A real-time optimization method for heat source temperature in large-scale central heating systems

Abstract

The present invention provides a method for real-time optimization of heat source temperature of a large-scale centralized heating system, comprising the following steps: step 1, obtaining historical heating data of centralized heating; step 2, calculating the maximum temperature conduction delay of the centralized heating network according to the historical heating data obtained in step 1; and constructing a heat load model; step 3, constructing a temperature conduction model of the centralized heating network according to the maximum temperature conduction delay of the centralized heating network obtained in step 2; step 4, calculating the expected primary-side water supply temperature of the heating network according to the historical heating data and the heat load model; step 5, calculating the optimal heat source temperature at the current moment according to the expected primary-side water supply temperature of the heating network obtained in step 4 and the temperature conduction model of the centralized heating network; the present invention can respond to outdoor temperature changes in real time by integrating deep learning technology and digital signal processing technology, thereby avoiding poor end-user experience or invalid heat consumption loss caused by mismatch between heat source heating and heat consumption load.

Description

Technical Field

[0001] The invention belongs to a heating system, and in particular relates to a real-time optimization method for heat source temperature of a large-scale centralized heating system. Background Art

[0002] At present, the heat source temperature of the central heating system is generally adjusted manually based on the experience of the operation and maintenance personnel. Usually, a suitable heat source temperature value is obtained by looking up the table according to the temperature of the day. This method is extremely dependent on the personal experience of the operation and maintenance personnel, and it is impossible to achieve timely and accurate adjustment. The existing automatic adjustment technology for the heat source temperature of the central heating system is mainly a quality and adjustment technology, that is, adjusting the heat source temperature and heat source flow at the same time, which is not suitable for central heating systems operating at a fixed flow rate. Summary of the invention

[0003] The purpose of the present invention is to provide a real-time optimization method for the heat source temperature of a large-scale centralized heating system, which solves the defects of low accuracy and low efficiency in the existing heat source temperature regulation technology of the centralized heating system.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is:

[0005] The present invention provides a method for real-time optimization of heat source temperature in a large-scale centralized heating system, comprising the following steps:

[0006] Step 1, obtaining historical heating data of central heating;

[0007] Step 2, according to the historical heating data obtained in step 1, the maximum temperature conduction delay of the central heating network is calculated; and a heat load model is constructed;

[0008] Step 3, constructing a temperature conduction model of the centralized heating network according to the maximum temperature conduction delay of the centralized heating network obtained in step 2;

[0009] Step 4, calculating the expected primary side water supply temperature of the heating network based on historical heating data and heat load model;

[0010] Step 5, calculate the optimal heat source temperature at the current moment according to the expected primary side water supply temperature of the heating network obtained in step 4 and the temperature conduction model of the centralized heating network.

[0011] Preferably, in step 1, the historical heating data includes the heating temperature of the heat source, data information of each heating station and the outdoor temperature;

[0012] The data information of each thermal power station includes primary side water supply temperature, primary side flow rate and primary side return water temperature.

[0013] Preferably, in step 2, according to the historical heating data obtained in step 1, the maximum temperature conduction delay of the centralized heating network is calculated, and the specific method is:

[0014] The time delay correlation coefficient between the heat source heating temperature time curve and the primary side heating temperature time curve of each thermal power station is calculated by the following formula:

[0015]

[0016] in, represent The heat source water supply temperature at the moment; Indicates the heat source and The temperature conduction delay of each thermal power station is The value of is a positive integer from 0 to 48;

[0017] The maximum value of the delay correlation coefficient corresponds to As the maximum temperature conduction delay of the central heating network .

[0018] Preferably, in step 2, a heat load model is constructed according to the historical heating data obtained in step 1, and the specific method is:

[0019] First, the model is established

[0020] Set the outdoor temperature during the set time period As input, the primary water supply temperature of the heating network within the set time period As an output, a linear regression model is constructed and used as a heat load model;

[0021] The expression of the heat load model is:

[0022] in, , All are coefficients;

[0023] Next, update the model

[0024] Use the outdoor temperature during the current time period and the primary water supply temperature of the heating network , update the model and obtain the updated heat load model.

[0025] Preferably, in step 3, a temperature conduction model of the centralized heating network is constructed according to the maximum temperature conduction delay of the centralized heating network obtained in step 2, and the specific method is:

[0026] Will Heat source water supply temperature at the moment As input time series;

[0027] The primary water supply temperature of the heating network at time n As output time series;

[0028] The first 70% of the input time series and the first 70% of the output time series are used as training data, and the rest are used as test data;

[0029] The LSTM model is used to train the training data to obtain the temperature conduction model of the central heating network;

[0030] The temperature conduction model of the central heating network is tested using the test data, and finally a trained temperature conduction model of the heating network is obtained;

[0031] The delay parameter t_step in the LSTM model is the maximum temperature conduction delay of the centralized heating network obtained in step 2.

[0032] Preferably, in step 4, the expected primary side water supply temperature of the heating network is calculated based on historical heating data and a heat load model, specifically:

[0033] for moment, will The outdoor temperature corresponding to each time Input the heat load model obtained in step 2 and calculate Expected primary water supply temperature of the heating network;

[0034] All the expected primary water supply temperatures of the heating network are combined into a vector, denoted as ;

[0035] in, Each moment represents the moment when the heat source water arrives at each heat exchange station.

[0036] Preferably, in step 5, the optimal heat source temperature at the current moment is calculated according to the expected primary side water supply temperature of the heating network obtained in step 4 and the temperature conduction model of the centralized heating network, specifically:

[0037] Take a positive integer with a minimum value of 50, a maximum value of 120, and an interval of 1 to generate a candidate heat source heating temperature set;

[0038] Select a value from the candidate heat source heating temperature set and The heating source heating temperatures corresponding to each moment are combined into a vector;

[0039] This vector is input into the temperature conduction model of the central heating network to obtain the primary side water supply temperature sequence of the heating network;

[0040] Calculate the primary water supply temperature series of each heating network and The mean square error of

[0041] Sort all the obtained mean square errors and get the minimum mean square error. As the optimal heat source temperature at the current moment;

[0042] in, It is an element in the set of candidate heat source heating temperatures.

[0043] A large-scale central heating system heat source temperature real-time optimization system, the system can run a large-scale central heating system heat source temperature real-time optimization method, including a data acquisition unit, a model building unit and a data processing unit, wherein:

[0044] The data acquisition unit is used to obtain historical heating data of central heating;

[0045] The model building unit is used to calculate the maximum temperature conduction delay of the central heating network according to the historical heating data obtained; and to build a heat load model; and to build a temperature conduction model of the central heating network according to the maximum temperature conduction delay of the central heating network obtained;

[0046] The data processing unit is used to calculate the expected primary side water supply temperature of the heating network based on historical heating data and a heat load model;

[0047] According to the desired primary-side water supply temperature of the heating network and the temperature conduction model of the centralized heating network, the optimal heat source temperature at the current moment is calculated.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention constructs a heat load model through deep learning technology, converts the heating load determined by the outdoor temperature into a form expressed by a controllable parameter (primary water supply temperature), so that the heating network can match the heat supply and heat load by adjusting the controllable parameter (primary water supply temperature). At the same time, the present invention constructs a temperature conduction model of a centralized heating network through deep learning technology, establishes the relationship between the heating temperature of the heat source and the primary water supply temperature of each thermal power station, so that the heating network can indirectly optimize the controllable parameter (primary water supply temperature) by optimizing the heat source temperature, and finally matches the heat supply of the entire heating network with the heat load. Due to the uncertainty of the temperature conduction delay between the remote thermal power station and the heat source of a large centralized heating system (the delay difference between each thermal power station may be greater than 10 hours), the model may greatly reduce the accuracy and effectiveness of the model by using an empirically based delay parameter. The present invention uses the time synchronization technology in digital signal processing, and determines the maximum temperature conduction delay of the entire heating network by calculating the mutual correlation coefficient of the primary water supply temperature time series, thereby determining the delay parameter in the centralized heating network temperature conduction model. The present invention provides a real-time optimization method for the heat source temperature of a large-scale centralized heating system, in order to achieve the purpose of energy saving and consumption reduction under the premise of meeting the heating load. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a structural diagram of the centralized heating network of the present invention. DETAILED DESCRIPTION

[0051] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0052] like Figure 1 As shown, the present invention provides a method for real-time optimization of heat source temperature in a large-scale centralized heating system, comprising the following steps:

[0053] Step 1: Obtain historical heating data for central heating

[0054] The historical heating data is collected from the measuring instruments of the heating network at a set sampling period T and stored in a computer system; wherein the sampling period T is set to 5 minutes; the historical heating data includes the heating temperature of the heat source, data information of each heating station and the outdoor temperature.

[0055] The data information of each thermal power station includes primary side water supply temperature, primary side flow rate and primary side return water temperature.

[0056] Step 2, according to the historical heating data obtained in step 1, the maximum temperature conduction delay of the central heating network is calculated; and a heat load model is constructed;

[0057] Specifically, the method for calculating the maximum temperature conduction delay of the district heating network is:

[0058] The time delay correlation coefficient between the heat source heating temperature time curve and the primary side water supply temperature time curve of each thermal power station is calculated by the following formula:

[0059]

[0060] in, represent The heat source water supply temperature at the moment; Indicates the heat source and The temperature conduction delay of a thermal power station is a positive integer between 0 and 48.

[0061] The maximum value of the delay correlation coefficient corresponds to As the maximum temperature conduction delay of the central heating network .

[0062] The specific method of establishing the heat load model is:

[0063] S201, Model Construction

[0064] Set the outdoor temperature during the set time period As input, the primary water supply temperature of the heating network within the set time period As output, a linear regression model is constructed and used as the heat load model; this function can be implemented using the sklearn library in the python language.

[0065] The expression of the heat load model is:

[0066] in, , All are coefficients.

[0067] S202, Model Update

[0068] To ensure the real-time performance of the model, the heat load model needs to be updated regularly. The update cycle is determined according to the actual heating network situation. For example, it can be set to update every day. Assuming that only the data of the latest week is taken, the data of the latest week can be taken again at 0:00 every day, and the data cleaning and model training can be re-performed. The heat load model obtained will replace the previously trained heat load model.

[0069] S203, adjust the model according to the complaint rate

[0070] The model can be expressed as: , it can be seen that when the complaint rate decreases, the value of intercept b is reduced; when the complaint rate increases, the value of intercept b is increased.

[0071] Step 3: construct a temperature conduction model of the central heating network according to the maximum temperature conduction delay of the central heating network obtained in step 2. Specifically:

[0072] Will Heat source water supply temperature at the moment As input time series;

[0073] The primary water supply temperature of the heating network at time n As output time series;

[0074] The first 70% of the input time series and the first 70% of the output time series are used as training data, and the rest are used as test data;

[0075] The LSTM model is used to train the training data to obtain the temperature conduction model of the central heating network;

[0076] The temperature conduction model of the centralized heating network is tested using the test data, and finally a trained temperature conduction model of the heating network is obtained and saved in a file format.

[0077] This function can be implemented using the keras library in Python.

[0078] The delay parameter t_step in the LSTM model is set to the maximum temperature conduction delay of the centralized heating network obtained in step 2.

[0079] The temperature conduction model of the district heating network includes an input layer, an LSTM layer, and an output layer. First, an LSTM layer with 6 neurons and a default activation function tanh is created. The remaining parameters of this layer are all default values. Then, an output layer with a single neuron and no activation function is created.

[0080] After creating the model, compile the model and specify the mean_squared_error loss function and the adam optimizer.

[0081] The primary water supply temperature of the heating network at time n is the weighted average of the primary water supply temperatures of all heating stations, that is,

[0082]

[0083] in, Indicates The primary water supply temperature of a thermal power station at time n; Indicates The primary side flow of a thermal power station at time n.

[0084] The primary return water temperature of the heating network at time n is the weighted average of the primary return water temperatures of all thermal power stations, that is,

[0085]

[0086] in, Indicates The primary return water temperature of a thermal power station at time n; Indicates The primary side flow of a thermal power station at time n.

[0087] Step 4: Calculate the expected primary water supply temperature of the heating network based on historical heating data and heat load model. Specifically:

[0088] for moment, will The outdoor temperature corresponding to each time Input the heat load model obtained in step 2 and calculate Expected primary water supply temperature of the heating network ;

[0089] Each moment represents the moment when the heat source water arrives at each heat exchange station.

[0090] All Combined into a vector, denoted as .

[0091] Step 5, calculating the optimal heat source temperature at the current moment according to the desired primary side water supply temperature of the heating network obtained in step 4, specifically:

[0092] Take a positive integer with a minimum value of 50, a maximum value of 120, and an interval of 1 to generate a candidate heat source heating temperature set , whose values ​​are {50, 51,52,…, 119, 120}.

[0093] from Pick any value in , ,and The heat source heating temperatures corresponding to each moment are combined into a vector, recorded as ,

[0094] This vector is input into the temperature conduction model of the central heating network, and the primary water supply temperature sequence of the heating network corresponding to t_source_candidate[j] is obtained, which is recorded as ;

[0095] .

[0096] calculate and The mean square error of

[0097] Sort all the obtained mean square errors and get the minimum mean square error. As the optimal heat source water supply temperature at the current moment, .

Claims

1. A method for real-time optimization of heat source temperature in a large-scale centralized heating system, characterized in that: The following steps are involved: Step 1, obtaining historical heating data of central heating; Step 2, calculating the maximum temperature conduction delay of the central heating network according to the historical heating data obtained in step 1; and build a heat load model; Step 3, constructing a temperature conduction model of the centralized heating network according to the maximum temperature conduction delay of the centralized heating network obtained in step 2; Step 4, calculating the expected primary side water supply temperature of the heating network based on historical heating data and heat load model; Step 5, calculate the optimal heat source temperature at the current moment according to the expected primary side water supply temperature of the heating network obtained in step 4 and the temperature conduction model of the centralized heating network.

2. A method for real-time optimization of heat source temperature in a large-scale centralized heating system according to claim 1, characterized in that: In step 1, the historical heating data includes the heating temperature of the heat source, the data information of each heating station and the outdoor temperature; The data information of each thermal power station includes primary side water supply temperature, primary side flow rate and primary side return water temperature.

3. A method for real-time optimization of heat source temperature in a large-scale centralized heating system according to claim 1, characterized in that: Step 2: Calculate the maximum temperature conduction delay of the central heating network based on the historical heating data obtained in step 1. The specific method is: The time delay correlation coefficient between the heat source heating temperature time curve and the primary side heating temperature time curve of each thermal power station is calculated by the following formula: in, represent The heat source water supply temperature at the moment; Indicates the heat source and The temperature conduction delay of each thermal power station is The value of is a positive integer from 0 to 48; The maximum value of the delay correlation coefficient corresponds to As the maximum temperature conduction delay of the central heating network .

4. A method for real-time optimization of heat source temperature in a large-scale centralized heating system according to claim 1, characterized in that: In step 2, a heat load model is constructed based on the historical heating data obtained in step 1. The specific method is: First, the model is established Set the outdoor temperature during the set time period As input, the primary water supply temperature of the heating network within the set time period As an output, a linear regression model is constructed and used as a heat load model; The expression of the heat load model is: in, , All are coefficients; Next, update the model Use the outdoor temperature during the current time period and the primary water supply temperature of the heating network , update the model and obtain the updated heat load model.

5. A method for real-time optimization of heat source temperature in a large-scale centralized heating system according to claim 1, characterized in that: In step 3, according to the maximum temperature conduction delay of the central heating network obtained in step 2, a temperature conduction model of the central heating network is constructed. The specific method is: Will Heat source water supply temperature at the moment As input time series; The water supply temperature of the primary side of the heating network at time n As output time series; The first 70% of the input time series and the first 70% of the output time series are used as training data, and the rest are used as test data; The LSTM model is used to train the training data to obtain the temperature conduction model of the central heating network; The temperature conduction model of the central heating network is tested using the test data, and finally a trained temperature conduction model of the heating network is obtained; The delay parameter t_step in the LSTM model is the maximum temperature conduction delay of the centralized heating network obtained in step 2.

6. A method for real-time optimization of heat source temperature in a large-scale centralized heating system according to claim 1, characterized in that: In step 4, the expected primary water supply temperature of the heating network is calculated based on the historical heating data and the heat load model. Specifically: for moment, will The outdoor temperature corresponding to each time Input the heat load model obtained in step 2 and calculate Expected primary water supply temperature of the heating network; All the expected primary water supply temperatures of the heating network are combined into a vector, denoted as ; in, Each moment represents the moment when the heat source water arrives at each heat exchange station.

7. A method for real-time optimization of heat source temperature in a large-scale centralized heating system according to claim 6, characterized in that: In step 5, the optimal heat source temperature at the current moment is calculated based on the expected primary water supply temperature of the heating network obtained in step 4 and the temperature conduction model of the centralized heating network. Specifically: Take a positive integer with a minimum value of 50, a maximum value of 120, and an interval of 1 to generate a candidate heat source heating temperature set; Select a value from the candidate heat source heating temperature set and The heating source heating temperatures corresponding to each moment are combined into a vector; This vector is input into the temperature conduction model of the central heating network to obtain the primary side water supply temperature sequence of the heating network; Calculate the primary water supply temperature series of each heating network and The mean square error of Sort all the obtained mean square errors and get the minimum mean square error. As the optimal heat source temperature at the current moment; in, It is an element in the set of candidate heat source heating temperatures.

8. A real-time optimization system for heat source temperature of a large-scale centralized heating system, characterized in that: The system is capable of running a real-time optimization method for heat source temperature of a large-scale centralized heating system according to any one of claims 1 to 7, comprising a data acquisition unit, a model building unit and a data processing unit, wherein: The data acquisition unit is used to obtain historical heating data of central heating; The model building unit is used to calculate the maximum temperature conduction delay of the central heating network according to the historical heating data obtained; and to build a heat load model; and to build a temperature conduction model of the central heating network according to the maximum temperature conduction delay of the central heating network obtained; The data processing unit is used to calculate the expected primary side water supply temperature of the heating network based on historical heating data and a heat load model; According to the desired primary-side water supply temperature of the heating network and the temperature conduction model of the centralized heating network, the optimal heat source temperature at the current moment is calculated.

Images