A method for automatic control of water mixing based on dew point temperature

Abstract

The present invention relates to the field of automation control technology, and in particular to a dew point temperature-based automatic water mixing control method, which predicts the future dew point temperature and a first prediction confidence set by analyzing a historical environmental data sequence and a historical dew point temperature sequence, predicts control data by combining historical environmental data, historical mixed water temperature and a predicted dew point temperature set, predicts control data and confidence data by combining historical environmental data, historical mixed water temperature, historical dew point temperature and an initial control model, and trains the initial control model according to the predicted control data and confidence data corresponding to different angles, so that a target control model can more efficiently and accurately output a suitable water mixing control strategy based on current data of a target water mixing system collected in real time by a sensor, thereby improving the accuracy of the control strategy and the efficiency of the control process, thereby ensuring that the mixed water temperature is slightly higher than the dew point temperature and preventing condensation in pipelines.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic control, and in particular to a water mixing automatic control method based on dew point temperature. Background Art

[0002] In industrial production, building HVAC, and other fields, controlling the mixing water temperature based on dew point temperature is crucial for preventing condensation in pipes. Traditional dew point temperature-based automated mixing water control methods often use fixed parameter control logic, such as presetting a fixed difference between the mixing water temperature and the dew point temperature, and adjusting the hot and cold water valves through simple threshold comparisons. This method relies only on a single or small amount of historical data, making it difficult to capture the dynamic changes in variables such as ambient temperature, humidity, and mixing water temperature. When environmental conditions such as seasonal changes, day and night temperature differences, or fluctuations in system operating conditions fluctuate, the control accuracy is low, which can easily lead to a large deviation between the mixing water temperature and the dew point temperature, resulting in pipe condensation or energy waste.

[0003] Furthermore, existing control models are mostly static and cannot optimize or update based on actual operating data. As environmental factors and system characteristics change over time, control effectiveness gradually declines, making it difficult to achieve long-term, stable, and efficient automated water mixing control. Furthermore, relying on large amounts of historical data to develop control strategies requires processing and analyzing this data, resulting in low control efficiency.

[0004] Therefore, how to ensure the accuracy of the control strategy of the mixed water temperature and the efficiency of the control process has become an urgent problem to be solved. Summary of the Invention

[0005] In response to the above technical problems, the technical solution adopted by the present invention is a water mixing automation control method based on dew point temperature, which includes the following steps:

[0006] S1. According to the historical environmental data sequence and historical dew point temperature sequence corresponding to each sample mixed water system at the ikth to ith historical time points, obtain the predicted dew point temperature set corresponding to each sample mixed water system at the i+1th historical time point and the first prediction confidence set corresponding to the predicted dew point temperature set, where i=k+1, k+2,...,M, M is the total number of historical time points, and k is the number of historical time points corresponding to the historical environmental data sequence and the historical dew point temperature sequence.

[0007] S2. Obtain the first prediction control data set corresponding to each sample mixing water system at the i-th historical time point based on the historical environmental data, historical mixing water temperature, and predicted dew point temperature set corresponding to the i+1-th historical time point for each sample mixing water system.

[0008] S3. Based on the historical environmental data, historical mixed water temperature, historical dew point temperature and initial control model corresponding to each sample mixed water system at the i-th historical time point, obtain the second prediction control data set corresponding to each sample mixed water system at the i-th historical time point and the second prediction confidence set corresponding to the second prediction control data set.

[0009] S4, traverse all sample mixed water systems and M historical time points, train the initial control model according to the first prediction control data set, first prediction confidence set, second prediction control data set and second prediction confidence set corresponding to each sample mixed water system at each historical time point, and obtain the trained target control model.

[0010] S5. When the real-time environmental data, real-time mixed water temperature and real-time dew point temperature corresponding to the target mixed water system at the current time point are obtained, the first valve and the second valve of the target mixed water system are controlled according to the real-time environmental data, real-time mixed water temperature, real-time dew point temperature and the trained target control model.

[0011] The present invention has at least the following beneficial effects: by analyzing the historical environmental data sequence and the historical dew point temperature sequence of the sample water mixing system from the ikth to the ith historical time point, the complex relationship between environmental factors and dew point temperature is comprehensively captured, and the accuracy of predicting future dew point temperature and the first prediction confidence set is improved; combining the historical environmental data, the historical mixed water temperature and the predicted dew point temperature set to obtain the first prediction control data set, combining the historical environmental data, the historical mixed water temperature, the historical dew point temperature and the initial control model to obtain the second prediction control data set and the second prediction confidence set, digging out the operating rules and experience of the water mixing system under different environmental conditions from different angles, and according to the prediction control data and confidence data corresponding to different angles Training the initial control model greatly simplifies the operational logic of the target control model, reduces the error accumulation caused by the intermediate prediction link, and enables the target control model to more efficiently and accurately output the appropriate mixing water control strategy based on the current data collected by the sensor in real time, thereby improving the accuracy of the control strategy and the efficiency of the control process. When the real-time environmental data, real-time mixing water temperature and real-time dew point temperature of the target mixing water system are obtained, the first valve and the second valve of the target mixing water system are controlled by the target control model to ensure that the mixing water temperature is slightly higher than the dew point temperature, thereby preventing condensation in the pipeline, realizing a fully automated and precise control process of the mixing water system from data collection to valve adjustment, and improving the accuracy of the control strategy and the efficiency of the control process. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A flow chart of an automatic water mixing control method based on dew point temperature provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar water mixing systems, and are not necessarily used to describe a specific order or sequence. It is understood that, where appropriate, the above-mentioned terms used to distinguish similar water mixing systems can be interchanged so that the present invention can also implement other embodiments other than the above-mentioned illustrated embodiments or described embodiments. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0016] This embodiment provides a method for automatically controlling water mixing based on dew point temperature, and the method includes the following steps: Figure 1 As shown:

[0017] S1. According to the historical environmental data sequence and historical dew point temperature sequence corresponding to each sample mixed water system at the ikth to ith historical time points, obtain the predicted dew point temperature set corresponding to each sample mixed water system at the i+1th historical time point and the first prediction confidence set corresponding to the predicted dew point temperature set, where i=k+1, k+2,...,M, M is the total number of historical time points, and k is the number of historical time points corresponding to the historical environmental data sequence and the historical dew point temperature sequence.

[0018] The sample water mixing system is a water mixing system entity used for model training and pattern discovery in the control system. From a physical perspective, the sample water mixing system can be a building HVAC water mixing system in a specific scenario, such as a central air conditioning water mixing system in different buildings such as office buildings, hospitals, or shopping malls. The sample water mixing system and the target water mixing system belong to the same specific water mixing system. Therefore, by collecting historical environmental data series and historical dew point temperature series of the sample water mixing system at different historical time points, the corresponding regular characteristics of the sample water mixing system can be discovered and used to train the control model, thereby providing data support for the precise control of the target water mixing system.

[0019] Historical time points are the time marks for data collection and recording, and are arranged in chronological order to form a time series. The specific collection frequency is set by the implementer based on actual needs and system characteristics.

[0020] Historical environmental data series includes a variety of environmental parameters collected by various sensors at various historical points in time, such as air temperature from temperature sensors and air humidity from humidity sensors. These interrelated and interactive environmental parameters jointly influence dew point temperature changes and serve as a crucial basis for predicting future dew point temperatures.

[0021] The historical dew point temperature sequence is a series of dew point temperature values ​​collected by the dew point temperature sensor at various historical time points. It intuitively shows the evolution of the dew point temperature of the environment in which the sample mixing water system is located within the historical time period, providing direct reference data for predicting future dew point temperatures.

[0022] The ikth to ith historical time points define the range of historical data used for analysis and modeling. If the k value is too small, it may not be able to fully capture the patterns of the data. If the k value is too large, it may introduce too much noise data, affecting the accuracy of the prediction. Therefore, the selection of the k value can be set by the implementer by comprehensively considering the stability and changing trend of the data.

[0023] The i+1th historical time point represents the future moment to be predicted based on the historical environmental data sequence and the historical dew point temperature sequence corresponding to the ikth to ith historical time points.

[0024] The predicted dew point temperature set includes several predicted dew point temperatures for each sample mixing system at the i+1th historical time point, serving as the basis for predictive control of the sample mixing system. The first prediction confidence set includes the confidence assessment value for each predicted dew point temperature in the corresponding predicted dew point temperature set. This helps determine the reliability of each predicted dew point temperature and enables more rational use of the prediction results in subsequent decision-making.

[0025] Specifically, under certain climatic conditions and system operating conditions, there is a stable regularity and trend relationship between changes in ambient temperature and humidity and dew point temperature. This embodiment mines the intrinsic correlation between environmental factors and dew point temperature in historical data, and estimates future dew point temperature by constructing a suitable prediction model, providing a data basis for the water mixing automation control system.

[0026] As mentioned above, by analyzing multiple sample mixing water systems and a large amount of historical data, it is possible to more comprehensively capture the complex relationship between environmental factors and dew point temperature, improve the accuracy of predicting future dew point temperature, and introduce the first prediction confidence level, so that decision makers can intuitively understand the reliability of each predicted dew point temperature, reduce the risk of decision-making errors due to prediction errors, and thus improve the reliability and stability of the entire mixing water automation control system.

[0027] In one embodiment, S1 includes the following steps:

[0028] S11, based on the historical environmental data sequence composed of k+1 historical environmental data corresponding to each sample mixed water system at the ikth to i-th historical time points and the preset second prediction model, obtain the first predicted dew point temperature corresponding to each sample mixed water system at the i+1th historical time point.

[0029] S12, input the historical dew point temperature sequence consisting of the k+1 historical dew point temperatures corresponding to each sample mixing water system at the ikth to i-th historical time points into the preset third prediction model, and obtain the second predicted dew point temperature corresponding to each sample mixing water system at the i+1th historical time point.

[0030] S13, based on the first predicted dew point temperature, the second predicted dew point temperature, the first weight corresponding to the second prediction model, the second weight corresponding to the third prediction model and the preset number of set elements N, obtain N-2 reference predicted dew point temperatures within the temperature range of the first predicted dew point temperature and the second predicted dew point temperature, as well as the first prediction confidence corresponding to the first predicted dew point temperature, the first prediction confidence corresponding to the second predicted dew point temperature and the first prediction confidence corresponding to each reference predicted dew point temperature.

[0031] S14, the first predicted dew point temperature, the second predicted dew point temperature and N-2 reference predicted dew point temperatures are combined in ascending order to form a predicted dew point temperature set corresponding to the i+1th historical time point for each sample mixed water system.

[0032] S15. In the order corresponding to the predicted dew point temperature set, a first prediction confidence set corresponding to the predicted dew point temperature set is formed according to the first prediction confidence corresponding to the first predicted dew point temperature, the first prediction confidence corresponding to the second predicted dew point temperature, and the first prediction confidence corresponding to each reference predicted dew point temperature.

[0033] In a specific embodiment, S11 includes the following steps:

[0034] S111: Inputting a historical environmental data sequence consisting of k+1 historical environmental data corresponding to each sample mixed water system at the ikth to ith historical time points into a preset second prediction model to obtain the predicted environmental data corresponding to each sample mixed water system at the i+1th historical time point;

[0035] S112 , obtaining a first predicted dew point temperature corresponding to the i+1th historical time point for each sample mixed water system based on the predicted environmental data corresponding to the i+1th historical time point for each sample mixed water system.

[0036] Among them, the second prediction model is a model constructed based on machine learning or time series analysis algorithms, which is used to mine long-term dependencies and complex patterns in historical environmental data sequences, thereby predicting future environmental data, so that the subsequent calculation of dew point temperature based on environmental data is more in line with the actual situation, avoiding relying solely on historical data and ignoring the errors caused by future environmental changes, and improving the accuracy of dew point temperature prediction. The second prediction model can be constructed by several convolutional layers and fully connected layers, wherein each convolutional layer is composed of multiple convolution kernels, each convolution kernel slides on the input sequence, extracts data features by weighted summation, and the upper convolution layer inputs the extracted data features to the next convolution layer, thereby performing multi-layer feature extraction on the input historical environmental data sequence, and inputting the last extracted data features to the fully connected layer, which maps the data features to the output space to obtain the corresponding predicted environmental data.

[0037] Based on the predicted environmental data corresponding to each sample mixing water system at the i+1th historical time point, the calculation formula related to dew point temperature is used to calculate the first predicted dew point temperature corresponding to each sample mixing water system at that time point, providing an important temperature prediction value for subsequent comprehensive prediction, thereby combining other prediction results to more comprehensively grasp the possible range of future dew point temperatures.

[0038] The third prediction model, also built using machine learning or time series analysis algorithms, is used to mine the changing patterns and trends of the historical dew point temperature series to predict a second predicted dew point temperature at a future time point. For example, by analyzing dew point temperature fluctuations over a period of time, it can predict the dew point temperature at a future time point. This prediction does not rely on indirect derivation from environmental data, but instead draws on the time series characteristics of the dew point temperature itself. This complements the first predicted dew point temperature prediction based on environmental data, providing another dimension of prediction and helping to more accurately assess the likelihood of future dew point temperatures. The third prediction model can also be constructed using several convolutional and fully connected layers.

[0039] A first weight and a second weight are introduced to reflect the reliability of different dew point temperature prediction methods. The specific values ​​of the first and second weights can be determined based on an analysis of the differences between the prediction results of the two dew point temperature prediction methods and the actual dew point temperatures corresponding to the respective time points. For example, the sum of the first differences between a number of first predicted dew point temperatures and the actual dew point temperatures corresponding to each first predicted temperature, as well as the sum of the second differences between a number of second predicted dew point temperatures and the actual dew point temperatures corresponding to each second predicted temperature, are statistically calculated. The sum of the differences is inversely proportional to the reliability. Therefore, the sum of the first differences / (the sum of the first differences + the sum of the second differences) is used as the first weight for the second prediction model, and the sum of the second differences / (the sum of the first differences + the sum of the second differences) is used as the second weight for the third prediction model.

[0040] By inserting the reference predicted dew point temperature between two predicted temperature values ​​and calculating the confidence level in combination with the weights, we can more comprehensively describe the distribution of future dew point temperatures and quantify the uncertainty of the prediction results.

[0041] As described above, the second prediction model predicts the dew point temperature from the dimension of environmental data affecting the dew point temperature, and the third prediction model predicts the dew point temperature from the dimension of the dew point temperature sequence itself. The two prediction methods complement each other to avoid the limitations of a single prediction perspective. The two prediction results are combined to set the reference predicted dew point temperature, and a confidence level is assigned to each predicted dew point temperature. The multiple possibilities of the prediction results are quantified, so that decision makers can intuitively understand the reliability of each predicted value when formulating the mixed water control strategy, fully reflect the actual situation, and reduce the prediction error.

[0042] In a specific embodiment, the predicted environmental data includes predicted relative humidity and predicted air temperature, and S112 includes the following steps:

[0043] S1121, based on the predicted air temperature of each sample water mixing system at the i+1th historical time point, obtain the predicted saturated water vapor pressure of each sample water mixing system at the i+1th historical time point;

[0044] S1122, obtaining the predicted actual water vapor pressure of each sample mixing system corresponding to the i+1th historical time point based on the predicted saturated water vapor pressure and predicted relative humidity of each sample mixing system at the i+1th historical time point;

[0045] S1123 , obtaining a first predicted dew point temperature corresponding to the i+1th historical time point for each sample mixing water system based on the predicted actual water vapor pressure corresponding to the i+1th historical time point for each sample mixing water system.

[0046] Among them, the predicted air temperature W corresponding to each sample mixing system at the i+1th historical time point is i+1 , substituted into the calculation formula of saturated water vapor pressure, and the predicted saturated water vapor pressure B corresponding to each sample mixed water system at the i+1th historical time point is obtained. i+1 =g×exp((p×W i+1 ) / (W i+1 +q)), where exp() is an exponential function with base e.

[0047] The predicted saturated vapor pressure B corresponding to the i+1th historical time point of each sample mixed water system i+1 and predicted relative humidity RH i+1 , substituted into the calculation formula of actual water vapor pressure, and the predicted actual water vapor pressure Y corresponding to each sample mixing system at the i+1th historical time point is obtained. i+1 =RH i+1 ×B i+1 .

[0048] The predicted actual water vapor pressure Y corresponding to the i+1th historical time point of each sample mixing system i+1 , substituted into the calculation formula of dew point temperature, and the first predicted dew point temperature L corresponding to each sample mixed water system at the i+1th historical time point is obtained. i+1 =(q×ln(Y i+1 / g)) / (p-ln(Y i+1 / g)), where ln() is the natural logarithm function.

[0049] Among them, g, p, and q are parameters obtained by fitting a large amount of experimental data and actual observations, so that the calculation formula of saturated water vapor pressure can more accurately describe the relationship between saturated water vapor pressure and air temperature, and the calculation formula of dew point temperature can more accurately describe the relationship between actual water vapor pressure and dew point temperature.

[0050] In one specific embodiment, in the calculation formula for the saturated water vapor pressure of the water surface, g=6.112, p=17.67, q=243.5; in the calculation formula for the saturated water vapor pressure of the ice surface, g=6.112, p=22.46, q=272.7.

[0051] In one embodiment, S13 includes the following steps:

[0052] S131 : Determine the first weight α1 corresponding to the second prediction model as the first prediction confidence corresponding to the first predicted dew point temperature of each sample mixed water system at the i+1th historical time point.

[0053] S132 : Determine the second weight α2 corresponding to the third prediction model as the first prediction confidence corresponding to the second predicted dew point temperature of each sample mixed water system at the i+1th historical time point.

[0054] In a specific embodiment, S13 further includes the following steps:

[0055] S133: The lower temperature of the first predicted dew point temperature and the second predicted dew point temperature corresponding to the i+1th historical time point of each sample mixed water system is determined as the lower temperature limit T i+1 1 , the larger temperature is determined as the upper temperature limit T i+1 2 .

[0056] S134: Obtain the jth reference dew point temperature T corresponding to the i+1th historical time point for each sample mixing system in ascending order. (i+1),j =T i+1 1 +((T i+1 2 -T i+1 1 ) / (N-1))×j, where j=1, 2,…, N-2.

[0057] S135: Based on the j-th reference dew point temperature T corresponding to the i+1-th historical time point of each sample mixing system, (i+1),j , lower limit of temperature T i+1 1 , upper temperature limit T i+1 2 , the first weight α1 corresponding to the second prediction model and the second weight α2 corresponding to the third prediction model, and obtain the j-th reference dew point temperature T corresponding to each sample mixed water system at the i+1-th historical time point (i+1),j The corresponding first prediction confidence Z (i+1),j =((T i+1 2 -T(i+1),j ) / (T i+1 2 -T i+1 1 ))×α1+((T (i+1),j -T i+1 1 ) / (T i+1 2 -T i+1 1 ))×α2.

[0058] The results from two different dew-point temperature prediction methods were initially integrated to determine the approximate range of possible future dew-point temperatures. Then, based on the principle of linear interpolation, reference predicted dew-point temperatures were generated at equal intervals within the determined temperature range. More possible temperature values ​​were added between the first and second predicted dew-point temperatures, enriching the prediction results and increasing the prediction dimension. This more comprehensively covers the potential range of future dew-point temperature values ​​to address the uncertainties in actual conditions. Furthermore, the newly added predicted values ​​were ensured to be distributed within a reasonable range, avoiding prediction results that deviate from realistic possibilities and ensuring the rationality and reliability of the prediction system.

[0059] Furthermore, based on the degree of confidence in the second and third prediction models (i.e., the first and second weights) and the relative positional relationships between the reference dew point temperatures and the first and second predicted dew point temperatures, the closer the reference dew point temperature is to the first predicted dew point temperature, i.e., the greater the influence of the second prediction model based on the environmental data, the higher the first weight α1 is assigned when calculating the second prediction confidence. Conversely, the closer the reference dew point temperature is to the second predicted dew point temperature, i.e., the greater the influence of the third prediction model based on the historical sequence of dew point temperatures, the higher the second weight α2 is assigned when calculating the first prediction confidence, thereby quantifying the reliability of each reference predicted dew point temperature.

[0060] As described above, by determining the upper and lower limits of the temperature and inserting the reference dew point temperature, a prediction set covering multiple possible temperature values ​​is constructed. It is no longer limited to the two original prediction values, and more comprehensively covers the potential value range of the future dew point temperature. It can more accurately reflect the multiple possibilities in the actual situation. Based on the degree of trust in the second prediction model and the third prediction model and the relative position relationship between the reference dew point temperature and the first predicted dew point temperature and the second predicted dew point temperature, the first prediction confidence corresponding to each predicted dew point temperature is obtained as the basis for obtaining control data, thereby improving the accuracy of the control strategy.

[0061] S2. Obtain the first prediction control data set corresponding to each sample mixing water system at the i-th historical time point based on the historical environmental data, historical mixing water temperature, and predicted dew point temperature set corresponding to the i+1-th historical time point for each sample mixing water system.

[0062] Among them, by exploring the intrinsic connections and laws between environmental changes, mixed water temperature adjustment and dew point temperature, combining the historical environmental data, historical mixed water temperature and predicted dew point temperature at future time points corresponding to each historical time point, the control measures to be taken at the i-th historical time point are inferred, thereby obtaining the first predicted control data set, which provides guidance for subsequent practical operations such as pipeline valve control, so that the mixed water temperature corresponding to the sample mixed water system reaches the corresponding target temperature at the i+1-th time point, that is, the mixed water temperature at the i+1-th time point is slightly higher than the predicted dew point temperature to prevent frost in the pipeline.

[0063] In a specific embodiment, S2 includes the following steps:

[0064] S21 , obtaining a target mixed water temperature set corresponding to each sample mixed water system at the i+1th historical time point based on the predicted dew point temperature set and the preset temperature increment corresponding to each sample mixed water system at the i+1th historical time point.

[0065] S22, input the historical environmental data, historical mixed water temperature and target mixed water temperature set corresponding to each sample mixed water system at the i-th historical time point into the preset first prediction model, and obtain the first prediction control data set corresponding to each sample mixed water system at the i-th historical time point.

[0066] Among them, the preset temperature increment is a fixed temperature value pre-set by the implementer based on the actual application scenario and the need to avoid condensation. For example, in order to ensure that there is no condensation in the pipeline, the preset temperature increment is set to 2°C, which means that the mixed water temperature is guaranteed to be 2°C higher than the dew point temperature.

[0067] By adding the preset temperature increment to each predicted dew point temperature in the predicted dew point temperature set, the target mixed water temperature set corresponding to each sample mixed water system at the i+1th historical time point is obtained. For example, if the predicted dew point temperature set for a sample mixed water system is {15°C, 16°C, 17°C} and the preset temperature increment is 2°C, then the target mixed water temperature set is {17°C, 18°C, 19°C}.

[0068] The first prediction model is also a model constructed based on machine learning or time series analysis algorithms. By studying historical data, it mines the intrinsic mapping relationship between the historical environmental data, historical mixed water temperature, target mixed water temperature, and the required control measures corresponding to each historical time point. Therefore, when new historical environmental data, historical mixed water temperature, and target mixed water temperature are input, the first prediction model determines what control measures need to be taken at the current historical time point in order to achieve the target mixed water temperature based on the learned rules, and thus outputs the corresponding control data. For example, in the historical data, the first prediction model learns that when the ambient humidity increases, the mixed water temperature is low, and the target mixed water temperature is high, the hot water valve opening needs to be increased and the cold water valve opening needs to be decreased. When new data is input, the first prediction model can make judgments and outputs based on this rule. The valve opening is the degree to which the valve is open.

[0069] The first prediction model can also be constructed by several convolutional layers and fully connected layers.

[0070] It can be understood that in order to ensure the generalization and output accuracy of the first prediction model, the second prediction model, and the third prediction model, the implementer can train the first prediction model, the second prediction model, and the third prediction model. The number of convolutional layers in the first prediction model, the second prediction model, and the third prediction model can be set by the implementer according to actual conditions. For example, the number of convolutional layers can be set to 3 or 5. The model structure and model training method of the first prediction model, the second prediction model, and the third prediction model are all existing technologies and will not be further described here.

[0071] As described above, by predicting the dew point temperature and combining it with the preset temperature increment to determine the target mixed water temperature set, an accurate temperature control target that meets the anti-condensation requirements is set for the mixed water system. With the help of the preset first prediction model, the multi-dimensional historical data and the target mixed water temperature are quickly converted into the first prediction control data set, which improves the adaptability and accuracy of the control target and ensures the safe operation of the mixed water system.

[0072] S3. Based on the historical environmental data, historical mixed water temperature, historical dew point temperature and initial control model corresponding to each sample mixed water system at the i-th historical time point, obtain the second prediction control data set corresponding to each sample mixed water system at the i-th historical time point and the second prediction confidence set corresponding to the second prediction control data set.

[0073] The initial control model is a machine learning algorithm model constructed based on the basic operating principles and preliminary settings of the water mixing system. It includes the basic rules and logic for operations such as hot and cold water mixing ratio adjustment and valve control. By inputting historical environmental data, historical mixed water temperature, and historical dew point temperature into the initial control model, the initial control model simulates the operation of the water mixing system under the input historical conditions based on its internal algorithm and learned rules. It then predicts several control measures that should be taken at the i-th historical time point, namely the second set of predicted control data. Simultaneously, by analyzing the control effect corresponding to each second set of predicted control data and evaluating its reliability, the corresponding second set of prediction confidence levels is generated.

[0074] The initial control model can also be constructed from several convolutional layers and fully connected layers. The number of convolutional layers in the initial control model can be set by the implementer based on actual conditions. For example, the number of convolutional layers can be set to 3 or 5. The structure of the initial control model is well known in the art and will not be further described here.

[0075] As mentioned above, by integrating historical environmental data, historical mixed water temperature and historical dew point temperature, fully utilizing the data generated by the system during historical operation, and inputting it into the initial control model for analysis, the operating rules and experience of the mixed water system under different environmental conditions are excavated, and the second predictive control data set and the corresponding second predictive confidence set are obtained, which provides more comprehensive information, so that when formulating the mixed water control strategy, more scientific and reasonable decisions can be made in combination with the confidence level, thereby improving the accuracy of the control strategy.

[0076] S4, traverse all sample mixed water systems and M historical time points, train the initial control model according to the first prediction control data set, first prediction confidence set, second prediction control data set and second prediction confidence set corresponding to each sample mixed water system at each historical time point, and obtain the trained target control model.

[0077] Among them, for each sample water mixing system and historical time point, the first predictive control data set and its corresponding first predictive confidence set are obtained based on the historical environmental data sequence and historical dew point temperature sequence corresponding to each sample water mixing system at the ikth to i-th historical time points, the historical environmental data and historical mixed water temperature corresponding to each sample water mixing system at the i-th historical time point, and the second predictive control data set and its corresponding second predictive confidence set are obtained based on the historical environmental data, historical mixed water temperature, historical dew point temperature and initial control model corresponding to the sample water mixing system at the i-th historical time point, reflecting from multiple angles the control strategy that should be adopted by the water mixing system in different environments and operating conditions and the reliability of the control strategy.

[0078] Furthermore, by traversing all sample mixed water systems and historical time points, the historical environmental data, historical mixed water temperature, and historical dew point temperature corresponding to the sample mixed water system at the i-th historical time point are used as input data, the corresponding second prediction control data set machine and the second prediction confidence set corresponding to the obtained second prediction control data set are used as the output result corresponding to the initial control model, and the first prediction control data set and its corresponding first prediction confidence set are used as the expected results corresponding to each sample mixed water system and historical time point. The initial control model is iteratively adjusted using an optimization algorithm (such as a gradient descent algorithm) to minimize the error between the model output result and the actual expected result, so that the initial control model can learn a more accurate control strategy prediction model, and finally a trained target control model is obtained.

[0079] The training of the initial control model enables the target control model to more accurately and efficiently predict the appropriate mixed water control strategy based on the data collected by the sensor. That is, there is no need to predict the dew point temperature after the current time point based on the environmental data and dew point temperature before the current time point, and then combine the environmental data and mixed water temperature at the current time point to predict the control data. Instead, it is only necessary to predict the control data based on the environmental data, mixed water temperature and dew point temperature at the current time point. This greatly simplifies the calculation logic of the target control model, reduces the error accumulation caused by the intermediate prediction link, and enables the target control model to more efficiently and accurately output the appropriate mixed water control strategy based on the current data collected by the sensor in real time, thereby improving the accuracy of the control strategy and the efficiency of the control process.

[0080] As mentioned above, the initial control model is trained by traversing a large number of sample mixed water systems and predictive control data sets and their prediction confidence sets at historical time points, providing training information for the model from multiple dimensions, allowing the initial control model to fully learn the operating laws and control strategies of the mixed water system, so that the target control model can more accurately and efficiently predict the appropriate mixed water control strategy based on the data collected by the sensor, thereby improving the accuracy of the control strategy and the efficiency of the control process.

[0081] In a specific embodiment, S4 includes the following steps:

[0082] S41. For any sample water mixing system, based on the first prediction control data set, the first prediction confidence set, the second prediction control data set and the second prediction confidence set corresponding to the current sample water mixing system at the i-th historical time point, the model sub-loss of the initial control model for the current sample water mixing system and the i-th historical time point is obtained.

[0083] S42, traverse all sample mixed water systems and historical time points from the k+1th to the Mth, and determine the sum of the model sub-losses of the initial control model for each sample mixed water system and each historical time point as the total model loss corresponding to the initial control model.

[0084] S43, updating the parameters of the initial control model according to the total model loss until the total model loss converges, and obtaining the trained target control model.

[0085] The first and second predictive control data sets represent the control strategy results obtained using different prediction methods, while the corresponding confidence sets reflect the reliability of the control data. By combining the control data and confidence data, we can more comprehensively assess the deviation between the model prediction and actual demand, deriving the model sub-loss to measure the model's prediction accuracy for that sample water mixing system and time point. The model sub-losses for each sample water mixing system at each historical time point are then accumulated to obtain the total model loss, which is then used to update the parameters of the initial control model.

[0086] Specifically, the total model loss is a function of the model parameters. By calculating the gradient of the loss function with respect to the model parameters, the direction and step size of parameter adjustment can be determined, so that the model parameters are updated in a direction that reduces the total model loss. When the total model loss converges, it reflects that the prediction error of the control model being trained on the current sample mixed water system dataset has been minimized. At this point, the control model is the trained target control model and can more accurately predict the appropriate mixed water control strategy based on the input data.

[0087] In a specific embodiment, S41 includes the following steps:

[0088] S411 , obtaining a first sub-loss of an initial control model for the current sample mixed water system and the i-th historical time point based on a first prediction control data set and a second prediction control data set corresponding to the current sample mixed water system at the i-th historical time point.

[0089] S412 , obtaining the second sub-loss of the initial control model for the current sample mixed water system and the i-th historical time point based on the first prediction confidence set and the second prediction confidence set corresponding to the current sample mixed water system at the i-th historical time point.

[0090] S413, based on the first sub-loss and second sub-loss of the initial control model for the current sample mixed water system and the i-th historical time point, as well as the third weight corresponding to the first sub-loss and the fourth weight corresponding to the second sub-loss, obtain the model sub-loss of the initial control model for the current sample mixed water system and the i-th historical time point.

[0091] The difference between the predicted control data at corresponding positions in the first and second predicted control data sets is calculated, and the mean square error is calculated. This is used as the first sub-loss of the initial control model for the current sample mixed water system and the i-th historical time point. The difference between the predicted confidence data at corresponding positions in the first and second prediction confidence sets is calculated, and the mean square error is calculated. This is used as the second sub-loss of the initial control model for the current sample mixed water system and the i-th historical time point, to assist in determining the accuracy of the initial control model's prediction for the sample mixed water system and time point.

[0092] The third weight represents the importance of the data in the predictive control dimension in characterizing the output accuracy of the initial control model. The fourth weight represents the importance of the data in the prediction confidence dimension in characterizing the output accuracy of the initial control model. The specific values ​​of the third and fourth weights can be set by the implementer based on actual conditions. For example, the third weight can be 0.6 and the fourth weight can be 0.4.

[0093] As mentioned above, the calculation of the model sub-loss is split into two dimensions: prediction error and prediction confidence, and then weighted integration is performed. Compared with a single calculation method, this method more carefully evaluates the performance of the model in specific sample mixed water systems and historical time points, can more accurately locate the problems of the model, provide a more accurate direction for subsequent targeted optimization, and improve the training efficiency and output accuracy of the target control model.

[0094] S5. When the real-time environmental data, real-time mixed water temperature and real-time dew point temperature corresponding to the target mixed water system at the current time point are obtained, the first valve and the second valve of the target mixed water system are controlled according to the real-time environmental data, real-time mixed water temperature, real-time dew point temperature and the trained target control model.

[0095] In a specific embodiment, S5 includes the following steps:

[0096] S51, input the real-time environmental data, real-time mixed water temperature and real-time dew point temperature into the trained target control model to obtain the target control data corresponding to the target mixed water system at the current time point, wherein the target control data includes the first valve opening and the second valve opening.

[0097] S52 : Control the opening of the first valve and the opening of the second valve of the target water mixing system according to the target control data corresponding to the target water mixing system at the current time point.

[0098] The target water mixing system is a water mixing system to be controlled that belongs to the same specific water mixing system as the sample water mixing system. The first valve and the second valve correspond to valves of a cold water pipe and a hot water pipe, respectively.

[0099] The actuator converts the target control data into physical actions, changing the opening of the first and second valves to adjust the flow of hot and cold water and control the mixed water temperature. This ensures that the mixed water temperature is higher than the dew point, and that the difference between the mixed water temperature and the dew point temperature meets the preset temperature increment, thereby preventing condensation in the pipes and ensuring safe operation of the equipment. The actuator includes a motor, transmission device, and other devices.

[0100] As mentioned above, by inputting real-time data into the target control model and executing output instructions, a fully automated and precise control process of the water mixing system from data acquisition to valve adjustment is realized, which improves the accuracy of the control strategy and the efficiency of the control process.

[0101] In the above, by analyzing the historical environmental data sequence and the historical dew point temperature sequence of the sample mixed water system from the ikth to the ith historical time point, the complex relationship between environmental factors and dew point temperature is fully captured, and the accuracy of predicting future dew point temperature and the first prediction confidence set is improved; the first prediction control data set is obtained by combining the historical environmental data, the historical mixed water temperature and the predicted dew point temperature set, and the second prediction control data set and the second prediction confidence set are obtained by combining the historical environmental data, the historical mixed water temperature, the historical dew point temperature and the initial control model, and the operating laws and experiences of the mixed water system under different environmental conditions are excavated from different angles, and the initial control is carried out according to the prediction control data and confidence data corresponding to different angles. The model is trained, which greatly simplifies the operation logic of the target control model and reduces the error accumulation caused by the intermediate prediction link, so that the target control model can more efficiently and accurately output the appropriate mixing water control strategy based on the current data collected by the sensor in real time, thereby improving the accuracy of the control strategy and the efficiency of the control process. When the real-time environmental data, real-time mixing water temperature and real-time dew point temperature of the target mixing water system are obtained, the first valve and the second valve of the target mixing water system are controlled by the target control model to ensure that the mixing water temperature is slightly higher than the dew point temperature, thereby preventing condensation in the pipeline, and realizing a fully automated and precise control process of the mixing water system from data collection to valve adjustment, thereby improving the accuracy of the control strategy and the efficiency of the control process.

[0102] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any form. Although the present invention has been disclosed as above in terms of preferred embodiments, they are not intended to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for automatic water mixing control based on dew point temperature, characterized in that: The automatic water mixing control method based on dew point temperature comprises the following steps: S1. Obtain, based on the historical environmental data sequence and the historical dew point temperature sequence corresponding to each sample water mixing system at the i+1th historical time point, a predicted dew point temperature set corresponding to each sample water mixing system and a first prediction confidence set corresponding to the predicted dew point temperature set, where i=k+1, k+2, ..., M, where M is the total number of historical time points and k is the number of historical time points corresponding to the historical environmental data sequence and the historical dew point temperature sequence; S2, obtaining a first prediction control data set corresponding to each sample mixing water system at the i-th historical time point based on the historical environmental data, historical mixing water temperature, and predicted dew point temperature set corresponding to the i+1-th historical time point for each sample mixing water system; S3, obtaining a second prediction control data set corresponding to each sample mixed water system at the i-th historical time point and a second prediction confidence set corresponding to the second prediction control data set based on the historical environmental data, historical mixed water temperature, historical dew point temperature, and initial control model corresponding to each sample mixed water system at the i-th historical time point; S4, traversing all sample mixed water systems and M historical time points, and training the initial control model according to the first prediction control data set, the first prediction confidence set, the second prediction control data set, and the second prediction confidence set corresponding to each sample mixed water system at each historical time point, to obtain a trained target control model; S5. When the real-time environmental data, real-time mixed water temperature and real-time dew point temperature corresponding to the target mixed water system at the current time point are obtained, the first valve and the second valve of the target mixed water system are controlled according to the real-time environmental data, the real-time mixed water temperature, the real-time dew point temperature and the trained target control model.

2. The automatic water mixing control method based on dew point temperature according to claim 1, characterized in that: S1 includes the following steps: S11, obtaining a first predicted dew point temperature corresponding to the i+1th historical time point for each sample mixed water system based on a historical environmental data sequence consisting of k+1 historical environmental data corresponding to the i-th to i-th historical time points for each sample mixed water system and a preset second prediction model; S12: Inputting a historical dew point temperature sequence consisting of the k+1 historical dew point temperatures corresponding to the ikth to ith historical time points of each sample mixed water system into a preset third prediction model to obtain a second predicted dew point temperature corresponding to the i+1th historical time point of each sample mixed water system; S13: Obtaining, based on the first predicted dew point temperature, the second predicted dew point temperature, the first weight corresponding to the second prediction model, the second weight corresponding to the third prediction model, and a preset number of set elements N, N-2 reference predicted dew point temperatures within a temperature range between the first predicted dew point temperature and the second predicted dew point temperature, a first prediction confidence corresponding to the first predicted dew point temperature, a first prediction confidence corresponding to the second predicted dew point temperature, and a first prediction confidence corresponding to each reference predicted dew point temperature; S14, combining the first predicted dew point temperature, the second predicted dew point temperature, and N-2 reference predicted dew point temperatures in ascending order to form a predicted dew point temperature set corresponding to the i+1th historical time point for each sample mixed water system; S15. In the order corresponding to the predicted dew point temperature set, a first prediction confidence set corresponding to the predicted dew point temperature set is formed according to the first prediction confidence corresponding to the first predicted dew point temperature, the first prediction confidence corresponding to the second predicted dew point temperature, and the first prediction confidence corresponding to each reference predicted dew point temperature.

3. The automatic water mixing control method based on dew point temperature according to claim 1, characterized in that: S11 includes the following steps: S111: Inputting a historical environmental data sequence consisting of k+1 historical environmental data corresponding to each sample mixed water system at the ikth to ith historical time points into a preset second prediction model to obtain the predicted environmental data corresponding to each sample mixed water system at the i+1th historical time point; S112 , obtaining a first predicted dew point temperature corresponding to the i+1th historical time point for each sample mixed water system based on the predicted environmental data corresponding to the i+1th historical time point for each sample mixed water system.

4. The automatic water mixing control method based on dew point temperature according to claim 3, characterized in that: The predicted environmental data includes predicted relative humidity and predicted air temperature, and S112 includes the following steps: S1121, based on the predicted air temperature of each sample water mixing system at the i+1th historical time point, obtain the predicted saturated water vapor pressure of each sample water mixing system at the i+1th historical time point; S1122, obtaining the predicted actual water vapor pressure of each sample mixing system corresponding to the i+1th historical time point based on the predicted saturated water vapor pressure and predicted relative humidity of each sample mixing system at the i+1th historical time point; S1123 , obtaining a first predicted dew point temperature corresponding to the i+1th historical time point for each sample mixing water system based on the predicted actual water vapor pressure corresponding to the i+1th historical time point for each sample mixing water system.

5. The automatic water mixing control method based on dew point temperature according to claim 2, characterized in that: S13 includes the following steps: S131, determining a first weight α1 corresponding to the second prediction model as a first prediction confidence corresponding to the first predicted dew point temperature of each sample mixed water system corresponding to the i+1th historical time point; S132: Determine the second weight α2 corresponding to the third prediction model as the first prediction confidence corresponding to the second predicted dew point temperature of each sample mixed water system at the i+1th historical time point.

6. The automatic water mixing control method based on dew point temperature according to claim 5, characterized in that: S13 further includes the following steps: S133: The lower temperature of the first predicted dew point temperature and the second predicted dew point temperature corresponding to the i+1th historical time point of each sample mixed water system is determined as the lower temperature limit T. i+1 1 , the larger temperature is determined as the upper temperature limit T i+1 2 ; S134: Obtain the jth reference dew point temperature T corresponding to the i+1th historical time point for each sample mixing system in ascending order. (i+1),j =T i+1 1 +((T i+1 2 -T i+1 1 ) / (N-1))×j, where j=1, 2, ..., N-2; S135: Based on the j-th reference dew point temperature T corresponding to the i+1-th historical time point of each sample mixing system, (i+1),j , the lower limit of temperature T i+1 1 , the upper temperature limit T i+1 2 , the first weight α1 corresponding to the second prediction model and the second weight α2 corresponding to the third prediction model, and obtain the j-th reference dew point temperature T corresponding to each sample mixed water system at the i+1-th historical time point (i+1),j The corresponding first prediction confidence Z (i+1),j =((T i+1 2 -T (i+1),j ) / (T i+1 2 -T i+1 1 ))×α1+((T (i+1),j -T i+1 1 ) / (T i+1 2 -T i+1 1 ))×α2.

7. The automatic water mixing control method based on dew point temperature according to claim 1, characterized in that: S2 includes the following steps: S21, obtaining a target mixed water temperature set corresponding to each sample mixed water system at the i+1th historical time point based on the predicted dew point temperature set and the preset temperature increment corresponding to each sample mixed water system at the i+1th historical time point; S22, input the historical environmental data, historical mixed water temperature and target mixed water temperature set corresponding to each sample mixed water system at the i-th historical time point into the preset first prediction model, and obtain the first prediction control data set corresponding to each sample mixed water system at the i-th historical time point.

8. The automatic water mixing control method based on dew point temperature according to claim 1, characterized in that: S4 includes the following steps: S41, for any sample water mixing system, obtaining a model sub-loss of the initial control model for the current sample water mixing system and the i-th historical time point based on the first prediction control data set, the first prediction confidence set, the second prediction control data set, and the second prediction confidence set corresponding to the current sample water mixing system at the i-th historical time point; S42, traversing all sample mixed water systems and historical time points (k+1) to (M), and determining the sum of the model sub-losses of the initial control model for each sample mixed water system and each historical time point as the total model loss corresponding to the initial control model; S43, updating the parameters of the initial control model according to the total loss of the model until the total loss of the model converges, thereby obtaining a trained target control model.

9. The automatic water mixing control method based on dew point temperature according to claim 8, characterized in that: S41 includes the following steps: S411, obtaining a first sub-loss of the initial control model for the current sample water mixing system and the i-th historical time point based on a first predictive control data set and a second predictive control data set corresponding to the current sample water mixing system at the i-th historical time point; S412, obtaining the second sub-loss of the initial control model for the current sample mixed water system and the i-th historical time point based on the first prediction confidence set and the second prediction confidence set corresponding to the current sample mixed water system at the i-th historical time point; S413, according to the first sub-loss and the second sub-loss of the initial control model for the current sample mixed water system and the i-th historical time point, as well as the third weight corresponding to the first sub-loss and the fourth weight corresponding to the second sub-loss, obtain the model sub-loss of the initial control model for the current sample mixed water system and the i-th historical time point.

10. The automatic water mixing control method based on dew point temperature according to claim 1, characterized in that: S5 includes the following steps: S51, inputting the real-time environmental data, the real-time mixed water temperature, and the real-time dew point temperature into the trained target control model to obtain target control data corresponding to the target mixed water system at the current time point, wherein the target control data includes a first valve opening and a second valve opening; S52 : Control the opening of the first valve and the opening of the second valve of the target water mixing system according to the target control data corresponding to the target water mixing system at the current time point.

Images