KR20220089853A - Method for Failure prediction and prognostics and health management of renewable energy generation facilities using machine learning technology - Google Patents

Abstract

The present invention relates to a failure prediction and soundness management method of a renewable energy power generation facility using machine learning technology.

Description

{Method for Failure prediction and prognostics and health management of renewable energy generation facilities using machine learning technology}

The present invention relates to the technical field mentioned in the title of the invention, and if the technology is applied and can be applied to other fields, the technical field of the present invention can be extended and applied to such an applied technology field.

As various new and renewable energy facilities such as solar power, wind power, and small hydro power are being distributed with private and government support, social concerns about the durability and safety of equipment are growing.

Renewable energy systems that have been disseminated or disseminated nationwide have increased maintenance costs due to simple breakdowns and replacement of parts. Since the number of cases of damage to human life/property due to natural disasters such as fire, typhoon, and heavy rain has continued, the disadvantages of new and renewable energy have been highlighted more than the advantages.

In the operation of new and renewable energy systems, it is urgent to introduce a system that can use IoT network and AI algorithm analysis to diagnose, identify, and maintain the system status in advance, reduce maintenance costs and prevent accidents in advance.

The technology underlying the invention and its detailed description are set forth together in the configuration for carrying out the invention. This specification is submitted by utilizing the claim deferral system, and in order to more clearly distinguish the present invention from the technology that is the background of the invention in the future, the content that is the background of the invention is included in the detailed content for carrying out the invention do.

An object of the present invention is to provide a technology capable of solving the above problems, a system containing the technology, a method using the technology, and the like. This may include not only the provided technology, but also various application methods applied by utilizing this technology, and various finished products utilizing the provided technology.

This specification is submitted using the claim deferral system and does not describe specific claims for means of solving the technology. Accordingly, the means for solving the problems include various means for solving the problems that can be interpreted with reference to the description and drawings of the present specification.

The present invention can thus be presented as a means of solving the problems mentioned. Such an effect is an effect obtained when the technique is implemented according to the implementation of the technique, but an effect that has not yet been discovered when the same technique is repeated can be further discussed. The addition of an effect cannot be viewed as a separate addition of a new matter if it is an effect that appears when the technical composition is implemented based on the contents described in the first specification.

The accompanying drawings are referred to for interpreting the present technology. Although the content of the technology shown in the drawings is not described, it can be seen as described in the present specification. If specific content is added later based on the content described only in the drawings, it does not correspond to the addition of new matters.

Hereinafter, embodiments disclosed in the present specification will be described with reference to the accompanying drawings.

In particular, even in the absence of reference numerals, components that can be inferred from the description in the specification may be interpreted according to the textual description.

Even if the word is used at the level of those skilled in the art, it may not be used incorrectly because it does not accurately reflect the actual function and role. In this case, it is a principle to be interpreted according to the intention of the inventor by analogy based on the content uniformly described in the specification. should be judged more flexibly by referring to the overall content.

Embodiments disclosed herein will be described in detail with reference to the accompanying drawings, but the same or similar components are assigned the same reference numbers regardless of reference numerals, and redundant description thereof will be omitted.

The suffixes "module" and "part" for components used in the following description are given or mixed in consideration of only the ease of writing the specification, and do not have distinct meanings or roles by themselves. In addition, in describing the embodiments disclosed in the present specification, if it is determined that detailed descriptions of related known technologies may obscure the gist of the embodiments disclosed in this specification, the detailed description thereof will be omitted. In addition, the accompanying drawings are only for easy understanding of the embodiments disclosed in the present specification, and the technical idea disclosed herein is not limited by the accompanying drawings, and all changes included in the spirit and scope of the present invention , should be understood to include equivalents or substitutes.

When an element is referred to as being “connected” or “connected” to another element, it is understood that it may be directly connected or connected to the other element, but other elements may exist in between. it should be On the other hand, when it is said that a certain element is "directly connected" or "directly connected" to another element, it should be understood that the other element does not exist in the middle.

The singular expression includes the plural expression unless the context clearly dictates otherwise. Also, throughout this specification, when a part "includes" a certain component, it means that other components may be further included, rather than excluding other components, unless specifically stated to the contrary.

It will be apparent to those skilled in the art that the present invention may be embodied in other specific forms without departing from the spirit and essential characteristics of the present invention.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. In the drawings, like reference numerals refer to like elements.

I. Overview of the system

Currently, a small hydro power system facility has been built in Wando, Jeollanam-do, and is currently in operation, and it is being conducted as a test bed for this project. Ajinsan has a potential energy of 38.2J and a possible fall of 5.4m, making it suitable for small hydro power facilities. By installing IoT sensors (temperature, vibration, flow rate, RPM), state data of small hydro power generation equipment is collected from HEMS (Hydro Electricity Managements System) through Modbus protocol (485 or Ethernet).

Referring to FIG. 1 , VPP (Virtual Power Plant) completes big data by collecting a large amount of energy and environmental data through a sensor network using IoT and IoT, and converges AI-based technology and integrated platform technology based on this. The purpose is to derive the optimal power generation environment and to eliminate the power plant operation risk by foreseeing and removing all possible risk factors and obstacles. In addition, by using the digital twin technology of renewable power generation facilities such as solar power, wind power, and small hydro power, it is possible to integrate operation and management of new and renewable energy power plants in each region.

Designing and developing an AI-based Prognostics and Health Management (PHM) system through a vast amount of data based on virtual power plants and digital twin technology and applied to new and renewable power generation facilities Condition-Based Maintenance (CBM) ) want to

1. Implementation of Prognostics and Health Management (PHM) technology

Referring to FIG. 2 , the failure prediction and soundness management technology is a diagnostic technology that monitors equipment or environmental data using various sensors and captures signs of failure and prediction of Remaining Useful Life (RUL) and effective prudential management techniques. In the case of fault diagnosis, it is also applied in industrial fields according to the development of sensor technology and IoT technology, but there is a limit to its usefulness because it is usually detected when a sufficient state change occurs due to the deterioration of performance already progressing to a certain extent. To supplement this, research on predictive technology that predicts the remaining useful life of mechanical equipment or systems based on the current state is being conducted, and it can be divided into two main methods according to the method of estimating the lifespan. Data-Driven Approach is a model that predicts future failures after training the relationship between input and damage using machine learning techniques. In the case of the physical model-based method, although the physical model varies depending on the application target, it is possible to predict relatively accurately the long-term damage behavior. At this time, the physical model must be verified, and statistical methods such as the statistical hypothesis test or the Dayesian method are used for the method.

2. Implementation of CBM (Condition-Based Maintenance) technology

When it becomes possible to predict the remaining useful life of power generation facilities and systems by using failure prediction and soundness management technology (PHM), accidents, risk factors, and failures can be prevented in advance and prompt response is possible even in situations where it is difficult for workers to access. In particular, the increase in maintenance costs and huge losses due to the shutdown of power generation facilities can be fatal.

CBM maintenance can be divided into three stages: real-time condition monitoring, big data processing, and maintenance timing and scope determination as shown in Figure 6. Monitoring the status of all production facilities and products in real time through IoT, extracting an appropriate health index through big data processing techniques such as signal processing, machine learning, and statistical analysis, and then predicting failures (Prognostics) Techniques can be used to predict the remaining useful life (RUL) and determine the optimal maintenance interval and extent. This enables users to manage overall health (Health Management)

Depending on the system to perform condition-based maintenance, measurement parameters and applied diagnostic technology may vary. After excluding the experience-based method that does not provide real-time monitoring as a method of fault diagnosis and predictive technology, there are two directions: a data-based method and a model-based method. In the case of data-based methods, it can be used to diagnose and predict failures for objects that are difficult to make physical models. After installation, the process of acquiring as much data as possible should be considered. On the other hand, when the model-based method is applied, the uncertainty and complexity in the parameter estimation and correlation between parameters of the physical model step by approaching a single machine and equipment unit rather than approaching the entire complex system. It is effective to lower

3. Real-time sensor data signal processing and monitoring system based on IoT network

Referring to FIG. 3 , HEMS (Hydro Electricity Management System) firmware: builds an IoT network system for collecting power generation facility data such as generators, inverters, and rectifiers and environmental data, and collects data of renewable power generation facilities such as generators, inverters, rectifiers, and meters and external data such as vibration, temperature, RPM, flow rate, etc. through the IoT sensor network, an interface is developed and transmitted in real time, and through this, the generator, inverter, rectifier, etc., which are power generation facilities, are maintained and monitored in real time 24 hours.

1) Implementation of cloud platform for real-time IoT data processing

By building a cloud platform to collect and process massive amounts of data collected from renewable energy power plants in each region in real time, data integration and maintenance

2) Development of pre-processing API for sensor data signal processing

Data collection through IoT is essential for integration, analysis, and maintenance through big data, and it is transmitted to the cloud in real time.

3) Implementation of User-Friendly front-end for real-time monitoring and real-time control

Front-end web for real-time action and control, monitoring analysis data, statistical data and failure prediction data through data integration and AI algorithm analysis of real-time data of new and renewable energy power plants and virtual power plants to which digital twin technology is applied build the page

4) IoT power generation system sensor and power line design and construction

Based on the data and know-how accumulated while operating small hydro power generation facilities and systems for the past 4 years, it is possible to design a power generation system considering the fall and flow rate, and to design and build a system through an agreement with Juam Telecommunication for power generation facilities and electricity related construction.

4. Implementation of AI-based 　PHM/CBM system

Referring to FIG. 4, using machine learning techniques (NeuralNetwork, DeepLearning, Relevance Vector Machine), data of new and renewable power generation facilities such as power generation, voltage, current, RPM, and ESS charge/discharge data loading, big data-based After analyzing and training the relationship, it is applied to online monitoring to predict future failure and exception data by extrapolation.

The optimal efficiency result of the deep learning technique is derived through the construction of an artificial neural network, and the effective data combination of big data is determined through the construction of a digital twin. Conduct simulation modeling and data-based modeling to abstract the digital twin model by judging the properties, processes, and dynamics of the actual energy facility. Diagnosis, prediction, and normative analysis including the causal relationship between the input and output controlled through the knowledge and data of the rules and actions above to determine the effective data combination based on the modeling data and the analysis and diagnostic data

1) Deep Learning development with Big Data-based Recurrent Neural Network (RNN. Recurrent Neural Network) configuration

Inspection and approval of Verification-Validation-Accreditation to ensure reliability of simulation and valid data combination. Strengthening the validation/validation stage of learning results through deep learning learning and empirical data evaluation

2) Energy generation, statistics, forecasting, warning, and recovery automation system development

Failure prediction, warning, and recovery automation system development/application through deep learning development of data collected through IoT (power generation, voltage, current, RPM, temperature, charge amount, discharge amount, weather, etc.)

3) Expansion of IoT network-based renewable energy (small power) power generation system

Proving the effectiveness of machine learning and PHM by applying this technology to at least two existing renewable energy facilities and collecting data for at least 4 months, and establishing the process for each stage

5. Construction of power generation facility system using IoT sensor network

Referring to FIG. 5 , data acquisition of power generation facilities through the establishment of RS-485 communication lines for each power generation facility is shown. Server API I/F program development for collection and real-time transmission of statistical data is in progress. Communication definition between power generation facility data and environmental data and I/F program development are in progress. The real-time data transmission service program is designed and developed by defining sensor data for each power generation facility and defining communication and interface.

1) Defect prediction and residual useful life (RUL) prediction using AI

Referring to FIG. 6 , the data collected by the IoT sensor through the data pre-processing process is processed into a data format suitable for AI learning. Preprocessing data with errors can lead to incorrect analysis results, so it is an important step that determines the performance of AI. Data cleaning to fill in missing values and check inconsistent data, data consolidation that integrates multiple databases and data files, data reduction to reduce data volume or the dimension of the object of analysis through sampling, and normalizing or grouping data Proceed with data pre-processing through operations such as data transformation

The process of selecting characteristic factors using pre-processed data proceeds. Extract data used for defect prediction or residual useful life prediction by identifying data that affects defect prediction or remaining useful life

If all variables are used, time and computing power requirements increase, and the problem of overfitting the model may occur.

The representative structures of deep learning models are classified into CNN and RNN structures, and the RNN structure is classified as a good model to reflect time-series characteristics.

Since the data input from the IoT sensor is time series data that is input according to the passage of time, predict the amount of power, remaining useful life, and normal/abnormal parts using analysis techniques such as LSTM and GRU, which are often used for time series data analysis.

LSTM is one of the transformation models to overcome the long-term dependencies of the RNN structure, and is mainly used for classification and prediction of time series data.

2) Virtual power plant (VPP) integrated operation platform development

Identify the requirements for monitoring and integrated control, and implement the planning and design of the web service

Implementation and development of user-friendly UI/publishing/interface

Cryptographic module and Server WCF API programming development

Statistical scheduling procedure development through SSMS based on MS-SQL

Developed a web integrated operation platform to manage the site data, statistics, and predictive operation of the power plant. Through the web page, the user can grasp various sites that are currently being developed at a glance, and furthermore, it plays the role of a control tower that predicts the amount of power generation and prevents risk factors in advance.

3) Application and verification to renewable energy generation sites

Developing an integrated operation platform to build a system using IoT sensor networking, analyze data through AI algorithms, and monitor and operate it

Introduce and build a renewable energy power generation facility demonstration site to predict the optimal power generation amount and remaining useful life for the relevant power generation site, and determine the optimal maintenance cycle and scope

4) Produce system schematic guidebooks, etc. and distribute them to applicable sites

Feasibility study for about 500 distribution in Jeollanam-do

Promotion of system introduction advantages, etc. and distribution of brochures

5) Management of business roles among participating companies

Cloud database construction for smooth communication between companies, data sharing, and big data construction

Data backup required for future commercialization and sharing of output and data with programming sources such as DLL module, encryption module and DB communication module through TFS service

II. Service Definition and Scope

Service contents

IoT Integrated System (HEMS: Hydro Electricity Management System) Firmware

Collecting power generation data such as generators, inverters, rectifiers, and meters that are power generation facilities

IoT Cat.M1 communication-based sensor data (temperature, RPM, flow, vibration) collection

Data transmission to cloud system based on LTE Cat.M1 communication

Design and develop embedded F/W program to transmit data in real-time without simplification of IoT data

AI-based power generation, RUL prediction service

Predict the amount of power generation and the remaining useful life (RUL) through PHM algorithm analysis based on data such as power generation facilities and environmental factors

Condition-based maintenance (CBM) service for failure prediction and prevention based on PHM algorithm

System and Network Configuration Fig. 7 is a system and network configuration diagram. Fig. 8 is a system layout.

The present invention is not limited by the above embodiments and the accompanying drawings. For those of ordinary skill in the art to which the present invention pertains, it will be apparent that the components according to the present invention can be substituted, modified and changed without departing from the technical spirit of the present invention.

In addition, since the content included in the present invention was conducted using the claim deferral application system, it can be extended and applied to improved content in the future. can be applied.

-

Claims (5)

a data acquisition unit for detecting anomalies;
a diagnostic unit for diagnosing defects and causes of defects and classifying their severity;
Prediction unit that predicts future defect progress and predicts remaining useful life;
A suggestion/action unit that suggests solutions to problems and provides key information for corrective actions; and
A method for predicting failure and soundness management of new and renewable energy power generation facilities using machine learning technology, including; a post-analysis unit that confirms problem solving and performs post-diagnosis status.
The method of claim 1,
The data of the data acquisition unit includes noise/vibration pattern data, system temperature/humidity data, and generator current/voltage data.
Failure prediction and soundness management method of renewable energy power generation facilities using machine learning technology including RPM data, inverter data, flow rate/flow rate, and external weather data.
The method of claim 1,
The diagnosis unit compares and analyzes the data with reference data to diagnose defects/symptoms, and to predict failure and health management of new and renewable energy power generation facilities using machine learning technology, characterized in that the cause is diagnosed.
The method of claim 1,
The prediction unit, probability evaluation and reliability analysis failure mode and impact evaluation (FMEA) damage start point / progress analysis
Failure prediction and soundness management method of new and renewable energy power generation facilities using machine learning technology, characterized in that short-term trend extrapolation and step-by-step change analysis are applied.
The method of claim 1,
The presentation/action unit is a failure prediction and soundness management method of a new and renewable energy power generation facility using machine learning technology, characterized in that the failure/defect type and severity estimation of the equipment parts, and recommended actions and costs are calculated.

Images