Unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociation regression model

Abstract

The present invention discloses a unit energy efficiency monitoring and diagnosis method based on information entropy and auto-association regression model, comprising: collecting historical data of the unit's full operating range and pre-processing the historical data; clustering condition parameters with the goal of minimizing decision variables to establish a state estimation model for the optimal operating condition area; determining the weight value of each condition parameter according to the information entropy weight of the condition parameter; inputting the condition parameters during the unit operation process into the state estimation model of the optimal operating condition area in real time, predicting the estimated value of the condition parameter, calculating the deviation between the estimated value and the measured value and combining the weight value, constructing the energy efficiency deviation, and obtaining the sliding average deviation after filtering; comparing the sliding average deviation with a low energy efficiency deviation threshold and a high energy efficiency deviation threshold to provide a unit energy efficiency monitoring and diagnosis scheme. The present invention is based on a fusion data mining model of auto-association regression estimation, data clustering and information entropy, and is used for unit energy efficiency monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of unit energy efficiency status monitoring, and in particular to a steam turbine energy efficiency monitoring and diagnosis method based on information entropy and an autoassociation regression model. Background Art

[0002] At present, the energy efficiency monitoring and diagnosis of most units are still at the level of qualitative diagnosis. However, with the deepening of the country's energy conservation and emission reduction work, the traditional method of judging the energy efficiency status of units based on operating experience can no longer meet the current development needs.

[0003] With the continuous improvement of domestic unit automation and information technology, data-driven unit energy efficiency information extraction and analysis technologies have made some progress. However, energy efficiency evaluation still largely relies on theoretical or design values, lacking dynamics. Furthermore, data-driven energy efficiency monitoring and evaluation still have problems such as not considering the impact differences between variables and not integrating energy efficiency monitoring with fault diagnosis.

[0004] Based on massive historical data, this paper obtains the space with better actual operating energy efficiency of the unit through cluster analysis, proposes a new information entropy weight method to allocate attribute weight calculation method to determine the weight of the variable, and uses multivariate autoassociation regression technology to monitor the current energy efficiency status, and proposes a comprehensive evaluation method combining energy efficiency monitoring and fault diagnosis. Summary of the Invention

[0005] The purpose of the present invention is to address the above-mentioned problems existing in the prior art and propose a unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociation regression model, and a fusion data mining model based on autoassociation regression estimation, data clustering and information entropy for energy efficiency monitoring of the unit.

[0006] In order to achieve the above-mentioned object, the present invention adopts the following technical solution: a unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model, which specifically includes the following steps:

[0007] Step 1: Collect historical data of the unit's full operating range and pre-process the historical data; the historical data is represented by the state vector [x j , x m ], where x j is the measured value of the conditional parameter, j is the index of the conditional parameter, j∈[1,m-1], x m is the decision variable, m is the total number of conditional parameters and decision variables;

[0008] Step 2: Divide the pre-processed full operating range into several local operating zones, cluster the condition parameters in each local operating zone with the goal of minimizing the decision variables, and obtain the optimal condition parameter set; divide the condition parameter set into a test set and an operating condition library;

[0009] Step 3: Use the autoassociative regression estimation method to establish a state estimation model for the optimal operating condition area of ​​the test set;

[0010] Step 4: Calculate the initial entropy weight and the secondary entropy weight according to the information entropy weight of the conditional parameters, and determine the weight value of each conditional parameter;

[0011] Step 5: Input the condition parameters during the unit operation into the state estimation model of the optimal operating condition area in real time, predict the estimated value of the condition parameters, calculate the deviation between the estimated value and the measured value, and combine it with the weight value of step 4 to construct the energy efficiency deviation. The energy efficiency deviation is filtered to obtain the sliding average deviation, and the low energy efficiency deviation threshold and the high energy efficiency deviation threshold are determined;

[0012] Step 6: Compare the sliding average deviation with the low energy efficiency deviation threshold and the high energy efficiency deviation threshold, and provide a unit energy efficiency monitoring and diagnosis plan.

[0013] Furthermore, the process of preprocessing historical data in step 1 is specifically as follows:

[0014] Step 11: Determine whether each condition parameter exceeds the threshold value based on the normal operation threshold value of each condition parameter in the historical data. If exceeded, delete the state vector corresponding to the condition parameter.

[0015] Step 12: For the conditional parameters in the retained state vector, determine whether the fluctuation rate of each conditional parameter at adjacent moments is less than a set value. If not, delete the state vector corresponding to the conditional parameter.

[0016] Furthermore, the state estimation model of the optimal operating condition area in step 3 is:

[0017]

[0018] Among them, x est，j is the estimated value of the j-th conditional parameter in the test set, x kj is the measured value of the jth condition parameter in the test set under the kth working condition, n is the total number of condition parameters under each working condition in the test set, q(k) is x kj The weight of , q(k) = ker(x, x′), ker() is the Gaussian kernel function, x is the test set, and x′ is the working condition library.

[0019] Furthermore, step 4 includes the following sub-steps:

[0020] Step 41: Calculate the initial entropy weight of each conditional parameter based on the information entropy of the conditional parameter.

[0021] Step 42: Calculate the secondary entropy weight based on the information entropy and the mean information entropy of the conditional parameters:

[0022] Step 43: According to the initial entropy weight and the secondary entropy weight, the weight value of each condition parameter is determined as

[0023] Among them, H j is the information entropy of the j-th conditional parameter, is the mean information entropy, H k is the information entropy of the kth conditional parameter, k≠j.

[0024] Furthermore, the construction process of the energy efficiency deviation is:

[0025]

[0026] Among them, dist i represents the energy efficiency deviation at the i-th moment, Indicates the weight value of the jth conditional parameter, x est，j is the estimated value of the jth conditional parameter, x j is the measured value of the jth conditional parameter.

[0027] Furthermore, the low energy efficiency deviation threshold is expressed as δ1=3.5σ, and the high energy efficiency deviation threshold is expressed as δ2=6σ, where σ represents the standard deviation of the energy efficiency deviation.

[0028] Furthermore, the process of obtaining the sliding average deviation in step 5 is as follows: filtering the energy efficiency deviation through sliding average filtering, setting the number of sliding averages to 1, then the sliding average deviation Among them, dist i is the energy efficiency deviation at the i-th moment.

[0029] Furthermore, the specific process of step 6 is as follows: when the sliding average deviation is lower than the low energy efficiency deviation threshold, it indicates that the condition parameters are operating reasonably; when the sliding average deviation is between the low energy efficiency deviation threshold and the high energy efficiency deviation threshold, the corresponding condition parameters are combined with the operation adjustment principle and experience to give a corresponding adjustment method; when the sliding average deviation exceeds the high energy efficiency deviation threshold, the corresponding condition parameters are adjusted in operation. If the operation adjustment can improve energy efficiency, the operation adjustment is performed. Otherwise, considering the situation of equipment failure, the corresponding symptom parameters are used in combination with expert diagnostic knowledge to perform corresponding reasoning to locate the faulty equipment.

[0030] Furthermore, when the sliding average deviation exceeds the high energy efficiency deviation threshold, the prerequisite for operating the corresponding condition parameters is: performing sliding average filtering on the deviation between the estimated value and the measured value, setting the number of sliding averages to 1, and obtaining the sliding average deviation If the sliding average deviation exceeds the control threshold of the corresponding condition parameter, the corresponding condition parameter deviates from the optimal operating range and the operating parameter needs to be adjusted; k Represents the deviation between the estimated value and the measured value, and i represents the i-th moment.

[0031] Furthermore, the control threshold of the condition parameter is determined by inputting the test set into the state estimation model of the optimal operating condition area to obtain an estimated value of the condition parameter, and statistically calculating the deviation between the estimated value and the measured value of the condition parameter.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. The steam turbine energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model of the present invention aims to minimize the heat loss of the steam turbine and clusters the optimal condition parameter set, thus avoiding the influence of human subjectivity and having a stronger scientific and theoretical basis.

[0034] 2. The steam turbine energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model of the present invention is based on information entropy theory and proposes a weight calculation method for conditional parameters, which makes the weight difference between important parameters and minor parameters in the conditional parameters more obvious. Combined with autoassociative regression estimation, the model prediction is more reliable.

[0035] 3. The steam turbine energy efficiency monitoring and diagnosis method based on information entropy and autoassociation regression model of the present invention is to meet the needs of real-time monitoring. It uses autoassociation regression estimation model to design a sliding average deviation index for real-time monitoring of energy efficiency indicators and improve the reliability of monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of the unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociation regression model of the present invention;

[0037] Figure 2 This is the effect diagram of parameter estimation under the optimal operating space of a unit with steam turbine heat consumption energy efficiency index, where: Figure 2 (a) is a comparison chart of the estimated and measured main steam pressure values. Figure 2 (b) is a comparison chart of the estimated and measured values ​​of the feed water temperature. Figure 2 (c) is a comparison chart of the estimated and measured condenser vacuum values;

[0038] Figure 3 This is the effect diagram of steam turbine energy efficiency index monitoring, among which, Figure 3 (a) is the deviation warning prompt diagram. Figure 3 (b) Condenser vacuum warning diagram. DETAILED DESCRIPTION

[0039] The technical solution of the present invention will be further explained below with reference to the accompanying drawings.

[0040] like Figure 1 The flowchart of the unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model of the present invention is as follows:

[0041] Step 1: Collect historical data of the unit's full operating range and pre-process the historical data; in the present invention, the historical data is represented by the state vector [x j , x m ], where x j is the measured value of the conditional parameter, j is the index of the conditional parameter, j∈[1,m-1], the conditional parameter is a parameter related to the energy efficiency index, and each conditional parameter has a measured value under each working condition; m is the decision variable, and m is the total number of conditional parameters and decision variables. Taking steam turbine energy efficiency monitoring as an example, the conditional parameters involved include: main steam temperature, reheat temperature, main steam pressure, reheat steam pressure, feedwater temperature, feedwater flow, feedwater pressure, condensate flow, and desuperheating water flow; the decision variable involved is the steam turbine heat rate.

[0042] The process of preprocessing historical data in the present invention is specifically as follows:

[0043] Step 11: Determine whether each condition parameter exceeds the threshold value based on the normal operation threshold value of each condition parameter in the historical data. If exceeded, delete the state vector corresponding to the condition parameter.

[0044] Step 12: For the conditional parameters in the retained state vector, determine whether the fluctuation rate of each conditional parameter at adjacent moments is less than a set value. If not, delete the state vector corresponding to the conditional parameter.

[0045] Due to the complexity of the unit system, the large amount of data with many measurement point parameter types, the high temperature and high pressure on-site production environment, coupled with electromagnetic interference, communication failure and other reasons, a few measurement values ​​may be abnormal. All collected historical data are tested through the above-mentioned threshold test and volatility test methods, and outliers are deleted to ensure the quality of historical data and create favorable conditions for subsequent modeling and analysis.

[0046] Step 2: Divide the pre-processed full operating range into several local operating areas, cluster the condition parameters in each local operating area with the goal of minimizing the decision variables, and obtain the optimal condition parameter set, avoiding the influence of human subjectivity and being more scientific and theoretical; and divide the condition parameter set into a test set and an operating condition library.

[0047] Step 3: Use the autoassociative regression estimation method to establish a state estimation model for the optimal operating condition range for the test set. The autoassociative regression estimation method is a non-parametric estimation method, which eliminates the need for manually selected hyperparameters in the state estimation model and ensures a scientific and reasonable calculation process. The autoassociative regression estimation method essentially uses the optimal set of conditional parameters obtained through clustering as the reference baseline operating condition set for operating state monitoring. The closer the state estimate is to the state vector within the operating condition set, the closer the current state is to, or even belongs to, the optimal operating condition set. Conversely, it indicates that its energy efficiency state has deviated from the optimal range and needs to be adjusted.

[0048] The process of establishing the state estimation model of the optimal operating condition area in the present invention is specifically as follows:

[0049]

[0050] Among them, x est，j is the estimated value of the j-th conditional parameter in the test set, x kj is the measured value of the jth condition parameter in the test set under the kth working condition, n is the total number of condition parameters under each working condition in the test set, q(k) is x kj The weight of , q(k) = ker(x, x′), ker() is the Gaussian kernel function, x is the test set, and x′ is the working condition library.

[0051] Take steam turbine as an example, Figure 2 is the result obtained by calculating the test set using the autoassociative regression estimation method. Figure 2 (a) is a comparison chart of the estimated and measured main steam pressure values. Figure 2 (b) is a comparison chart of the estimated and measured values ​​of the feed water temperature. Figure 2 (c) is a comparison chart of the estimated and measured condenser vacuum values. Figure 2 As can be seen from (a)-(c) in the figure, the estimated values ​​of the above three conditional parameters are the results calculated using autoassociative regression estimation. The estimated values ​​are very close to the measured values ​​with a small deviation, indicating that the test data set belongs to a better working condition and is consistent with the selected results, verifying that the state estimation model is effective.

[0052] Step 4. Considering the different degrees of influence of various conditional parameters on energy efficiency, the importance of different parameters should be objectively considered when calculating energy efficiency related indicators. The information entropy of each conditional parameter should be calculated to determine the corresponding weight. However, when calculating with the traditional entropy weight calculation formula, there will be an unreasonable situation where the entropy value gap is different but the entropy weight calculation value is the same. Therefore, a two-step method for determining entropy weight is designed. The initial entropy weight and the secondary entropy weight are calculated in turn according to the information entropy weight of the conditional parameter. The sum of the two constitutes the total weight value of the parameter, which is conducive to distinguishing the importance of the parameters. It specifically includes the following sub-steps:

[0053] Step 41: Calculate the initial entropy weight of each conditional parameter based on the information entropy of the conditional parameter.

[0054] Step 42: Calculate the secondary entropy weight based on the information entropy and the mean information entropy of the conditional parameters:

[0055] Step 43: According to the initial entropy weight and the secondary entropy weight, the weight value of each condition parameter is determined as

[0056] Among them, H j is the information entropy of the j-th conditional parameter, is the mean information entropy, H k is the information entropy of the kth conditional parameter, k≠j.

[0057] Step 5: Input the conditional parameters during the unit operation into the state estimation model of the optimal operating condition area in real time, predict the estimated value of the conditional parameters, calculate the deviation between the estimated value and the measured value, and combine it with the weight value of step 4 to construct the energy efficiency deviation. The energy efficiency deviation covers the information of all conditional parameters. The energy efficiency deviation within the normal threshold range actually shows that the current state vector belongs to the optimal energy efficiency state space. Once the deviation exceeds the threshold, there must be a certain degree of deviation in the conditional parameters, which is conducive to analysis and adjustment by operators. During the real-time monitoring process, in order to improve the reliability of the deviation index, the energy efficiency deviation is filtered to obtain the sliding average deviation, and the low energy efficiency deviation threshold and high energy efficiency deviation threshold are determined.

[0058] The construction process of the energy efficiency deviation in the present invention is:

[0059]

[0060] Among them, dist i represents the energy efficiency deviation at the i-th moment, Indicates the weight value of the jth conditional parameter, x est，j is the estimated value of the jth conditional parameter, x j is the measured value of the jth conditional parameter.

[0061] The process of sliding average deviation in the present invention is as follows: filtering the energy efficiency deviation by sliding average filtering, setting the number of sliding averages to 1, then the sliding average deviation Among them, dist i is the energy efficiency deviation at the i-th moment.

[0062] In the present invention, the low energy efficiency deviation threshold is expressed as δ1=3.5σ, and the high energy efficiency deviation threshold is expressed as δ2=6σ, where σ represents the standard deviation of the energy efficiency deviation. Here, the reduction in energy efficiency during unit operation is divided into two situations. The first is that the energy efficiency is reduced to a certain extent. The causes include changes in external conditions, disturbances, etc., which are all normal situations. The magnitude of the reduction and fluctuations are generally not too large, and the energy efficiency can be improved as much as possible through operational adjustments; the second reason for the reduction in energy efficiency is caused by equipment or system failures. The energy efficiency may decrease significantly or suddenly, and a comprehensive judgment is needed to determine the cause of the failure to avoid the escalation of the situation. Therefore, setting a low energy efficiency deviation threshold and a high energy efficiency deviation threshold respectively to preliminarily judge the cause of the energy efficiency reduction is conducive to distinguishing different situations and handling them in a timely manner.

[0063] Step 6. Compare the sliding average deviation with the low energy efficiency deviation threshold and the high energy efficiency deviation threshold, and provide a unit energy efficiency monitoring and diagnosis plan. Specifically: when the sliding average deviation is lower than the low energy efficiency deviation threshold, it indicates that the condition parameters are operating reasonably; when the sliding average deviation is between the low energy efficiency deviation threshold and the high energy efficiency deviation threshold, the corresponding condition parameters are combined with the operation adjustment principle and experience to provide a corresponding adjustment method; when the sliding average deviation exceeds the high energy efficiency deviation threshold, the corresponding condition parameters are adjusted in operation. If the operation adjustment can improve energy efficiency, then the operation adjustment is performed. Otherwise, considering the situation of equipment failure, the corresponding symptom parameters are used in combination with expert diagnostic knowledge to perform corresponding reasoning and locate the faulty equipment.

[0064] When the sliding average deviation exceeds the high energy efficiency deviation threshold, the prerequisite for operating the corresponding condition parameters is: perform sliding average filtering on the deviation between the estimated value and the measured value, set the number of sliding averages to 1, and obtain the sliding average deviation If the sliding average deviation exceeds the control threshold of the corresponding condition parameter, the corresponding condition parameter deviates from the optimal operating range and the operating parameter needs to be adjusted; k represents the deviation between the estimated value and the measured value, and i represents the i-th moment. The control threshold of the condition parameter in the present invention is determined by inputting the test set into the state estimation model of the optimal operating condition area to obtain the estimated value of the condition parameter, and then statistically calculating the deviation between the estimated value and the measured value of the condition parameter.

[0065] Figure 3 The effect diagram of the energy efficiency index monitoring of the steam turbine using the energy efficiency monitoring and diagnosis method of the present invention can be seen from the monitoring process of the heat consumption increase caused by the deterioration of the condenser vacuum. Figure 3 As can be seen from (a) in the figure, the energy efficiency deviation index has a high sensitivity. After a certain moment, its value shows an abnormal upward trend, indicating that the current operating state has a low energy efficiency. At this time, the condenser pressure parameter calculated by the model also shows an abnormal increase, such as Figure 3 As shown in the "early warning" in (b), however, the on-site operator screen did not prompt the condenser vacuum alarm at this time. After a period of time, the DCS alarm prompt appeared on the DCS operator screen, which verified that the energy efficiency monitoring of the present invention is effective and has the effect of early warning.

[0066] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociation regression model, characterized in that: The specific steps include: Step 1: Collect historical data of the unit's full operating range and pre-process the historical data; the historical data is represented by the state vector The set consists of , among which, is the measured value of the conditional parameter, j is the index of the conditional parameter, j∈[1,m-1], is the decision variable, m is the total number of conditional parameters and decision variables; Step 2: Divide the pre-processed full operating range into several local operating zones, cluster the condition parameters in each local operating zone with the goal of minimizing the decision variables, and obtain the optimal condition parameter set; divide the condition parameter set into a test set and an operating condition library; Step 3: Use the autoassociative regression estimation method to establish a state estimation model for the optimal operating condition area of ​​the test set; Step 4: Calculate the initial entropy weight and the secondary entropy weight according to the information entropy weight of the conditional parameters, and determine the weight value of each conditional parameter; Step 5: Input the condition parameters during the unit operation into the state estimation model of the optimal operating condition area in real time, predict the estimated value of the condition parameters, calculate the deviation between the estimated value and the measured value, and combine it with the weight value of step 4 to construct the energy efficiency deviation. The energy efficiency deviation is filtered to obtain the sliding average deviation, and the low energy efficiency deviation threshold and the high energy efficiency deviation threshold are determined; Step 6: Compare the sliding average deviation with the low energy efficiency deviation threshold and the high energy efficiency deviation threshold to provide a unit energy efficiency monitoring and diagnosis plan. The specific process is as follows: When the sliding average deviation is lower than the low energy efficiency deviation threshold, it indicates that the condition parameters are operating reasonably. When the sliding average deviation is between the low energy efficiency deviation threshold and the high energy efficiency deviation threshold, the corresponding condition parameters are adjusted according to the operation adjustment principles and experience. When the sliding average deviation exceeds the high energy efficiency deviation threshold, the corresponding condition parameters are adjusted. If the operation adjustment can improve energy efficiency, the operation adjustment is performed. Otherwise, considering the possibility of equipment failure, the corresponding symptom parameters are used in combination with expert diagnostic knowledge to perform corresponding reasoning and locate the faulty equipment. When the sliding average deviation exceeds the high energy efficiency deviation threshold, the prerequisite for operating the corresponding condition parameters is: perform sliding average filtering on the deviation between the estimated value and the measured value, and set the number of sliding averages to l Sliding mean deviation If the sliding mean deviation exceeds the control threshold of the corresponding condition parameter, the corresponding condition parameter deviates from the optimal operating range and the operating parameter needs to be adjusted; represents the deviation between the estimated value and the measured value, i Indicates the i time.

2. The unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model according to claim 1 is characterized in that: The specific process of preprocessing historical data in step 1 is as follows: Step 11: Determine whether each condition parameter exceeds the threshold value based on the normal operation threshold value of each condition parameter in the historical data. If exceeded, delete the state vector corresponding to the condition parameter. Step 12: For the conditional parameters in the retained state vector, determine whether the fluctuation rate of each conditional parameter at adjacent moments is less than a set value. If not, delete the state vector corresponding to the conditional parameter.

3. The unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model according to claim 1 is characterized in that: The state estimation model of the optimal operating condition area in step 3 is: in, For the test set j The estimated values ​​of the conditional parameters, For the test set j The measured value of the conditional parameter under the kth working condition, n is the total amount of conditional parameters under each working condition in the test set, for The weight of , ker () is the Gaussian kernel function, For the test set, For the working condition library.

4. The steam turbine energy efficiency monitoring and diagnosis method based on information entropy and autoassociation regression model according to claim 1 is characterized in that: Step 4 includes the following sub-steps: Step 41: Calculate the initial entropy weight of each conditional parameter based on the information entropy of the conditional parameter. , Step 42: Calculate the secondary entropy weight based on the information entropy and the mean information entropy of the conditional parameters: , Step 43: According to the initial entropy weight and the secondary entropy weight, the weight value of each condition parameter is determined as , Among them, H j For the j The information entropy of the conditional parameters, is the mean information entropy, For the k The information entropy of the conditional parameters, k≠j.

5. The unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociation regression model according to claim 1 is characterized in that: The construction process of the energy efficiency deviation is as follows: in, Indicates the i The energy efficiency deviation at a moment, Indicates the j The weight value of the conditional parameter, For the j The estimated value of the conditional parameters, For the j The measured value of a conditional parameter.

6. The unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model according to claim 1 is characterized in that: The low energy efficiency deviation threshold is expressed as , the high energy efficiency deviation threshold is expressed as ,in, Indicates the standard deviation of energy efficiency deviation.

7. The unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model according to claim 1 is characterized in that: The process of obtaining the sliding average deviation in step 5 is as follows: filtering the energy efficiency deviation by sliding average filtering, setting the number of sliding averages to l The sliding mean deviation ,in, For the i The energy efficiency deviation at each moment.

8. The unit energy efficiency monitoring and diagnosis method based on information entropy and autoassociative regression model according to claim 1 is characterized in that: The control threshold of the condition parameter is determined by inputting the test set into the state estimation model of the optimal operating condition area to obtain an estimated value of the condition parameter, and statistically analyzing the deviation between the estimated value and the measured value of the condition parameter.

Images