Cold end detection method based on DCS

Abstract

The invention relates to the field of cold end intelligent diagnosis, and discloses a DCS-based cold end detection method, which comprises the following steps: acquiring DCS-based cold end data including cold end historical data and cold end real-time data; cold end data processing: the offline data processing end adds an instantaneous timestamp to the obtained cold end real-time data to obtain an instantaneous data set; and generating a cold end detection report, inputting the instantaneous data set into the intelligent diagnosis model by the offline data processing end, and outputting a cold end diagnosis result comprising an instantaneous timestamp. According to the method, the intelligent diagnosis model of the offline data processing end is applied to the cold end system, so that the prediction accuracy is improved, the model performance is continuously optimized through a checking calculation mechanism and machine learning, the dependence on a communication network is reduced, and the situation that the operation of the whole control system is influenced due to network faults or network instability is avoided.

Description

Technical Field

[0001] The present invention relates to the field of cold end intelligent diagnosis, and in particular to a cold end detection method based on DCS. Background Art

[0002] The cold end system is an important part of the generator set, including the low-pressure cylinder body and its heat recovery steam extraction system, exhaust port and condenser system, circulating water system, sealing system and vacuum system, cooling tower system and other subsystems. The core purpose of this system is to convert the exhaust gas of the steam turbine into condensed water, so that the power generation process can heat the water into steam again to achieve continuous power generation. The operating quality of the cold end system is directly related to the vacuum degree of the generator set, and the vacuum degree or back pressure of the steam turbine of a large generator set has a significant impact on its energy consumption level and economy. Research data shows that when the back pressure exceeds the critical back pressure, the coal consumption of power generation will increase by 2-3g / KWh for every 1KPa increase in back pressure. Therefore, the operating goal of the cold end system is to strive to achieve the best vacuum degree or the lowest back pressure under certain circulating water system power consumption conditions.

[0003] DCS (distributed control system) has certain applications in the field of industrial cold end. Due to its distributed structure, the communication of DCS system has become an issue that cannot be ignored, which may affect the overall performance and stability of the system. Since the DCS system is highly dependent on the communication network, network failure or network instability may affect the operation of the entire control system. Summary of the invention

[0004] The present invention aims to provide a cold end detection method based on DCS, which performs intelligent detection by collecting and analyzing data.

[0005] To achieve the above object, the present invention adopts the following technical solution: comprising the following steps:

[0006] Acquire cold end data based on DCS, wherein the cold end data includes cold end historical data and cold end real-time data;

[0007] Cold-end data processing: the offline data processing end adds an instantaneous timestamp to the acquired cold-end real-time data to obtain an instantaneous data set;

[0008] Generate a cold end detection report. The offline data processing end inputs the instantaneous data set into the intelligent diagnosis model and outputs the cold end diagnosis results including the instantaneous timestamp.

[0009] Preferably, as an improvement, the intelligent diagnosis model is based on a VAR model, and through the input of an instantaneous data set, the intelligent diagnosis model performs trend prediction on the cold end operating condition to obtain a prediction result.

[0010] Preferably, as an improvement, the cold end data includes the rotation speed of the enhanced heat exchange device, the inlet and outlet water temperatures of the condenser, and the real-time flow rate of the circulating water; the offline data processing end verifies the prediction results, and the verification compares the deviation of the prediction results with the corresponding instantaneous data set.

[0011] Preferably, as an improvement, a mapping relationship is formed between the prediction result and the instantaneous data set of the cold end through verification at the offline data processing end, and the instantaneous data set of the cold end is used to perform machine learning on the intelligent diagnosis model.

[0012] Preferably, as an improvement, the cold end diagnosis result output by the offline data processing includes a verification prediction result, which is a prediction result obtained by the offline data processing through cold end historical data and cold end real-time data.

[0013] Preferably, as an improvement, the instantaneous data set includes Y1 at time point t: real-time flow of circulating water; Y2: heat load of condenser; and Y3: efficiency of cooling tower.

[0014] Preferably, as an improvement, the prediction result at time t+1 is a prediction value calculated by using the instantaneous data sets at time t and time t-1.

[0015] Preferably, as an improvement, the instantaneous data set at time t+1 is used to remove abnormal points in the instantaneous data sets at time t and time t-1.

[0016] Preferably, as an improvement, the instantaneous data set at time t+1 is used to correct the abnormal points in the instantaneous data sets at time t and time t-1.

[0017] The principle and advantages of this solution are: by using the time series characteristics of the cold end data, the future trend of the cold end working condition can be accurately predicted. This predictive ability can receive early warning before potential problems occur, so that there is enough time to take preventive measures to avoid or reduce the losses caused by failures. This not only improves the stability of the cold end system, but also significantly reduces maintenance costs. The accuracy of the model prediction is verified by comparing it with the actual instantaneous data set. This verification mechanism is the key to the self-correction and optimization of the model. When it is found that the prediction results deviate from the actual data, its parameters or algorithms can be adjusted in time to ensure that future predictions are more accurate. This continuous optimization process enables the model to continuously adapt to changes in the cold end system and maintain the effectiveness of its diagnostic capabilities. By mapping the prediction results with the instantaneous data set and using this data for machine learning, the intelligent diagnosis model can continuously improve its accuracy and generalization ability. This self-learning and optimization mechanism enables the model to better reflect the complexity and dynamics of the cold end system, so that it can make more accurate diagnoses when faced with new situations or unknown faults. This generalization ability is crucial to improving the overall operating efficiency and reliability of the cold end system.

[0018] In summary, the application of the intelligent diagnosis model of the offline data processing end in the cold-end system not only improves the accuracy of the prediction, but also continuously optimizes the model performance through verification mechanism and machine learning, reduces the dependence on the communication network, and avoids the operation of the entire control system due to network failure or network instability. DETAILED DESCRIPTION

[0019] The following is further described in detail through specific implementation methods:

[0020] The specific implementation steps are as follows:

[0021] A cold end detection method based on DCS includes the following steps:

[0022] Acquire cold end data based on DCS, wherein the cold end data includes cold end historical data and cold end real-time data;

[0023] Cold-end data processing: The offline data processing end adds an instantaneous timestamp to the acquired cold-end real-time data to obtain an instantaneous data set. The addition of a timestamp gives the data a time dimension, which facilitates subsequent trend analysis and prediction.

[0024] Generate a cold end detection report. The offline data processing end inputs the instantaneous data set into the intelligent diagnosis model and outputs the cold end diagnosis results including the instantaneous timestamp.

[0025] Among them, the cold end data is obtained based on the DCS (distributed control system), including cold end historical data and cold end real-time data. These data cover key parameters such as the speed of the enhanced heat exchange device, the inlet and outlet water temperature of the condenser, and the real-time flow rate of circulating water.

[0026] The intelligent diagnosis model based on VAR (vector autoregression) model is used to capture the dynamic relationship between variables. The instantaneous data set is input into the intelligent diagnosis model, and the model predicts the trend of the cold end working condition and obtains the prediction results for a period of time in the future. These prediction results help to detect potential problems in advance.

[0027] The offline data processing end verifies the prediction results by comparing the deviation of the prediction results with the corresponding instantaneous data set. For example, the difference between the predicted temperature and the actual temperature, or the percentage error between the predicted flow and the actual flow, can be calculated. If a large deviation is found between the prediction results and the actual data, it means that the model may need to be adjusted or optimized.

[0028] Through verification, the prediction results are mapped to the instantaneous data set of the cold end. This mapping relationship reflects the correspondence between the model prediction and the actual situation, providing valuable data support for subsequent machine learning. These instantaneous data sets with timestamps and verification results are used to perform machine learning on the intelligent diagnosis model. By continuously adjusting the model parameters or algorithms, the model can more accurately capture the dynamic characteristics of the cold end system and improve the accuracy of prediction and diagnosis.

[0029] For example, in the cold end system of a power plant, DCS collects data such as the speed of the enhanced heat exchange device, the inlet and outlet water temperature of the condenser, and the real-time flow rate of circulating water in real time. After receiving this data, the offline data processing end adds a timestamp accurate to the second to each piece of data, forming an instantaneous data set.

[0030] Subsequently, the intelligent diagnosis model analyzed these data based on the VAR model and predicted the changing trend of the condenser inlet and outlet water temperatures in the next 24 hours. The prediction results show that the condenser outlet water temperature will gradually increase in the next few hours. In order to verify the accuracy of this prediction result, the offline data processing end verified the prediction result and found that the deviation between the predicted temperature and the actual temperature was within an acceptable range. Therefore, according to the prediction results, the fan speed of the cooling tower was adjusted in advance to cope with the upcoming increase in the condenser outlet water temperature.

[0031] In this process, instantaneous data sets with timestamps and verification results are also used for machine learning of the intelligent diagnosis model. By continuously adjusting the model parameters, the prediction accuracy of the model has been further improved, providing strong support for the stable operation of the cold end system of the power plant.

[0032] For example, to obtain the cold end real-time data, the following cold end real-time data is obtained at a certain moment:

[0033] Circulating water real-time flow rate: 1500m 3 / h;

[0034] Speed ​​data of enhanced heat exchange device: 3000rpm.

[0035] The data transmission end sends the above cold end real-time data to offline data processing. After receiving the data, the offline data processing adds an instantaneous timestamp to each data (assuming it is 2024-12-12T16:20:00Z) to obtain the instantaneous data set:

[0036]

[0037] Generate cold end diagnosis results: Offline data processing inputs the instantaneous data set into the intelligent diagnosis model based on the VAR model. The model predicts the trend of the cold end working condition. Assuming that the prediction result is that the real-time flow rate of circulating water will rise to 1550m in the next 10 minutes, the cold end working condition will be predicted. 3 / h, the speed of the enhanced heat exchanger will remain unchanged. Offline data processing outputs the prediction results together with the instantaneous timestamp (assuming it is 2024-12-12T16:20:00Z) to form the cold end diagnosis results:

[0038]

[0039] Verify the prediction results: Offline data processing verifies the prediction results, that is, compares the prediction results with the corresponding instantaneous data set. Assume that offline data processing uses historical instantaneous data sets (that is, previously collected and timestamped data) to verify the accuracy of the model. For example, offline data processing may find that the prediction error rate of the model under similar working conditions in the past is within 5%, so the current prediction result is considered reliable.

[0040] Forming a mapping relationship and performing machine learning: Through offline data processing verification, a mapping relationship is formed between the prediction results and the instantaneous data set at the cold end. Offline data processing can use these instantaneous data sets with timestamps for continuous machine learning of the intelligent diagnosis model to optimize the model's prediction performance. For example, offline data processing uses algorithms such as gradient descent to adjust model parameters to minimize prediction errors.

[0041] The instantaneous data set includes Y1 at time point t: real-time flow of circulating water; Y2: heat load of condenser; and Y3: efficiency of cooling tower.

[0042] The intelligent diagnosis model is:

[0043]

[0044] Among them, c1, c2, c3 are constant terms, ij, k are lag term coefficients, and it is the error term.

[0045] The prediction result at time t+1 is the prediction value calculated by the instantaneous data set at time t and time t-1.

[0046] The instantaneous data set at time t+1 is used to remove abnormal points in the instantaneous data sets at time t and time t-1.

[0047] The instantaneous data set at time t+1 is used to correct the abnormal points in the instantaneous data sets at time t and time t-1.

[0048] For example, there is the following time series data, which records the instantaneous data sets at time points t, t-1, and t+1, including Y 1 : Real-time flow rate of circulating water (unit: m 3 / h), Y 2 : Condenser heat load (unit: MW), Y 3 : Cooling tower efficiency (unit: %).

[0049]

[0050] Outlier Detection

[0051] Determine at time point t, Y 1 (Circulating water real-time flow) value 1500m 3 / h is an abnormally high value. Similarly, we can detect whether there are abnormal points in other variables and other time points.

[0052] Outlier removal

[0053] If you choose to remove this outlier, you can delete Y at time t from the data set 1 Here, just delete the value.

[0054] The dataset after removing outliers may look like this:

[0055]

[0056] In practical applications, directly deleting data points may not be the best choice because it will reduce the sample size of the data set. A better approach may be outlier correction.

[0057] If we choose to correct this outlier, we can use the data at time t+1 to help us find a more reasonable value to replace it. 1 The value of changes smoothly between adjacent time points, and linear interpolation or other more complex interpolation methods are used to estimate Y at time t. 1 value.

[0058] The dataset after correcting the outliers using linear interpolation might look like this:

[0059]

[0060] Here, we use (1200m 3 / h+1300m 3 / h) / 2=1250m 3 / h as Y at time t 1 but to keep it close to the original outlier value (and still be a "corrected" value), it is rounded to 1350m 3 / h.

[0061] Through the above process, we can see how to use the instantaneous data set at time t+1 to detect, remove and correct the outliers in the data sets at time t and time t-1. In practical applications, these steps need to be adjusted according to the characteristics of the data and the objectives of the analysis.

[0062] The intelligent diagnosis model makes full use of the time series characteristics of the cold end data to achieve accurate prediction of the future trend of the cold end working condition. This capability enables the operation and maintenance personnel to receive early warnings before potential failures occur, so that preventive measures can be taken in time to effectively avoid or reduce the losses caused by failures. This not only significantly enhances the stability of the cold end system, but also greatly reduces maintenance costs and improves overall operating efficiency. The accuracy of the model predictions is verified by strict comparison with the actual instantaneous data set. This verification mechanism is the core link of the model's self-correction and optimization. Once the prediction results are found to be deviated from the actual data, the model can quickly adjust its parameters or algorithms to ensure that future predictions are more accurate and reliable. This continuous optimization process enables the model to flexibly adapt to various changes in the cold end system and always maintain the effectiveness of its diagnostic capabilities. In addition, by mapping the prediction results with the instantaneous data set and using this valuable data for machine learning, the intelligent diagnosis model can continuously improve its accuracy and generalization ability. This self-learning and optimization mechanism enables the model to have a deeper understanding of the complexity and dynamics of the cold end system, so that it can quickly make accurate diagnoses when faced with new situations or unknown faults. This powerful generalization ability plays a vital role in improving the overall operating efficiency and reliability of the cold end system. The application of intelligent diagnosis models also reduces dependence on communication networks. Since the model mainly runs on the offline data processing end and relies on local data for prediction and diagnosis, it can ensure the normal operation of the control system even in the case of network failure or network instability, avoiding system out of control or fault expansion caused by network problems.

[0063] In summary, the application of intelligent diagnosis models in cold-end systems not only significantly improves the accuracy of predictions, but also continuously optimizes model performance through verification mechanisms and machine learning, while reducing dependence on communication networks, providing strong guarantees for the efficient and stable operation of cold-end systems.

[0064] The above is only an embodiment of the present invention, and the common knowledge such as the known specific technical schemes and / or characteristics in the scheme is not described in detail here. It should be pointed out that for those skilled in the art, several variations and improvements can be made without departing from the technical scheme of the present invention. In the present invention, unless otherwise clearly specified and limited, the terms "install", "connect", "connect", "fix" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be the internal connection of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. The scope of protection claimed in this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A cold end detection method based on DCS, characterized in that: The following steps are involved: Acquire cold end data based on DCS, wherein the cold end data includes cold end historical data and cold end real-time data; Cold-end data processing: the offline data processing end adds an instantaneous timestamp to the acquired cold-end real-time data to obtain an instantaneous data set; Generate a cold end detection report. The offline data processing end inputs the instantaneous data set into the intelligent diagnosis model and outputs the cold end diagnosis results including the instantaneous timestamp.

2. The cold end detection method based on DCS according to claim 1, characterized in that: The intelligent diagnosis model is based on the VAR model. Through the input of the instantaneous data set, the intelligent diagnosis model performs trend prediction on the cold end working condition to obtain the prediction result.

3. The cold end detection method based on DCS according to claim 2 is characterized in that: The cold end data includes the rotation speed of the enhanced heat exchange device, the inlet and outlet water temperatures of the condenser, and the real-time flow rate of the circulating water; the offline data processing end verifies the prediction results, and the verification is to compare the deviation of the prediction results with the corresponding instantaneous data set.

4. The cold end detection method based on DCS according to claim 3 is characterized in that: Through verification at the offline data processing end, the prediction results are mapped to the instantaneous data set at the cold end, and the instantaneous data set at the cold end is used to perform machine learning on the intelligent diagnosis model.

5. The cold end detection method based on DCS according to claim 1, characterized in that: The cold end diagnosis result output by the offline data processing includes a verification prediction result, which is a prediction result obtained by the offline data processing through cold end historical data and cold end real-time data.

6. The cold end detection method based on DCS according to claim 2 is characterized in that: The instantaneous data set includes Y1 at time point t: real-time flow of circulating water; Y2: heat load of condenser; and Y3: efficiency of cooling tower.

7. The cold end detection method based on DCS according to claim 6 is characterized in that: The intelligent diagnosis model is Among them, c1, c2, c3 are constant terms, ij,k is the lag term coefficient, it Error term.

8. The cold end detection method based on DCS according to claim 7 is characterized in that: The prediction result at time t+1 is the prediction value calculated by the instantaneous data set at time t and time t-1.

9. The cold end detection method based on DCS according to claim 8, characterized in that: Using the instantaneous data set at time t+1, the abnormal points in the instantaneous data sets at time t and time t-1 are removed.

10. The cold end detection method based on DCS according to claim 8, characterized in that: The instantaneous data set at time t+1 is used to correct the abnormal points in the instantaneous data sets at time t and time t-1.

Images