A kiln data monitoring and early warning method based on reinforcement learning

Abstract

The application relates to the technical field of data processing, in particular to a kiln data monitoring and early warning method based on reinforcement learning, which comprises the following steps: acquiring intelligent gas meter readings at each collection time in each cycle during the operation of a rotary kiln and temperatures at each collection time at each monitoring point; calculating heat space diffusivity at each monitoring point at each collection time in each cycle, calculating heat time dissipation and heat propagation index at each monitoring point at each collection time in each cycle; constructing heat abnormality degree at each monitoring point; calculating energy consumption abnormality degree of the rotary kiln; calculating heat preservation degree at each collection time in each cycle; constructing comprehensive abnormality degree of the rotary kiln; constructing a prediction autoregressive term number according to the comprehensive abnormality degree of the rotary kiln, acquiring a predicted energy consumption of the rotary kiln, and monitoring abnormal conditions of rotary kiln operation energy consumption in combination with a preset energy consumption. The application aims to improve the reliability of rotary kiln operation energy consumption abnormality monitoring.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method for monitoring and early warning of kiln data based on reinforcement learning. Background Technology

[0002] Rotary kilns are common furnaces in industrial production, widely used in industries such as cement, lime, and metallurgy. However, rotary kilns consume a significant amount of energy, making energy reduction a pressing issue. Furthermore, rotary kilns are primarily used for clinker production, and during operation, abnormal phenomena may occur, such as unstable operation, excessively high temperatures, and excessive motor load. If these problems are not addressed promptly, they can lead to equipment damage, production stoppages, and safety accidents. Therefore, real-time monitoring and early warning of energy consumption during rotary kiln operation are of paramount practical importance.

[0003] The autoregressive difference moving average model can predict the energy consumption during the operation of a rotary kiln. By comparing the predicted energy consumption with the preset energy consumption, it can determine whether the energy consumption of the rotary kiln is abnormal. However, the number of autoregressive terms used in the model is set manually, which results in low reliability of monitoring abnormal energy consumption during the operation of the rotary kiln. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a kiln data monitoring and early warning method based on reinforcement learning, thereby resolving the existing issues.

[0005] The kiln data monitoring and early warning method based on reinforcement learning proposed in this application adopts the following technical solution:

[0006] One embodiment of this application provides a kiln data monitoring and early warning method based on reinforcement learning, the method comprising the following steps:

[0007] The readings of the smart gas meter and the temperature at each monitoring point at each sampling time are obtained during each cycle of the rotary kiln operation.

[0008] Based on the temperature differences and distances between each monitoring point and adjacent monitoring points at each sampling time within each cycle, the spatial diffusion of heat at each monitoring point at each sampling time within each cycle is calculated. Based on the temperature differences between each monitoring point and adjacent sampling times at each sampling time within each cycle, the temporal dissipation of heat at each monitoring point at each sampling time within each cycle is calculated. Based on the temporal dissipation and spatial diffusion of heat at each monitoring point at each sampling time within each cycle, the heat propagation index at each monitoring point at each sampling time within each cycle is constructed. Based on the heat propagation index at each monitoring point at all sampling times across all cycles, the degree of heat anomaly at each monitoring point is constructed. Based on the smart gas meter readings at all sampling times, the degree of energy consumption anomaly of the rotary kiln is calculated. Based on the heat propagation index at all monitoring points at each sampling time within each cycle, the insulation level at each sampling time within each cycle is calculated. Based on the insulation level at all sampling times, the smart gas meter readings, the degree of heat anomaly at all monitoring points, and the degree of energy consumption anomaly of the rotary kiln, the comprehensive degree of anomaly of the rotary kiln is constructed.

[0009] The number of predicted autoregressive terms in a time series model is constructed based on the overall anomaly level of the rotary kiln. The predicted energy consumption of the rotary kiln is obtained using the time series model based on the number of predicted autoregressive terms. Anomalies in the operating energy consumption of the rotary kiln are monitored based on the difference between the predicted energy consumption and the preset energy consumption.

[0010] Furthermore, the calculation of the spatial diffusion of heat at each monitoring point at each acquisition time within each cycle includes:

[0011] For each monitoring point at each collection time within each cycle, calculate the absolute value of the temperature difference between the monitoring point and the adjacent previous monitoring point, calculate the distance between the monitoring point and the adjacent previous monitoring point, and calculate the ratio of the absolute value of the temperature difference to the distance, which is recorded as the first ratio. For the monitoring point and the adjacent next monitoring point, use the calculation method of the first ratio between the monitoring point and the adjacent previous monitoring point to obtain the ratio between the monitoring point and the adjacent next monitoring point, which is recorded as the second ratio. The sum of the first ratio and the second ratio is used as the heat spatial diffusion at the monitoring point.

[0012] Furthermore, the calculation of the heat dissipation over time at each monitoring point within each cycle includes:

[0013] For each collection time at each monitoring point within each cycle, calculate the absolute value of the temperature difference between the collection time and the adjacent previous collection time, and record it as the first absolute value of the temperature difference. Calculate the absolute value of the temperature difference between the collection time and the adjacent subsequent collection time, and record it as the second absolute value of the temperature difference. Calculate the sum of the first absolute value of the temperature difference and the second absolute value of the temperature difference. Calculate the product of 2 and the collection time interval. The ratio of the sum to the product is used as the heat dissipation over time at the collection time.

[0014] Furthermore, the construction of the heat transfer index at each monitoring point and at each collection time within each cycle includes:

[0015] Calculate the product of heat time dissipation and heat spatial diffusion at each monitoring point and at each collection time within each cycle. Take the negative of the product as the exponent of an exponential function with the natural constant as the base. Take the difference between 1 and the calculated result of the exponential function as the heat propagation index at each monitoring point and at each collection time within each cycle.

[0016] Furthermore, the determination of the degree of thermal anomaly at each monitoring point includes:

[0017] The heat propagation indices at each monitoring point within each cycle are arranged in the order of the collection time to form the heat propagation time series for each monitoring point in each cycle, and the Hearst exponent for each heat propagation time series is calculated.

[0018] For each monitoring point, the sum of the absolute values ​​of the differences between the Hurst exponents of all two periods of heat propagation time series at the monitoring point is calculated, and the mean of the Hurst exponents of all periods of heat propagation time series at the monitoring point is calculated. The ratio of the sum to the mean is taken as the degree of heat anomaly at the monitoring point.

[0019] Furthermore, the calculation of the abnormal energy consumption level of the rotary kiln based on the smart gas meter readings at all collected times includes:

[0020] The difference between the smart gas meter readings at each sampling time and the next adjacent sampling time is taken as the energy consumption during the time period between each sampling time and the next adjacent sampling time. The energy consumption during the time period between all sampling times except the last sampling time and the next adjacent sampling time is arranged in chronological order to form an energy consumption time series. The standard deviation of all elements in the energy consumption time series is taken as the degree of energy consumption anomaly of the rotary kiln.

[0021] Furthermore, the calculation of the heat preservation degree at each sampling moment within each cycle includes: taking the average heat transfer index of all monitoring points at each sampling moment within each cycle as the heat preservation degree at each sampling moment within each cycle.

[0022] Furthermore, the overall degree of anomaly in the construction of the rotary kiln includes:

[0023] The insulation levels at all collected times are arranged in chronological order to form an insulation level time series. The Pearson correlation coefficient between the insulation level time series and the energy consumption time series is calculated. The mean of the heat anomaly at all monitoring points is calculated. The product of the Pearson correlation coefficient and the mean is calculated. The sum of the product and the energy consumption anomaly of the rotary kiln is taken as the comprehensive anomaly of the rotary kiln.

[0024] Furthermore, the number of predictive autoregressive terms in the time series model constructed based on the comprehensive anomaly degree of the rotary kiln includes:

[0025] The negative of the overall anomaly degree of the rotary kiln is used as the exponent of an exponential function with the natural constant as the base. The difference between 1 and the calculation result of the exponential function is calculated. The product of the difference and a preset adjustment value greater than 0 is calculated. The sum of the preset initial value of the number of autoregressive terms and the product is calculated. The result of rounding down the sum is used as the predicted number of autoregressive terms of the time series model.

[0026] Furthermore, the monitoring of abnormal energy consumption during rotary kiln operation based on the difference between the predicted energy consumption and the preset energy consumption includes:

[0027] When the absolute value of the difference between the predicted energy consumption and the preset energy consumption is greater than the preset proportion of the preset energy consumption, it indicates that the rotary kiln's operating energy consumption is abnormal; otherwise, it indicates that the rotary kiln's operating energy consumption is normal.

[0028] This application has at least the following beneficial effects:

[0029] By comprehensively considering the temporal and spatial changes in heat at various monitoring points on the surface of the rotary kiln, and based on the spatial diffusion and temporal dissipation of heat within each cycle, the degree of heat anomaly at the monitoring points on the surface of the rotary kiln is obtained, thus improving the accuracy of calculating the degree of heat anomaly.

[0030] Based on the fluctuation range of energy consumption in different time periods, the degree of abnormality in energy consumption was obtained, and the possibility of abnormal energy consumption in rotary kilns was analyzed. Since energy consumption can only be calculated by the change in the reading of smart gas meters over a period of time, and the insulation status of rotary kilns is directly related to the energy consumption of rotary kilns, the degree of abnormality in energy consumption was corrected based on the degree of heat abnormality at each monitoring point on the surface of rotary kilns, and the comprehensive degree of abnormality was obtained, which improved the accuracy of estimating the degree of abnormality in rotary kiln operation.

[0031] The number of predicted autoregressive terms in the ARIMA model is adaptively determined based on the comprehensive degree of anomaly, enabling real-time monitoring and early warning of abnormal energy consumption data in rotary kiln operation, thereby improving the reliability of monitoring abnormal energy consumption in rotary kiln operation. Attached Figure Description

[0032] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 A flowchart illustrating the steps of a kiln data monitoring and early warning method based on reinforcement learning provided in this application;

[0034] Figure 2 This is a schematic diagram illustrating the process of obtaining the overall anomaly level. Detailed Implementation

[0035] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a kiln data monitoring and early warning method based on reinforcement learning proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0037] The following description, in conjunction with the accompanying drawings, details a specific scheme for a kiln data monitoring and early warning method based on reinforcement learning provided in this application.

[0038] This application provides an embodiment of a kiln data monitoring and early warning method based on reinforcement learning. Specifically, it provides the following method for kiln data monitoring and early warning based on reinforcement learning. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps:

[0039] Step S001: Obtain the smart gas meter readings and the temperature at each monitoring point at each sampling time during each cycle of the rotary kiln operation.

[0040] A rotary kiln is a cylindrical device. Its working principle involves placing raw materials inside the drum. The rotation of the drum and high-temperature drying process cause a chemical reaction in the raw materials, ultimately yielding the desired product. During operation, the transmission device drives the cylinder to rotate, ensuring uniform heating of the material. Therefore, the temperature distribution along the various generatrices of the cylinder is approximately the same. A generatrice is randomly selected, and M monitoring points are evenly distributed along it. Temperature sensors are placed at these monitoring points on the surface of the rotary kiln. The value of M can be set by the implementer; in this embodiment, M is 10.

[0041] It should be noted that: each T-hour period is defined as a cycle, and temperature data and smart gas meter readings are collected every t minutes within each cycle, for a total of N cycles. The values ​​of T, t, and N can be set by the implementer; in this embodiment, T is 12, t is 15, and N is 7.

[0042] Step S002: Based on the changes in heat at each monitoring point on the surface of the rotary kiln in time and space, obtain the degree of heat anomaly; based on the fluctuation range of energy consumption in different time periods, obtain the degree of energy consumption anomaly; then, based on the degree of heat anomaly at each monitoring point on the surface of the rotary kiln, correct the degree of energy consumption anomaly and obtain the comprehensive anomaly degree.

[0043] Since energy consumption can only be calculated from the changes in the readings of smart gas meters over a period of time, and the insulation status of the rotary kiln is directly related to its energy consumption, the number of autoregressive terms p of the ARIMA model can be determined based on the temperature changes at various monitoring points inside the rotary kiln, combined with the degree of abnormality in energy consumption. This allows for the prediction of energy consumption during the operation of the rotary kiln, improving the accuracy of the prediction and thus enhancing the reliability of monitoring abnormal energy consumption during rotary kiln operation.

[0044] Industrial rotary kilns serve as both combustion and heat transfer devices. During operation, combustion and heat transfer play crucial roles in energy utilization. Within the kiln, heat transfer occurs through conduction, convection, and radiation, resulting in varying temperatures across different parts of the kiln. Temperature changes at adjacent monitoring points at the same sampling time can reflect the heat propagation within the kiln and determine the degree of heat diffusion in space. The spatial diffusion of heat at the j-th monitoring point at the k-th sampling time within the i-th period is shown. It can be represented as follows:

[0045]

[0046] in, The spatial diffusion of heat at the j-th monitoring point at the k-th acquisition time within the i-th cycle. , , These represent the temperatures at the (j-1)th, jth, and (j+1)th monitoring points at the kth time interval within the i-th period. The distance between the (j-1)th monitoring point and the jth monitoring point is... Let be the distance between the j-th monitoring point and the (j+1)-th monitoring point. Let this be the first ratio, and... This is denoted as the second ratio.

[0047] When the distance between adjacent monitoring points is constant, a greater temperature difference indicates a greater difference in the spatial distribution of heat and a greater spatial diffusion value of heat. When the temperature difference between adjacent monitoring points is constant, a smaller distance indicates a greater difference in the spatial distribution of heat and a more drastic change, and a greater spatial diffusion value of heat.

[0048] Because the temperature of the rotary kiln is much higher than the temperature of the outside cold air, heat is lost from the outer wall of the kiln during operation. This results in different temperatures at adjacent sampling times at the same monitoring point, reflecting the heat dissipation at that point. The degree of heat dissipation over time on the outer wall of the rotary kiln can be obtained by measuring the temperature changes at adjacent sampling times at the same monitoring point. The heat dissipation over time at the j-th monitoring point at the k-th sampling time within the i-th period is shown. It can be represented as follows:

[0049]

[0050] in, The heat dissipation over time at the j-th monitoring point during the i-th period at the k-th data collection time is given by [reference to a specific time period]. , , These represent the temperatures at the (k-1), k, and (k+1)th data collection times at the j-th monitoring point within the i-th period. This represents the time interval between data collection points. Let the absolute value of the first difference be denoted as , and This is denoted as the absolute value of the second difference.

[0051] When the interval between data collection times is constant, the greater the temperature difference between adjacent data collection times at the same monitoring point on the surface of the rotary kiln, the faster the heat is dissipated from the outer wall of the rotary kiln, and the greater the heat dissipation over time value. When the temperature difference between adjacent data collection times at the same monitoring point is constant, the shorter the interval between data collection times, the faster the heat is dissipated from the outer wall of the rotary kiln, and the greater the heat dissipation over time value.

[0052] By considering the temporal and spatial changes in heat at various monitoring points on the rotary kiln surface, and based on the spatial diffusion and temporal dissipation of heat, the corresponding heat propagation index is obtained. The heat propagation index at the j-th monitoring point at the k-th sampling time within the i-th cycle is shown. It can be represented as follows:

[0053]

[0054] in, Let be the heat transfer index at the j-th monitoring point at the k-th data collection time within the i-th cycle. It is an exponential function with the natural constant as its base. The heat spatial diffusion at the j-th monitoring point at the k-th sampling time within the i-th cycle represents the degree of heat diffusion within the rotary kiln. The heat dissipation time at the j-th monitoring point within the i-th cycle at the k-th acquisition time represents the degree of heat dissipation on the outer wall of the rotary kiln over time.

[0055] The higher the degree of heat diffusion in the space inside the rotary kiln, and the higher the degree of heat dissipation on the outer wall of the rotary kiln over time, the faster the heat transfer speed of the rotary kiln and the larger the heat transfer index value.

[0056] Under normal circumstances, the change in surface temperature of the rotary kiln exhibits self-similarity and long-term dependence. By arranging the heat propagation index at each monitoring point for each period in chronological order of the acquisition time, a heat propagation time series for each monitoring point for each period is formed. Then, the Hurst exponent of the heat propagation time series at each monitoring point for each period is calculated to obtain the corresponding long-term dependence. The stronger the long-term dependence, the more normal the temperature change trend and the lower the degree of abnormality.

[0057] The degree of temperature anomaly at each monitoring point on the rotary kiln surface is determined by the consistency of the Hurst exponent in the heat propagation time series for each cycle. The degree of heat anomaly at the j-th monitoring point is then calculated. It can be represented as follows:

[0058]

[0059] in, Let N represent the degree of heat anomaly at the j-th monitoring point, and N be the number of periods. Let be the Hearst exponent of the heat propagation time series at the j-th monitoring point during the i-th period. For the j-th monitoring point, the th The Hearst exponent for a heat transfer time series with a period of 100 cycles. Let be the mean of the Hearst exponent for all cycles of heat propagation time series at the j-th monitoring point.

[0060] The larger the Hurst exponent of the heat propagation time series of each cycle at the monitoring point on the surface of the rotary kiln, the more long-term dependent the change in the surface temperature of the rotary kiln is, the more normal the temperature change trend is, the lower the degree of abnormality is, and the smaller the value of heat anomaly is. The smaller the difference between the Hurst exponents of the heat propagation time series of any two cycles at the monitoring point, the higher the consistency of the temperature change in each cycle at that monitoring point is, and the smaller the value of heat anomaly is.

[0061] Since the insulation status of the rotary kiln is directly related to its energy consumption, the degree of energy consumption anomaly can be corrected based on the degree of heat anomaly at each monitoring point on the surface of the rotary kiln to obtain the comprehensive anomaly degree, which is used to determine the number of autoregressive terms p in the ARIMA model. The weight of the degree of heat anomaly is determined by the correlation between the insulation status and energy consumption on the surface of the rotary kiln in each time period.

[0062] The difference between the smart gas meter readings at each sampling time and the subsequent sampling time is used as the energy consumption during the time interval between each sampling time and the subsequent sampling time. Assume the smart gas meter reading at the k-th sampling time in the i-th cycle is... The reading of the smart gas meter at the (k+1)th sampling time within the i-th period is The energy consumption during the time interval between the k-th and (k+1)-th acquisition times in the i-th period is: The heat preservation degree at the k-th sampling time within the i-th period is the average of the heat transfer index at all monitoring points at the corresponding sampling time, i.e. .

[0063] Arrange the insulation levels at all sampling times in chronological order to form a time series of insulation levels. The energy consumption over all time periods is arranged in chronological order to form an energy consumption time series. The Pearson correlation coefficient between the time series of insulation performance and the time series of energy consumption was obtained as follows: .

[0064] During the operation of a rotary kiln, the energy consumption should be approximately equal over equal time intervals. Therefore, the degree of energy consumption anomaly can be determined by the fluctuation range of energy consumption over different time intervals. The standard deviation of all elements in the energy consumption time series can then be used as the degree of energy consumption anomaly in the rotary kiln. .

[0065] Then the overall degree of abnormality It can be represented as follows:

[0066]

[0067] in, The overall degree of abnormality of the rotary kiln, The number of monitoring points The degree of heat anomaly at the j-th monitoring point. The abnormal level of energy consumption in rotary kilns, The Pearson correlation coefficient is used to represent the time series of insulation performance and the time series of energy consumption.

[0068] The higher the degree of energy consumption anomaly and the higher the degree of heat anomaly at each monitoring point on the rotary kiln surface, the greater the overall anomaly value. The higher the correlation between the insulation time series and the energy consumption time series, the higher the reliability of reflecting the degree of energy consumption anomaly through the degree of heat anomaly, and the higher the confidence level of the heat anomaly at each monitoring point on the rotary kiln surface. A schematic diagram of the process for obtaining the overall anomaly degree is shown below. Figure 2 As shown.

[0069] Step S003: Adaptively determine the number of predicted autoregressive terms of the ARIMA model based on the comprehensive degree of anomaly, and conduct real-time monitoring and early warning of abnormal energy consumption data during rotary kiln operation.

[0070] When using the ARIMA model for prediction, the number of autoregressive terms This indicates the number of historical data points used in the prediction. A larger value means that the prediction is based on more historical data and has higher reliability. A smaller value means that the prediction is based on fewer historical data and has higher prediction efficiency.

[0071] In this embodiment, an initial value for the number of autoregressive terms is preset. Preset adjustment value greater than 0 The initial value for the number of autoregressive terms is 3. and adjustment value The value of can be set by the implementer, thus predicting the number of autoregressive terms. It can be represented as follows:

[0072]

[0073] in, This represents the number of predicted autoregressive terms in the ARIMA model. This is the floor function. To preset the initial value of the number of autoregressive terms, It is an exponential function with the natural constant as its base. The overall degree of abnormality of the rotary kiln, This is a preset adjustment value greater than 0.

[0074] The higher the overall anomaly level, the larger the number of predicted autoregressive terms p should be. This allows for predictions based on more observations, avoiding inaccurate energy consumption predictions due to individual anomalies.

[0075] In this embodiment, the difference order d is set to 2, and the number of moving average terms q is set to 3. The implementer can set the values ​​of the difference order d and the number of moving average terms q themselves. Based on historical energy consumption data, the ARIMA model is used to predict the energy consumption in the next time period during the operation of the rotary kiln. When the absolute value of the difference between the predicted energy consumption and the preset energy consumption is greater than the preset energy consumption... If the rotary kiln's energy consumption is abnormal, it is determined that the rotary kiln is in an abnormal state and must be shut down and repaired immediately; otherwise, the rotary kiln's energy consumption is considered normal. The value can be set by the implementer. In this embodiment, The value is 10. The preset energy consumption is determined by the rated power and operating time of the rotary kiln. The rated power is related to the rotary kiln model and can be obtained by querying the relevant parameters of the rotary kiln.

[0076] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0077] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0078] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some of the technical features, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for monitoring and early warning of kiln data based on reinforcement learning, characterized in that, The method includes the following steps: The readings of the smart gas meter and the temperature at each monitoring point at each sampling time are obtained during each cycle of the rotary kiln operation. For each monitoring point at each sampling time within each cycle, calculate the absolute value of the temperature difference between the monitoring point and the adjacent previous monitoring point, calculate the distance between the monitoring point and the adjacent previous monitoring point, and calculate the ratio of the absolute value of the temperature difference to the distance, denoted as the first ratio. For the monitoring point and the adjacent subsequent monitoring point, use the same method as calculating the first ratio to obtain the ratio between the monitoring point and the adjacent previous monitoring point, denoted as the second ratio. The sum of the first ratio and the second ratio is taken as the spatial diffusion of heat at the monitoring point. For each sampling time at each monitoring point within each cycle, calculate the absolute value of the temperature difference between the sampling time and the adjacent previous sampling time, denoted as the first absolute value of the temperature difference, calculate the absolute value of the temperature difference between the sampling time and the adjacent subsequent sampling time, denoted as the second absolute value of the temperature difference, calculate the sum of the first absolute value of the temperature difference and the second absolute value of the temperature difference, and calculate the product of 2 and the sampling time interval. The ratio of the sum to the product is taken as the heat dissipation at the time of collection; the product of heat dissipation at each time of collection at each monitoring point within each cycle and heat spatial diffusion is calculated, and the negative of the product is taken as the exponent of an exponential function with the natural constant as the base. The difference between 1 and the calculated result of the exponential function is taken as the heat propagation index at each time of collection at each monitoring point within each cycle; the degree of heat anomaly at each monitoring point is constructed based on the heat propagation index at all times of collection in all cycles at each monitoring point; the degree of energy consumption anomaly of the rotary kiln is calculated based on the smart gas meter readings at all times of collection; the insulation degree is calculated based on the heat propagation index at all monitoring points within each cycle; and the comprehensive anomaly degree of the rotary kiln is constructed based on the insulation degree, smart gas meter readings, the degree of heat anomaly at all monitoring points, and the degree of energy consumption anomaly of the rotary kiln. The number of predicted autoregressive terms in a time series model is constructed based on the overall anomaly level of the rotary kiln. The predicted energy consumption of the rotary kiln is obtained using the time series model based on the number of predicted autoregressive terms. Anomalies in the operating energy consumption of the rotary kiln are monitored based on the difference between the predicted energy consumption and the preset energy consumption.

2. The kiln data monitoring and early warning method based on reinforcement learning as described in claim 1, characterized in that, The determination of the degree of heat anomaly at each monitoring point includes: The heat propagation indices at each monitoring point within each cycle are arranged in the order of the collection time to form the heat propagation time series for each monitoring point in each cycle, and the Hearst exponent for each heat propagation time series is calculated. For each monitoring point, the sum of the absolute values ​​of the differences between the Hurst exponents of all two periods of heat propagation time series at the monitoring point is calculated, and the mean of the Hurst exponents of all periods of heat propagation time series at the monitoring point is calculated. The ratio of the sum to the mean is taken as the degree of heat anomaly at the monitoring point.

3. The kiln data monitoring and early warning method based on reinforcement learning as described in claim 1, characterized in that, The calculation of the abnormal energy consumption level of the rotary kiln based on the smart gas meter readings at all collected times includes: The difference between the smart gas meter readings at each sampling time and the next adjacent sampling time is taken as the energy consumption during the time period between each sampling time and the next adjacent sampling time. The energy consumption during the time period between all sampling times except the last sampling time and the next adjacent sampling time is arranged in chronological order to form an energy consumption time series. The standard deviation of all elements in the energy consumption time series is taken as the degree of energy consumption anomaly of the rotary kiln.

4. The kiln data monitoring and early warning method based on reinforcement learning as described in claim 1, characterized in that, The calculation of the heat preservation degree at each collection time within each cycle includes: taking the average heat transfer index of all monitoring points at each collection time within each cycle as the heat preservation degree at each collection time within each cycle.

5. The kiln data monitoring and early warning method based on reinforcement learning as described in claim 3, characterized in that, The overall degree of anomaly in the construction of the rotary kiln includes: The insulation levels at all collected times are arranged in chronological order to form an insulation level time series. The Pearson correlation coefficient between the insulation level time series and the energy consumption time series is calculated. The mean of the heat anomaly at all monitoring points is calculated. The product of the Pearson correlation coefficient and the mean is calculated. The sum of the product and the energy consumption anomaly of the rotary kiln is taken as the comprehensive anomaly of the rotary kiln.

6. The kiln data monitoring and early warning method based on reinforcement learning as described in claim 1, characterized in that, The number of predictive autoregressive terms in the time series model constructed based on the comprehensive anomaly degree of the rotary kiln includes: The negative of the overall anomaly degree of the rotary kiln is used as the exponent of an exponential function with the natural constant as the base. The difference between 1 and the calculation result of the exponential function is calculated. The product of the difference and a preset adjustment value greater than 0 is calculated. The sum of the preset initial value of the number of autoregressive terms and the product is calculated. The result of rounding down the sum is used as the predicted number of autoregressive terms of the time series model.

7. The kiln data monitoring and early warning method based on reinforcement learning as described in claim 1, characterized in that, The method of monitoring abnormal energy consumption during rotary kiln operation based on the difference between the predicted energy consumption and the preset energy consumption includes: When the absolute value of the difference between the predicted energy consumption and the preset energy consumption is greater than the preset energy consumption... When the time is right, it indicates that the rotary kiln's operating energy consumption is abnormal; otherwise, it indicates that the rotary kiln's operating energy consumption is normal. It is a pre-defined natural number greater than 1.

Images