CN-121980920-A - Fan performance safety margin assessment method

CN121980920ACN 121980920 ACN121980920 ACN 121980920ACN-121980920-A

Abstract

The invention discloses a fan performance safety margin assessment method, which relates to the technical field of wind power equipment safety assessment and comprises the steps of obtaining operation parameters of a fan in a preset monitoring period, processing the operation parameters to form standardized operation parameters, generating a preliminary margin assessment sequence according to margin assessment thresholds of the fan design specification preset performance parameters and the relative deviation of the standardized operation parameters and the margin assessment thresholds, constructing a long-period memory network model combined with an attention mechanism, training the long-period memory network model by utilizing historical fan operation parameters, inputting the preliminary margin assessment sequence into the trained long-period memory network model, outputting comprehensive safety margin indexes, and determining the final performance safety level of the fan according to comparison results of the comprehensive safety margin indexes and the preset safety margin thresholds. The invention can improve the accuracy, dynamic adaptability and engineering interpretability of the fan performance safety evaluation.

Inventors

WANG HU
LU GANG
ZHU DEXING
REN GANG
TIAN WENHUI
HE KAIXIN
LIU JUN
GUO YONGQING

Assignees

新疆华电天山绿色能源有限公司

Dates

Publication Date: 20260505
Application Date: 20251226

Claims (9)

1. A fan performance safety margin assessment method is characterized by comprising the steps of, Acquiring operation parameters of the fan in a preset monitoring period, and processing the operation parameters to form standardized operation parameters, wherein the operation parameters comprise vibration, output power and gearbox oil temperature; generating a preliminary margin evaluation sequence according to a margin evaluation threshold value of a preset performance parameter of a fan design specification and the relative deviation of a standardized operation parameter and the margin evaluation threshold value; And constructing a long-period memory network model combined with an attention mechanism, training the long-period memory network model by utilizing the operation parameters of the historical fan, inputting a preliminary margin evaluation sequence into the trained long-period memory network model, outputting a comprehensive safety margin index, and determining the final performance safety level of the fan according to the comparison result of the comprehensive safety margin index and a preset safety margin threshold.
2. The method for evaluating a fan performance safety margin according to claim 1, wherein said processing the operating parameters comprises: obtaining vibration, output power and gearbox oil temperature of a fan in a preset monitoring period, and uniformly adding an acquisition time mark to each record; According to the acquisition time mark, mapping to the same continuous time reference, eliminating sampling frequency difference by interpolation and resampling, and performing time alignment; And performing denoising processing on the operation parameters after time alignment, and performing unified scale processing on the operation parameters after denoising to form standardized operation parameters.
3. The method for evaluating the safety margin of fan performance according to claim 1, wherein generating the preliminary margin evaluation sequence comprises: According to fan structural design specifications, material strength constraint and long-term operation experience, quantitatively describing the vibration amplitude, the power fluctuation level and the safety margin of the gearbox oil temperature under the design working condition, and defining the safety margin as a margin evaluation threshold; Comparing the actual value of each parameter in the standardized operation parameters with a margin evaluation threshold value at a corresponding moment, and quantifying the degree of deviation of the operation parameters from a safety boundary through relative deviation, wherein the degree is expressed as follows: ; Wherein, the Represent the first The individual operating parameters being at the moment The degree of deviation of the threshold value with respect to the margin evaluation, Represent the first The individual operating parameters being at the moment The actual values after the normalization process are used, Represent the first The individual operating parameters being at the moment A margin evaluation threshold of (2); And continuously calculating and arranging the relative deviation of all the operation parameters in the monitoring period at each moment to form a preliminary margin evaluation sequence.
4. The method for evaluating the safety margin of fan performance according to claim 1, wherein the long-term memory network model combined with the attention mechanism comprises a time sequence feature coding module, an attention weighting fusion module and a safety margin mapping module; the time sequence feature coding module comprises an input mapping layer, a first long-period and short-period memory network layer and a second long-period and short-period memory network layer which are sequentially arranged; the attention weighted fusion module is connected with the second long-short-term memory network layer and comprises a correlation calculation unit, a weight normalization unit and a characteristic weighted convergence unit; The safety margin mapping module is connected with the attention weighted fusion module and comprises a first full-connection layer, a second full-connection layer and a safety margin output layer which are sequentially connected, and finally outputs comprehensive safety margin indexes.
5. The method for evaluating fan performance safety margin of claim 4, wherein the attention weighted fusion module comprises: calculating a correlation score between a hidden state vector output by the long-term memory network at a time t and a current safety evaluation target, wherein the correlation score is expressed as follows: ; Wherein, the Represents a relevance score between the hidden state and the current security assessment objective, The attention score vector is represented as such, Representing the transposed symbol, And Representing the parameter matrix and bias term for the correlation map respectively, Indicating the time of the long-short-period memory network The outputted hidden state vector; the relevance score is converted to a normalized attention weight, expressed as: ; Wherein, the Representing the normalized attention weight, The length of time that the input sequence is represented, Representing a time index; And carrying out weighted summation on the hidden states by using the normalized attention weights to obtain a weighted and fused global time sequence feature representation, wherein the weighted and fused global time sequence feature representation is expressed as follows: ; Wherein, the Representing the global timing characteristic representation after weighted fusion via an attention mechanism.
6. The method for evaluating a fan performance safety margin as set forth in claim 1, wherein training the long-term and short-term memory network model using the historical fan operating parameters comprises: selecting a preliminary margin evaluation sequence corresponding to historical fan operation data as an input sample, and taking a fan safety state result reflected by a corresponding maintenance record as a supervision tag; And carrying out joint optimization on the long-short-period memory network element parameters and the attention mechanism parameters by minimizing the deviation between the comprehensive safety margin index output by the model and the supervision label, so as to obtain the trained long-short-period memory network model.
7. The method for evaluating a performance safety margin of a wind turbine according to claim 6, wherein said determining a final performance safety level of the wind turbine comprises: presetting a safety margin threshold according to historical operation statistical results and operation and maintenance experience of the fan, and determining the final performance safety level of the fan according to a comparison result of the comprehensive safety margin index and the preset safety margin threshold; The performance safety level comprises three levels of safety, early warning and danger, wherein the performance safety level is judged to be dangerous when the comprehensive safety margin index is smaller than a first preset safety margin threshold value, the performance safety level is judged to be early warning when the comprehensive safety margin index is located between the first preset safety margin threshold value and a second preset safety margin threshold value, and the performance safety level is judged to be safe when the comprehensive safety margin index is larger than the second preset safety margin threshold value.
8. A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the computer program is characterized in that the processor executes the computer program to implement the steps of the fan performance safety margin assessment method according to any one of claims 1-7.
9. A computer readable storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the steps of a fan performance safety margin assessment method according to any one of claims 1 to 7.

Description

Fan performance safety margin assessment method Technical Field The invention relates to the technical field of wind power equipment safety evaluation, in particular to a fan performance safety margin evaluation method. Background With the continuous increase of the installed capacity of wind power generation, fans gradually develop to large-scale, clustered and long-term continuous operation directions, and the operation safety and reliability of the fans become key factors for restricting the economy and stability of a wind power system. The existing fan running state evaluation technology is mainly developed around single or multi-parameter monitoring means such as vibration monitoring, power analysis and temperature monitoring, and the state sensing of core components such as a bearing, a gear box and a transmission system is realized by collecting running data of key components through an online monitoring system. In recent years, with the improvement of sensing technology and data acquisition capability, the joint analysis of multi-source operation parameters gradually becomes a research hot spot, and some methods begin to introduce statistical analysis, threshold judgment or machine learning models to comprehensively evaluate the operation state of a fan so as to improve the timeliness and accuracy of fault early warning. However, the related art still has significant drawbacks. On the one hand, most evaluation methods focus on abnormality detection of single parameters or instantaneous states, usually adopt fixed thresholds or empirical thresholds to judge, and are difficult to reflect dynamic change characteristics of safety margin of a fan under different loads and working conditions, and false alarm or missing alarm is easy to cause. On the other hand, the model based on the data driving partly introduces a time sequence analysis method, but the description of the multi-parameter safety margin evolution process is still not fine enough, and key time slices and redundant information cannot be effectively distinguished, so that the physical interpretability and engineering applicability of the model output result are insufficient. In addition, in the prior art, fault probability or health index is used as an output result, and a quantization index directly corresponding to a safety margin concept in design specifications is lacked, so that clear basis is difficult to provide for grading operation and risk decision. Disclosure of Invention The invention is provided in view of the problems of the existing fan performance safety margin assessment method. Therefore, the problem to be solved by the present invention is how to provide a fan performance safety margin assessment method. In order to solve the technical problems, the invention provides the following technical scheme: In a first aspect, the invention provides a fan performance safety margin assessment method, which comprises the steps of obtaining operation parameters of a fan in a preset monitoring period, and processing the operation parameters to form standardized operation parameters, wherein the operation parameters comprise vibration, output power and gearbox oil temperature; generating a preliminary margin evaluation sequence according to a margin evaluation threshold value of a preset performance parameter of a fan design specification and the relative deviation of a standardized operation parameter and the margin evaluation threshold value; And constructing a long-period memory network model combined with an attention mechanism, training the long-period memory network model by utilizing the operation parameters of the historical fan, inputting a preliminary margin evaluation sequence into the trained long-period memory network model, outputting a comprehensive safety margin index, and determining the final performance safety level of the fan according to the comparison result of the comprehensive safety margin index and a preset safety margin threshold. As a preferable scheme of the fan performance safety margin assessment method, the method comprises the following steps: obtaining vibration, output power and gearbox oil temperature of a fan in a preset monitoring period, and uniformly adding an acquisition time mark to each record; According to the acquisition time mark, mapping to the same continuous time reference, eliminating sampling frequency difference by interpolation and resampling, and performing time alignment; And performing denoising processing on the operation parameters after time alignment, and performing unified scale processing on the operation parameters after denoising to form standardized operation parameters. As a preferable scheme of the fan performance safety margin assessment method, the generating the preliminary margin assessment sequence comprises the following steps: According to fan structural design specifications, material strength constraint and long-term operation experience, quantitatively describin