An intelligent heating load control system and control method based on big data

Abstract

The present application relates to the technical field of data processing, and especially relates to an intelligent heating load control system and control method based on big data, which comprises the following steps: collecting and storing and backing up heating data in real time, sequentially performing cleaning, denoising and feature extraction operations on the heating data, and training a deep learning model according to the features; using the trained deep learning model to predict the heating load situation to output a load prediction result, triggering an alarm and making corresponding adjustments when the load prediction result is abnormal; determining the control stability of the heating load based on the packet loss rate of the heating data in the collection cycle; if the control stability does not meet the requirements, adjusting the collection interval length of the heating data; if the prediction accuracy does not meet the requirements, adjusting the backup ratio of the heating data. The present application improves the control stability of the heating load.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to an intelligent heating load control system and method based on big data. BACKGROUND

[0002] In the prior art, with the acceleration of urbanization and the continuous improvement of people's requirements for living quality, the stable, efficient and intelligent operation of the heating system as an important part of modern urban infrastructure is increasingly critical. Traditional heating methods often rely on manual experience and simple timing control, which is difficult to adapt to complex and variable heating demands and the requirements of energy saving and efficient use. In today's digital age, big data technology is booming and widely penetrating into various fields, and the heating industry has ushered in a new opportunity for change. A large amount of data is generated during the operation of the heating system, including meteorological data (such as outdoor temperature, wind speed, solar radiation, etc.), thermal data (supply water temperature, return water temperature, heating flow, pressure, etc.), user data (indoor temperature, heating habit, etc.) and building data (building type, area, enclosure structure information). These data contain rich information, which reflects the operation state of the heating system, user demand and the influence of the external environment on heating in real time.

[0003] Chinese Patent Publication No. CN117875515A discloses a time series driven heating system heat load prediction method, comprising: obtaining heating system historical data, the historical data including heating system data in different time periods and historical influence factors influencing the heating system heat load; establishing a historical heat load data set according to the heating system historical data; fitting and analyzing the historical heat load data set to obtain a heat load time series model; training the heat load time series model based on the historical heat load data set to adjust the model parameters; and using the trained heat load time series model to predict the heat load in the future period. As can be seen, the time series driven heating system heat load prediction method has the problem that due to the aging of internal components of the sensor after long-term use, part of the data is not collected, which leads to incomplete collected data and thus causes the control stability of the heating load to decrease. SUMMARY

[0004] Therefore, the present application provides an intelligent heating load control system and method based on big data to overcome the problem in the prior art that due to the aging of internal components of the sensor after long-term use, part of the data is not collected, which leads to incomplete collected data and thus causes the control stability of the heating load to decrease.

[0005] To achieve the above object, the application provides an intelligent heating load control method based on big data, which comprises the following steps: collecting and storing and backing up heating data in real time, sequentially performing cleaning, denoising and feature extraction operations on the heating data, and training a deep learning model according to the features; using the trained deep learning model to predict the heating load condition to output a load prediction result, triggering an alarm and making corresponding adjustments when the load prediction result is abnormal; obtaining the packet loss rate of the heating data in a plurality of collection periods; determining the control stability of the heating load based on the packet loss rate of the heating data in the plurality of collection periods; if the control stability does not meet the requirements, adjusting the collection interval length of the heating data, or determining the prediction accuracy of the heating data based on the missing rate of the heating data; if the prediction accuracy does not meet the requirements, adjusting the backup ratio of the heating data, or adjusting the learning rate of the deep learning model based on the accuracy rate of the heating temperature.

[0006] Further, the control stability of the heating load is determined, comprising:

[0007] The packet loss rate of the heating data in a plurality of collection periods is compared with a preset first packet loss rate.

[0008] If the packet loss rate of the heating data in the plurality of collection periods is greater than the preset first packet loss rate, it is determined that the control stability of the heating load does not meet the requirements.

[0009] Further, the prediction accuracy of the heating data is determined, comprising:

[0010] The packet loss rate of the heating data in the plurality of collection periods is compared with the preset first packet loss rate and a preset second packet loss rate, respectively.

[0011] If the packet loss rate of the heating data in the plurality of collection periods is greater than the preset first packet loss rate and less than or equal to the preset second packet loss rate, it is preliminarily determined that the prediction accuracy of the heating data does not meet the requirements, and it is determined whether the prediction accuracy of the heating data meets the requirements according to the missing rate of the heating data.

[0012] Further, the collection interval length of the heating data is adjusted, comprising:

[0013] The packet loss rate of the heating data in the plurality of collection periods is compared with the preset second packet loss rate.

[0014] If the packet loss rate of the heating data in the plurality of collection periods is greater than the preset second packet loss rate, the collection interval length of the heating data is increased.

[0015] Further, the increase range of the collection interval length of the heating data is determined by the difference between the packet loss rate of the heating data in the plurality of collection periods and the preset second packet loss rate.

[0016] Further, the backup ratio of the heat supply data is adjusted, comprising:

[0017] Comparing the missing rate of the heat supply data with a preset first missing rate and a preset second missing rate respectively;

[0018] If the missing rate of the heat supply data is greater than the preset first missing rate, it is determined that the prediction accuracy of the heat supply data does not meet the requirements;

[0019] If the missing rate of the heat supply data is greater than the preset first missing rate and less than or equal to the preset second missing rate, the backup ratio of the heat supply data is increased;

[0020] If the missing rate of the heat supply data is greater than the preset second missing rate, it is preliminarily determined that the training effectiveness of the deep learning model does not meet the requirements, and the training effectiveness of the deep learning model is determined according to the accuracy rate of the heat supply temperature.

[0021] Further, the increase range of the backup ratio of the heat supply data is determined by the difference between the missing rate of the heat supply data and the preset first missing rate.

[0022] Further, the learning rate of the deep learning model is adjusted, comprising:

[0023] Comparing the accuracy rate of the heat supply temperature with a preset accuracy rate;

[0024] If the accuracy rate of the heat supply temperature is less than the preset accuracy rate, it is determined that the training effectiveness of the deep learning model does not meet the requirements, and the learning rate of the deep learning model is reduced.

[0025] Further, the reduction range of the learning rate of the deep learning model is determined by the difference between the preset accuracy rate and the accuracy rate of the heat supply temperature.

[0026] The application also provides an intelligent heat supply load control system based on big data, comprising:

[0027] A data acquisition module is used to acquire heat supply data in real time;

[0028] A data processing module is connected to the data acquisition module and comprises a preprocessing unit used to preprocess the heat supply data to output optimized data and a feature extraction unit connected to the preprocessing unit and used to extract the features of the optimized data;

[0029] a model training module connected with the data processing module, comprising a model training unit connected with the feature extraction unit to train a deep learning model according to the features, and a prediction unit connected with the deep learning model to predict the heat supply load condition to output a load prediction result;

[0030] an alarm module connected with the model training module to issue an alarm when the load prediction result is abnormal;

[0031] a storage module connected with the data acquisition module, the data processing module, the model training module and the alarm module respectively to store the heat supply data, the optimization data, the features, the load prediction result and the alarm information respectively;

[0032] a control module connected with the data acquisition module, the data processing module, the model training module, the alarm module and the storage module respectively to determine the collection interval length of the heat supply data according to the packet loss rate of the heat supply data in a plurality of collection periods, or determine the backup ratio of the heat supply data according to the missing rate of the heat supply data, and determine the learning rate of the deep learning model according to the accuracy rate of the heat supply temperature.

[0033] Compared with the prior art, the method has the beneficial effects that the collection interval length of the heat supply data is adjusted according to the packet loss rate of the heat supply data in a plurality of collection periods, since the internal elements of the sensor are aged after long-term use, part of the data is not collected, so that the collected data is incomplete, by increasing the collection interval length of the heat supply data, the sensor has more time to complete data collection and processing, so that relatively complete data is obtained, the backup ratio of the heat supply data is adjusted according to the missing rate of the heat supply data, since a series of dynamically changing data are generated when automatic adjustment is performed, the fluctuation of these data may be normal adjustment feedback, but is mistakenly considered as abnormal data, so that it is cleaned up, by increasing the backup ratio of the heat supply data, the possibility of data recovery is increased, there is more opportunity to recover the mistakenly cleaned data, so that the influence of data loss on the accuracy of the heat supply load control is reduced, the learning rate of the deep learning model is adjusted according to the accuracy rate of the heat supply temperature, since most of the data in the heat supply data set may be collected under similar weather conditions and heat supply conditions, the change rule of the heat supply load under various complex conditions cannot be fully displayed, the model will overfit these limited patterns, so that it performs poorly when encountering different heat supply scenes, by reducing the learning rate of the deep learning model, the model training process is more stable, it is more conducive to finding the commonality of the change of the heat supply load under different scenes, the over-reliance on specific similar scene patterns is reduced, and the control stability of the heat supply load is improved.

[0034] Further, the method adjusts the collection interval length of the heat supply data by setting the preset first packet loss rate and the preset second packet loss rate. Since the internal elements of the sensor are aged after long-term use, part of the data is not collected, so that the collected data is incomplete. By increasing the collection interval length of the heat supply data, the sensor is given more time to complete data collection and processing, so that relatively complete data is obtained, and the control stability of the heat supply load is further improved.

[0035] Further, the method adjusts the backup ratio of the heat supply data by setting the preset first missing rate and the preset second missing rate. Since a series of dynamically changing data are generated when automatic adjustment is performed, the fluctuation of the data may be normal adjustment feedback, but is mistakenly considered as abnormal data and is cleaned up. By increasing the backup ratio of the heat supply data, the possibility of data recovery is increased, and there is more opportunity to recover the mistakenly cleaned data, so that the influence of data loss on the control accuracy of the heat supply load is reduced, and the control stability of the heat supply load is further improved.

[0036] Further, the method adjusts the learning rate of the deep learning model by setting the preset accuracy rate. Since most of the data in the heat supply data set may be collected under similar meteorological conditions and heat supply conditions, the change rule of the heat supply load under various complex conditions cannot be fully displayed, the model will overfit these limited patterns, and the model will not perform well when encountering different heat supply scenes. By reducing the learning rate of the deep learning model, the model training process is more stable, which is more conducive to finding the commonness of the change of the heat supply load under different scenes, reducing the excessive dependence on specific similar scene patterns, and further improving the control stability of the heat supply load. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The overall flowchart of the intelligent heat supply load control method based on big data of the embodiment of the application is shown in the figure.

[0038] Figure 2 The overall structure block diagram of the intelligent heat supply load control system based on big data of the embodiment of the application is shown in the figure.

[0039] Figure 3 The logic flowchart of the intelligent heat supply load control method based on big data of the embodiment of the application is shown in the figure.

[0040] Figure 4 The specific flowchart of the process of adjusting the collection interval length of the heat supply data of the intelligent heat supply load control method based on big data of the embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0041] In order to make the objects, technical schemes and advantages of the present application clearer, the following further describes the present application with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0042] The preferred embodiments of the present application are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present application and are not used to limit the protection scope of the present application.

[0043] It should be noted that, in the description of the present application, the terms indicating the direction or positional relationship such as "upper", "lower", "left", "right", "inner", "outer" and the like are based on the direction or positional relationship shown in the drawings, which is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0044] In addition, it should be noted that, in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integrally connected; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or the internal communication of two elements. Those skilled in the art can understand the specific meaning of the above terms in the present application according to the specific circumstances.

[0045] Please refer to Figure 1 , Figure 2 , Figure 3 and Figure 4 , which are respectively the overall flowchart, the overall structure block diagram, the logic flowchart and the specific flowchart of the process of adjusting the collection interval length of heating data of the embodiment of the present application based on big data intelligent heating load control system and control method. The present application is a kind of intelligent heating load control method based on big data, comprising:

[0046] Step S1, real-time collection of heating data and storage and backup, sequentially cleaning, denoising and extracting features on the heating data, and training a deep learning model according to the features;

[0047] Step S2, using the trained deep learning model to predict the heating load condition to output the load prediction result, triggering an alarm and making corresponding adjustments when the load prediction result appears abnormal;

[0048] Step S3, obtaining the packet loss rate of the heating data in a plurality of collection periods;

[0049] Step S4, determining the control stability of the heating load based on the packet loss rate of the heating data in the plurality of collection periods;

[0050] Step S5, if the control stability does not meet the requirements, adjusting the collection interval length of the heating data, or determining the prediction accuracy of the heating data based on the missing rate of the heating data;

[0051] Step S6, if the prediction accuracy does not meet the requirements, adjusting the backup ratio of the heating data, or adjusting the learning rate of the deep learning model based on the accuracy rate of the heating temperature.

[0052] Specifically, the heating data includes outdoor temperature, outdoor wind speed, and water supply temperature.

[0053] Specifically, the features include the diurnal water supply and return temperature difference change rate, the indoor temperature fluctuation amplitude, and the heating peak period and valley period.

[0054] Specifically, the deep learning model includes a deep neural network, a convolutional neural network, and a long short-term memory network.

[0055] Specifically, the heating load conditions include real-time heating load, diurnal variation of heating load, and regional distribution difference of heating load.

[0056] Specifically, the load prediction result includes heating load quantity prediction in the next few hours, heating load peak and valley prediction in the next day to a week, and total load prediction in the entire heating season.

[0057] Specifically, the abnormal situation of the load prediction result includes the maximum heating load exceeding the maximum load threshold, the minimum heating load being lower than the minimum load threshold, and the change rate of the heating load exceeding the change threshold, wherein the maximum load threshold, the minimum load threshold, and the change threshold are all set according to actual conditions.

[0058] In the implementation, the method adjusts the collection interval length of the heating data according to the packet loss rate of the heating data in a plurality of collection periods. Due to the aging of internal elements of the sensor after long-term use, part of the data is not collected, resulting in incomplete collected data. By increasing the collection interval length of the heating data, more time is given to the sensor to complete data collection and processing, so as to obtain relatively complete data. The backup ratio of the heating data is adjusted according to the missing rate of the heating data. Due to a series of dynamically changing data generated during automatic adjustment, the fluctuations of these data may be normal adjustment feedback, but are mistakenly considered as abnormal data and are cleaned up. By increasing the backup ratio of the heating data, the possibility of data recovery is increased, and there is more opportunity to recover the mistakenly cleaned data, thereby reducing the influence of data loss on the accuracy of the heating load control. The learning rate of the deep learning model is adjusted according to the accuracy rate of the heating temperature. Since most of the data in the heating data set may be collected under similar meteorological conditions and heating working conditions, the change law of the heating load under various complex conditions cannot be fully displayed, the model will overfit these limited patterns, and thus performs poorly when encountering different heating scenes. By reducing the learning rate of the deep learning model, the model training process is more stable, which is more conducive to finding the commonality of the change of the heating load in different scenes, reducing the excessive dependence on specific similar scene patterns, and improving the control stability of the heating load.

[0059] Specifically, determining the control stability of the heating load comprises:

[0060] obtaining a packet loss rate of heating data in a plurality of collection periods;

[0061] comparing the packet loss rate of the heating data in the plurality of collection periods with a preset first packet loss rate;

[0062] If the packet loss rate of the heating data in the plurality of collection periods is greater than the preset first packet loss rate, it is determined that the control stability of the heating load does not meet the requirements.

[0063] Specifically, determining the prediction accuracy of the heating data comprises:

[0064] comparing the packet loss rate of the heating data in the plurality of collection periods with the preset first packet loss rate and a preset second packet loss rate, respectively;

[0065] If the packet loss rate of the heating data in the plurality of collection periods is greater than the preset first packet loss rate and less than or equal to the preset second packet loss rate, it is preliminarily determined that the prediction accuracy of the heating data does not meet the requirements, and it is determined whether the prediction accuracy of the heating data meets the requirements according to the missing rate of the heating data.

[0066] It can be understood that the three intervals divided by the preset first packet loss rate and the preset second packet loss rate correspond to three situations respectively:

[0067] The first interval is that the packet loss rate of the heat supply data in the plurality of collection periods is less than or equal to the preset first packet loss rate, and the corresponding situation is that it is determined that the control stability of the heat supply load meets the requirements;

[0068] The second interval is that the packet loss rate of the heat supply data in the plurality of collection periods is greater than the preset first packet loss rate and less than or equal to the preset second packet loss rate, and the corresponding situation is that a series of dynamic change data are generated when automatic adjustment is performed, and the fluctuation of the data may be normal adjustment feedback but is mistakenly considered as abnormal data and thus is cleaned up;

[0069] The third interval is that the packet loss rate of the heat supply data in the plurality of collection periods is greater than the preset second packet loss rate, and the corresponding situation is that the internal elements of the sensor are aged after long-term use, so that part of the data is not collected, and thus the collected data is incomplete.

[0070] In practice, the preset first packet loss rate is generally selected in the range of [18%, 22%], and the preset second packet loss rate is generally selected in the range of [23%, 27%].

[0071] Preferably, the preferred embodiment of the preset first packet loss rate is 20%, and the preferred embodiment of the preset second packet loss rate is 25%.

[0072] In implementation, the method provided by the application determines the control stability of the heat supply load by setting the preset first packet loss rate and the preset second packet loss rate, reduces the influence of the decrease in the control accuracy of the heat supply load caused by inaccurate determination of the control stability of the heat supply load, and further improves the control stability of the heat supply load.

[0073] Specifically, the collection interval duration of the heat supply data is adjusted, including:

[0074] The packet loss rate of the heat supply data in the plurality of collection periods is compared with the preset second packet loss rate;

[0075] If the packet loss rate of the heat supply data in the plurality of collection periods is greater than the preset second packet loss rate, the collection interval duration of the heat supply data is increased.

[0076] Specifically, the increase range of the collection interval duration of the heat supply data is determined by the difference between the packet loss rate of the heat supply data in the plurality of collection periods and the preset second packet loss rate.

[0077] Specifically, if the difference between the packet loss rate of the heating data in the plurality of collection periods and the preset second packet loss rate is within 3%, the collection interval length of the heating data is increased to 1.1 times of the original; when the difference between the packet loss rate of the heating data in the plurality of collection periods and the preset second packet loss rate exceeds 5%, the collection interval length of the heating data is increased by 2 minutes for each excess of 2% on the basis of being increased to 1.1 times of the original, for example, if the difference between the packet loss rate of the heating data in the plurality of collection periods and the preset second packet loss rate is 7%, the current collection interval length of the heating data is 10 minutes, and the increased collection interval length of the heating data is 10*1.1+2*2=15 minutes.

[0078] In implementation, the method provided by the application adjusts the collection interval length of the heating data by setting the preset first packet loss rate and the preset second packet loss rate. Since the internal components of the sensor are aged after long-term use, part of the data is not collected, resulting in incomplete collected data. By increasing the collection interval length of the heating data, the sensor is given more time to complete data collection and processing, so that relatively complete data is obtained, and the control stability of the heating load is further improved.

[0079] Specifically, the backup ratio of the heating data is adjusted, including:

[0080] The missing data amount of the heating data and the total data amount of the heating data in a plurality of processing periods are respectively acquired, and the missing rate of the heating data is calculated;

[0081] The missing rate of the heating data is compared with the preset first missing rate and the preset second missing rate respectively;

[0082] If the missing rate of the heating data is greater than the preset first missing rate, it is determined that the prediction accuracy of the heating data does not meet the requirements;

[0083] If the missing rate of the heating data is greater than the preset first missing rate and less than or equal to the preset second missing rate, the backup ratio of the heating data is increased;

[0084] If the missing rate of the heating data is greater than the preset second missing rate, it is preliminarily determined that the training effectiveness of the deep learning model does not meet the requirements, and the training effectiveness of the deep learning model is determined according to the accuracy of the heating temperature.

[0085] It can be understood that the three intervals divided by the preset first missing rate and the preset second missing rate correspond to three situations respectively:

[0086] The first interval is that the missing rate of the heating data is less than or equal to the preset first missing rate, and the corresponding situation is that it is determined that the prediction accuracy of the heating data meets the requirements;

[0087] The second interval is that the missing rate of the heating data is greater than the preset first missing rate and less than or equal to the preset second missing rate, and the corresponding case is that, due to a series of dynamic changes in automatic adjustment, the fluctuation of the data may be normal adjustment feedback, but is mistakenly considered as abnormal data and is cleaned up.

[0088] The third interval is that the missing rate of the heating data is greater than the preset second missing rate, and the corresponding case is that, due to the fact that most of the data in the heating data set may be collected under similar weather conditions and heating working conditions, the change rule of the heating load under various complex conditions cannot be fully displayed, the model will overfit these limited patterns, and thus performs poorly when encountering different heating scenes.

[0089] In practice, the preset first missing rate is generally selected in the range of [3%, 7%], and the preset second missing rate is generally selected in the range of [8%, 12%].

[0090] Preferably, the preferred embodiment of the preset first missing rate is 5%, and the preferred embodiment of the preset second missing rate is 10%.

[0091] Specifically, the missing rate of the heating data is the ratio of the missing data amount of the heating data in a plurality of processing periods to the total data amount of the heating data.

[0092] In implementation, the method determines the prediction accuracy of the heating data by setting the preset first missing rate and the preset second missing rate, reduces the influence of the decline in the control stability of the heating load caused by inaccurate determination of the prediction accuracy of the heating data, and further improves the control stability of the heating load.

[0093] Specifically, the increasing range of the backup ratio of the heating data is determined by the difference between the missing rate of the heating data and the preset first missing rate.

[0094] Specifically, when the difference between the missing rate of the heating data and the preset first missing rate is within 5%, the backup ratio of the heating data is increased to 1.2 times of the original; when the difference between the missing rate of the heating data and the preset first missing rate exceeds 5%, the backup ratio of the heating data is increased by 2% for each excess of 2% on the basis of being increased to 1.2 times of the original, for example, the difference between the missing rate of the heating data and the preset first missing rate is 9%, the current backup ratio of the heating data is 70%, and the increased backup ratio of the heating data is 70x1.2+2x2=88%.

[0095] In the implementation, the method adjusts the backup ratio of the heat supply data by setting a preset first missing rate and a preset second missing rate. Since a series of dynamically changing data are generated during automatic adjustment, the fluctuations of the data can be normal adjustment feedback but are mistakenly considered as abnormal data and thus are cleaned. By increasing the backup ratio of the heat supply data, the possibility of data recovery can be increased, and there is more opportunity to recover the mistakenly cleaned data, thereby reducing the influence of data loss on the accuracy of heat load control and further improving the control stability of the heat load.

[0096] Specifically, the learning rate of the deep learning model is adjusted, including:

[0097] The number of areas that actually reach the heat supply temperature during the heat supply process and the total number of areas that are supplied with heat are acquired respectively, and the accuracy of the heat supply temperature is calculated.

[0098] The accuracy of the heat supply temperature is compared with a preset accuracy.

[0099] If the accuracy of the heat supply temperature is less than the preset accuracy, it is determined that the training effectiveness of the deep learning model does not meet the requirements, and the learning rate of the deep learning model is reduced.

[0100] It can be understood that the two intervals of the preset accuracy correspond to two situations respectively.

[0101] The first interval is that the accuracy of the heat supply temperature is less than the preset accuracy, and the corresponding situation is that: since most of the data in the heat supply data set can be collected under similar meteorological conditions and heat supply conditions, the change law of the heat load under various complex situations cannot be fully displayed, the model can overfit these limited patterns, and thus the model performs poorly when encountering different heat supply scenes.

[0102] The second interval is that the accuracy of the heat supply temperature is greater than or equal to the preset accuracy, and the corresponding situation is that: it is determined that the training effectiveness of the deep learning model meets the requirements.

[0103] In practice, the preset accuracy is generally selected in the range of [93%, 97%].

[0104] Preferably, the preferred embodiment of the preset accuracy is 95%.

[0105] Specifically, the accuracy of the heat supply temperature is the ratio of the number of areas that actually reach the heat supply temperature during the heat supply process to the total number of areas that are supplied with heat.

[0106] In the implementation, the method determines the training effectiveness of the deep learning model by setting the preset accuracy rate, reduces the influence of the determination of the training effectiveness of the deep learning model on the control stability of the heating load, and further improves the control stability of the heating load.

[0107] Specifically, the reduction range of the learning rate of the deep learning model is determined by the difference between the preset accuracy rate and the accuracy rate of the heating temperature.

[0108] Specifically, when the difference between the preset accuracy rate and the accuracy rate of the heating temperature is within 5%, the learning rate of the deep learning model is reduced to 0.9 times of the original learning rate; when the difference between the preset accuracy rate and the accuracy rate of the heating temperature exceeds 5%, the learning rate of the deep learning model is reduced by 0.002 for each excess of 2% on the basis of being reduced to 0.9 times of the original learning rate, for example, when the difference between the preset accuracy rate and the accuracy rate of the heating temperature is 9% and the current learning rate of the deep learning model is 0.1, the reduced learning rate of the deep learning model is 0.1*0.9-0.02*2=0.05.

[0109] In the implementation, the method adjusts the learning rate of the deep learning model by setting the preset accuracy rate. Since most of the data in the heating data set may be collected under similar meteorological conditions and heating working conditions, the change rule of the heating load under various complex conditions cannot be fully demonstrated, the model may overfit these limited patterns, and thus the model may not perform well when encountering different heating scenes. By reducing the learning rate of the deep learning model, the model training process can be more stable, it is more conducive to finding the commonality of the change of the heating load under different scenes, the excessive dependence on specific similar scene patterns is reduced, and the control stability of the heating load is further improved.

[0110] An intelligent heating load control system based on big data, comprising:

[0111] A data acquisition module is configured to acquire heating data in real time.

[0112] A data processing module is connected to the data acquisition module and comprises a preprocessing unit configured to preprocess the heating data to output optimized data and a feature extraction unit connected to the preprocessing unit and configured to extract features of the optimized data.

[0113] A model training module is connected to the data processing module and comprises a model training unit connected to the feature extraction unit and configured to train a deep learning model according to the features and a prediction unit connected to the deep learning model and configured to predict a heating load condition to output a load prediction result.

[0114] an alarm module connected with the model training module, configured to send an alarm when the load prediction result is abnormal;

[0115] a storage module connected with the data collection module, the data processing module, the model training module and the alarm module, respectively, configured to store the heat supply data, the optimization data, the features, the load prediction result and the alarm information, respectively;

[0116] a control module connected with the data collection module, the data processing module, the model training module, the alarm module and the storage module, configured to determine the collection interval length of the heat supply data according to the packet loss rate of the heat supply data in a plurality of collection periods, or to determine the backup ratio of the heat supply data according to the missing rate of the heat supply data, and to determine the learning rate of the deep learning model according to the accuracy rate of the heat supply temperature.

[0117] Specifically, the preprocessing includes cleaning and denoising operations.

[0118] Specifically, the optimization data includes the water supply temperature after missing value processing, the heat supply flow after error value removal, and the indoor temperature after denoising.

[0119] Specifically, the alarm information includes over-high load alarm, over-low load alarm, and sharp rise in load change rate alarm.

[0120] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after these changes or replacements will all fall within the protection scope of the present application.

Claims

1. A big data-based intelligent heating load control method, characterized in that, The method comprises the following steps: real-time collection, storage and backup of heating data, cleaning, denoising and feature extraction of the heating data, and training of a deep learning model based on the features; using the trained deep learning model to predict the heating load to output a load prediction result, triggering an alarm and making corresponding adjustments when the load prediction result is abnormal; obtaining the packet loss rate of the heating data in a plurality of collection periods; determining the control stability of the heating load based on the packet loss rate of the heating data in the plurality of collection periods; if the control stability does not meet the requirements, adjusting the collection interval of the heating data, or determining the prediction accuracy of the heating data based on the missing rate of the heating data; if the prediction accuracy does not meet the requirements, adjusting the backup ratio of the heating data, or adjusting the learning rate of the deep learning model based on the accuracy of the heating temperature; adjusting the collection interval of the heating data, comprising: comparing the packet loss rate of the heating data in the plurality of collection periods with a preset second packet loss rate; if the packet loss rate of the heating data in the plurality of collection periods is greater than the preset second packet loss rate, increasing the collection interval of the heating data; adjusting the backup ratio of the heating data, comprising: comparing the missing rate of the heating data with a preset first missing rate and a preset second missing rate respectively; if the missing rate of the heating data is greater than the preset first missing rate, determining that the prediction accuracy of the heating data does not meet the requirements; if the missing rate of the heating data is greater than the preset first missing rate and less than or equal to the preset second missing rate, increasing the backup ratio of the heating data; if the missing rate of the heating data is greater than the preset second missing rate, preliminarily determining that the training effectiveness of the deep learning model does not meet the requirements, and determining the training effectiveness of the deep learning model based on the accuracy of the heating temperature; adjusting the learning rate of the deep learning model, comprising: comparing the accuracy of the heating temperature with a preset accuracy; if the accuracy of the heating temperature is less than the preset accuracy, determining that the training effectiveness of the deep learning model does not meet the requirements, and reducing the learning rate of the deep learning model. 2.The big data based intelligent heating load control method according to claim 1, wherein, determining the control stability of the heating load, comprising: comparing the packet loss rate of the heating data in the plurality of collection periods with a preset first packet loss rate; if the packet loss rate of the heating data in the plurality of collection periods is greater than the preset first packet loss rate, determining that the control stability of the heating load does not meet the requirements. 3.The big data based intelligent heating load control method according to claim 2, wherein, determining the prediction accuracy of the heating data, comprising: comparing the packet loss rate of the heating data in the plurality of collection periods with the preset first packet loss rate and the preset second packet loss rate respectively; if the packet loss rate of the heating data in the plurality of collection periods is greater than the preset first packet loss rate and less than or equal to the preset second packet loss rate, preliminarily determining that the prediction accuracy of the heating data does not meet the requirements, and determining whether the prediction accuracy of the heating data meets the requirements based on the missing rate of the heating data. 4.The big data based intelligent heating load control method according to claim 3, wherein, The increase range of the collection interval of the heating data is determined by the difference between the packet loss rate of the heating data in the plurality of collection periods and the preset second packet loss rate. 5.The big data based intelligent heating load control method according to claim 4, wherein, The increase range of the backup ratio of the heating data is determined by a difference between the missing rate of the heating data and a preset first missing rate. 6.The big data based intelligent heating load control method according to claim 5, wherein, The decrease range of the learning rate of the deep learning model is determined by a difference between the preset accuracy rate and an accuracy rate of the heating temperature. 7.A big data-based intelligent heating load control system using the big data-based intelligent heating load control method according to any one of claims 1 to 6, characterized by, The method comprises the steps of: collecting the heating data in real time by a data collection module; preprocessing the heating data to output optimized data by a preprocessing unit of a data processing module connected with the data collection module, and extracting features of the optimized data by a feature extraction unit connected with the preprocessing unit; training a deep learning model according to the features by a model training unit connected with the feature extraction unit, and predicting a heating load condition to output a load prediction result by a prediction unit connected with the deep learning model; issuing an alarm when the load prediction result is abnormal by an alarm module connected with the model training module; storing the heating data, the optimized data, the features, the load prediction result and the alarm information by a storage module connected with the data collection module, the data processing module, the model training module and the alarm module respectively; and determining a collection interval of the heating data according to a packet loss rate of the heating data in a plurality of collection periods, or determining a backup ratio of the heating data according to a missing rate of the heating data, and determining a learning rate of the deep learning model according to an accuracy rate of the heating temperature by a control module connected with the data collection module, the data processing module, the model training module, the alarm module and the storage module.

Images