0030] Optionally, the dynamic weighted integrated fault judgment algorithm is as follows:

[0031]

[0032] Fault=Predict(Model,features)

[0033] In the formula, Let w be the comprehensive evaluation value of the fault state at time t, N be the number of models, and w be the value of the fault state at time t. i (t) represents the weights of the i-th trained fault classification model at time t, S i (t) represents the state estimate of the i-th trained fault classification model at time t, where Fault is the fault state, Predict is the prediction function, and features is the set of feature values.

[0034] Furthermore, to achieve the above objectives, the present invention also proposes a fault diagnosis system for power equipment, the fault diagnosis system for power equipment comprising:

[0035] The acquisition module is used to acquire all raw power signals during the operation of the drive motor;

[0036] The processing module is used to preprocess multiple raw power signals to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal.

[0037] The calculation module is used to select multiple excellent power signals from multiple normalized power signals based on the time domain feature information, the frequency domain feature information, and the time-frequency feature information using an improved genetic algorithm internal fitness evaluation formula;

[0038] The calculation module is also used to determine an optimized feature set based on multiple excellent power signals;

[0039] The training module is used to train multiple fault classification models based on the optimized feature set.

[0040] The diagnostic module is used to diagnose faults in power equipment by dynamically weighting and integrating fault judgment algorithms based on multiple trained fault classification models.

[0041] Furthermore, to achieve the above objectives, the present invention also proposes a fault diagnosis device for power equipment, the device comprising: a memory, a processor, and a fault diagnosis program for power equipment stored in the memory and executable on the processor, the fault diagnosis program for power equipment being configured to implement the steps of the fault diagnosis method for power equipment as described above.

[0042] Furthermore, to achieve the above objectives, the present invention also proposes a storage medium storing a fault diagnosis program for power equipment, wherein when the fault diagnosis program for power equipment is executed by a processor, the program implements the steps of the fault diagnosis method for power equipment as described above.

[0043] This invention first collects all raw power signals during the operation of the drive motor. Then, it preprocesses these raw power signals to obtain time-domain, frequency-domain, and time-frequency characteristic information for each normalized power signal. Based on these characteristics, a modified genetic algorithm with an internal fitness evaluation formula is used to select several superior power signals from the normalized power signals. An optimized feature set is then determined based on these superior power signals. Finally, multiple fault classification models are trained based on the optimized feature set. A dynamic weighted ensemble fault diagnosis algorithm is then used to diagnose faults in power equipment based on these trained fault classification models. Compared to existing technologies that rely on fixed threshold settings and expert experience, which are prone to missed or false fault detections and have poor adaptability to new faults or system changes, making it difficult to reflect the true state of the system in a timely manner, this invention achieves high efficiency and accuracy in fault diagnosis by introducing adaptive parameter adjustment, improved selection, and mutation strategies. This provides crucial protection for the safe operation of power systems, reduces power outages caused by equipment failures, and ultimately improves the operational efficiency and reliability of power systems. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the structure of a fault diagnosis device for power equipment in the hardware operating environment involved in the embodiments of the present invention;

[0045] Figure 2This is a flowchart illustrating the first embodiment of the fault diagnosis method for power equipment of the present invention;

[0046] Figure 3 This is a structural block diagram of the first embodiment of the fault diagnosis system for power equipment of the present invention.

[0047] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0048] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0049] Reference Figure 1 , Figure 1 This is a schematic diagram of the fault diagnosis device structure of the power equipment in the hardware operating environment involved in the embodiments of the present invention.

[0050] like Figure 1 As shown, the fault diagnosis device for this power equipment may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be high-speed random access memory (RAM) or stable non-volatile memory (NVM), such as a disk storage device. Optionally, the memory 1005 may also be a storage system independent of the aforementioned processor 1001.

[0051] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the fault diagnosis equipment for electrical equipment, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0052] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a network communication module, a user interface module, and a fault diagnosis program for power equipment.

[0053] exist Figure 1In the power equipment fault diagnosis device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the power equipment fault diagnosis device of the present invention can be set in the power equipment fault diagnosis device, and the power equipment fault diagnosis device calls the power equipment fault diagnosis program stored in the memory 1005 through the processor 1001 and executes the power equipment fault diagnosis method provided in the embodiment of the present invention.

[0054] This invention provides a fault diagnosis method for power equipment, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the fault diagnosis method for power equipment of the present invention.

[0055] In this embodiment, the fault diagnosis method for power equipment includes the following steps:

[0056] Step S10: Collect all raw power signals during the operation of the drive motor.

[0057] It is easy to understand that the executing entity of this embodiment can be a fault diagnosis system for power equipment with functions such as data processing, network communication and program execution, or other computer equipment with similar functions. This embodiment does not limit it.

[0058] In practical implementation, it is necessary to collect the drive motor power signal, i.e. the original power signal P(t), during the operation of the drive motor through high-precision sensors and monitoring equipment.

[0059] The frequency and time window for data acquisition should be set to capture all raw power signals P(t) during motor operation to ensure the comprehensiveness and accuracy of the data.

[0060] Step S20: Perform data preprocessing on multiple raw power signals to obtain time domain feature information, frequency domain feature information, and time-frequency feature information corresponding to each normalized power signal.

[0061] Furthermore, the data preprocessing of multiple raw power signals to obtain the time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal is performed as follows: denoising is performed on multiple raw power signals to obtain denoised power signals corresponding to each raw power signal; normalization is performed on multiple denoised power signals to obtain multiple normalized power signals; feature extraction is performed on multiple normalized power signals to obtain the time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal.

[0062] It should be understood that wavelet transform is used to denoise the original power signal P(t), eliminating high-frequency noise and interference, and the denoised signal, i.e., the denoised power signal, is expressed as follows:

[0063]

[0064] In the formula, For the denoised power signal, c k Let ψ be the wavelet coefficient, ψ be the wavelet basis function, and a be the wavelet coefficient. k b is the scale parameter k These are the translation parameters.

[0065] The denoised power signal is normalized to obtain the normalized signal P. n (t):

[0066]

[0067] In the formula, P n (t) represents the normalized power signal, where t is the specific time point of the signal in the time domain. This represents the minimum value of the power signal after denoising. This represents the maximum value of the power signal after denoising.

[0068] In this embodiment, it is also necessary to extract feature information from the normalized power signal, including time-domain feature information, frequency-domain feature information, and time-frequency feature information.

[0069] It should also be noted that the time-domain feature information includes the mean, standard deviation, and peak value of the power signal; the frequency-domain feature information includes the spectral energy and main frequency components; and the time-domain feature information includes time-frequency feature values.

[0070] Time-domain feature information:

[0071] Mean μ:

[0072]

[0073] In the formula, μ is the mean of the power signal, and T is the total time length.

[0074] Standard deviation σ:

[0075]

[0076] In the formula, σ is the standard deviation of the power signal.

[0077] Peak value p:

[0078] p = max(P) n (t))

[0079] In the formula, p is the peak value of the power signal.

[0080] Frequency domain characteristic information:

[0081] Spectral energy E:

[0082]

[0083] In the formula, E is the spectral energy, X(f) is the Fourier transform result of the normalized power signal, and f is the frequency.

[0084] Main frequency component f peak :

[0085]

[0086] Time-frequency characteristic information:

[0087] The time-frequency features W(t,f) are obtained through wavelet transform:

[0088]

[0089] In the formula, W(t,f) represents the time-frequency characteristic, and P n (τ) represents the normalized power signal, τ is the time variable used for integration in the time-frequency analysis, and ψ * is the conjugate of the wavelet basis functions, t is the time position parameter, and s is the scale parameter.

[0090] Step S30: Select multiple excellent power signals from multiple normalized power signals based on the time domain feature information, the frequency domain feature information, and the time-frequency feature information using an improved genetic algorithm internal fitness evaluation formula.

[0091] Furthermore, the process of selecting multiple excellent power signals from multiple normalized power signals based on time-domain feature information, frequency-domain feature information, and time-frequency feature information using an improved genetic algorithm fitness evaluation formula is as follows: calculate the individual fitness corresponding to each normalized power signal based on the time-domain feature information, frequency-domain feature information, and time-frequency feature information using an improved genetic algorithm fitness evaluation formula; and select multiple excellent power signals from multiple normalized power signals based on the individual fitness.

[0092] Population = {P1, P2, ..., P} N}

[0093] In the formula, Population represents the initial population, and P i This represents the i-th normalized power signal (i.e., the i-th individual).

[0094] The fitness assessment formula is:

[0095] F(P i ) = F(features(P i ))

[0096]

[0097] In the formula, F is the fitness function, F(P) i P represents the individual fitness corresponding to the i-th normalized power signal. i For the i-th normalized power signal, features(P) i Let f(peak) be the individual feature value corresponding to the i-th normalized power signal, μ be the mean of the power signal, σ be the standard deviation of the power signal, p be the peak value of the power signal, E be the spectral energy, f(peak) be the main frequency component, and W be the peak value of the power signal. mean W is the mean of the time-frequency characteristics. var denoted as the variance of the time-frequency feature, ∈ is a constant (to prevent the denominator from being zero), and w1, w2, w3, w4, w5, w6, w7 and w8 are weighting coefficients.

[0098] It should also be noted that, This ratio reflects the stability and distribution characteristics of the power signal. The mean represents the central tendency of the power signal, while the standard deviation represents the dispersion of the signal. This ratio can effectively distinguish the changes in the signal under normal and fault conditions.

[0099] The logarithmic transformation of log(1+p) can reduce the impact of excessive peak values ​​while preserving the relative differences in peak value information, making it suitable for detecting sudden faults.

[0100] E 2 The square of the spectral energy can amplify energy differences in the frequency domain, which helps in identifying periodic faults. For certain types of faults, the signal energy will increase significantly within a specific frequency range.

[0101] Sigmoid function of the main frequency components: Processing the main frequency component f using the Sigmoid function peak This smooths out the impact of extreme values ​​on fitness, preventing them from interfering with fitness. The main frequency components reflect the most significant frequency information in the signal, which is crucial for detecting frequency anomalies.

[0102] Square of time-frequency characteristics The square of the mean of the time-frequency characteristics can amplify the central trend information of the signal in the time-frequency domain and adapt to dynamically changing fault characteristics.

[0103] The ratio of variance to standard deviation of time-frequency characteristics This ratio reflects the fluctuation of the signal in the time and frequency domain, and when combined with the standard deviation, it further reflects the fluctuation characteristics of the signal.

[0104] Square root of mean and peak This interaction term combines information from the mean and peak values, using square root transformation to mitigate the impact of excessively large values ​​while preserving their relative relationship.

[0105] The logarithmic interaction term between spectral energy and the main frequency components, E·log(1+f peak This interaction term combines information from spectral energy and major frequency components, and further enhances the complexity and discriminative power of the fitness function through logarithmic transformation and product operations.

[0106] It should be understood that the improved fitness function described above has the following advantages: First, the fitness function combines time domain, frequency domain, and time-frequency domain characteristics, comprehensively reflecting the multidimensional characteristics of the signal, adapting to different types of fault characteristics, and improving the comprehensiveness of fault diagnosis; Second, through the combination of various nonlinear transformations and interaction terms, the robustness of the fitness function is improved, reducing the probability of false detection and missed detection; Third, the parameters in the fitness function can be adjusted according to the actual situation, adapting to different application scenarios and fault modes, and have strong adaptive capabilities; Fourth, through the combination of polynomial terms, logarithmic terms, exponential terms, and interaction terms, the fitness function's ability to distinguish different fault modes is enhanced, ensuring accuracy and stability in complex environments.

[0107] It should also be noted that, in order to further enhance the diversity and evolutionary capacity of the population, the following content is introduced into the information entropy-based selection process:

[0108] Dynamically adjust weights: The weights in the information entropy calculation are dynamically adjusted according to the population diversity during the evolutionary process to ensure that appropriate diversity is maintained at different evolutionary stages.

[0109] Hybrid selection strategy: Combining information entropy selection with other selection strategies (such as roulette wheel selection or tournament selection) to improve the robustness of selection through multi-strategy mixing.

[0110] Fitness normalization: Before calculating information entropy, the fitness values ​​are normalized to ensure that the contributions of different features to information entropy are balanced.

[0111] Calculate information entropy:

[0112]

[0113] In the formula, w j p represents the weight of the j-th feature (dynamically adjusted based on population diversity). ij Let be the probability of the i-th individual on the j-th feature.

[0114]

[0115] In the formula, f ij Let be the fitness value of the i-th individual on the j-th feature.

[0116] Normalization is performed (to ensure that different features contribute equally to information entropy and to prevent any single feature from having an excessive influence on the selection process):

[0117]

[0118] Calculate the probability of selection:

[0119]

[0120] Hybrid selection strategies (which ensure individual diversity and prevent the population from getting trapped in local optima; by mixing with other selection strategies (such as roulette wheel selection or tournament selection), they improve the robustness of selection and ensure that the population has good evolutionary capacity):

[0121] P final =αP i +(1-α)P other

[0122] P other It represents the selection probability of other selection strategies (such as roulette wheel selection or tournament selection), and α is the mixing coefficient, which controls the weight of information entropy selection compared to other selection strategies.

[0123] In practice, in the early stages of evolution, different features can be assigned higher weights to increase exploration capabilities; in the later stages of evolution, feature weights can be reduced to increase development capabilities, focusing on features that perform better.

[0124] In this embodiment, information entropy is used to measure the diversity of individuals, ensuring that the selection process can retain diverse individuals and enhance the population's exploratory capabilities; feature weights are dynamically adjusted to adapt to the needs of different evolutionary stages and balance exploration and development capabilities; and multiple selection strategies are combined to improve the robustness of selection and ensure that the population can evolve stably in complex environments.

[0125] Furthermore, the processing method for selecting multiple excellent power signals from multiple normalized power signals based on individual fitness is as follows: sort the multiple normalized power signals according to the rule of descending from the individual fitness to obtain the fitness ranking result, and select a preset number of excellent power signals in sequence according to the fitness ranking result; or select multiple excellent power signals from multiple normalized power signals whose individual fitness is greater than a preset fitness threshold.

[0126] P selected=Select(Population,F)

[0127] In the formula, P selected For excellent power signals, Select is the selection function.

[0128] It should be understood that the preset quantity and preset fitness threshold are user-defined settings and are not limited in this embodiment.

[0129] Step S40: Determine the optimized feature set based on multiple excellent power signals.

[0130] Furthermore, the processing method for determining the optimal feature set based on multiple excellent power signals is as follows: cross-operation is performed on multiple excellent power signals through a cross function to obtain a cross individual set, which contains multiple cross individuals; a mutation operation is performed on multiple cross individuals through a mutation function to obtain a mutated individual set; when the termination condition is met, the feature information corresponding to each mutated individual in the mutated individual set is obtained; and an optimal feature set is constructed based on the feature information corresponding to each mutated individual.

[0131] P new =Crossover(P selected )

[0132] In the formula, P new Let Crossover be the set of crossover individuals, and Crossover be the crossover function.

[0133] P mutated =Mutate(P new )

[0134] In the formula, P mutated Let Mutate be the set of mutated individuals, and Mutate be the mutation function.

[0135] Termination condition: The algorithm terminates based on fitness convergence or by reaching the maximum number of iterations.

[0136] Terminateifmax(F)-min(F)<∈

[0137] or

[0138] Iterations>Max_Iterations

[0139] In the formula, ∈ is the fitness convergence threshold, Iterations is the current iteration number, and Max_Iterations is the maximum iteration number.

[0140] Step S50: Train multiple fault classification models based on the optimized feature set.

[0141] It should be noted that the optimized feature set is the set of feature values ​​optimized by improving the genetic algorithm. Then, multiple fault classification models are trained using the optimized feature set. The fault classification models can be support vector machines (SVM), random forests, neural networks, etc.

[0142] Model=Train(features,labels)

[0143] In the formula, Model is the trained classification model, Train is the training function, features is the set of feature values, and labels is the set of fault labels.

[0144] It should also be understood that cross-validation is used to verify the performance of the trained classification model, ensuring its reliability and accuracy in practical applications. When the accuracy value of the performance is greater than the preset accuracy threshold, the trained classification model can be put into use.

[0145] Accuracy=CrossValidate(Model,features,labels)

[0146] In the formula, Accuracy represents the accuracy of the model, and CrossValidate is the cross-validation function.

[0147] Step S60: Based on the trained multiple fault classification models, perform fault diagnosis of power equipment using a dynamic weighted integrated fault judgment algorithm.

[0148] In the specific implementation, the power data of the drive motor is collected in real time by the sensor, the data is preprocessed and the features are extracted, and the extracted features are input into the trained fault classification model. The dynamic weighted integrated fault judgment algorithm is run in real time to estimate the state of the drive motor and update the weights of the base model, and the fault state of the disconnecting switch is judged in real time.

[0149] The dynamic weighted integrated fault judgment algorithm is as follows:

[0150]

[0151] Fault=Predict(Model,features

[0152] In the formula, Let w be the comprehensive evaluation value of the fault state at time t, N be the number of models, and w be the value of the fault state at time t. i (t) represents the weights of the i-th trained fault classification model at time t, S i (t) represents the state estimate of the i-th trained fault classification model at time t, where Fault is the fault state, Predict is the prediction function, and features is the set of feature values.

[0153] It should also be noted that the feature set is input into multiple trained fault classification models. Each model identifies the corresponding fault type and fault state estimate. Then, a dynamic weighted ensemble learning algorithm calculates the comprehensive fault state evaluation value based on the weights of each model and the fault state estimate. The comprehensive evaluation value is then determined to fall within a specific fault state range. Based on this determination, the final fault type is identified. Fault types include three-phase asynchrony defects, incomplete opening / closing, and jamming defects. The time and characteristic information of the fault are recorded, and a fault diagnosis report is automatically generated, detailing the fault type, characteristics, and recommended maintenance measures.

[0154] In practice, the classification model is updated regularly, new fault data is added, and the accuracy and robustness of the model are continuously improved. The status of the sensors and data acquisition system is also checked regularly to ensure their normal operation and avoid data loss and false alarms.

[0155] In this embodiment, all raw power signals during the operation of the drive motor are first collected. Then, the data of multiple raw power signals are preprocessed to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal. Based on the time-domain feature information, frequency-domain feature information, and time-frequency feature information, multiple excellent power signals are selected from the multiple normalized power signals using an improved genetic algorithm internal fitness evaluation formula. Then, an optimized feature set is determined based on the multiple excellent power signals. Finally, multiple fault classification models are trained based on the optimized feature set, and fault diagnosis of power equipment is performed using a dynamic weighted ensemble fault judgment algorithm based on the trained multiple fault classification models. Compared with the existing technology that relies on fixed threshold settings and expert experience, which is prone to missed or false fault detection, and traditional methods have poor adaptability to new faults or system changes, making it difficult to reflect the true state of the system in a timely manner, this embodiment introduces adaptive parameter adjustment, improved selection, and mutation strategies to achieve high efficiency and accuracy in fault diagnosis, thereby providing an important guarantee for the safe operation of the power system, reducing power outages caused by equipment failures, and thus improving the operating efficiency and reliability of the power system.

[0156] Reference Figure 3 , Figure 3 This is a structural block diagram of the first embodiment of the fault diagnosis system for power equipment of the present invention.

[0157] like Figure 3 As shown, the fault diagnosis system for power equipment proposed in this embodiment of the invention includes:

[0158] The acquisition module 3001 is used to acquire all the raw power signals during the operation of the drive motor.

[0159] Processing module 3002 is used to perform data preprocessing on multiple raw power signals respectively to obtain time domain feature information, frequency domain feature information and time-frequency feature information corresponding to each normalized power signal;

[0160] The calculation module 3003 is used to select multiple excellent power signals from multiple normalized power signals based on the time domain feature information, the frequency domain feature information and the time-frequency feature information through an improved genetic algorithm internal fitness evaluation formula;

[0161] The calculation module 3003 is also used to determine an optimized feature set based on multiple excellent power signals;

[0162] Training module 3004 is used to train multiple fault classification models based on the optimized feature set;

[0163] The diagnostic module 3005 is used to diagnose faults in power equipment using a dynamic weighted ensemble learning algorithm based on multiple trained fault classification models.

[0164] In this embodiment, all raw power signals during the operation of the drive motor are first collected. Then, the data of multiple raw power signals are preprocessed to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal. Based on the time-domain feature information, frequency-domain feature information, and time-frequency feature information, multiple excellent power signals are selected from the multiple normalized power signals using an improved genetic algorithm internal fitness evaluation formula. Then, an optimized feature set is determined based on the multiple excellent power signals. Finally, multiple fault classification models are trained based on the optimized feature set, and fault diagnosis of power equipment is performed using a dynamic weighted ensemble fault judgment algorithm based on the trained multiple fault classification models. Compared with the existing technology that relies on fixed threshold settings and expert experience, which is prone to missed or false fault detection, and traditional methods have poor adaptability to new faults or system changes, making it difficult to reflect the true state of the system in a timely manner, this embodiment introduces adaptive parameter adjustment, improved selection, and mutation strategies to achieve high efficiency and accuracy in fault diagnosis, thereby providing an important guarantee for the safe operation of the power system, reducing power outages caused by equipment failures, and thus improving the operating efficiency and reliability of the power system.

[0165] Other embodiments or specific implementations of the fault diagnosis system for power equipment of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0166] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0167] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0168] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0169] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A fault diagnosis method for power equipment, characterized in that, The fault diagnosis method for the power equipment includes the following steps: Collect all raw power signals during the operation of the drive motor; Multiple raw power signals are preprocessed to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information for each normalized power signal. Based on the time-domain feature information, the frequency-domain feature information, and the time-frequency feature information, multiple excellent power signals are selected from multiple normalized power signals using an improved genetic algorithm internal fitness evaluation formula. Determine the optimal feature set based on multiple excellent power signals; Multiple fault classification models are trained based on the optimized feature set; Fault diagnosis of power equipment is performed by using a dynamic weighted ensemble fault judgment algorithm based on multiple trained fault classification models.

2. The method as described in claim 1, characterized in that, The step of performing data preprocessing on multiple raw power signals to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal includes: Denoising is performed on multiple original power signals to obtain the denoised power signals corresponding to each original power signal. Multiple denoised power signals are normalized to obtain multiple normalized power signals; Feature extraction is performed on multiple normalized power signals to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal.

3. The method as described in claim 1 or 2, characterized in that, The step of selecting multiple excellent power signals from multiple normalized power signals based on the time-domain feature information, the frequency-domain feature information, and the time-frequency feature information using an improved genetic algorithm internal fitness evaluation formula further includes: Based on the time-domain feature information, the frequency-domain feature information, and the time-frequency feature information, the individual fitness corresponding to each normalized power signal is calculated using the improved genetic algorithm's fitness evaluation formula. Based on the individual fitness, multiple excellent power signals are selected from multiple normalized power signals.

4. The method as described in claim 3, characterized in that, The fitness evaluation formula is as follows: F(P i )=F(features(P i )) In the formula, F is the fitness function, F(P) i P represents the individual fitness corresponding to the i-th normalized power signal. i For the i-th normalized power signal, features(P) i Let f(peak) be the individual feature value corresponding to the i-th normalized power signal, μ be the mean of the power signal, σ be the standard deviation of the power signal, p be the peak value of the power signal, E be the spectral energy, f(peak) be the main frequency component, and W be the peak value of the power signal. mean W is the mean of the time-frequency characteristics. var Let be the variance of the time-frequency characteristics, ∈ be a constant, and w1, w2, w3, w4, w5, w 6、 w7 and w8 are weighting coefficients.

5. The method as described in claim 4, characterized in that, The step of selecting multiple excellent power signals from multiple normalized power signals based on the individual fitness includes: Multiple normalized power signals are sorted according to their individual fitness scores from highest to lowest to obtain a fitness ranking result. A preset number of excellent power signals are then selected sequentially based on this fitness ranking result; or Select multiple excellent power signals from multiple normalized power signals whose individual fitness is greater than a preset fitness threshold.

6. The method as described in claim 3, characterized in that, The step of determining the optimized feature set based on multiple excellent power signals includes: By performing cross-operation on multiple excellent power signals using a cross-function, a set of cross-individuals is obtained, which contains multiple cross-individuals. By performing mutation operations on multiple crossover individuals using a mutation function, a set of mutated individuals is obtained; Obtain the feature information corresponding to each mutated individual within the set of mutated individuals; An optimized feature set is constructed based on the characteristic information corresponding to each mutated individual.

7. The method as described in claim 3, characterized in that, The dynamic weighted integrated fault judgment algorithm is as follows: Fault=Predict(Model,features In the formula, Let w be the comprehensive evaluation value of the fault state at time t, N be the number of models, and w be the value of the fault state at time t. i (t) represents the weights of the i-th trained fault classification model at time t, S i (t) represents the state estimate of the i-th trained fault classification model at time t, where Fault is the fault state, Predict is the prediction function, and features is the set of feature values.

8. A fault diagnosis system for power equipment, characterized in that, The fault diagnosis system for the power equipment includes: The acquisition module is used to acquire all raw power signals during the operation of the drive motor; The processing module is used to preprocess multiple raw power signals to obtain time-domain feature information, frequency-domain feature information, and time-frequency feature information corresponding to each normalized power signal. The calculation module is used to select multiple excellent power signals from multiple normalized power signals based on the time domain feature information, the frequency domain feature information, and the time-frequency feature information using an improved genetic algorithm internal fitness evaluation formula; The calculation module is also used to determine an optimized feature set based on multiple excellent power signals; The training module is used to train multiple fault classification models based on the optimized feature set. The diagnostic module is used to diagnose faults in power equipment by dynamically weighting and integrating fault judgment algorithms based on multiple trained fault classification models.

9. A fault diagnosis device for power equipment, characterized in that, The device includes: a memory, a processor, and a fault diagnosis program for a power device stored in the memory and executable on the processor, the fault diagnosis program for the power device being configured to implement the steps of the fault diagnosis method for a power device as claimed in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a fault diagnosis program for power equipment, which, when executed by a processor, implements the steps of the fault diagnosis method for power equipment as described in any one of claims 1 to 7.

Images