Oilfield heating furnace abnormality detection method and system and related equipment
Through the automatic training of semi-supervised self-coding network and isolated forest algorithm, the high cost and low accuracy of abnormal data detection of oilfield heating furnace sensors is solved, and more efficient abnormal data recognition and normal operation of the heating furnace are achieved.
Patent Information
- Application Number
- CN202010910892.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-09-02
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2040-09-02
AI Technical Summary
The sensor abnormal data detection of oilfield heating furnaces has problems such as high manual operation and maintenance costs and low detection accuracy. Especially due to the system sealing and the differences between sensors of different heating furnaces, it is difficult to achieve effective abnormal detection.
An abnormal detection is performed using an automatic training semi-supervised self-coding network. By acquiring and dividing the operating data of the time period, using an isolated forest algorithm to filter and self-coding network to train prediction models, abnormal detection is performed on the current data, reducing manual maintenance costs and improving detection Accuracy.
It has achieved the reduction of manual operation and maintenance costs, improved the accuracy of abnormal data detection, ensured the normal operation of the heating furnace, higher data extraction accuracy and recognition, and stronger algorithm adaptability.
Smart Images

Figure CN111967532B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of abnormality detection of oilfield heating furnaces, and in particular to an abnormality detection method and system for oilfield heating furnaces and related equipment. Background Art
[0002] The oilfield heating furnace is the most widely used oilfield special equipment in the processing and transportation links of the oil and gas gathering and transportation system. Its function is to heat crude oil, natural gas, oil and gas mixtures to the temperature required by the process to meet the requirements of oil and gas gathering and transportation technology and processing technology. It is a very important oilfield production facility. The sensor abnormal data detection of the oilfield heating furnace plays an important role in the normal operation of the heating furnace. Because the oilfield system is closed and cannot be remotely controlled, it is difficult to mark data, the algorithm maintenance cost is high, and the sensors of different heating furnaces are also different, which poses a challenge to the abnormal detection of oilfield heating furnaces. Summary of the invention
[0003] The purpose of the present invention is to provide an oilfield heating furnace anomaly detection method, which uses an automatically trained semi-supervised autoencoder network for anomaly detection, reduces manual operation and maintenance costs, improves the accuracy of abnormal data detection, and ensures the normal operation of the heating furnace. The purpose of the present invention is also to provide a corresponding oilfield heating furnace anomaly detection system and related equipment.
[0004] To achieve the above purpose, the technical solution adopted by the present invention is as follows.
[0005] In a first aspect, a method for detecting abnormalities in an oilfield heating furnace is provided, comprising: obtaining operating data of the oilfield heating furnace in the previous time period, and dividing the operating data according to time periods to obtain at least two data sets belonging to different time periods; filtering each data set using an isolation forest algorithm to obtain a set of training data; using each set of training data to train an autoencoder network to obtain prediction models corresponding to different time periods; for the current operating data to be detected, using the prediction model of the corresponding time period to perform abnormality detection
[0006] In one possible implementation, the use of a prediction model for a corresponding time period for anomaly detection includes: setting different detection standards for operating data in different time periods; using the prediction model corresponding to the time period to which the current operating data belongs to predict the current operating data to obtain a difference value; when the difference value exceeds the detection standard for the corresponding time period, confirming that the current operating data is abnormal.
[0007] In a possible implementation, the filtering using the isolation forest algorithm for each data set includes: for each data set, further dividing it into two subsets according to the efficiency of the heating furnace; for the subset with low efficiency, using the isolation forest algorithm to remove edge data according to a first ratio; for the subset with high efficiency, using the isolation forest algorithm to remove edge data according to a second ratio; wherein the first ratio is greater than the second ratio.
[0008] In a possible implementation, the operating data is divided according to time periods to obtain at least two data sets belonging to different time periods, including: based on the large difference in data between day and night, the operating data of several days in the previous time period are classified according to day and night, and divided into a data set S1 belonging to the day and a data set S2 belonging to the night.
[0009] In a possible implementation, the method of using each set of training data to train the autoencoder network to obtain prediction models corresponding to different time periods includes: using normalized training data to train the autoencoder network to obtain the corresponding prediction model, and using the Hyperopt algorithm to tune the parameters of the prediction model.
[0010] In a second aspect, an oilfield heating furnace anomaly detection system is provided, comprising: an acquisition module, used to acquire the operating data of the oilfield heating furnace in the previous time period, and divide the operating data according to time periods to obtain at least two data sets belonging to different time periods; a processing module: used to filter each data set using an isolation forest algorithm to obtain a set of training data; a training module: used to train an autoencoder network using each set of training data to obtain prediction models corresponding to different time periods; a testing module: used to perform anomaly detection on the current operating data to be detected using the prediction model of the corresponding time period.
[0011] In one possible implementation, the test module is specifically used to: set different detection standards for operating data in different time periods; use the prediction model corresponding to the time period to which the current operating data belongs to predict the current operating data to obtain a difference value; when the difference value exceeds the detection standard of the corresponding time period, confirm that the current operating data is abnormal.
[0012] In a possible implementation, the processing module is specifically used to: for each data set, further divide it into two subsets according to the efficiency of the heating furnace; for the subset with low efficiency, use the isolation forest algorithm to remove edge data according to a first ratio; for the subset with high efficiency, use the isolation forest algorithm to remove edge data according to a second ratio; wherein the first ratio is greater than the second ratio.
[0013] In a third aspect, a computer device is provided, comprising a processor and a memory, wherein the memory stores a program, the program comprising computer execution instructions, and when the computer device is running, the processor executes the computer execution instructions stored in the memory, so that the computer device executes the heating furnace abnormality detection method as described in the first aspect.
[0014] In a fourth aspect, a computer-readable storage medium storing one or more programs is provided, wherein the one or more programs include computer execution instructions, and when the computer execution instructions are executed by a computer device, the computer device executes the heating furnace abnormality detection method as described in the first aspect.
[0015] It can be seen from the above technical solutions that the embodiments of the present invention have the following advantages:
[0016] Through automated training, that is, automatically acquiring training data for retraining and adjusting algorithm parameters according to time periods, the manual maintenance cost of the model can be reduced, the accuracy of abnormal data detection can be improved, and the normal operation of the heating furnace can be ensured.
[0017] By dividing the operating data into data sets and using the isolation forest algorithm to remove edge data, training data and training models are extracted and tested according to the data sets, so that the data extraction accuracy is improved, thereby making the recognition of abnormal data of the heating furnace higher and the application more adaptable. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments and 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.
[0019] Figure 1 is a flow chart of an oilfield heating furnace abnormality detection method provided by an embodiment of the present invention;
[0020] Figure 2 is a structural diagram of an autoencoding network in an embodiment of the present invention;
[0021] Figure 3 is a flow chart of model training in an embodiment of the present invention;
[0022] Figure 4 is a schematic diagram of an automatic training process in an embodiment of the present invention;
[0023] Figure 5 is a structural diagram of an oilfield heating furnace abnormality detection system provided by an embodiment of the present invention;
[0024] Figure 6 It is a structural diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.
[0026] The terms "first", "second", "third", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.
[0027] The following describes the embodiments in detail.
[0028] Please refer to Figure 1 One embodiment of the present invention provides an oilfield heating furnace anomaly detection method. The method uses an automatically trained semi-supervised autoencoder network for anomaly detection, reduces manual operation and maintenance costs, improves the accuracy of abnormal data detection, and ensures the normal operation of the heating furnace.
[0029] The method may include the following steps:
[0030] Step S1, extracting data: obtaining the operating data of the oilfield heating furnace in the previous time period, and dividing the operating data according to time periods to obtain at least two data sets belonging to different time periods; and filtering each data set using an isolation forest algorithm to obtain a set of training data;
[0031] Step S2, model training: using each set of training data to train the autoencoder network respectively to obtain prediction models corresponding to different time periods;
[0032] Step S3, model testing: for the current running data to be tested, anomaly detection is performed using the prediction model of the corresponding time period.
[0033] Optionally, in some implementations, the operation data is divided according to time periods to obtain at least two data sets belonging to different time periods, which may include: based on the data differences between different time periods, for example, the data differences between daytime and nighttime are relatively large, the operation data of several days of the previous time period may be classified according to daytime and nighttime, and divided into data set S1 belonging to daytime and data set S2 belonging to nighttime. For example, the time period to which data set S1 belongs may be from 8 am to 8 pm, and the time period to which data set S2 belongs may be from 9 pm to 7 am the next day, but it should be understood that this time division is only an example, and the specific division may be based on needs.
[0034] Optionally, in some implementations, the filtering using the Isolation Forest algorithm for each data set may include: for each data set, further dividing it into two subsets according to the efficiency of the heating furnace; for example, dividing the data with an efficiency higher than 85% into one subset, and dividing the data with an efficiency not higher than 85% into another subset; for the subset with low efficiency (for example, not higher than 85%), using the Isolation Forest algorithm to remove edge data (i.e., outlier data) according to a first ratio; for the subset with high efficiency (for example, higher than 85%), using the Isolation Forest algorithm to remove edge data (i.e., outlier data) according to a second ratio; wherein the first ratio is greater than the second ratio, for example, the first ratio is 30%, and the second ratio is 10%.
[0035] Optionally, in some implementations, the use of the prediction model for the corresponding time period for anomaly detection may include: setting different detection standards for operating data in different time periods; using the prediction model corresponding to the time period to which the current operating data belongs to predict the current operating data to obtain a difference value; when the difference value exceeds the detection standard for the corresponding time period, confirming that the current operating data is abnormal. The difference value may be the absolute value of the difference between the predicted value (or the predicted result) and the original data. For example, different detection standards may be set for operating data during the day and at night, which may be denoted as mae1 and mae2, and mae1 and mae2 may be set based on empirical data.
[0036] Optionally, in some implementations, the method of using each group of training data to train the autoencoder network to obtain prediction models corresponding to different time periods includes: using normalized training data to train the autoencoder network to obtain the corresponding prediction model, and using the Hyperopt algorithm to tune the parameters of the prediction model.
[0037] The method of this embodiment is further described below in conjunction with an embodiment in a specific application scenario.
[0038] 1. Extract data when the heating furnace is operating normally.
[0039] 1. Because the data in the oil field operation area is quite different between morning and evening, the operation data that needs to detect abnormal features can be divided into two data sets S1 and S2. For example, S1 can be the data from 8 am to 8 pm, and S2 can be the data from 9 pm to 7 am the next day.
[0040] 2. Since heating furnace anomalies often occur in lower furnace efficiency areas, each data set in S1 and S2 can be divided into two subsets, for example, one subset has furnace efficiency below 85% and the other subset has furnace efficiency above 85%.
[0041] 3. Use the Isolation Fortest algorithm to remove marginal data from the datasets S1 and S2. For the subset with furnace efficiency below 85%, 30% of the data with larger anomaly scores can be filtered out, and for the subset with furnace efficiency above 85%, 10% of the data with larger anomaly scores can be filtered out. The remaining running data can be used as autoencoder training data.
[0042] The Isolation Fortest algorithm can be divided into two stages. The first stage is to train isolated trees to form an isolation forest. The second stage is to bring each sample point (running data) into each isolated tree in the isolation forest, calculate the average height, and then calculate the anomaly score of each sample point.
[0043] The steps include:
[0044] 1) X = {x1, ....., xn} is a given data set. For any xi∈X, xi = (xi1, ....., xid), a subset X consisting of ψ sample points is randomly selected from it and placed in the root node.
[0045] 2) Randomly specify a dimension q from the d dimensions and randomly generate a cutting point p in the current data.
[0046] 3) This cutting point p generates a hyperplane that divides the current data space into two subspaces: sample points with a specified dimension less than p are placed in the left child node, and those greater than or equal to p are placed in the right child node.
[0047] min(x ij , j = q, x ij ∈X′)<p<max(x ij , j = q, x ij ∈X′)
[0048] 4) Recursively repeat steps 2) and 3) until all leaf nodes have only one sample point or the isolated tree has reached the specified height.
[0049] 5) Repeat steps 1) to 4) until t isolated trees (iTree) are generated.
[0050] 6) For each data point xi, traverse each isolated tree, calculate the average height h(xi) of the point in the forest, normalize the average height of all points, and calculate the outlier score using the formula:
[0051]
[0052] 4. Finally, the two sets of training data (still recorded as S1 and S2) can be normalized. The normalization formula can be: Min is the smallest value in this column of features, and max is the largest value in this column of features.
[0053] 2. The structure of the model.
[0054] 1. Please refer to Figure 2 , is the structure of the autoencoder network, including an encoder and a decoder.
[0055] 2. The encoder consists of LSTM1, Layer1, Layer2 and other layers. Among them, LSTM1 (connected to input) is used to extract time features, and Layer1 and Layer2 are used to compress the extracted features to the output bottleneck.
[0056] 3. The decoder consists of Layer3, Layer4, LSTM2 and other layers, among which Layer3, Layer4, LSTM2 (connecting input and output) are used to restore the output layer bottleneck.
[0057] 3. Model training.
[0058] 1. Please refer to Figure 3 , is the flowchart of model training.
[0059] 2. The total number of features is n, and the selected neuron parameters of the model change with the number of features.
[0060] 3. The LSTM1 layer is set to 2*n or 1.8*n neurons, dropout = 0.2; the Layer1 layer is set to 1.6*n or 1.4*n neurons; the Layer2 layer is set to 1.2*n or n neurons.
[0061] 4. The bottleneck layer is set to 0.8*n or 0.6*n neurons.
[0062] 5. Layer3 is set to 1.2*n or n neurons; Layer4 is set to 1.6*n or 1.4*n neurons; LSTM2 is set to 2*n or 1.8*n neurons, dropout = 0.2.
[0063] 6. Configure the model's learning process compile. The learning process compile parameters include optimizer optimizer, loss function loss, evaluation criteria metrics, optimizer optimizer, and loss function loss.
[0064] 7. Set the batch size batch_size to between 1-100, the number of iterations epoch to between 1-10000, the optimizer optimizer to 'Adam', 'RMSprop', 'Adagrad', and the loss function loss to 'mse'.
[0065] 8. Finally, the Hyperopt algorithm can be used to search for the optimal parameters. The number of iterations is, for example, 50 times, to optimize the parameters of the trained autoencoder network and obtain the optimized prediction model.
[0066] 4. Automatic model training.
[0067] 1. Please refer to Figure 4 , is a schematic diagram of the automatic training process.
[0068] 2. You can set a timer on the Windows or Linux system to start running the model file once a month according to the set time period. The model re-reads data (operation data of the previous time period), trains, and adjusts parameters every month to keep synchronization with the field data.
[0069] V. Effect of automatically trained autoencoder network for anomaly detection.
[0070] 1. If the current running data a belongs to daytime, corresponding to S1 category, the prediction result is a', and the abnormal value is calculated as the absolute value of a'-a. If the abnormal value is greater than mae1, it is abnormal data, otherwise it is normal.
[0071] If the current running data a belongs to the business, corresponding to the S2 category, the abnormal value is calculated as the absolute value of a'-a. If the abnormal value is greater than mae2, it is abnormal data, otherwise it is normal.
[0072] Among them, mae1 and mae2 are detection standards corresponding to different time periods.
[0073] 2. Verification: One month's data from the previous cycle is used for training, and another month's data is used for prediction. Anomaly detection is performed for half a year, and the final AUC (Area Under Curve) can be around 95%.
[0074] 3. It should be noted that the abnormal data mentioned in this article include some comprehensive anomalies such as abnormal gas flow, abnormal oxygen content, sudden changes in some sensor points, and abnormal furnace conditions.
[0075] The above is a description of the oilfield heating furnace anomaly detection method provided by the embodiment of the present invention. The key to this method lies in the entire model training and testing process, as well as the data processing process in the data extraction stage before training.
[0076] To facilitate the implementation of the present invention, corresponding systems and related devices are also provided below.
[0077] Please refer to Figure 5 One embodiment of the present invention provides an oilfield heating furnace abnormality detection system, the system comprising:
[0078] The acquisition module 51 is used to acquire the operation data of the oilfield heating furnace in the previous time period, and divide the operation data according to time periods to obtain at least two data sets belonging to different time periods;
[0079] Processing module 52: for filtering each data set using an isolation forest algorithm to obtain a set of training data;
[0080] Training module 53: used to train the autoencoder network using each set of training data to obtain prediction models corresponding to different time periods;
[0081] Testing module 54: used to perform anomaly detection on the current operating data to be detected using the prediction model of the corresponding time period.
[0082] Optionally, the test module 54 is specifically used to: set different detection standards for operating data in different time periods; use the prediction model corresponding to the time period to which the current operating data belongs to predict the current operating data to obtain a difference value; when the difference value exceeds the detection standard of the corresponding time period, confirm that the current operating data is abnormal.
[0083] Optionally, the processing module 52 is specifically used to: for each data set, further divide it into two subsets according to the efficiency of the heating furnace; for the subset with low efficiency, use the isolation forest algorithm to remove edge data according to a first ratio; for the subset with high efficiency, use the isolation forest algorithm to remove edge data according to a second ratio; wherein the first ratio is greater than the second ratio.
[0084] Please refer to Figure 6According to one embodiment of the present invention, a computer device 60 is provided, including a processor 61 and a memory 62. The memory 62 stores a program, and the program includes computer execution instructions. When the computer device 60 is running, the processor 61 executes the computer execution instructions stored in the memory 62, so that the computer device 60 executes the heating furnace abnormality detection method as described above.
[0085] One embodiment of the present aspect also provides a computer-readable storage medium storing one or more programs, wherein the one or more programs include computer execution instructions, and when the computer execution instructions are executed by a computer device, the computer device executes the heating furnace abnormality detection method as described above.
[0086] In summary, the oilfield heating furnace abnormality detection method and system and related equipment provided by the embodiment of the present invention are described. From the above technical scheme, it can be seen that the embodiment of the present invention has the following advantages:
[0087] Through automated training, that is, automatically acquiring training data for retraining and adjusting algorithm parameters according to time periods, the manual maintenance cost of the model can be reduced, the accuracy of abnormal data detection can be improved, and the normal operation of the heating furnace can be ensured.
[0088] By dividing the operating data into data sets according to time periods, for different data sets: the isolation forest algorithm is used to remove edge data, extract training data, train the autoencoder network, and obtain the corresponding prediction model; and the prediction model of the corresponding time period is used to detect anomalies in the current operating data; so that: the data extraction accuracy is improved, the recognition of abnormal data of the heating furnace is higher, and the algorithm application adaptability is stronger.
[0089] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0090] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. A person skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for detecting abnormality of an oilfield heating furnace, characterized in that: include: Acquire the operation data of the oilfield heating furnace in the previous time period, and divide the operation data according to time periods to obtain at least two data sets belonging to different time periods; For each data set, the isolation forest algorithm is used to filter and obtain a set of training data; including: For each data set, it is further divided into two subsets according to the efficiency of the heating furnace; For the subset with low furnace efficiency, the isolation forest algorithm is used to remove marginal data according to the first ratio; For the subset with high furnace efficiency, the isolation forest algorithm is used to remove marginal data according to the second ratio; wherein the first ratio is greater than the second ratio; Use each set of training data to train the autoencoder network to obtain the prediction models corresponding to different time periods; including: Use the normalized training data to train the autoencoder network to obtain the corresponding prediction model, and use the Hyperopt algorithm to tune the parameters of the prediction model; Specifically: The prediction model includes two parts: an encoder and a decoder; The encoder includes LSTM1 and a first fully connected layer; wherein the first fully connected layer includes Layer1 and Layer2, which are used to extract and compress features of input data; The decoder includes LSTM2 and a second fully connected layer; wherein the second fully connected layer includes Layer3 and Layer4, which are used to restore the output of the encoder; Setting the number of neurons in each layer of the prediction model, wherein the number of neurons is dynamically adjusted according to the number of features of the input data; Using the processed data as training data, the prediction model is trained through an optimizer and a loss function; wherein the optimizer includes Adam, RMSprop, and Adagrad; and the loss function includes mean square error (mse); For the current running data to be detected, use the prediction model of the corresponding time period to perform anomaly detection; including: Set different testing standards for operating data in different time periods; Use the prediction model corresponding to the time period to which the current operation data belongs to predict the current operation data to obtain the difference value; When the difference value exceeds the detection standard of the corresponding time period, it is confirmed that the current operating data is abnormal.
2. The method according to claim 1, characterized in that The operation data is divided according to time periods to obtain at least two data sets belonging to different time periods, including: Since the data difference between daytime and nighttime is large, the operation data of several days in the previous time period are classified into daytime and nighttime data sets S1 belonging to daytime and S2 belonging to nighttime.
3. An oilfield heating furnace abnormality detection system, characterized in that: include: Acquisition module: used to acquire the operation data of the oilfield heating furnace in the previous time period, and divide the operation data according to time periods to obtain at least two data sets belonging to different time periods; Processing module: used to filter each data set using the isolation forest algorithm to obtain a set of training data; Training module: used to train the autoencoder network using each set of training data to obtain prediction models corresponding to different time periods; including: Use the normalized training data to train the autoencoder network to obtain the corresponding prediction model, and use the Hyperopt algorithm to tune the parameters of the prediction model; Specifically: The prediction model includes two parts: an encoder and a decoder; The encoder includes LSTM1 and a first fully connected layer; wherein the first fully connected layer includes Layer1 and Layer2, which are used to extract and compress features of input data; The decoder includes LSTM2 and a second fully connected layer; wherein the second fully connected layer includes Layer3 and Layer4, which are used to restore the output of the encoder; Setting the number of neurons in each layer of the prediction model, wherein the number of neurons is dynamically adjusted according to the number of features of the input data; Using the processed data as training data, the prediction model is trained through an optimizer and a loss function; wherein the optimizer includes Adam, RMSprop, and Adagrad; and the loss function includes mean square error (mse); Testing module: used to perform anomaly detection on the current operating data to be detected using the prediction model of the corresponding time period; the testing module is specifically used to set different detection standards for the operating data of different time periods; Use the prediction model corresponding to the time period to which the current operation data belongs to predict the current operation data to obtain the difference value; When the difference value exceeds the detection standard of the corresponding time period, it is confirmed that the current operating data is abnormal.
4. The system according to claim 3, characterized in that The processing module is specifically used to: for each data set, further divide it into two subsets according to the efficiency of the heating furnace; For the subset with low furnace efficiency, the isolation forest algorithm is used to remove marginal data according to the first ratio; For the subset with high furnace efficiency, the isolation forest algorithm is used to remove marginal data according to the second ratio; The first ratio is greater than the second ratio.
5. A computer device, comprising a processor and a memory, wherein the memory stores a program, the program comprising computer execution instructions, and when the computer device is running, the processor executes the computer execution instructions stored in the memory so that the computer device executes the heating furnace abnormality detection method as described in claim 1.
6. A computer-readable storage medium storing one or more programs, wherein the one or more programs include computer-executable instructions, and when the computer-executable instructions are executed by a computer device, the computer device executes the heating furnace abnormality detection method according to claim 1.
