Sensor signal processing method for gas turbine electronic controller board fault simulation

Abstract

The present invention provides a sensor signal processing method for simulating a gas turbine electronic controller board fault, and relates to the technical field of gas turbine power plant fault diagnosis. The method comprises: obtaining the original process signal of the sensor from the gas turbine controller system, performing data preprocessing, and forming a normal signal data set; superimposing the simulation signal of the fault on the signal data set to obtain a fault-containing simulation signal; inputting the fault-containing simulation signal into a health calculation model to obtain the health value of each sensor signal; weighted summing the health value of each sensor as the health value of the IO board, judging whether the health value of the IO board exceeds the preset normal range, and determining whether the IO board enters a fault state accordingly. Using this method, a large number of IO board fault state samples can be quickly and conveniently generated, providing a simulation environment for identifying IO board faults, and providing important decision support for gas turbine control system loop fault diagnosis.

Description

Technical Field

[0001] The invention relates to the technical field of fault diagnosis of gas turbine power plants, and in particular to a sensor signal processing method for fault simulation of a gas turbine electronic controller board. Background Art

[0002] With the increasing maturity of gas turbines and their combined cycle technology, as well as the massive development of natural gas resources worldwide and the increasingly severe environmental pressure on the earth, gas turbine power generation is not only used as emergency backup power and peak load, but also as clean energy, distributed energy and basic load to transmit electricity to the power grid. To this day, gas turbines are growing and developing rapidly in the field of power generation; gas turbine power generation using hydrogen as fuel has received unprecedented attention and great development.

[0003] The gas turbine control system is a complex nonlinear dynamic system composed of a large number of components that are assembled in a certain way, function and requirement. The electronic controller (DCS system) in the gas turbine control system is the core and key of the entire gas turbine control system. How to effectively diagnose the electronic controller of the gas turbine control system has always been a difficult problem in the industry. The traditional electronic controller fault diagnosis method based on BIT technology requires adding a large number of redundant diagnostic circuits in the electronic components, which increases the system cost on the one hand and adds new electronic fault points on the other hand. Different gas turbine control system manufacturers have built-in many professional system diagnostic functions in their respective DCS systems. These functions must be viewed in the engineer operation station with corresponding permissions, and the information acquisition process is passive. The operator needs to find various alarm information in different locations by himself, and then combine it with various paper drawings and materials, and even need to check the hardware indication status in the DCS cabinet to determine the fault point of the system. To further determine the cause of the failure, it is more about the personal ability and experience of the engineer. This seriously restricts the rapid judgment of gas turbine power plants when system failures occur. As a result, unplanned shutdowns caused by false alarms of the electronic controller system often occur, causing great economic losses to the power plant.

[0004] At present, the method adopted for maintaining the industrial equipment in the factory is "prevention first, planned maintenance first, and temporary repair as a supplement". Specifically, the factory sets a period of three to five years according to the new and old conditions of the industrial equipment to carry out a major overhaul of all the industrial equipment in the factory, that is, to shut down, dismantle, maintain and reinstall all the industrial equipment, without considering whether a certain industrial equipment needs maintenance. Generally, a major overhaul lasts for more than two months, requiring more than 100 professional maintenance personnel and nearly 1,000 construction personnel of all kinds. During the two major overhauls, the industrial equipment is repaired once every one to two years, that is, the auxiliary working equipment in the factory, such as the steam turbine generator set in the power plant and most of the industrial equipment other than other large core equipment, is dismantled, maintained and reinstalled, generally for a period of more than ten days to two months. At the same time, while the factory is carrying out "planned maintenance", it is assisted by "temporary repair", that is, temporary repair of industrial equipment that has failed and cannot operate.

[0005] The above-mentioned maintenance method for industrial equipment in factories is widely used in all domestic heavy industrial enterprises such as electric power, metallurgy and petrochemical. However, this maintenance method for industrial equipment in factories also has two disadvantages: First, it is obvious that this industrial equipment maintenance method will cause a huge waste of resources. Since the planned maintenance completely ignores the specific operating conditions of each specific industrial equipment, a unified maintenance method is used to treat all industrial equipment in the factory, resulting in unnecessary huge investment in personnel, time and economy; second, this industrial equipment maintenance method will cause some of the industrial equipment to fail and age faster, and it cannot be guaranteed that each industrial equipment in the factory will operate reliably and smoothly during the two overhauls, and temporary emergency repairs are also needed.

[0006] In the gas turbine power generation industry, this routine maintenance work mode is not suitable for electronic controller systems represented by DCS. The DCS system is composed of high-precision and high-density electronic components / chips. The working principles and failure mechanisms of various components are completely different. Their failure modes are often "sudden". It is impossible to effectively solve the fault detection problem by relying solely on daily inspection and testing methods. In the daily operation of the gas turbine, the temperature of the combustion chamber is the most important control and monitoring parameter. The stability and consistency of this group of parameters (9F unit is 31 circularly arranged measurement points) is directly related to whether the unit can continue to operate; once the measurement value is distorted due to the failure of the thermocouple IO board, it may be necessary to shut down immediately, and the consequences are very serious.

[0007] When studying data-driven fault diagnosis methods for electronic controller IO boards, the lack of actual fault samples is the main bottleneck restricting the development and application of diagnostic algorithms. How to effectively obtain fault samples with practical value has always been a difficult problem in the industry. Summary of the invention

[0008] The object of the present invention is to provide a sensor signal processing method for gas turbine electronic controller board fault simulation in view of the above-mentioned deficiencies in the prior art, so as to solve the problem that fault samples are difficult to obtain in the fault diagnosis of the thermocouple IO board of the electronic controller of the gas turbine control system, thereby resulting in the inability to effectively verify the fault diagnosis method.

[0009] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0010] The present invention provides a sensor signal processing method for gas turbine electronic controller board fault simulation, the method is used for performing fault simulation on a thermocouple IO board of a gas turbine electronic controller, the method comprising:

[0011] From the gas turbine electronic controller system, the original temperature process signal of each temperature sensor is obtained, and data preprocessing is performed to form an original normal signal data set;

[0012] Preset multiple fault simulation signals are superimposed on the signals in the original normal signal data set to obtain fault simulation signals for thermocouple IO board fault simulation;

[0013] Input the faulty simulation signal into the sensor signal health calculation model to obtain the health value of each temperature sensor signal;

[0014] The health values ​​of each temperature sensor are weighted and summed as the health value of the thermocouple IO board. It is judged whether the health value of the thermocouple IO board exceeds the preset health value normal interval range boundary, and whether the thermocouple IO board enters a fault state is determined based on the judgment result, thereby realizing the state simulation of the thermocouple IO board failure.

[0015] Optionally, the original temperature process signal of each temperature sensor is acquired from the gas turbine electronic controller system using the OPC UA industrial protocol.

[0016] Optionally, real-time temperature process signals of all temperature sensors associated with the thermocouple IO board are obtained from the gas turbine electronic controller DCS system using an OPC UA protocol interface, and data preprocessing includes redundant sample removal operations, abnormal sample removal operations, and data normalization operations.

[0017] Optionally, the preset multiple faults include: deviation fault, impact fault, drift fault, periodic fault and random interference fault.

[0018] Optionally, in the process of superimposing simulation signals of multiple preset faults on the signals in the original normal signal data set, a step signal is used to simulate a deviation fault in the temperature sensor signal; a pulse signal is used to simulate an impact fault in the temperature sensor signal; a linear signal is used to simulate a drift fault in the temperature sensor signal; a sinusoidal wave signal is used to simulate a periodic fault in the temperature sensor signal; and a Gaussian white noise signal is used to simulate a random interference fault in the temperature sensor signal.

[0019] Optionally, the fault-containing simulation signal is input into a sensor signal health calculation model to obtain the health value of each temperature sensor signal, including:

[0020] The weighted sum of squares of the sensor residual sequence based on the unscented Kalman filter is used as the health value of the temperature sensor signal.

[0021]

[0022] ∑=(diag(σ)) 2

[0023] Among them, R i is the weighted sum of the residual sequence of the temperature sensor corresponding to the ith channel of the thermocouple IO board. Based on the unscented Kalman filter, a filter corresponding to the m sensors associated with the thermocouple IO board is established. The input of each filter is m-1 measurement parameter values. The input of the ith filter is the measurement parameters of the remaining m-1 sensors except the ith sensor. σ is the standard deviation of the measurement parameters. (i) is the measurement parameter of the ith filter, is the predicted output of the i-th filter nonlinear model; m is the total number of temperature sensors associated with the thermocouple IO board.

[0024] Optionally, the health value of the thermocouple IO board is as follows:

[0025]

[0026] Among them, R IO-card is the health value of the thermocouple IO board, k i is the weighted weight of the i-th channel of the thermocouple IO board.

[0027] Optionally, whether the thermocouple IO board fails is determined according to the following formula:

[0028] F IOcard =1, if R IO-card >R high limit

[0029] Among them, F IOcardIt is the fault judgment flag of the thermocouple IO board. IOcard =1, it is determined that the thermocouple IO board fails;

[0030] R high limit The health value of the thermocouple IO card exceeds the preset health value normal range boundary, that is, R IO-card >R high limit In the case of IOcard =1.

[0031] Optionally, R high limit The value of is equal to 3σ.

[0032] The beneficial effects of the present invention include:

[0033] The sensor signal processing method for gas turbine electronic controller board fault simulation provided by the present invention comprises: obtaining the original temperature process signal of each temperature sensor from the gas turbine electronic controller system, performing data preprocessing, and forming an original normal signal data set; performing simulation signal superposition of preset multiple faults on the signal in the original normal signal data set, and obtaining a fault-containing simulation signal for thermocouple IO board fault simulation; inputting the fault-containing simulation signal into the sensor signal health calculation model to obtain the health value of each temperature sensor signal; performing weighted summation on the health value of each temperature sensor as the health value of the thermocouple IO board, judging whether the health value of the thermocouple IO board exceeds the preset health value normal interval range boundary, and determining whether the thermocouple IO board enters a fault state according to the judgment result, thereby realizing the state simulation of the thermocouple IO board fault. The method can quickly and conveniently generate a large number of IO board fault state samples for the research of fault diagnosis algorithm for IO board, provide an experimental simulation environment for timely identification of IO board faults of electronic controllers, and provide important decision support for gas turbine control system loop fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 A schematic flow chart of a sensor signal processing method for simulating a gas turbine electronic controller board fault provided by an embodiment of the present invention is shown;

[0036] Figure 2The figure shows a 'step type' fault signal synthesis flow chart provided by an embodiment of the present invention;

[0037] Figure 3 The figure shows a 'pulse type' fault signal synthesis flow chart provided by an embodiment of the present invention;

[0038] Figure 4 The flowchart of the synthesis of the 'time-varying temperature drift type' fault signal provided by the embodiment of the present invention is shown;

[0039] Figure 5 The 'periodic type' fault signal synthesis flow chart provided by the embodiment of the present invention is shown;

[0040] Figure 6 The flowchart of synthesizing a 'white noise type' fault signal provided by an embodiment of the present invention is shown;

[0041] Figure 7 The actual operation flow diagram of the sensor signal processing method for gas turbine electronic controller board fault simulation provided by the embodiment of the present invention is shown. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] The electronic controller in the gas turbine control system is a complex system composed of highly integrated electronic components. The fault diagnosis method based on BIT is not only costly to implement, but also the newly added fault diagnosis loop increases the complexity of the system and also increases the probability of system failure. Data-based fault diagnosis methods are a recent research hotspot. However, data-based fault diagnosis methods require a large number of high-quality fault samples for algorithm training and verification. Industrial sites cannot generate enough fault samples (this is easy to understand. If the fault samples in industrial sites are rich, it means that the process method is immature and it is impossible to obtain widespread industrial application). This results in data-based fault diagnosis methods often only performing well in laboratory environments and cannot be truly used in industrial sites.

[0044] In order to solve the above problems, the present invention provides a convenient and feasible thermocouple IO board fault simulation method.

[0045] Figure 1 FIG. 1 is a flow chart showing a sensor signal processing method for simulating a gas turbine electronic controller board fault provided by an embodiment of the present invention. Figure 1 As shown, the sensor signal processing method for gas turbine electronic controller board fault simulation provided by the present invention is used to perform fault simulation on the thermocouple IO board of the gas turbine electronic controller, and the method comprises:

[0046] Step 101: Obtain the original temperature process signal of each temperature sensor from the gas turbine electronic controller system, perform data preprocessing, and form an original normal signal data set. For example, obtain the original process signal of the gas turbine combustion chamber temperature sensor from the gas turbine electronic controller system of the 9F unit.

[0047] Step 102: Preset multiple fault simulation signals are superimposed on the signals in the original normal signal data set to obtain a fault simulation signal for thermocouple IO board fault simulation.

[0048] Step 103: input the fault-containing simulation signal into the sensor signal health calculation model to obtain the health value of each temperature sensor signal;

[0049] Step 104: perform a weighted summation of the health values ​​of the temperature sensors as the health value of the thermocouple IO board, determine whether the health value of the thermocouple IO board exceeds the preset health value normal interval range boundary, and determine whether the thermocouple IO board enters a fault state based on the judgment result, thereby realizing the state simulation of the thermocouple IO board failure.

[0050] This method can be used to quickly and easily generate a large number of IO board fault status samples, which can be used to study the fault diagnosis algorithm of IO boards, provide an experimental simulation environment for timely identification of IO board faults of electronic controllers, and provide important decision-making support for gas turbine control system loop fault diagnosis.

[0051] Optionally, the gas turbine electronic controller system is connected to an OPC Server data source of the gas turbine electronic controller system by using the OPC UA industrial protocol to obtain the original temperature process signals of each temperature sensor.

[0052] Optionally, the real-time temperature process signals of all temperature sensors associated with the thermocouple IO board are obtained from the gas turbine electronic controller DCS system using the OPC UA protocol interface. The data preprocessing includes operations of removing redundant samples, removing abnormal samples (removing data under equipment shutdown conditions), and data normalization operations to convert the data into a Gaussian distribution form with a mean of 0 and a variance of 1.

[0053] Optionally, the preset multiple faults include: deviation fault, impact fault, drift fault, periodic fault and random interference fault.

[0054] Optionally, in the process of superimposing simulation signals of preset multiple faults on the signals in the original normal signal data set, a step signal is used to simulate a deviation fault of the temperature sensor signal, and the function of the signal is as follows:

[0055] J(t)+B

[0056] Among them, J(t) is the fault-free signal, and B is the amplitude of the step fault signal.

[0057] In the process of superimposing the simulated signals of multiple preset faults on the signals in the original normal signal data set, a pulse signal is used to simulate the impact fault of the temperature sensor signal. The function of the signal is as follows:

[0058] J(t)+C

[0059] Among them, J(t) is the fault-free signal, and C is the amplitude of the pulse fault signal. Compared with the step fault, the duration of the pulse fault is shorter.

[0060] In the process of superimposing the simulated signals of multiple preset faults on the signals in the original normal signal data set, a linear signal is used to simulate the drift fault of the temperature sensor signal. The function of the signal is as follows:

[0061] J(t)+At+D

[0062] Wherein, J(t) is the fault-free signal, A is the slope of the time-varying fault signal, and D is the initial offset of the time-varying fault signal.

[0063] In the process of superimposing the simulated signals of multiple preset faults on the signals in the original normal signal data set, a sine wave signal is used to simulate the periodic fault of the temperature sensor signal (such as interference from high-power electrical equipment). The function of the signal is as follows:

[0064] J(t)+Esinω0t

[0065] Among them, J(t) is the fault-free signal, E is the amplitude of the periodic fault, and ω0 is the angular frequency of the fault.

[0066] In the process of superimposing the simulated signals of multiple preset faults on the signals in the original normal signal data set, Gaussian white noise signals are used to simulate the random interference faults of the temperature sensor signals (such as poor grounding / poor contact, etc.). Gaussian noise refers to the noise whose probability density function obeys the normal distribution. Gaussian distribution, denoted as N(μ,σ 2 ), where μ is the mean (mathematical expectation) of the Gaussian distribution, σ 2 is the variance of the Gaussian distribution. When μ=0, σ 2=1, the distribution is called the standard normal distribution. The one-dimensional probability density of the Gaussian distribution can be expressed as:

[0067]

[0068] The data with the fault signal superimposed thereon are input into the weighted sum of squares calculation model of the sensor signal residual sequence to obtain the weighted sum of squares of the residual sequence of each temperature sensor;

[0069] Optionally, the fault-containing simulation signal is input into a sensor signal health calculation model to obtain the health value of each temperature sensor signal, including:

[0070] The weighted sum of squares of the sensor residual sequence based on the unscented Kalman filter is used as the health value of the temperature sensor signal.

[0071]

[0072] ∑=(diag(σ)) 2

[0073] Among them, R i is the weighted sum of the residual sequence of the temperature sensor corresponding to the ith channel of the thermocouple IO board. Based on the unscented Kalman filter, a filter corresponding to the m sensors associated with the thermocouple IO board is established. The input of each filter is m-1 measurement parameter values. The input of the ith filter is the measurement parameters of the remaining m-1 sensors except the ith sensor. σ is the standard deviation of the measurement parameters. (i) is the measurement parameter of the ith filter, is the predicted output of the i-th filter nonlinear model; m is the total number of temperature sensors associated with the thermocouple IO board.

[0074]

[0075] From the above formula, we can see that the unscented Kalman filter can be used according to the measured parameters y of the gas turbine k Predicting output with nonlinear models The residual is used to estimate the health parameter p, and by updating the health parameters of the nonlinear model, the output parameters of the model are made to track the output of the real gas turbine, thereby establishing a gas path fault diagnosis system based on the unscented Kalman filter and realizing the estimation of the health parameters of the gas turbine.

[0076] Optionally, the health value of the thermocouple IO board is as follows:

[0077]

[0078] Among them, R IO-card is the health value of the thermocouple IO board, ki is the weighted weight of the i-th channel of the thermocouple IO board.

[0079] Optionally, whether the thermocouple IO board fails is determined according to the following formula:

[0080] F IOcard =1, if R IO-card >R high limit

[0081] Among them, F IOcard It is the fault judgment flag of the thermocouple IO board. IOcard =1, it is determined that the thermocouple IO board fails;

[0082] R high limit The health value of the thermocouple IO card exceeds the preset health value normal range boundary, that is, R IO-card >R high limit In the case of IOcard =1.

[0083] Optionally, R high limit The value of is equal to 3σ.

[0084] Figure 2 The figure shows a 'step type' fault signal synthesis flow chart provided by an embodiment of the present invention; Figure 3 The figure shows a 'pulse type' fault signal synthesis flow chart provided by an embodiment of the present invention; Figure 4 The synthesis flow chart of the 'time-varying temperature drift type' fault signal provided by the embodiment of the present invention is shown; Figure 5 shows a 'periodic type' fault signal synthesis flow chart provided by an embodiment of the present invention; Figure 6 The flowchart of synthesizing a 'white noise type' fault signal provided by an embodiment of the present invention is shown; Figure 7 The actual operation flow diagram of the sensor signal processing method for gas turbine electronic controller board fault simulation provided by the embodiment of the present invention is shown.

[0085] In summary, when the thermocouple IO board of the electronic controller fails, the sensor signal state collected by the board will be abnormal, such as a sudden jump in the value (step fault), a sudden jump in the value and then returning to normal (pulse fault), a continuous increase or decrease in the value (time-varying / temperature drift fault), a periodicity in the value (often due to the strong electric induction signal in the board), or a deterioration in signal stability (white noise interference, often caused by poor contact or abnormal grounding). The first reaction to the failure of the IO board is the abnormality of the above-mentioned sensor signals. By superimposing the fault signal in the sensor signal data, the morphology of the IO board in the fault state is simulated, so that a large number of fault samples can be easily generated, which solves the problem of missing samples in the fault diagnosis of the electronic controller IO board and provides a basis for further IO board fault diagnosis.

[0086] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it. They cannot be used to limit the protection scope of the present invention. All equivalent changes or modifications made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A sensor signal processing method for gas turbine electronic controller board fault simulation, characterized in that: The method is used to perform fault simulation on a thermocouple IO board of a gas turbine electronic controller, and the method comprises: From the gas turbine electronic controller system, the original temperature process signal of each temperature sensor is obtained, and data preprocessing is performed to form an original normal signal data set; Preset multiple fault simulation signals are superimposed on the signals in the original normal signal data set to obtain a fault simulation signal for thermocouple IO board fault simulation; Inputting the fault-containing simulation signal into a sensor signal health calculation model to obtain health values ​​of each temperature sensor signal; The health values ​​of each temperature sensor are weighted and summed as the health value of the thermocouple IO board, and it is judged whether the health value of the thermocouple IO board exceeds the preset health value normal interval range boundary, and whether the thermocouple IO board enters a fault state is determined according to the judgment result, so as to realize the state simulation of the thermocouple IO board fault; Inputting the faulty simulation signal into the sensor signal health calculation model to obtain the health value of each temperature sensor signal includes: The weighted sum of squares of the sensor residual sequence based on the unscented Kalman filter is used as the health value of the temperature sensor signal. ， Among them, R i It is the weighted sum of the residual sequence of the temperature sensor corresponding to the ith channel of the thermocouple IO board. Based on the unscented Kalman filter, a filter corresponding to the m sensors associated with the thermocouple IO board is established. The input of each filter is m-1 measurement parameter values. The input of the ith filter is the measurement parameters of the remaining m-1 sensors except the ith sensor. is the standard deviation of the measured parameter, is the measurement parameter of the ith filter, is the predicted output of the i-th filter nonlinear model; m is the total number of temperature sensors associated with the thermocouple IO board; The health values ​​of the thermocouple IO board are as follows: ， in, is the health value of the thermocouple IO board, k i is the weighted weight of the ith channel of the thermocouple IO board; Determine whether the thermocouple IO board fails according to the following formula: ， in, This is the fault judgment flag of the thermocouple IO board. When the thermocouple IO board is determined to be faulty; is the upper limit of the preset normal interval range of the health value. When the health value of the thermocouple IO board exceeds the preset normal interval range of the health value, that is, In the case of .

2. The sensor signal processing method for gas turbine electronic controller board fault simulation according to claim 1, characterized in that: The original temperature process signals of various temperature sensors are acquired from the gas turbine electronic controller system using the OPC UA industrial protocol.

3. The sensor signal processing method for gas turbine electronic controller board fault simulation according to claim 2 is characterized in that: From the gas turbine electronic controller system, the real-time temperature process signals of all temperature sensors associated with the thermocouple IO board are obtained from the gas turbine electronic controller DCS system using the OPC UA protocol interface, and the data preprocessing includes redundant sample removal operations, abnormal sample removal operations, and data normalization operations.

4. The sensor signal processing method for gas turbine electronic controller board fault simulation according to claim 1 is characterized in that: The preset multiple faults include: deviation fault, impact fault, drift fault, periodic fault and random interference fault.

5. The sensor signal processing method for gas turbine electronic controller board fault simulation according to claim 4 is characterized in that: In the process of superimposing simulation signals of multiple preset faults on the signals in the original normal signal data set, a step signal is used to simulate a deviation fault of the temperature sensor signal; a pulse signal is used to simulate an impact fault of the temperature sensor signal; a linear signal is used to simulate a drift fault of the temperature sensor signal; a sinusoidal wave signal is used to simulate a periodic fault of the temperature sensor signal; and a Gaussian white noise signal is used to simulate a random interference fault of the temperature sensor signal.

6. The sensor signal processing method for gas turbine electronic controller board fault simulation according to claim 1 is characterized in that: The value is equal to .

Images