Secondary network water supply temperature optimization method based on physical model and data-driven model

Abstract

The application relates to a secondary network water supply temperature optimization method based on a physical model and a data-driven model, which comprises the following steps: S01) obtaining operation data and meteorological data based on historical heating data; S02) building a data-driven model; S03) determining an initial water supply temperature based on a physical model, and obtaining discrete water supply temperatures after discretization; S04) substituting the discrete water supply temperatures into the data-driven model to obtain a plurality of indoor temperature prediction values and return water temperature prediction values; S05) designing a controller to evaluate the cost of the indoor temperature prediction values and the return water temperature prediction values, searching for the optimal combination of the indoor temperature prediction values and the return water temperature prediction values through an optimizer, and outputting an actual water supply temperature through the controller; the application can obtain an optimized water supply temperature value by balancing various factors, can solve the problems of excessive heating or insufficient heating, and can optimize the heating mode.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of central heating, and particularly relates to a secondary network water supply temperature optimization method based on a physical model and a data-driven model. BACKGROUND

[0002] In a central heating system, due to the complex heat demand of users and the influence of multiple factors such as heating network layout, building maintenance structure and climate, most heating enterprises determine the water supply temperature according to experience. The judgment result obtained by determining the water supply temperature according to experience is greatly influenced by subjective factors, and lacks a global perspective for the water supply temperature judgment under the comprehensive action of multiple factors, which is prone to cause the situation that the indoor temperature does not meet the standard or excessive heating causes energy waste. SUMMARY

[0003] The application provides a secondary network water supply temperature optimization method based on a physical model and a data-driven model, which can balance multiple factors to obtain an optimized water supply temperature value and solve the problems of excessive heating or insufficient heating.

[0004] The secondary network water supply temperature optimization method based on a physical model and a data-driven model comprises the following steps:

[0005] S01) Based on the historical heating data of the secondary network, the running data of the secondary network in the temperature control time interval and the meteorological data are obtained; S02) a data-driven model is built, the driving model comprising: an indoor temperature prediction model for predicting the indoor temperature of the user end, and a return water temperature prediction model for predicting the return water temperature; S03) the initial water supply temperature is determined based on the physical model, the determined initial water supply temperature is discretized to obtain a plurality of discrete water supply temperatures; S04) the discrete water supply temperature is substituted into the data-driven model to obtain a plurality of indoor temperature prediction values and return water temperature prediction values; S05) a controller is designed to evaluate the cost of the indoor temperature prediction value and the return water temperature prediction value, and an optimizer is used to search for the optimal combination of the indoor temperature prediction value and the return water temperature prediction value, and the actual water supply temperature is output based on the optimal combination of the indoor temperature prediction value and the return water temperature prediction value by the controller.

[0006] Further, the running data comprises the water supply temperature, the return water temperature and the indoor temperature; and the meteorological data comprises the indoor-outdoor temperature difference, the weather condition, the wind direction, the wind power, the relative humidity and the outdoor temperature prediction value.

[0007] Further, in the step S02), the indoor temperature prediction model is built based on an LSTM neural network, and the return water temperature prediction model is built based on a BP neural network.

[0008] Further, when outputting the indoor temperature prediction value at a target time point by the indoor temperature prediction model, the parameters input into the indoor temperature prediction model include: a current time point value, time point values corresponding to five temperature regulation time intervals before the current time point, and outdoor temperatures, indoor-outdoor temperature differences, weather conditions, wind directions, wind forces, relative humidities, water supply temperatures, return water temperatures at a previous time point, and outdoor temperature prediction values corresponding to the aforementioned six time points respectively; wherein the target time point is a time point corresponding to one temperature regulation time interval after the current time point.

[0009] Further, when outputting the return water temperature prediction value at a target time point by the return water temperature prediction model, the parameters input into the return water temperature prediction model include the outdoor temperature at the target time point and the water supply temperature at the target time point.

[0010] Further, the physical model in the step S03) is established based on a steady-state heat balance equation, and the formula is as follows:

[0011] ;

[0012] In the formula, t g is an initial water supply temperature, and the unit is ℃; t n is the indoor temperature of the user, and the unit is ℃; t’ g is the water supply temperature of the secondary network to the user, and the unit is ℃; t’ h is the return water temperature collected by the secondary network, and the unit is ℃; b is a radiator heat transfer index; is a relative heating heat load ratio.

[0013] Further, the controller is an MPC, and the objective function of the MPC is as follows:

[0014] ;

[0015] In the formula, J is an optimization target, is a penalty factor, tn is an indoor temperature prediction value, tnaim is a set target temperature, Tg is an initial water supply temperature, Th is a return water temperature prediction value, is a slack variable, N is a prediction time point, n is a current time point, and the subscript in the function indicates a time step index; the result of the optimization target is an actual water supply temperature value.

[0016] Further, the objective function constrains the initial water supply temperature, and the constraint interval is [31, 51] ℃.

[0017] Due to the adoption of the above technical solutions, the application has the following beneficial effects:

[0018] 1. The application uses physical model localization, data-driven model solution, and controller result optimization. Compared with the existing technology that only relies on physical model calculation of water supply temperature, the technical solution adopted by the application can balance meteorological data and pipe network operation data, make the actual water supply temperature independent of subjective worker experience, and make the water supply temperature setting value more objective.

[0019] 2. After the initial water supply temperature generated by the physical model, the data-driven model is used to predict the discrete water supply temperature to realize the interaction between the physical model and the data-driven model, and finally the controller is used to evaluate and optimize the result after the interaction. The multi-dimensional data and the optimized result can reduce the fluctuation of indoor temperature, make the indoor temperature tend to be stable near the target value, and also facilitate temperature control, thereby reducing the temperature difference between the water supply temperature and the return water temperature and improving the heating efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0020] The drawings described herein are used to provide further understanding of the application, and form a part of the application. The illustrative embodiments of the application and their descriptions are used to explain the application, and do not constitute an improper limitation on the application. In the drawings:

[0021] Figure 1 A flowchart of the secondary network water supply temperature optimization method based on the physical model and the data-driven model is provided. DETAILED DESCRIPTION

[0022] As described in the background, the secondary network, as an important part of the central heating system, is the link between the user and the heat source, and the high or low of the secondary network water supply temperature directly affects the user's heating experience and also concerns the energy loss of the heat source. The water supply temperature determined by experience is subjective and difficult to balance various factors. For example, when the wind direction, wind power and outdoor temperature change greatly at the same time, it is difficult for workers to determine which factor to dominate to regulate the water supply temperature, and inappropriate temperature regulation will affect comfort and may cause energy waste.

[0023] In view of the above, as shown in the accompanying Figure 1 The application provides a secondary network water supply temperature optimization method based on a physical model and a data-driven model, comprising the following steps:

[0024] S01) Based on the historical heating data of the secondary network, the operation data and meteorological data of the secondary network in the temperature regulation time interval are obtained.

[0025] Preferably, the operation data comprises water supply temperature, return water temperature and indoor temperature; and the meteorological data comprises indoor-outdoor temperature difference, weather condition, wind direction, wind force, relative humidity and outdoor temperature prediction value.

[0026] The operation data can be exported by the heating system management platform, and after the data is exported, the missing values and abnormal values of the data should be processed. In addition, the data also needs to be divided to determine the training set and the test set for the subsequent steps.

[0027] In one specific embodiment of the present application, the operation data and meteorological data of a certain community in 2021 and 2022 are used, and after the data is preprocessed, the data is divided into a training set and a data set in a ratio of 8:2.

[0028] S02) Building a data-driven model, the driving model comprises: an indoor temperature prediction model for predicting the indoor temperature of the user end, and a return water temperature prediction model for predicting the return water temperature.

[0029] Preferably, the indoor temperature prediction model is built based on the LSTM neural network, and the return water temperature prediction model is built based on the BP neural network.

[0030] In one specific embodiment of the present application, in the indoor temperature prediction model built based on the LSTM neural network, a predicted indoor temperature prediction value at a certain time is needed to be output, and the time to be predicted is called the target time. The parameters input into the indoor temperature prediction model include: the current time value, the time values corresponding to the previous five temperature control time intervals respectively, and the outdoor temperature, the indoor-outdoor temperature difference, the weather condition, the wind direction, the wind force, the relative humidity, the water supply temperature, the return water temperature and the outdoor temperature prediction value corresponding to the aforementioned six times respectively; wherein the target time is the time corresponding to one temperature control time interval backward from the current time. The temperature control time interval of a certain community is 1h, that is, in the above parameters, the time corresponding to the return water temperature of the water supply temperature is different by 1h.

[0031] During training, the model is trained by selecting appropriate features, and finally the indoor temperature prediction value is output.

[0032] In one specific embodiment of the present application, in the return water temperature prediction model built based on the BP neural network, when the return water temperature prediction model outputs a return water temperature prediction value at a target time, the parameters input into the return water temperature prediction model include the outdoor temperature at the target time and the water supply temperature at the target time.

[0033] When the temperature control time interval is 1h, the predicted return water temperature is the return water temperature 1h backward from the target time.

[0034] It should be noted that the present application does not limit a specific deep learning method, and an output model capable of outputting indoor temperature prediction values and return water temperature prediction values can also be established based on an LSTM neural network.

[0035] S03) determining an initial water supply temperature based on a physical model, discretizing the determined initial water supply temperature to obtain a plurality of discrete water supply temperatures.

[0036] Preferably, the physical model is established based on a steady-state heat balance equation, and the formula is as follows:

[0037] ;

[0038] In the formula, t g is the initial water supply temperature, and the unit is ℃; t n is the indoor temperature of the user, and the unit is ℃; t’ g is the water supply temperature of the secondary network to the user, and the unit is ℃; t’ h is the return water temperature collected by the secondary network, and the unit is ℃; b is the radiator heat transfer index; is the relative heating heat load ratio.

[0039] In the above formula, The calculation formula of is as follows:

[0040] ;

[0041] In the formula, t w is the outdoor temperature under the operating condition, t’ w is the heating outdoor calculation temperature, is the relative flow ratio.

[0042] It should be noted that when the relative flow ratio is 1, can be derived by t w , t’ w and t n The above physical model is derived, and in specific implementation, the parameters of the above formula can select appropriate values according to actual conditions.

[0043] In one specific embodiment of the present application, t n is taken as 24.2 ℃, and the unit is ℃; t’ g is taken as 45 ℃; t’ h is taken as 40 ℃; b is taken as 1.274; t’w Taking -5.91℃, after calculation, the initial water supply temperature is discretized, and then expanded upwards and downwards by 1.5℃. Therefore, the set formed by the discretized water supply temperatures is: [ t g - 1.5, t g -1, t g - 0.5, t g - 1.5, t g , t g + 0.5, t g + 1, t g + 1.5];

[0044] S04) Substitute the discrete water supply temperature into the data-driven model to obtain several indoor temperature prediction values ​​and return water temperature prediction values.

[0045] Substitute the discrete water supply temperature into the data-driven model built in step S02) to output the predicted indoor temperature and the predicted return water temperature.

[0046] S05) The controller is designed to evaluate the cost of the indoor temperature prediction and the return water temperature prediction, search for the optimal combination of the indoor temperature prediction and the return water temperature prediction through the optimizer, and output the actual water supply temperature through the controller.

[0047] In one specific embodiment of this application, the controller is MPC, which stands for Model Predictive Control. The objective of MPC is to minimize the heat consumption of the heating system while minimizing the deviation between the indoor temperature and the set temperature across all future prediction intervals. Therefore, the objective function of MPC established in this embodiment is as follows:

[0048] ;

[0049] In the formula, J To optimize the objective, As a penalty factor, tn This is the predicted indoor temperature value. tnaim For the set target temperature, Tg For water supply temperature, Th This is the predicted return water temperature. Let N be the slack variable and N be the prediction time. nFor the current time, when N takes 2, the value representing the optimization index is not only related to the state of the next time but also related to the states of the next three times, and the subscript in the function represents the time step index; the result of the optimization target is the actual water supply temperature value.

[0050] In the embodiment, the objective function needs to be constrained to adapt to the actual working condition. The specific constraint is that the initial water supply temperature is constrained, and the constraint interval is [31, 51] ℃; the penalty factor takes 1000 to ensure that the indoor temperature prediction value is close to the set target temperature; the target temperature is set to 24.2 ℃ which is more comfortable in terms of body feeling; N takes 2 to represent that the optimization target is related to the states of the next three times; the slack variable takes 50, indicating that a small violation of the water supply temperature is allowed; when N takes 2, the prediction interval takes 3h; the control interval takes 1h, although the building has thermal inertia, but 3 hours of time is also enough to meet the temperature change of most cases, and too large prediction interval will also lead to too large calculation amount, and the control interval takes the time of 1h adjusted by artificial experience, and too small control interval will also lead to the increase of calculation amount.

[0051] After the objective function is used to evaluate each indoor temperature prediction value and return water temperature prediction value, the ESM optimizer searches for the optimal combination of temperature prediction value and return water temperature prediction value. ESM is an algorithm for generating and testing, which finds the best candidate option by searching all candidate objects. After the solution (actual water supply temperature) of the MPC controller is discretized, the ESM optimizer will generate m candidate spaces at each time. Since the prediction interval of MPC is 3h (3 temperature regulation time intervals), the number of candidate solutions for the entire prediction interval is m x m x m, and the optimizer determines the optimal candidate combination from the search space. Since the control interval takes 1h, in the specific implementation, only the solution of the objective function at the first time in the optimal combination is implemented, and this solution is the finally determined actual water supply temperature.

[0052] The places not described in the application can be implemented or referred to the existing technology. Each embodiment in the specification is described in a progressive manner, and the same and similar parts between each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments.

[0053] The above only describes the embodiments of the application and is not used to limit the application. The application can have various changes and modifications for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the scope of claims of the application.

Claims

1. A method for optimizing the temperature of secondary network water supply based on physical models and data-driven models, characterized in that, Includes the following steps: S01) Based on the historical heating data of the secondary network, obtain the operating data of the secondary network and meteorological data during the temperature control time interval; S02) Build a data-driven model, the driving model including: an indoor temperature prediction model for predicting the indoor temperature at the user end, and a return water temperature prediction model for predicting the return water temperature. In step S02), an indoor temperature prediction model is built based on an LSTM neural network, and a return water temperature prediction model is built based on a BP neural network. When the indoor temperature prediction model outputs a predicted value of the indoor temperature at a target time, the parameters input into the indoor temperature prediction model include: the current time value, the time values ​​corresponding to the previous five temperature control time intervals, and the outdoor temperature, indoor-outdoor temperature difference, weather conditions, wind direction, wind force, relative humidity, water supply temperature, return water temperature at the previous time, and the predicted outdoor temperature value corresponding to the aforementioned six times; wherein, the target time is the time corresponding to the next temperature control time interval from the current time; When the return water temperature prediction model outputs a predicted value of the return water temperature at a target time, the parameters input into the return water temperature prediction model include the outdoor temperature at the target time and the water supply temperature at the target time. S03) Determine the initial water supply temperature based on the physical model, and discretize the determined initial water supply temperature to obtain several discrete water supply temperatures. The physical model in step S03) is based on the steady-state thermal equilibrium equation, as shown in the following formula: ； In the formula, t g This is the initial water supply temperature, in °C. t n The indoor temperature for the user is expressed in °C. t’ g The water supply temperature from the secondary network to users is expressed in °C. t’ h The temperature of the return water collected by the secondary network is expressed in °C. b The heat transfer index of the radiator; This refers to the relative heating load ratio; S04) Substitute the discrete water supply temperature into the data-driven model to obtain several indoor temperature prediction values ​​and return water temperature prediction values. S05) The controller is designed to evaluate the cost of the indoor temperature prediction and the return water temperature prediction, and searches for the optimal combination of the indoor temperature prediction and the return water temperature prediction through the optimizer. Based on the optimal combination of the indoor temperature prediction and the return water temperature prediction, the controller outputs the actual water supply temperature.

2. The method for optimizing secondary network water supply temperature based on a physical model and a data-driven model according to claim 1, characterized in that, The operational data includes: water supply temperature, return water temperature, and indoor temperature; the meteorological data includes: indoor and outdoor temperature difference, weather conditions, wind direction, wind force, relative humidity, and predicted outdoor temperature.

3. The method for optimizing secondary network water supply temperature based on a physical model and a data-driven model according to claim 1, characterized in that, The controller is an MPC, and the objective function of the MPC is as follows: ； In the formula, J To optimize the objective, As a penalty factor, tn This is the predicted indoor temperature value. tnaim For the set target temperature, Tg The initial water supply temperature, Th This is the predicted return water temperature. The variables are slack variables, N is the prediction time, n is the current time, and the subscript in the function represents the time step index; the result of the optimization objective is the actual water supply temperature value.

4. The method for optimizing secondary network water supply temperature based on a physical model and a data-driven model according to claim 3, characterized in that, The objective function constrains the initial water supply temperature within the range of [31, 51]℃.

Images