CN117232097B

CN117232097B - Central air conditioner refrigerating station optimal control method and system based on self-learning fusion model

Info

Publication number: CN117232097B
Application number: CN202311482558.3A
Authority: CN
Inventors: 樊融; 陈向阳; 宋志刚; 严雯静; 黄梦遥
Original assignee: Shanghai Qinghuan Energy Technology Co ltd
Current assignee: Shanghai Qinghuan Energy Technology Co ltd
Priority date: 2023-11-09
Filing date: 2023-11-09
Publication date: 2024-02-20
Anticipated expiration: 2043-11-09
Also published as: CN117232097A

Abstract

The invention relates to the field of air conditioning, and provides a central air conditioning refrigeration station optimization control method and system based on a self-learning fusion model, wherein the method comprises the following steps: step S1: constructing a mechanism and data fusion model of the water chilling unit; step S2: constructing a mechanism and data fusion model of a chilled water pump, and constructing a mechanism and data fusion model of a cooling water pump; step S3: constructing a cooling tower mechanism and data fusion model, a cooling tower heat exchange model and a refrigeration load prediction model; step S4: and all the models adopt a self-learning architecture, an overall energy consumption model of the refrigerating station is established according to an actual design structure, and overall energy efficiency optimization of a refrigerating station system is performed. The invention introduces a mechanism model to ensure the reliability and accuracy of the model, and can effectively play a role even under the condition of limited data.

Description

Central air conditioner refrigerating station optimal control method and system based on self-learning fusion model

Technical Field

The invention relates to the field of central air conditioning system optimal control, in particular to a central air conditioning refrigerating station optimal control method and system based on a self-learning fusion model.

Background

Central air conditioning systems play a key role in modern economies and are widely used in public buildings, industrial plants, data centers and other places. These locations often require stable temperature and comfortable environments for large areas of space or constant temperature cooling water for industrial equipment. With the continuous development of economy, the energy consumption problem of the central air conditioning system in various fields is increasing, and the energy efficiency of the central air conditioning system is becoming urgent in facing the increasingly severe energy pressure and environmental problems.

Central air conditioning systems are generally composed of a refrigeration station and an air conditioning terminal. The refrigerating station is a core part of the central air conditioning system and generally consists of a plurality of water chilling units, a chilled water pump, a cooling tower and other devices, and is a main source of energy consumption of the central air conditioning system, so that the control strategy of the refrigerating station directly determines the energy consumption of the central air conditioning system. The control means of the central air-conditioning refrigerating station mainly comprises a traditional group control method and an intelligent control method. The group control method refers to the integral control and adjustment of the refrigerating station through a centralized control system, such as setting a uniform temperature set point, starting and stopping time and the like. The method is relatively simple, but severely depends on manual operation level, cannot be adjusted in real time according to the change of the demand side, and has low comprehensive energy efficiency in actual operation; the intelligent control method is based on advanced sensor technology, data analysis and artificial intelligent algorithm, monitors and analyzes the running condition of the refrigerating station in real time, intelligently adjusts according to actual requirements, dynamically adjusts the running parameters of the refrigerating station, realizes accurate control and energy consumption optimization, and is widely applied in recent years.

However, when the intelligent control method is adopted, a large amount of historical operation data is often required for building an accurate equipment performance model, and the data is not easy to obtain and needs to be subjected to long-time data collection and arrangement work; in addition, when the device state of the refrigeration station changes or the external environment changes, the existing model may not accurately predict the performance and behavior of the new device. Such model hysteresis can lead to a decrease in control efficiency and an increase in energy consumption. Therefore, the intelligent control system which can be rapidly on line under the condition of lacking historical data and can be continuously self-learned and optimized in the actual operation process is constructed, and is a key for realizing efficient and intelligent operation of the central air-conditioning refrigeration station.

Patent document CN113821902B discloses an active disturbance rejection control system for static optimization of a central air-conditioning refrigeration station, the air-conditioning load prediction unit being used for load prediction through a BP neural network; the central air-conditioning refrigeration station model calculation unit is used for executing one of a single model modeling mode and a multi-model fusion modeling mode; the system working point optimizing unit is used for calculating the optimal working point of the system by utilizing an optimizing algorithm; and the central control refrigerating station control unit is used for tracking control by adopting the active disturbance rejection controller. However, the invention does not introduce a mechanism model to ensure the reliability and accuracy of the model.

Disclosure of Invention

Aiming at the defects in the prior art, the invention aims to provide a central air conditioning refrigeration station optimization control method and system based on a self-learning fusion model.

The invention provides a central air-conditioning refrigeration station optimization control method based on a self-learning fusion model, which comprises the following steps:

step S1: constructing a mechanism and data fusion model of the water chilling unit;

step S2: constructing a mechanism and data fusion model of a chilled water pump, and constructing a mechanism and data fusion model of a cooling water pump;

step S3: constructing a cooling tower mechanism and data fusion model, a cooling tower heat exchange model and a refrigeration load prediction model;

step S4: and all the models adopt a self-learning architecture, an overall energy consumption model of the refrigerating station is established according to an actual design structure, and overall energy efficiency optimization of a refrigerating station system is performed.

Preferably, in said step S1:

the equipment modeling of the refrigeration station adopts a fusion model combining a theoretical model and a neural network model, a chiller theoretical model adopts an AHRI 10 coefficient model or a 4-order fitting model, and an AHRI 10 coefficient model is used:

wherein,is the energy efficiency ratio of the water unit; />Is the temperature of cooling water; />Is the chilled water temperature; />The fitting coefficient is obtained through the computing software of the compressor of the refrigerating unit; if the model of the water chilling unit cannot be obtained or the coefficient difficulty in the model of the water chilling unit with the coefficient of 10 is greater than a preset standard, calculating the energy efficiency ratio of the water chilling unit by using a 4-order fitting model:

Wherein PLR is the load factor of the water chilling unit;obtaining running data for fitting coefficients;

the theoretical calculation formula of the power consumption of the refrigerating unit is as follows:

wherein,the refrigerating power of the water chilling unit; />The power consumption of the water chilling unit; />Is the flow of the chilled water; h (T) is the enthalpy of the chilled water at temperature T;

the fusion model of the water chiller is constructed by adopting a neural network: the input layer parameters include: chilled water flow G; return water temperature of chilled waterThe method comprises the steps of carrying out a first treatment on the surface of the Chilled water outlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water flow->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water inlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water outlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the If no cooling water flowmeter is installed in the cooling water system of the refrigerating station, the flow rate of cooling water is +.>Estimating, wherein all input parameters are normalized before being input;

the hidden layer 1 neural network adopts a double-layer structure, 16 nodes are arranged on each layer, and a ReLU activation function is used as an activation function; setting Dropout for each layer to prevent over fitting;

the mechanism model is a theoretical model of the chiller, the model parameters are subjected to inverse normalization before being input, and the output of the model is subjected to normalization again;

the connecting layer connects the output of the data model and the theoretical model together, and an output result vector is formed by adopting a splicing connection mode;

After the hidden layer 2 is connected with the layer, the output results of the data model and the theoretical model are further fused, the hidden layer 2 adopts a single-layer 4-node structure, and an activation function adopts a ReLU;

the output layer is an output result of the fusion model, and for the water chilling unit, the output parameter is the power consumption Pc of the water chilling unit;

the water chiller fusion model is expressed as:

。

preferably, in said step S2:

and (3) a mechanism and data fusion model of the chilled water pump:

the performance of the chilled water pump was calculated using a semi-empirical formula:

wherein,consuming electric power for the chilled water pump; />For water density, 1000 kg/-of depicting>；/>Gravitational acceleration of 9.8 m/-j>；/>Is the lift of the water pump; />The efficiency of the variable-frequency water pump is achieved; />Is the rotation speed ratio or the frequency ratio of the chilled water pump; />Is the actual frequency of the chilled water pump; />Is the rated frequency of the chilled water pump; />And->Determining fitting coefficients according to the running history data of the chilled water pump;

the frozen water pump fusion model adopts a cold machine fusion model structure, wherein the input layer is the frequency of the frozen water pumpChilled Water flow->The method comprises the steps of carrying out a first treatment on the surface of the The output layer is the power of the chilled water pump>The method comprises the steps of carrying out a first treatment on the surface of the The hidden layer 1 neural network adopts 8 single-layer nodes, and the ReLU activation function is used as the activation function;

the chilled water pump fusion model is expressed as:

And (3) a cooling water pump mechanism and data fusion model:

the cooling water pump fusion model adopts the same structure as the chilled water pump fusion model, and is expressed as:

wherein,consuming electric power for the cooling water pump; />The frequency of the cooling water pump; />For cooling water flow, if a cooling water flow meter is not installed in a cooling water system of the refrigerating station, estimating;

the cooling water pump fusion model structure is consistent with the chilled water pump fusion model.

Preferably, in said step S3:

cooling tower mechanism and data fusion model:

the performance of the cooling tower was calculated using a semi-empirical formula:

wherein,power consumption for the cooling tower; />Determining historical data of the operation of the cooling tower as a fitting coefficient; />For cooling air flow, estimated by:

wherein,is the rated air flow of the cooling tower; />Is the cooling tower operating frequency; />Rated frequency for the cooling tower; />To fit toA number determined from historical data of cooling tower operation;

the cooling tower fusion model adopts a cold machine fusion model structure, wherein an input layer comprises: cooling tower operating frequencyCooling water flow->Cooling water inlet temperature->Cooling water outlet temperature->Ambient temperature->Ambient relative humidity- >The method comprises the steps of carrying out a first treatment on the surface of the The output layer is cooling tower power->The method comprises the steps of carrying out a first treatment on the surface of the The hidden layer 2 neural network adopts 8 single-layer nodes, and the ReLU activation function is used as the activation function;

the cooling tower fusion model is expressed as:

cooling tower heat exchange model:

the heat exchange amount is calculated according to a heat transfer unit number model:

in the method, in the process of the invention,for the heat exchange of the cooling tower->For the heat exchange efficiency of the cooling tower>For cooling air flow>And->Is a function of the enthalpy of saturated air and the enthalpy of ambient air, and is obtained by ambient temperature +.>And relative humidity->Calculation of->And (3) withEnthalpy values at the cooling water outlet temperature and the inlet temperature, respectively;

the cooling tower heat exchange efficiency was calculated using the following formula:

wherein:

wherein,is the average specific heat of air->For the average specific heat capacity of water>And n is the relevant performance parameter of the cooling tower, and is provided by a manufacturer or fitted with historical data;

refrigeration load prediction model:

the cold load prediction adopts a neural network model, and the input layer comprises:

x1 to X24, 24 parameters respectively corresponding to one-hot transforms of 24 time periods in a day;

x25 represents whether it is weekdays or weekends;

x26 represents whether or not it is holiday;

x27 represents the corresponding environmental temperature at the time t, historical data adopted in model training is an actual measurement value, and a predicted value provided by an hour weather forecast is adopted in prediction;

X28 represents relative humidity corresponding to the moment t, historical data adopted in model training is an actual measurement value, and a predicted value provided by an hour weather forecast is adopted in prediction;

the cold load model comprises a day-ahead prediction and a day-ahead prediction, wherein the day-ahead prediction result is used for making a starting plan, and the constraint condition of unit operation is adjusted in advance one day; the daytime prediction is used for real-time optimization, and the running unit and the optimal setting parameters are determined in advance in a preset hour;

the predictive model of the cooling load is expressed as:

wherein,to predict the refrigerating capacity +.>Is the hour value of the predicted time.

Preferably, in said step S4:

self-learning architecture of model:

in the self-learning architecture, each model comprises an online model and an offline model, wherein the online model is a model currently in use, and the latest target parameters are predicted by inputting real-time data; the structure and super parameter setting of the offline model and the online model are the same, and the offline model does not participate in a real-time prediction task; each time the running data is accumulated to a preset degree or every preset fixed time, the self-learning framework automatically collects the latest data as a test data set, the online model and the offline model respectively predict target parameters, the prediction result is compared with the actual value and the RMSE is calculated, if the RMSE of the online model is smaller than the offline model, the online model is continuously used for executing a prediction task, and meanwhile, the test data set is combined into a training set of the original model to retrain the offline model; if the RMSE of the offline model is smaller than the online model, using the offline model to replace the online model;

Overall energy consumption model of refrigeration station:

the overall electricity consumption model of the refrigeration station is as follows:

wherein,for the whole power consumption of refrigerating station, < >>The start-stop state of each device is 1, the running state is 0 and the stop state is 0 +.>Is a water chilling unit power consumption model>For the power consumption model of the chilled water pump, < >>For the power consumption model of the cooling water pump, < >>A cooling tower power consumption model;

the comprehensive energy efficiency of the central air-conditioning refrigerating station is as follows:

wherein,comprehensive energy efficiency of the refrigerating station system; />For the cooling load of the refrigeration station system, the total flow of chilled water is +.>And the enthalpy value of the chilled jellyfish pipe water supply of the refrigeration station +.>Enthalpy value of backwater->Calculating; />The total power consumption of the refrigeration station is determined by main control parameters of all equipment of the refrigeration station, wherein the parameters comprise: chilled water set temperature +.>The frequency of the chilled water pump->Cooling water pump frequency->Cooling tower heat exchange fan frequency->；

Overall energy efficiency optimization of refrigeration station system:

the optimal control goal of the central air-conditioning refrigeration station system is to minimize the total energy consumption of the whole system and maximize the efficiency of the integrated system by optimizing control strategies and adjusting operation parameters under the condition of meeting the refrigeration load demand and the normal operation of equipment:

optimizing an objective function:

the constraint conditions include:

Energy balance:

the total refrigerating capacity of the chilled water of the refrigerating station is equal to the sum of the refrigerating capacities of the chilled water of all water chilling units; wherein,for the cooling load of the refrigeration station system, +.>The water chilling unit is in a start-stop state, an operation state is 1, and a stop state is 0; i is the number of the water chilling units, and N is the number of the water chilling units in the refrigerating station system; />Is the flow of the chilled water of the ith water chiller,and->Enthalpy values of chilled water flowing through the ith water chiller at backwater temperature and water supply temperature respectively; />For the total flow of chilled water of the refrigeration station system, +.>And->Enthalpy values of the refrigeration station freezing water main pipe backwater temperature and the water supply temperature are respectively;

mass balance:

the total flow of the freezing water main pipe of the refrigerating station is equal to the sum of the flow of the freezing water of each water chilling unit; wherein,the total flow of chilled water of the refrigeration station system; />The water chilling unit is in a start-stop state, an operation state is 1, and a stop state is 0; i is the number of the water chilling units, and N is the number of the water chilling units in the refrigerating station system; />The flow rate of the chilled water of the ith water chilling unit;

heat dissipation balance:

the above indicates that the heat dissipation capacity of the cooling tower system is equal to the heat exchange capacity of the cooling water system; wherein, The total heat dissipation capacity of the cooling tower system of the refrigeration station; />The cooling tower is in a start-stop state of the ith cooling tower, the running state is 1, and the stop state is 0; i is the number of the cooling tower, N is the refrigerating stationThe number of cooling towers in the system; />For the heat exchange efficiency of the ith cooling tower, < >>Cooling air flow of the i-th cooling tower, +.>And->Respectively at ambient temperature->And relative humidity->The enthalpy of saturated air and the enthalpy of unsaturated air under the condition; />Cooling water flow of the i-th cooling tower, +.>And->Enthalpy values of the i-th cooling tower at cold water backwater temperature and water supply temperature are respectively obtained;

operating parameter boundaries:

wherein,and->The temperature of the water supply and return of the chilled water of the ith cooling machine is respectively +>And->The upper limit and the lower limit of the water chilling unit on the requirements of chilled water supply and return water temperature are respectively set; />And->The temperatures of the cooling water supply and return water of the cooling tower system of the ith station are respectively +.>And->The upper limit and the lower limit of the cooling water supply and return water temperature requirements of the cooling tower system are respectively met; />、/>And->The set frequencies of the chilled water pump, the cooling water pump and the cooling tower fan are respectively +.>And->An upper limit and a lower limit set for the frequency, respectively;

optimizing target parameters:

optimal starting combination:

optimal chilled water outlet temperature of water chiller:

Optimal chilled water pump frequency:

optimal cooling water pump frequency:

optimum cooling tower radiator fan frequency:

wherein,for the start-up combination of the water chilling unit and the matched system, < + >>Setting the optimal temperature of chilled water of each water chilling unit, < >>Setting frequency for optimizing each chilled water pump, +.>The frequency is set for the optimization of each cooling water pump,for each cooling tower fanIs set at the optimal temperature; the model is optimized by adopting a genetic algorithm.

The invention provides a central air-conditioning refrigerating station optimization control system based on a self-learning fusion model, which comprises the following components:

module M1: constructing a mechanism and data fusion model of the water chilling unit;

module M2: constructing a mechanism and data fusion model of a chilled water pump, and constructing a mechanism and data fusion model of a cooling water pump;

module M3: constructing a cooling tower mechanism and data fusion model, a cooling tower heat exchange model and a refrigeration load prediction model;

module M4: and all the models adopt a self-learning architecture, an overall energy consumption model of the refrigerating station is established according to an actual design structure, and overall energy efficiency optimization of a refrigerating station system is performed.

Preferably, in said module M1:

Wherein,is the energy efficiency ratio of the water unit; />Is the temperature of cooling water; />Is the chilled water temperature; />The fitting coefficient is obtained through the computing software of the compressor of the refrigerating unit; if the model of the water chiller cannot be obtained or the coefficient difficulty in obtaining the 10-coefficient model is greater than a preset standard, a 4-order fitting model meter is usedCalculating the energy efficiency ratio of the water chilling unit:

the water chiller fusion model is expressed as:

。

preferably, in said module M2:

and (3) a mechanism and data fusion model of the chilled water pump:

the chilled water pump fusion model is expressed as:

and (3) a cooling water pump mechanism and data fusion model:

Preferably, in said module M3:

cooling tower mechanism and data fusion model:

wherein,is the rated air flow of the cooling tower; />Is the cooling tower operating frequency; />Rated frequency for the cooling tower; / >Determining historical data of the operation of the cooling tower as a fitting coefficient;

the cooling tower fusion model adopts a cold machine fusion model structure, wherein an input layer comprises: cooling tower operating frequencyCooling water flow->Cooling water inlet temperature->Cooling water outlet temperature->Ambient temperature->Ambient relative humidity->The method comprises the steps of carrying out a first treatment on the surface of the The output layer is cooling tower power->The method comprises the steps of carrying out a first treatment on the surface of the The hidden layer 2 neural network adopts 8 single-layer nodes, and the ReLU activation function is used as the activation function;

the cooling tower fusion model is expressed as:

cooling tower heat exchange model:

wherein:

refrigeration load prediction model:

x25 represents whether it is weekdays or weekends;

x26 represents whether or not it is holiday;

the predictive model of the cooling load is expressed as:

Preferably, in said module M4:

self-learning architecture of model:

Overall energy consumption model of refrigeration station:

Overall energy efficiency optimization of refrigeration station system:

optimizing an objective function:

the constraint conditions include:

Energy balance:

mass balance:

the total flow of the freezing water main pipe of the refrigerating station is equal to the sum of the flow of the freezing water of each water chilling unit; wherein,chilled water supply for refrigeration station systemsA flow rate; />The water chilling unit is in a start-stop state, an operation state is 1, and a stop state is 0; i is the number of the water chilling units, and N is the number of the water chilling units in the refrigerating station system; />The flow rate of the chilled water of the ith water chilling unit;

heat dissipation balance:

the above indicates that the heat dissipation capacity of the cooling tower system is equal to the heat exchange capacity of the cooling water system; wherein, The total heat dissipation capacity of the cooling tower system of the refrigeration station; />The cooling tower is in a start-stop state of the ith cooling tower, the running state is 1, and the stop state is 0; i is the number of cooling towers, and N is the number of cooling towers in the refrigerating station system; />For the heat exchange efficiency of the ith cooling tower, < >>Cooling air flow of the i-th cooling tower, +.>And->Respectively at ambient temperature->And relative humidity->The enthalpy of saturated air and the enthalpy of unsaturated air under the condition; />Cooling water flow of the i-th cooling tower, +.>And->Enthalpy values of the i-th cooling tower at cold water backwater temperature and water supply temperature are respectively obtained;

operating parameter boundaries:

wherein,and->The temperature of the water supply and return of the chilled water of the ith cooling machine is respectively +>And->The upper limit and the lower limit of the water chilling unit on the requirements of chilled water supply and return water temperature are respectively set; />And->Respectively are provided withFor the temperature of the cooling water supply and return water of the ith cooling tower system, < >>And->The upper limit and the lower limit of the cooling water supply and return water temperature requirements of the cooling tower system are respectively met; />、/>And->The set frequencies of the chilled water pump, the cooling water pump and the cooling tower fan are respectively +.>And->An upper limit and a lower limit set for the frequency, respectively;

optimizing target parameters:

optimal starting combination:

optimal chilled water outlet temperature of water chiller:

Optimal chilled water pump frequency:

optimal cooling water pump frequency:

optimum cooling tower radiator fan frequency:

wherein,for the start-up combination of the water chilling unit and the matched system, < + >>Setting the optimal temperature of chilled water of each water chilling unit, < >>Setting frequency for optimizing each chilled water pump, +.>The frequency is set for the optimization of each cooling water pump,setting the temperature for optimization of each cooling tower fan; the model is optimized by adopting a genetic algorithm.

Compared with the prior art, the invention has the following beneficial effects:

1. the invention introduces a mechanism model to ensure the reliability and accuracy of the model, and can effectively play a role even under the condition of limited data;

2. the self-learning architecture enables the invention to interact with actual operation data, and updates and adjusts the model according to the continuously accumulated data, thereby improving the precision and adaptability of the model;

3. by continuous optimization and adaptive control, the invention enables the refrigeration station to achieve an optimal operating condition under constantly changing environmental and load conditions.

Drawings

Other features, objects and advantages of the present invention will become more apparent upon reading of the detailed description of non-limiting embodiments, given with reference to the accompanying drawings in which:

FIG. 1 is a schematic diagram of a fusion model of a chiller;

FIG. 2 is a schematic diagram of a neural network architecture;

FIG. 3 is a schematic diagram of the working principle of the self-learning architecture;

FIG. 4 is a schematic diagram of an apparatus for a large refrigeration station;

fig. 5 is a schematic diagram of a specific flow of optimization control of a refrigeration station system.

Detailed Description

The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the present invention, but are not intended to limit the invention in any way. It should be noted that variations and modifications could be made by those skilled in the art without departing from the inventive concept. These are all within the scope of the present invention.

Example 1:

in order to solve the problems in the existing intelligent control technology of the central air-conditioning refrigeration station, the invention provides a control method with a self-learning framework. The method captures the complex relationship between the operation mechanism and key parameters of the refrigeration station in a comprehensive mode through the fusion of the mechanism model and the data model. The reliability and the accuracy of the model can be ensured by introducing the mechanism model, and the function can be ensured under the condition of limited data. Through the self-learning architecture, the model is interacted with actual operation data, and the model is continuously updated and adjusted according to the continuously accumulated data, so that the accuracy and adaptability of the model are further improved. By means of continuous optimization and adaptive control, the refrigeration station is able to achieve an optimal operating state under constantly changing environmental and load conditions.

According to the invention, the central air-conditioning refrigerating station optimization control method based on the self-learning fusion model, as shown in fig. 1-5, comprises the following steps:

specifically, in the step S1:

wherein,is a water chilling unitA cold power; />The power consumption of the water chilling unit; />Is the flow of the chilled water; h (T) is the enthalpy of the chilled water at temperature T;

the fusion model of the water chiller is constructed by adopting a neural network: the input layer parameters include: chilled water flow G; return water temperature of chilled water The method comprises the steps of carrying out a first treatment on the surface of the Chilled water outlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water flow->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water inlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water outlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the If no cooling water flowmeter is installed in the cooling water system of the refrigerating station, the flow rate of cooling water is +.>Estimating, wherein all input parameters are normalized before being input;

the water chiller fusion model is expressed as:

。

Specifically, in the step S2:

and (3) a mechanism and data fusion model of the chilled water pump:

the chilled water pump fusion model is expressed as:

and (3) a cooling water pump mechanism and data fusion model:

specifically, in the step S3:

cooling tower mechanism and data fusion model:

wherein,is the rated air flow of the cooling tower; />Is the cooling tower operating frequency; />Rated frequency for the cooling tower; />Determining historical data of the operation of the cooling tower as a fitting coefficient;

the cooling tower fusion model is expressed as:

cooling tower heat exchange model:

wherein:

refrigeration load prediction model:

x25 represents whether it is weekdays or weekends;

x26 represents whether or not it is holiday;

The predictive model of the cooling load is expressed as:

Specifically, in the step S4:

self-learning architecture of model:

Overall energy consumption model of refrigeration station:

wherein,is a refrigerating stationBody electricity consumption>The start-stop state of each device is 1, the running state is 0 and the stop state is 0 +.>Is a water chilling unit power consumption model>For the power consumption model of the chilled water pump, < >>For the power consumption model of the cooling water pump, < >>A cooling tower power consumption model;

wherein,comprehensive energy efficiency of the refrigerating station system; />For the cooling load of the refrigeration station system, the total flow of chilled water is +.>And the enthalpy value of the chilled jellyfish pipe water supply of the refrigeration station +.>Enthalpy value of backwater->Calculating; />For the whole electricity consumption of the refrigerating station, the main control of each device of the refrigerating stationAnd determining parameters, wherein the parameters comprise: chilled water set temperature +.>The frequency of the chilled water pump->Cooling water pump frequency->Cooling tower heat exchange fan frequency->；

Overall energy efficiency optimization of refrigeration station system:

optimizing an objective function:

the constraint conditions include:

Energy balance:

mass balance:

the total flow of the freezing water main pipe of the refrigerating station is equal to the sum of the flow of the freezing water of each water chilling unit; wherein,the total flow of chilled water of the refrigeration station system; />The water chilling unit is in a start-stop state, an operation state is 1, and a stop state is 0; i is the number of the water chilling units, and N is the number of the water chilling units in the refrigerating station system; />Is the ithChilled water flow of the water chiller;

heat dissipation balance:

operating parameter boundaries:

optimizing target parameters:

optimal starting combination:

optimal chilled water outlet temperature of water chiller:

Optimal chilled water pump frequency:

optimal cooling water pump frequency:

optimum cooling tower radiator fan frequency:

Example 2:

example 2 is a preferable example of example 1 to more specifically explain the present invention.

The invention also provides a central air-conditioning refrigerating station optimizing control system based on the self-learning fusion model, which can be realized by executing the flow steps of the central air-conditioning refrigerating station optimizing control method based on the self-learning fusion model, namely, a person skilled in the art can understand the central air-conditioning refrigerating station optimizing control method based on the self-learning fusion model as a preferred implementation mode of the central air-conditioning refrigerating station optimizing control system based on the self-learning fusion model.

specifically, in the module M1:

/>

wherein,the refrigerating power of the water chilling unit; />The power consumption of the water chilling unit;/>is the flow of the chilled water; h (T) is the enthalpy of the chilled water at temperature T;

the fusion model of the water chiller is constructed by adopting a neural network: the input layer parameters include: chilled water flow G; return water temperature of chilled waterThe method comprises the steps of carrying out a first treatment on the surface of the Chilled water outlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water flow- >The method comprises the steps of carrying out a first treatment on the surface of the Cooling water inlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water outlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the If no cooling water flowmeter is installed in the cooling water system of the refrigerating station, the flow rate of cooling water is +.>Estimating, wherein all input parameters are normalized before being input;

the water chiller fusion model is expressed as:

。

specifically, in the module M2:

And (3) a mechanism and data fusion model of the chilled water pump:

wherein,consuming electric power for the chilled water pump; />For water density, 1000 kg/-of depicting>；/>Gravitational acceleration of 9.8 m/-j>；/>Is the lift of the water pump; />The efficiency of the variable-frequency water pump is achieved; />For freezingRotation speed ratio or frequency ratio of the water pump; />Is the actual frequency of the chilled water pump; />Is the rated frequency of the chilled water pump; />And->Determining fitting coefficients according to the running history data of the chilled water pump;

the chilled water pump fusion model is expressed as:

and (3) a cooling water pump mechanism and data fusion model:

specifically, in the module M3:

cooling tower mechanism and data fusion model:

the cooling tower fusion model is expressed as:

cooling tower heat exchange model:

in the method, in the process of the invention,for the heat exchange of the cooling tower- >For the heat exchange efficiency of the cooling tower>For cooling air flow>And->Is a function of the enthalpy of saturated air and the enthalpy of ambient air, and is obtained by ambient temperature +.>And relative humidity->Calculation of->And (3) withEnthalpy values at the cooling water outlet temperature and the inlet temperature, respectively;

/>

wherein:

refrigeration load prediction model:

x25 represents whether it is weekdays or weekends;

x26 represents whether or not it is holiday;

The predictive model of the cooling load is expressed as:

Specifically, in the module M4:

self-learning architecture of model:

Overall energy consumption model of refrigeration station:

wherein,for the whole power consumption of refrigerating station, < >>For the start-stop state of each device, the operationState 1, stop state 0, +.>Is a water chilling unit power consumption model>For the power consumption model of the chilled water pump, < >>For the power consumption model of the cooling water pump, < >>A cooling tower power consumption model;

Overall energy efficiency optimization of refrigeration station system:

optimizing an objective function:

the constraint conditions include:

Energy balance:

mass balance:

the total flow of the freezing water main pipe of the refrigerating station is equal to the sum of the flow of the freezing water of each water chilling unit; wherein,the total flow of chilled water of the refrigeration station system; />The water chilling unit is in a start-stop state, an operation state is 1, and a stop state is 0; i is the number of the water chilling units, and N is the number of the water chilling units in the refrigerating station system; />The flow rate of the chilled water of the ith water chilling unit; />

Heat dissipation balance:

operating parameter boundaries:

optimizing target parameters:

optimal starting combination:

optimal chilled water outlet temperature of water chiller:

Optimal chilled water pump frequency:

optimal cooling water pump frequency:

optimum cooling tower radiator fan frequency:

Example 3:

example 3 is a preferable example of example 1 to more specifically explain the present invention.

The invention adopts a control method with a self-learning architecture to improve the optimal control effect of the refrigeration station. The method captures complex correlations between the operation mechanism and key parameters of the refrigeration station through a comprehensive mechanism model and a data model.

The invention comprises the following steps:

step 1: mechanism and data fusion model of water chilling unit

In order to ensure that the model is rapidly on line under the condition of small data volume, the invention adopts a fusion model combining a theoretical model and a neural network model for modeling main equipment of the refrigeration station. Wherein the basis of the theoretical model is the equipment operation mechanism, so even if the data lacks extrapolation capability capable of ensuring the final fusion model; along with the accumulation of the operation data, the data model can ensure that the final fusion model has enough precision.

In the invention, a theoretical model adopted by a water chiller adopts an AHRI 10 coefficient model or a 4-order fitting model, and the AHRI 10 coefficient model is taken as an example:

wherein COP is the energy efficiency ratio of the water chilling unit; te' is the temperature of cooling water; tc is the chilled water temperature; alpha is a fitting coefficient and can be obtained through the computing software of the compressor of the refrigerating unit. If the model of the water chilling unit is unknown or the coefficients in the 10-coefficient model are difficult to obtain, the 4-order fitting model can be used for calculating the energy efficiency ratio of the water chilling unit:

wherein PLR is the load factor of the water chilling unit; c is a fitting coefficient, and can be obtained by using a small amount of operation data.

wherein Qc is the refrigerating power of the water chilling unit; pc is the power consumption of the water chilling unit; g is the flow of chilled water; h (T) is the enthalpy of the chilled water at temperature T.

The fusion model of the water chiller is constructed by adopting a neural network, and the structure of the model is shown in fig. 1:

the input layer includes 6 parameters, which are respectively: g, refrigerating water flow; tc', chilled water return temperature; tc '', chilled water outlet temperature; mw, cooling water flow; te', cooling water inlet temperature; te '', cooling water outlet temperature; if no cooling water flow meter is installed in the cooling water system of the refrigeration station, the cooling water flow rate mw in the formula can be estimated by the equation in the step 5. All input parameters need to be normalized before they are input.

The hidden layer 1 neural network adopts a double-layer structure, 16 nodes are arranged on each layer, and a ReLU activation function is used as an activation function; dropout=0.7-0.8 is also set for each layer to prevent overfitting.

The mechanism model is the theoretical model of the water chiller, the model parameters need to be subjected to inverse normalization before being input, and the model output needs to be subjected to normalization again.

The connection layer is used for connecting the output of the data model and the output of the theoretical model together, and an output result vector is formed by adopting a splicing connection (connection) mode.

After the connection layer, the hidden layer 2 is used for further fusing the output results of the data model and the theoretical model together, so that the model can learn and capture the relation between the features better, and the performance and generalization capability of the model are improved. The hidden layer 2 adopts a single-layer 4-node structure, and the activation function also adopts a ReLU.

The output layer is the output result of the fusion model, and for the water chiller, the final output parameter is the power consumption Pc of the water chiller.

The chiller fusion model may ultimately be expressed as:

step 2: mechanism and data fusion model of chilled water pump

The performance of the chilled water pump can be calculated using a semi-empirical formula:

wherein Pcp is the electric power consumed by the chilled water pump; ρ is the density of water, 1000kg/m3; g is gravity acceleration, 9.8m/s2 is taken; h is the lift of the water pump; η is the efficiency of the variable frequency water pump; r is the rotation speed ratio or frequency ratio of the chilled water pump; f is the actual frequency of the chilled water pump; f0 is the rated frequency of the chilled water pump; a and b are fitting coefficients, which can be determined by using chilled water pump operation history data.

The frozen water pump fusion model also adopts a similar structure of the intercooler fusion model in the step 1, wherein the input layer is the frozen water pump frequency fcp and the frozen water flow G; the output layer is the power Pcp of the chilled water pump; the hidden layer 1 neural network adopts a single layer of 8 nodes, and the ReLU activation function is used as the activation function.

The chilled water pump fusion model can ultimately be expressed as:

step 3: cooling water pump mechanism and data fusion model

The cooling water pump fusion model adopts the same structure as the chilled water pump fusion model, and can be expressed as:

wherein Pep is the electric power consumed by the cooling water pump; fep is the cooling water pump frequency; mw is the cooling water flow rate, and if no cooling water flow meter is installed in the cooling water system of the refrigeration station, the estimation can be performed through the equation in the step 5.

And (3) the cooling water pump fusion model structure is consistent with the cooling water pump fusion model in the step (2).

Step 4: cooling tower mechanism and data fusion model

The performance of the cooling tower can be calculated using a semi-empirical formula:

wherein Pct is the power consumption of the cooling tower; c is a fitting coefficient, and can be determined through historical data of the operation of the cooling tower; ma is the cooling air flow, which can be estimated by:

wherein ma,0 is the rated air flow of the cooling tower; fct is the cooling tower operating frequency; fct,0 is the cooling tower nominal frequency; k is a fitting coefficient, which can be determined from historical data of cooling tower operation.

The cooling tower fusion model adopts a similar structure to the intercooler fusion model in the step 1, wherein the input layer comprises three parameters, namely cooling tower running frequency fct, cooling water flow mw, cooling water inlet temperature Te ', cooling water outlet temperature Te ' ', environment temperature T0 and environment relative humidity RH0; the output layer is the cooling tower power Pct; the hidden layer 2 neural network adopts a single layer of 8 nodes, and the ReLU activation function is used as the activation function.

The cooling tower fusion model can ultimately be expressed as:

step 5: cooling tower heat exchange model

The cooling tower heat exchange process is to make the cooling water temperature approach to the outdoor wet bulb temperature through heat transfer by cooling water and outdoor air, the temperature difference between the two can represent the influence of the cooling water and the outdoor wet bulb temperature on the heat dissipation effect, and the heat exchange quantity can be calculated according to a classical heat transfer unit number (epsilon-NTU) model:

wherein Qe is the heat exchange amount of the cooling tower, epsilon a is the heat exchange efficiency of the cooling tower, ma is the cooling air flow, ha, v and ha are the enthalpy function of saturated air and the enthalpy function of ambient air, and can be calculated through the ambient temperature T0 and the relative humidity RH 0.

The cooling tower heat exchange efficiency can be calculated by the following formula:

wherein:

wherein, cp, a is the average specific heat of air, cp, w is the average specific heat of water, c and n are the relevant performance parameters of the cooling tower, and can be provided by manufacturers or fitted by historical data.

Step 6: refrigeration load prediction model

Since the cold start preparation time is relatively long and the time from start-up to steady operation exceeds 15-30 minutes, future cold load demands need to be predicted in advance to be prepared in advance. In the invention, a neural network model is adopted for cold load prediction, and the structure of the neural network is shown in figure 2:

the input layer includes 28 parameters, which are respectively:

X1-X24-the 24 parameters respectively correspond to one-hot conversion of 24 time periods in one day, for example, the Xt parameter corresponding to the time t is equal to 1, and other parameters are equal to 0;

x25-whether weekday or weekend;

x26-whether holiday;

the environment temperature corresponding to the moment X27-t, the historical data adopted in the training of the model are actual measurement values, and the prediction value provided by the hour weather forecast is adopted in the prediction;

the relative humidity corresponding to the moment X28-t is measured, the historical data adopted in model training is measured, and the predicted value provided by the hour weather forecast is adopted in prediction.

The cold load model comprises a day-ahead prediction and a day-ahead prediction, wherein the day-ahead prediction result is mainly used for making a starting plan and adjusting constraint conditions of unit operation in advance of one day; the daytime prediction is mainly used for real-time optimization, and the running unit and the optimal setting parameters are determined 1-3 hours in advance.

The predictive model of the cooling load can ultimately be expressed as:

step 7: model self-learning Xi Jiagou

In order to enable the model to be rapidly in line to adapt to the continuously-changed environment and data characteristics, and update and adjust the model according to the continuously-accumulated data, the accuracy and the adaptability of the model are further improved, and all the models in the invention adopt a self-learning architecture. The working principle of the self-learning architecture is shown in fig. 3:

in a self-learning architecture, each model includes two components, an online model and an offline model. The online model is a model which is currently in use, and predicts the latest target parameters by inputting real-time data; the offline model is identical to the online model in structure and super parameter settings, but does not participate in the real-time predictive task. Each time the running data is accumulated to a certain degree or every other fixed time, the self-learning framework automatically collects the latest data as a test data set, the online model and the offline model respectively predict target parameters, the prediction result is compared with the actual value and the RMSE is calculated, if the RMSE of the online model is smaller than the offline model, the online model is continuously used for executing the prediction task, and meanwhile, the test data set is combined into a training set of the original model to retrain the offline model; if the RMSE of the offline model is less than the online model, the offline model is used in place of the online model.

The evaluation period of each equipment model of the cold station in the invention is 1 day.

Step 8: integral energy consumption model of refrigeration station

In order to optimize the overall energy consumption of the refrigeration station, an overall energy consumption model of the refrigeration station must be established according to the actual design structure. This model will take into account the operational characteristics of the individual energy consuming devices and the influence of each other in combination in order to better understand and evaluate the energy consumption situation of the refrigeration station and to provide guidance for energy consumption optimization.

The number and configuration of the large refrigeration stations varies from one type to another, but generally involves four main types of energy consuming equipment, namely a chiller, chilled water pump, cooling tower and chilled water pump, as exemplified in fig. 4.

The model established for each energy consumption device by adopting the steps is adopted, and the overall electricity consumption model of the refrigeration station is as follows:

wherein, pcs is the whole electricity consumption of the refrigerating station, xi is the start-stop state (the running state is 1, the stop state is 0) of each device, pc, i is the water chilling unit electricity consumption model, pcp, i is the chilled water pump electricity consumption model, pep, i is the cooling water pump electricity consumption model, pct, i is the cooling tower electricity consumption model.

wherein COPCs is the comprehensive energy efficiency of the refrigerating station system; qcs is the cold load of the refrigerating station system, and can be calculated through the total flow Gcs of the chilled water and the enthalpy value of water supply and return of the chilled jellyfish pipe of the refrigerating station; pcs is the whole electricity consumption of the refrigerating station, and is determined by main control parameters (chilled water temperature Tcs '', chilled water pump frequency fcp, cooling water pump frequency fep and cooling tower heat exchange fan frequency fct) of each device of the refrigerating station.

Step 9: overall energy efficiency optimization for refrigeration station systems

namely:

optimizing an objective function:

constraint conditions:

1. energy balance:

the total refrigerating capacity of the chilled water of the refrigerating station is equal to the sum of the refrigerating capacities of the chilled water of all water chilling units.

2. Mass balance:

the total flow of the freezing water main pipe of the refrigerating station is equal to the sum of the flow of the freezing water of each water chilling unit.

3. Heat dissipation balance:

the above indicates that the heat dissipation capacity of the cooling tower system is equal to the heat exchange capacity of the cooling water system.

4. Operating parameter boundaries:

wherein Tc, min and Tc, max are respectively the upper limit and the lower limit of the water chilling unit for the requirements of chilled water supply and return water temperature; tct, min and Tct, max are the upper limit and the lower limit of the cooling tower system for the temperature requirements of the cooling water supply and return water respectively; hzmin and hzmax are respectively the upper limit and the lower limit set by the frequency of the water pump.

Optimizing target parameters:

1. optimal starting combination:

2. optimal chilled water outlet temperature of water chiller:

3. Optimal chilled water pump frequency:

4. optimal cooling water pump frequency:

5. optimum cooling tower radiator fan frequency:

/>

the model belongs to the optimization problem with complex nonlinear characteristics, and genetic algorithm is generally adopted to optimize the problem, such as differential evolution method, particle swarm algorithm and the like. The genetic algorithm can effectively search and find the optimal control strategy so as to achieve the aims of minimizing the total energy consumption of the system and improving the comprehensive efficiency.

The specific implementation of the refrigeration station system optimization control can be according to the flow shown in fig. 5:

those skilled in the art will appreciate that the invention provides a system and its individual devices, modules, units, etc. that can be implemented entirely by logic programming of method steps, in addition to being implemented as pure computer readable program code, in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Therefore, the system and various devices, modules and units thereof provided by the invention can be regarded as a hardware component, and the devices, modules and units for realizing various functions included in the system can also be regarded as structures in the hardware component; means, modules, and units for implementing the various functions may also be considered as either software modules for implementing the methods or structures within hardware components.

The foregoing describes specific embodiments of the present invention. It is to be understood that the invention is not limited to the particular embodiments described above, and that various changes or modifications may be made by those skilled in the art within the scope of the appended claims without affecting the spirit of the invention. The embodiments of the present application and features in the embodiments may be combined with each other arbitrarily without conflict.

Claims

1. The utility model provides a central air conditioning refrigerating station optimal control method based on self-learning fusion model, which is characterized by comprising the following steps:

step S4: all models adopt a self-learning architecture, an overall energy consumption model of the refrigeration station is established according to an actual design structure, and overall energy efficiency optimization of a refrigeration station system is performed;

in the step S1:

Wherein,is the energy efficiency ratio of the water unit; />Is the temperature of cooling water; />Is the chilled water temperature; />The fitting coefficient is obtained through the computing software of the compressor of the refrigerating unit; if the model of the water chilling unit cannot acquire or the coefficient difficulty in acquiring the AHRI 10 coefficient model is larger than a preset standard, calculating the energy efficiency ratio of the water chilling unit by using a 4-order fitting model:

the fusion model of the water chiller is constructed by adopting a neural network: the input layer parameters include: chilled water flow G; return water temperature of chilled waterThe method comprises the steps of carrying out a first treatment on the surface of the Chilled water outlet temperature/>The method comprises the steps of carrying out a first treatment on the surface of the Cooling water flow->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water inlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water outlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the If no cooling water flowmeter is installed in the cooling water system of the refrigerating station, the flow rate of cooling water is +.>Estimating, wherein all input parameters are normalized before being input;

the water chiller fusion model is expressed as:

in the step S4:

self-learning architecture of model:

Overall energy consumption model of refrigeration station:

wherein,for the whole power consumption of refrigerating station, < >>The starting and stopping states of the devices are 1, the running state is 0,is a water chilling unit power consumption model>For the power consumption model of the chilled water pump, < >>For the power consumption model of the cooling water pump, < >>A cooling tower power consumption model;

Overall energy efficiency optimization of refrigeration station system:

optimizing an objective function:

the constraint conditions include:

Energy balance:

the total refrigerating capacity of the chilled water of the refrigerating station is equal to the sum of the refrigerating capacities of the chilled water of all water chilling units; wherein,for the cooling load of the refrigeration station system, +.>The water chilling unit is in a start-stop state, an operation state is 1, and a stop state is 0; i is the number of the water chilling units, and N is the number of the water chilling units in the refrigerating station system; />For the chilled water flow of the ith water chiller, < +.>And->Enthalpy values of chilled water flowing through the ith water chiller at backwater temperature and water supply temperature respectively; />For the total flow of chilled water of the refrigeration station system, +.>And->Enthalpy values of the refrigeration station freezing water main pipe backwater temperature and the water supply temperature are respectively;

mass balance:

heat dissipation balance:

operating parameter boundaries:

optimizing target parameters:

optimal starting combination:

optimal chilled water outlet temperature of water chiller:

Optimal chilled water pump frequency:

optimal cooling water pump frequency:

optimum cooling tower radiator fan frequency:

wherein,for the start-up combination of the water chilling unit and the matched system, < + >>Setting the optimal temperature of chilled water of each water chilling unit, < >>Setting frequency for optimizing each chilled water pump, +.>Setting frequency for optimizing each cooling water pump, +.>Setting the temperature for optimization of each cooling tower fan; the model is optimized by adopting a genetic algorithm.

2. The optimization control method of a central air conditioning refrigeration station based on a self-learning fusion model according to claim 1, wherein in the step S2:

and (3) a mechanism and data fusion model of the chilled water pump:

the frozen water pump fusion model adopts a cold machine fusion model structure, wherein the input layer is the frequency of the frozen water pump Chilled Water flow->The method comprises the steps of carrying out a first treatment on the surface of the The output layer is the power of the chilled water pump>The method comprises the steps of carrying out a first treatment on the surface of the The hidden layer 1 neural network adopts 8 single-layer nodes, and the ReLU activation function is used as the activation function;

the chilled water pump fusion model is expressed as:

and (3) a cooling water pump mechanism and data fusion model:

3. The optimization control method of a central air conditioning refrigeration station based on a self-learning fusion model according to claim 1, wherein in the step S3:

cooling tower mechanism and data fusion model:

wherein,is the rated air flow of the cooling tower; />Is the cooling tower operating frequency; />Rated frequency for the cooling tower; Determining historical data of the operation of the cooling tower as a fitting coefficient;

the cooling tower fusion model is expressed as:

cooling tower heat exchange model:

in the method, in the process of the invention,for the heat exchange of the cooling tower->For the heat exchange efficiency of the cooling tower>For cooling air flow>And->Is a function of the enthalpy of saturated air and the enthalpy of ambient air, and is obtained by ambient temperature +.>And relative humidity->Calculation of->And->Enthalpy values at the cooling water outlet temperature and the inlet temperature, respectively;

wherein:

refrigeration load prediction model:

x25 represents whether it is weekdays or weekends;

x26 represents whether or not it is holiday;

the predictive model of the cooling load is expressed as:

4. Central air conditioning refrigerating station optimizing control system based on self-learning fusion model, which is characterized by comprising:

module M4: all models adopt a self-learning architecture, an overall energy consumption model of the refrigeration station is established according to an actual design structure, and overall energy efficiency optimization of a refrigeration station system is performed;

in the module M1:

The fusion model of the water chiller is constructed by adopting a neural network: the input layer parameters include: chilled water flow G; return water temperature of chilled waterThe method comprises the steps of carrying out a first treatment on the surface of the Chilled water outlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water flow->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water inlet temperature->The method comprises the steps of carrying out a first treatment on the surface of the Cooling water outletTemperature->The method comprises the steps of carrying out a first treatment on the surface of the If no cooling water flowmeter is installed in the cooling water system of the refrigerating station, the flow rate of cooling water is +.>Estimating, wherein all input parameters are normalized before being input;

The water chiller fusion model is expressed as:

in the module M4:

self-learning architecture of model:

overall energy consumption model of refrigeration station:

wherein,for the whole power consumption of refrigerating station, < >>The starting and stopping states of the devices are 1, the running state is 0, Is a water chilling unit power consumption model>For the power consumption model of the chilled water pump, < >>For the power consumption model of the cooling water pump, < >>A cooling tower power consumption model;

Overall energy efficiency optimization of refrigeration station system:

optimizing an objective function:

the constraint conditions include:

energy balance:

the total refrigerating capacity of the chilled water of the refrigerating station is equal to the sum of the refrigerating capacities of the chilled water of all water chilling units; wherein,for the cooling load of the refrigeration station system, +.>The water chilling unit is in a start-stop state, an operation state is 1, and a stop state is 0; i is the number of the water chilling units, and N is the number of the water chilling units in the refrigerating station system; / >For the chilled water flow of the ith water chiller, < +.>And->Enthalpy values of chilled water flowing through the ith water chiller at backwater temperature and water supply temperature respectively; />For the total flow of chilled water of the refrigeration station system, +.>And->Enthalpy values of the refrigeration station freezing water main pipe backwater temperature and the water supply temperature are respectively;

mass balance:

heat dissipation balance:

the above indicates that the heat dissipation capacity of the cooling tower system is equal to the heat exchange capacity of the cooling water system; wherein,the total heat dissipation capacity of the cooling tower system of the refrigeration station; />The cooling tower is in a start-stop state of the ith cooling tower, the running state is 1, and the stop state is 0; i is the number of cooling towers, and N is the number of cooling towers in the refrigerating station system; />For the heat exchange efficiency of the ith cooling tower, < >>Cooling air flow of the i-th cooling tower, +.>And->Respectively at ambient temperature->And relative humidity- >The enthalpy of saturated air and the enthalpy of unsaturated air under the condition; />Cooling water flow of the i-th cooling tower, +.>And->Enthalpy values of the i-th cooling tower at cold water backwater temperature and water supply temperature are respectively obtained;

operating parameter boundaries:

optimizing target parameters:

optimal starting combination:

optimal chilled water outlet temperature of water chiller:

optimal chilled water pump frequency:

optimal cooling water pump frequency:

optimum cooling tower radiator fan frequency:

wherein,is a water chilling unitStarting up combination of matched system>Setting the optimal temperature of chilled water of each water chilling unit, < >>Setting frequency for optimizing each chilled water pump, +.>Setting frequency for optimizing each cooling water pump, +. >Setting the temperature for optimization of each cooling tower fan; the model is optimized by adopting a genetic algorithm.

5. The self-learning fusion model based central air conditioning and cooling station optimization control system according to claim 4, wherein in the module M2:

and (3) a mechanism and data fusion model of the chilled water pump:

wherein,consuming electric power for the chilled water pump; />For water density, 1000 kg/-of depicting>；/>By gravity addingSpeed of 9.8 m/o>；/>Is the lift of the water pump; />The efficiency of the variable-frequency water pump is achieved; />Is the rotation speed ratio or the frequency ratio of the chilled water pump; />Is the actual frequency of the chilled water pump; />Is the rated frequency of the chilled water pump; />And->Determining fitting coefficients according to the running history data of the chilled water pump;

the chilled water pump fusion model is expressed as:

and (3) a cooling water pump mechanism and data fusion model:

6. The self-learning fusion model based central air conditioning and cooling station optimization control system according to claim 4, wherein in the module M3:

cooling tower mechanism and data fusion model:

wherein,is the rated air flow of the cooling tower; />Is the cooling tower operating frequency; />Rated frequency for the cooling tower;determining historical data of the operation of the cooling tower as a fitting coefficient;

the cooling tower fusion model adopts a cold machine fusion model structure, wherein an input layer comprises: cooling tower operating frequencyCooling water flow->Cooling water inlet temperature->Cooling water outlet temperature->Ambient temperature->Ambient relative humidity->The method comprises the steps of carrying out a first treatment on the surface of the The output layer is cooling tower power- >The method comprises the steps of carrying out a first treatment on the surface of the The hidden layer 2 neural network adopts 8 single-layer nodes, and the ReLU activation function is used as the activation function;

the cooling tower fusion model is expressed as:

cooling tower heat exchange model:

in the method, in the process of the invention,for the heat exchange of the cooling tower->For the heat exchange efficiency of the cooling tower>For cooling air flow>And->As a function of the enthalpy of saturated air and the enthalpy of ambient airValue function by ambient temperature->And relative humidity->Calculation of->And->Enthalpy values at the cooling water outlet temperature and the inlet temperature, respectively;

wherein:

/>

refrigeration load prediction model:

x25 represents whether it is weekdays or weekends;

x26 represents whether or not it is holiday;

the predictive model of the cooling load is expressed as:

wherein,to predict the refrigerating capacity +.>Is the hour value of the predicted time. />