KR-102964311-B1 - Wind power generator operation management method and system

KR102964311B1KR 102964311 B1KR102964311 B1KR 102964311B1KR-102964311-B1

Abstract

A method and system for managing the operation of a wind turbine are provided. A wind turbine management method according to an embodiment of the present invention predicts the fire risk of a wind turbine by utilizing a neural network trained to predict fire risk by analyzing multiple input variables regarding the state of the wind turbine, predicts the fire risk of the wind turbine through fuzzy inference from multiple input variables regarding the state of the wind turbine, and calculates the fire risk of the wind turbine based on the predicted risks. Accordingly, the wind power generation facility can be safely operated by predicting the fire risk of the wind turbine through artificial intelligence and fuzzy inference.

Inventors

조수형
김종현
김대환

Assignees

한국전자기술연구원
디엑스랩즈 주식회사

Dates

Publication Date: 20260513
Application Date: 20231018

Claims (12)

A first prediction step for predicting the fire risk of a wind turbine using a neural network trained to predict the fire risk by analyzing multiple input variables regarding the state of the wind turbine; A second prediction step for predicting the fire risk of a wind turbine through fuzzy inference from multiple input variables regarding the state of the wind turbine; A step of calculating the fire risk of a wind turbine based on the risk predicted in the first prediction step and the risk predicted in the second prediction step; A wind turbine management method characterized by including the step of determining the optimal Pitch control value and the optimal Yaw control value of the wind turbine through fuzzy control from wind speed and wind direction.
In claim 1, In the first prediction step, the input variables are, A wind turbine management method characterized by including at least one of generator temperature, cooling fan temperature, PCS temperature, wind speed, pitch angle, rotor speed, gearbox temperature, and bearing temperature.
In claim 1, In the first prediction step, the input variables are, A wind turbine management method characterized by including at least one of vibration, flame, gas, and thermal imaging temperature.
A first prediction step for predicting the fire risk of a wind turbine using a neural network trained to predict the fire risk by analyzing multiple input variables regarding the state of the wind turbine; A second prediction step for predicting the fire risk of a wind turbine through fuzzy inference from multiple input variables regarding the state of the wind turbine; A step of calculating the fire risk of a wind turbine based on the risk predicted in the first prediction step and the risk predicted in the second prediction step; and In the first prediction step, the input variables are, It includes at least one of vibration, flame, gas, and thermal imaging temperature, Vibration and thermal imaging temperature are, A wind turbine management method characterized by indicating a dangerous condition occurring when a gearbox is worn.
In claim 3, The second prediction stage is, A wind turbine management method characterized by predicting fire risk by performing fuzzy inference from fuzzy rules representing fire risk according to at least one condition among vibration, flame, gas, and thermal imaging temperature.
delete
In claim 1, The decision stage is, A wind turbine management method characterized by determining an optimal Pitch control value and an optimal Yaw control value through fuzzy inference from fuzzy rules representing the maximum efficiency of a wind turbine according to conditions of a Pitch control value that can maximize the electricity production of the wind turbine according to wind speed and a Yaw control value that can maximize the electricity production of the wind turbine according to wind direction.
In claim 1, A wind turbine management method characterized by further including the step of visualizing the status of the wind turbine in real time using a SCADA (Supervisory Control And Data Acquisition) platform.
In claim 8, A method for managing a wind turbine, characterized by further including the step of predicting and displaying power generation based on weather, air density, altitude, and blade length.
A processor that predicts the fire risk of a wind turbine using a neural network trained to predict fire risk by analyzing multiple input variables regarding the state of the wind turbine, predicts the fire risk of the wind turbine through fuzzy inference from multiple input variables regarding the state of the wind turbine, calculates the fire risk of the wind turbine based on the predicted risks, and determines the optimal Pitch control value and the optimal Yaw control value of the wind turbine through fuzzy control from wind speed and wind direction; and A wind turbine management system characterized by including a storage unit that provides storage space required for a processor.
A step of predicting the fire risk of a wind turbine through fuzzy inference from multiple input variables regarding the state of the wind turbine; A wind turbine management method characterized by including the step of determining the optimal Pitch control value and the optimal Yaw control value of the wind turbine through fuzzy control from wind speed and wind direction.
A processor that predicts the fire risk of a wind turbine through fuzzy inference from multiple input variables regarding the state of the wind turbine, and determines the optimal Pitch control value and optimal Yaw control value of the wind turbine through fuzzy control from wind speed and wind direction; and A wind turbine management system characterized by including a storage unit that provides storage space required for a processor.

Description

Wind power generator operation management method and system The present invention relates to wind turbine operation management technology, and more specifically, to technology for fire prevention, optimal control, and maintenance of wind turbines. Various mechanical devices are installed inside the nacelle of a wind turbine to convert energy obtained from the wind into electrical energy. Since natural light enters and temperature fluctuations vary significantly depending on the season, as well as being further amplified by the operation of internal devices, applying standard fire thresholds results in a high probability of failure to detect fires early or false alarms. Furthermore, wind turbines are installed at a distance from land, making initial suppression difficult. Additionally, accessing towers 100 meters high is challenging, making fire prevention in wind turbines extremely important. Furthermore, since the output of wind power generation can fluctuate significantly depending on wind conditions and potentially affect the power grid, controlling the blade pitch and yaw is crucial for maximizing electricity production; therefore, it is necessary to explore methods for optimal control. Figures 1 and 2 are drawings for explaining neural network-based wind turbine fire risk prediction, FIGS. 3 to 5 are drawings for explaining fuzzy inference-based wind turbine fire risk prediction, FIGS. 6 to 10 are drawings for explaining wind turbine fuzzy control, FIGS. 11 and 12 are drawings for explaining wind turbine maintenance, FIG. 13 is a drawing for explaining a wind turbine operation management system. The present invention will be described in more detail below with reference to the drawings. 1. Neural Network-Based Prediction of Wind Turbine Fire Risk In a method for managing the operation of a wind turbine according to one embodiment of the present invention, the fire risk of a wind turbine is predicted using a neural network trained to predict the fire risk by analyzing a plurality of input variables regarding the state of the wind turbine. Here, the plurality of input variables regarding the state of the wind turbine include, as shown in FIG. 1, 1) generator temperature, 2) cooling fan temperature, 3) PCS temperature, 4) wind speed, 5) pitch angle, 6) rotor speed, 7) gearbox temperature, and 8) bearing temperature. A neural network for predicting fire risk in wind turbines can be implemented as a deep learning network using CNN, RNN, MLP, etc., and is trained using wind power big data as shown in Fig. 2. 2. Prediction of Wind Turbine Fire Risk Based on Fuzzy Inference In the wind turbine operation management method according to an embodiment of the present invention, the fire risk of the wind turbine is predicted through fuzzy inference from a plurality of input variables regarding the state of the wind turbine. Here, the plurality of input variables regarding the state of the wind turbine include 1) vibration, 2) flame, 3) gas, and 4) thermal imaging temperature, as shown in FIG. 3. Since flames are significantly affected by illumination, errors may occur during the day where actual flames are classified as normal. Regarding gases, errors may occur where surrounding garbage odors or suspended gases are mistaken for carbon dioxide. Meanwhile, spark discharges (arcs) generated during gearbox bearing wear can ignite the gear oil, potentially causing a fire. Therefore, in the embodiments of the present invention, to detect this risk early, probabilistic conditions for determining the conditions for a wind turbine fire are utilized by employing vibration and thermal imaging temperatures, which allow for an objective assessment of the hazardous state resulting from gearbox wear. Fuzzy rules are generated by connecting the presented fuzzy conditions using AND, OR, and NOT operations, and the fire risk of the wind turbine is predicted through fuzzy inference derived from these rules. Specifically, input data required for fuzzy logic is created, and the generated data can generally be represented in the IF-THEN format. Fuzzy inference refers to a series of processes that infer new relationships or facts from given rules using max-min reasoning. The conclusion of the FUZZY generation rule has two or more different belief values. In such cases, the function used to recalculate the belief value of the conclusion is the belief value combination function. For example, if the probability of fire risk in a wind turbine is set to 0.7, the normal probability of no fire occurring based on conventional mathematical statistics becomes 0.3. However, conventional probability-based statistics cannot definitively conclude that the probability of fire risk is necessarily 0.7 simply because the normal probability of no fire occurring is 0.3. This is because the actual probability of a fire occurring in a wind turbine could be 0.5, 0.8, or 0.6. This is because wind turbine fires can be caused by various sensor data errors. The res