KR20210097412A - Building Energy Failure Diagnosis and Analysis System with integrated Virtual sensor and Deep learning, and Building Energy Failure Diagnosis and Analysis Method - Google Patents

Abstract

The present invention relates to a building energy failure diagnosis and analysis system with integrated virtual sensor and deep learning, and a building energy failure diagnosis and analysis method using the same. In accordance with the present invention, the building energy failure diagnosis and analysis system with the integrated virtual sensor and deep learning comprises: a database which collects and stores the building temperature data measured by a temperature sensor unit in a building with one or more air conditioning apparatus; a virtual sensor unit which estimates an air conditioning apparatus virtual data by using the building temperature data, and generates a first failure diagnosis result through the air conditioning apparatus virtual data; and a management server which includes a deep learning unit which makes a deep learning model learn the building temperature data and the first failure diagnosis result of the virtual sensor unit, and generates a second failure diagnosis result through a simulation verification by using the deep learning model which has learned, and a failure diagnosis integration unit which integrates the first and second failure diagnosis results and generates a final diagnosis result. In addition, the building energy failure diagnosis and analysis method comprises: (a) a virtual sensor diagnosis step to use the building temperature data measured by the temperature sensor unit, estimate an air conditioning apparatus virtual data, and generate the first failure diagnosis result through the air conditioning apparatus virtual data; (b) a deep learning diagnosis step to make the deep learning model learn the building temperature data and the first failure diagnosis result, and generate the second failure diagnosis result through the simulation verification by using the deep learning model which has learned; and (c) a failure diagnosis integration step to integrate the first and second failure diagnosis results and generate the final diagnosis result. The present invention aims to provide a building energy failure diagnosis and analysis system with integrated virtual sensor and deep learning, and a building energy failure diagnosis and analysis method using the same, which are able to efficiently maintain and repair an air conditioning apparatus in a building.

Description

Building Energy Failure Diagnosis and Analysis System with integrated Virtual sensor and Deep learning, and Building Energy Failure Diagnosis and Analysis Method using the same

The present invention relates to a building energy failure diagnosis and analysis system integrating a virtual sensor and deep learning, and a building energy failure diagnosis and analysis method using the same. By automatically diagnosing and analyzing failures using deep learning technology with and an analysis system, and a building energy failure diagnosis and analysis method using the same.

With the recent rapid increase in energy demand, the importance of energy system management for commercial buildings is increasing day by day.

In 2010, electricity consumption in commercial buildings in the United States accounted for 35% of total electricity consumption in the United States, and air conditioners in commercial buildings accounted for about 30% of electricity consumption in buildings.

In addition, governments around the world are actively conducting research on smart cities as an attempt to pursue urban advancement and sustainability.

Globally, the building sector accounts for about 36% of energy consumption, and building energy-related greenhouse gas emissions account for about 39% of greenhouse gas emissions.

Korea also accounts for 30% of total energy consumption in buildings, which is lower than the global average, but in terms of unit area, it consumes much more energy than advanced countries such as the United States or Japan.

It is necessary not only to improve the energy efficiency of the building itself, but also to efficiently manage various facilities used inside the building. In particular, in the case of commercial buildings, energy used to maintain a pleasant indoor environment accounts for about 75% of building energy consumption. It is presumed that it will be possible

Accordingly, hardware and software technologies related to architecture, machinery, and control systems have been developed for various systems constituting a building for the past 10 years, but have been focused on unit technology development.

However, since the building is composed of various systems, the optimal control panel according to the objective function and situational awareness through the mutual information exchange of building facilities is required to improve the actual energy efficiency.

Therefore, in terms of specific technology development, it is necessary to collect, process, analyze, and predict state information for situational awareness, and to determine the optimal control direction according to the final predicted state and goal.

In order to solve the above problems, the present invention collects building temperature data with a general-purpose sensor and utilizes deep learning technology based on data estimated through a virtual sensor referring to thermodynamic and hydrodynamic information, thereby providing a proactive predictive formula A building energy failure diagnosis and analysis system that integrates a virtual sensor and deep learning that can efficiently maintain/repair the air conditioning system of a building through failure diagnosis and optimize energy use through this, and building energy failure diagnosis and An object of the present invention is to provide an analysis method.

In order to solve the above problems, a building energy failure diagnosis and analysis system integrating a virtual sensor and deep learning according to an embodiment of the present invention is a building temperature data measured from a temperature sensor unit in a building in which at least one air conditioner is installed. a database that collects and stores; a virtual sensor unit for estimating virtual data of the air conditioner by using the building temperature data and generating a first failure diagnosis result through the virtual data for the air conditioner; A deep learning unit that learns the building temperature data and the first failure diagnosis result of the virtual sensor unit in a deep learning model, and generates a second failure diagnosis result through simulation verification using the learned deep learning model, and the first and first 2 It is possible to provide a building energy failure diagnosis and analysis system that integrates deep learning and a virtual sensor including a management server including a failure diagnosis integrator that generates a final diagnosis result by integrating the failure diagnosis results.

Here, the virtual sensor unit includes a virtual sensor for estimating virtual data of the air conditioner using the building temperature data; a virtual classification error unit for calculating a classification error value through the air conditioner virtual data; and a virtual fault diagnosis unit configured to generate diagnostic information by determining whether there is a failure based on the classification error value, and a virtual diagnosis output unit configured to generate a first fault diagnosis result based on the air conditioner virtual data and diagnosis information.

In addition, the temperature sensor unit evaporating temperature (T _eva ), condensing temperature (T _cond ), intake line temperature (Suction line temperature, T _suc ) of the air conditioner as the building temperature data, liquid line temperature (Liquid line temperature, T _liquid ), Discharging line temperature (T _dis ), Condenser Inlet dry temperature (T _{cond.in ), Condenser Oulet} dry temperature (T cond.out ) , characterized by measuring at least one of an evaporator inlet temperature (Evaporator Inlet dry temperature, T _eva.in ) and an evaporator outlet temperature (Evaporator _{Oulet dry temperature, T eva.out ).}

In addition, the virtual sensor may include: a compressor power sensor for estimating a compressor power usage using the evaporation temperature and the condensation temperature; a refrigerant mass flow sensor estimating a refrigerant mass flow rate through the building temperature data; At least one of a gas flow sensor estimating a gas volume flow rate of at least one of a condenser and an evaporator through the building temperature data, and a refrigerant charge amount sensor estimating a refrigerant charge amount using the evaporation temperature, condensation temperature, intake line temperature, and liquid line temperature may include.

In addition, the deep learning unit learns the building temperature data stored in the database in the deep learning model and the first failure diagnosis result of the virtual sensor unit, and a model selection unit for selecting the learned deep learning model; a deep learning verification unit that performs energy saving simulation verification through the selected deep learning model; A deep learning fault diagnosis unit that compares the verification result obtained from the deep learning verification unit with the actual building energy information to determine whether there is a failure and generates diagnosis information, and a deep learning diagnosis output that generates a second failure diagnosis result through the diagnosis information may include wealth.

In addition, the model selection unit is characterized in that by determining the prediction accuracy of the selected deep learning model, and selecting the deep learning model again when out of the confidence range.

In addition, the fault diagnosis integration unit determines that when both the first and second fault diagnosis results are faulty, it is determined as a fault state, and when only one of the first and second fault diagnosis results is a fault determination, it is determined as a warning state, When both the first and second fault diagnosis results are determined to be no faults, it is determined that no faults are present and a final diagnosis result is generated.

In addition, the management server may further include a failure recognition unit for checking whether repair and replacement has been performed for a certain period of time when repair and replacement are required according to the final diagnosis result generated by the failure diagnosis integration unit.

In addition, the building energy failure diagnosis and analysis method using the building energy failure diagnosis and analysis system integrated with the virtual sensor and deep learning according to an embodiment of the present invention is (a) air conditioning using the building temperature data measured from the temperature sensor unit after estimating the device virtual data, a virtual sensor diagnosis step of generating a first failure diagnosis result through the air conditioner virtual data; (b) a deep learning diagnosis step of learning the building temperature data and the first failure diagnosis result in the deep learning model, and generating a second failure diagnosis result through simulation verification using the learned deep learning model, and (c) the first It is possible to provide a building energy failure diagnosis and analysis method including a failure diagnosis integration step of generating a final diagnosis result by integrating the first and second failure diagnosis results.

Here, the step (a) may include an estimation step of estimating virtual data for an air conditioner using the building temperature data; an error calculation step of calculating a classification error value through the air conditioner virtual data; The method may include a virtual diagnosis step of generating diagnostic information by determining whether there is a failure based on the classification error value, and a result output step of generating a first failure diagnosis result using the air conditioner virtual data and diagnostic information.

In addition, the step (b) is a model selection step of learning the building temperature data stored in the database in the deep learning model and the first failure diagnosis result of the virtual sensor unit, and selecting the learned deep learning model; A deep learning verification step of performing energy saving simulation verification through the selected deep learning model; It may include a deep learning diagnosis step of comparing the verification result with the actual building energy information to determine whether there is a failure and generating diagnostic information, and a result output step of generating a second failure diagnosis result through the diagnosis information.

In addition, after step (c), if maintenance and replacement are required according to the generated final diagnosis result, the method may further include a management step of confirming whether the maintenance and replacement have been performed for a certain period of time.

The building energy failure diagnosis and analysis system that integrates the virtual sensor and deep learning according to the embodiment of the present invention as described above, and the building energy failure diagnosis and analysis method using the same, include a failure diagnosis and analysis algorithm having the concept of a virtual sensor and an artificial By integrating the failure diagnosis and analysis algorithm with the concept of a neural network, it is possible to efficiently maintain/repair the air conditioning system of the building through optimized energy utilization.

In other words, it collects building temperature data from the temperature sensor unit and uses deep learning technology to automatically diagnose and analyze failures accurately and in a short time based on the data estimated through the virtual sensor referring to thermodynamic and hydrodynamic information. The cause of the failure can be intuitively identified, allowing the administrator to respond quickly.

Accordingly, time wasted in the maintenance/repair process can be saved, manager satisfaction is improved, and building energy can be easily managed under optimal conditions.

1 is a block diagram illustrating a building energy failure diagnosis and analysis system integrating a virtual sensor and deep learning according to an embodiment of the present invention.
FIG. 2 is a conceptual diagram illustrating the temperature of the air conditioner measured by the temperature sensor of FIG. 1 .
Fig. 3 is a block diagram showing the configuration of the management server of Fig. 1;
FIG. 4 is a block diagram illustrating the configuration of the virtual sensor unit of FIG. 3 .
5 is a graph of the result of obtaining classification error values by comparing the refrigerant mass flow rates estimated from the compressor map virtual sensor, the energy balance virtual sensor, and the expansion device virtual sensor of the refrigerant mass flow sensor.
6 is a block diagram showing the configuration of the deep learning unit of FIG.
7 is a flowchart illustrating a building energy failure diagnosis and analysis method using a building energy failure diagnosis and analysis system integrating a virtual sensor and deep learning according to an embodiment of the present invention.
8 is a flowchart sequentially illustrating step S1 of FIG. 7 .
9 is a flowchart sequentially illustrating steps S2 of FIG. 7;

Hereinafter, the description of the present invention with reference to the drawings is not limited to specific embodiments, and various modifications may be made and various embodiments may be provided. In addition, it should be understood that the contents described below include all conversions, equivalents, and substitutes included in the spirit and scope of the present invention.

In the following description, terms such as first and second are terms used to describe various components, meanings are not limited thereto, and are used only for the purpose of distinguishing one component from other components.

Like reference numbers used throughout this specification refer to like elements.

As used herein, the singular expression includes the plural expression unless the context clearly dictates otherwise. In addition, terms such as "comprises", "comprising" or "have" described below are intended to designate the existence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification. It should be construed as not precluding the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. does not

In addition, in the description with reference to the accompanying drawings, the same components are given the same reference numerals regardless of the reference numerals, and the overlapping description thereof will be omitted. In describing the present invention, if it is determined that a detailed description of a related known technology may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the accompanying drawings 1 to 9.

1 is a block diagram illustrating a building energy failure diagnosis and analysis system integrating a virtual sensor and deep learning according to an embodiment of the present invention, and FIG. 2 is a diagram illustrating the temperature of the air conditioner measured by the temperature sensor unit of FIG. It is a conceptual diagram, Figure 3 is a block diagram showing the configuration of the management server of Figure 1, Figure 4 is a block diagram showing the configuration of the virtual sensor unit of Figure 3, Figure 5 is a compressor map virtual sensor of the refrigerant mass flow sensor, It is a graph as a result of obtaining a classification error value by comparing the refrigerant mass flow estimated from the energy balance virtual sensor and the expansion device virtual sensor, and FIG. 6 is a block diagram illustrating the configuration of the deep learning unit of FIG. 3 .

The building energy failure diagnosis and analysis system (hereinafter referred to as 'building energy failure diagnosis and analysis system') that integrates a virtual sensor and deep learning according to an embodiment of the present invention is applied to a building in which one or more air conditioners are installed, It enables the efficient maintenance/repair of the air conditioning system of the building by accurately predicting the failure of the building and identifying the cause of the failure in a short time. Through this, building energy can be easily managed under optimal conditions.

Referring to FIG. 1 , the building energy failure diagnosis and analysis system may include a temperature sensor unit 1 , a management server 2 , and a manager terminal 3 .

First, the temperature sensor unit 1 may include one or more general-purpose temperature sensors, measure the temperature of the air conditioner in a building in which one or more air conditioners are installed, obtain building temperature data, and transmit it to the management server 2 . there is.

The temperature sensor unit 1 is the building temperature data, evaporating temperature (T _eva ), condensing temperature (T _cond ), intake line temperature (Suction line temperature, T _suc ) of the air conditioner, liquid line temperature _{(Liquid line temperature, T liquid)} , the discharge line temperature _{(discharging line temperature, T dis)} , the condenser inlet temperature (inlet dry condenser temperature, T _cond.in), the condenser outlet temperature (condenser Oulet dry temperature, T _cond.out ), evaporator inlet dry temperature (T _eva.in ), and evaporator outlet temperature (Evaporator _Oulet dry temperature, T eva.out ) can be measured, and it is most preferable to measure all of them it doesn't happen As shown in Figure 2, the temperatures refer to the temperature of the refrigerant at each location.

The management server (2) receives the building temperature data from the temperature sensor unit (1), uses the building temperature data to diagnose the failure of the air conditioner through an algorithm that integrates the virtual sensor and deep learning, and determines the cause of the failure It is possible to generate a final diagnosis result by analyzing, and transmit it to the manager terminal 3 .

Referring to FIG. 3 , the management server 2 may include a database 20 , a virtual sensor unit 21 , a deep learning unit 22 , a fault diagnosis integration unit 23 , and a fault recognition unit 24 . .

Specifically, the database 20 may collect and store building temperature data measured by the temperature sensor unit in a building in which at least one air conditioner is installed. In addition, the database 20 includes the air conditioner virtual data generated from the virtual sensor unit 21 , the deep learning unit 22 and the fault diagnosis integration unit 23 , the first fault diagnosis result, the second fault diagnosis result, and the final diagnosis. You can save the results.

In addition, the database 20 may acquire and store actual building energy information, but is not limited thereto, and may store various information necessary for the system.

The virtual sensor unit 21 implements a failure diagnosis and analysis algorithm having the concept of a virtual sensor. After estimating the air conditioner virtual data using the building temperature data, the first failure diagnosis result through the air conditioner virtual data can create

4 , the virtual sensor unit 21 may include a virtual sensor 210 , a virtual classification error unit 211 , a virtual fault diagnosis unit 212 , and a virtual diagnosis output unit 213 .

The virtual sensor 210 may estimate the air conditioner virtual data using the building temperature data.

Here, the air conditioner virtual data may include one or more of compressor power usage, refrigerant mass flow rate, gas volume flow rate, refrigerant charge amount, etc. In order to estimate this, the virtual sensor 210 includes a compressor power sensor, a refrigerant mass flow sensor, It may include one or more of a gas flow sensor and a refrigerant charge sensor.

)을 추정할 수 있다.The compressor power sensor (Virtual Compressor Power (VCP) sensor) estimates the compressor power usage by using the evaporation temperature and the condensation temperature among the building temperature data, and the compressor power usage (

) can be estimated.

[Equation 1]

Here, T _c is the condensation temperature, T _e is the evaporation temperature, and a ₀ to a ₈ are coefficients of the independent variable.

Refrigerant mass flow sensor (Virtual Refriderant Mass Flow (VRMF) sensor) estimates refrigerant mass flow through building temperature data, and includes a compressor map virtual sensor, an energy balance virtual sensor, and an expansion device virtual sensor, at least one of them can be used to estimate the refrigerant mass flow, respectively.

)을 추정할 수 있다.First, the compressor map virtual sensor is a virtual sensor that estimates the refrigerant mass flow rate of the compressor based on the evaporation temperature and the condensation temperature, and the refrigerant mass flow rate (

) can be estimated.

[Equation 2]

는 냉매의 체적 유량, c₁ 내지 c₅는 독립변수의 계수들이다.where T _c is the condensation temperature, T _e is the evaporation temperature, ρ _suc,rated is the density of the intake line refrigerant gas,

is the volume flow rate of the refrigerant, and c ₁ to c ₅ are coefficients of the independent variable.

)을 추정할 수 있다.Second, the energy balance virtual sensor is a virtual sensor developed based on the principle of the energy balance equation and thermodynamic state applied to the refrigeration cycle. After calculating each enthalpy, the refrigerant mass flow rate is estimated using the energy balance equation. That is, the refrigerant mass flow rate (

) can be estimated.

[Equation 3]

는 압축기 전력 사용량, α_loss는 압축기 손실 계수, h_dis 배출 라인 냉매 가스의 엔탈피, T_dis는 배출 라인 온도, P_dis는 배출 라인 압력, h_suc는 흡기 라인 냉매 가스의 엔탈피, T_suc는 흡기 라인 온도, P_suc는 흡기 라인 압력이다.here,

is the compressor power usage, α _loss is the compressor loss factor, h _dis is the enthalpy of the discharge line refrigerant gas, T _dis is the discharge line temperature, P _dis is the discharge line pressure, h _suc is the enthalpy of the intake line refrigerant gas, T _suc is the intake line The temperature, P _suc, is the intake line pressure.

Finally, the expansion device virtual sensor is a virtual sensor mainly used to check failures such as leakage of expansion device valves using evaporation temperature, condensation temperature, intake line temperature, and liquid line temperature.

)을 하기 수학식 4를 통해 추정할 수 있다.The mass flow rate is determined by controlling the flow rate through the relationship that the cross-sectional area through which the refrigerant flows according to the position of the valve is changed, so the refrigerant mass flow rate (

) can be estimated through Equation 4 below.

[Equation 4]

는 팽창장치밸브를 지나는 냉매 질량 유량, D_orifice는 팽창장치밸브의 최대 개도, D_pin는 팽창장치밸브의 폐쇄 지름,

는 팽창장치밸브를 지나는 최대 냉매 질량 유량, C_O, C₁은 독립변수의 계수들, A_orifice는 팽창장치밸브의 최대 개도 면적,

는 냉매 밀도, K는 팽창장치밸브 형상별 계수, Pc는 응축기 압력, P_e는 증발기 압력이다.here,

is the refrigerant mass flow through the expansion device valve, D _orifice is the maximum opening _{of the expansion device valve, D pin} is the closing diameter of the expansion device valve,

is the maximum refrigerant mass flow passing through the expansion device valve, C _{O and} C ₁ are the coefficients of the independent variable, A _orifice is the maximum opening area of the expansion device valve,

is the refrigerant density, K is the coefficient for each expansion device valve shape, Pc is the condenser pressure, and P _e is the evaporator pressure.

On the other hand, the expansion device valve serves to lower the pressure to facilitate evaporation in the evaporator, and there are fixed orifices and variable expansion valves as types. The expansion device virtual sensor may be applied to a variable expansion valve.

A virtual air flow (VAF) sensor estimates the gas volume flow rate of one or more of a condenser and an evaporator through building temperature data. It is a virtual sensor that uses the flow rate to estimate the gas volume flow rate.

First, the condenser converts the refrigerant gas from the compressor into a liquid state for easy evaporation to a high-temperature, high-pressure refrigerant liquid through a state change process.

)을 추정할 수 있다.Accordingly, the gas flow sensor uses the heat transfer equilibrium between the energy emitted from the refrigerant gas passing through the condenser and the energy absorbed by the air to estimate the gas volume flow rate, so that the condenser inlet temperature, the condenser outlet temperature, the principle of the thermodynamic state and the refrigerant Using the mass flow rate, the gas volume flow rate through the condenser is

) can be estimated.

[Equation 5]

는 추정된 냉매 질량 유량, h_c,out는 응축기 출구 냉매 액체의 엔탈피, P_c는 응축기 압력, T_c,in은 응축기 입구 온도, T_c,out는 응축기 출구 온도, v_a는 공기의 비체적, T_a,out은 응축기 출구 공기의 온도, T_a,in은 응축기 입구 공기의 온도, c_p,a는 공기 비열이다.here,

is the estimated refrigerant mass flow rate, h _c,out is the enthalpy of the condenser outlet refrigerant liquid, P _c is the condenser pressure, T _c,in is the condenser inlet temperature, T _c,out is the condenser outlet temperature, and v _a is the specific volume of air , Ta _,out is the temperature of the condenser outlet air, T _a,in is the temperature of the condenser inlet air, and c _p,a is the specific heat of air.

In addition, in the evaporator, the low-temperature, low-pressure refrigerant discharged through the expansion valve is vaporized through heat exchange from the indoor air, and as a result, the room temperature is lowered so that the cooling effect appears.

)을 냉매측과의 에너지 평형을 이용하여 하기 수학식 6을 통해 추정할 수 있다.Accordingly, the gas flow sensor detects the gas volume flow rate of the air passing through the evaporator (

) can be estimated through Equation 6 below using the energy balance with the refrigerant side.

[Equation 6]

는 추정된 냉매 질량 유량, h_liq는 액체 라인 냉매 액체의 엔탈피, h_e,out는 증발기 출구 냉매 기체의 엔탈피, P_c는 응축기 압력, P_e는 증발기 압력, T_liq는 액체 라인 온도, T_e,out는 증발기 출구 온도, v_a는 공기의 비체적, h_a,out은 증발기 출구 공기의 엔탈피, h_a,in은 증발기 입구 공기의 엔탈피이다.here,

is the estimated refrigerant mass flow rate, h _liq is the enthalpy of the liquid line refrigerant liquid, h _e,out is the enthalpy of the evaporator outlet refrigerant gas, P _c is the condenser pressure, P _e is the evaporator pressure, T _liq is the liquid line temperature, T _{e ,out} is the evaporator outlet temperature, v _a is the specific volume of air, h _a,out is the enthalpy of the evaporator outlet air, and h _a,in is the enthalpy of the evaporator inlet air.

)을 추정할 수 있다.The refrigerant charge sensor (Virtual Refrigerant charge (VRC) sensor) uses the evaporation temperature, condensation temperature, intake line temperature, and liquid line temperature among the building temperature data to calculate the amount of refrigerant charge (

) can be estimated.

[Equation 7]

는 이상적인 냉매 충전량, T_sc는 응축기 출구 과냉도, T_sc,rated는 이상적인 냉매 충전량일 때 응축기 출구 과냉도, T_sh는 증발기 출구 과열도, T_sh,rated는 이상적인 냉매 충전량일 때 증발기 출구 과열도, x_evap,in는 증발기 입구 건도, x_{evap,in,rated}는 이상적인 냉매 충전량일 때 증발기 입구 건도, K_sc, K_{sh ,} K_x은 냉동 시스템에 대한 독립 변수 계수들이다.here,

is the ideal refrigerant charge, T _sc is the condenser outlet subcooling, T _sc,rated is the condenser outlet subcooling at the ideal refrigerant charge, T _sh is the evaporator outlet superheating, T _sh,rated is the evaporator outlet superheating at the ideal refrigerant charge , x _evap,in is the evaporator inlet dryness, x _{evap,in,rated} is the evaporator inlet dryness at the ideal refrigerant charge, and K _sc , K _{sh ,} K _x are the independent variable coefficients for the refrigeration system.

In the above, the degree of superheat means that the refrigerant is consumed in the evaporator so as not to be in a liquid state and is appropriately superheated, and the degree of supercooling means that the refrigerant is statically supercooled so that there is no gaseous state in the condenser.

If the refrigerant charge is too large, the condenser outlet is filled with the supercooled refrigerant liquid, which reduces the condensation effect, reducing the condensation capacity and increasing the compression ratio. That is, the degree of subcooling increases and the degree of superheat decreases. Conversely, when the refrigerant filling strategy is small, the outlet of the evaporator is not used for evaporation, and the evaporator area is reduced by serving as a suction pipe, and the suction pipe is lengthened, thereby increasing the loss of suction pressure. That is, the degree of superheating increases and the degree of subcooling decreases.

As described above, the virtual sensor 210 may acquire the air conditioner virtual data by estimating the compressor power consumption, the refrigerant mass flow rate, the gas volume flow rate of the condenser, the gas volume flow rate of the evaporator, the refrigerant charge amount, and the like.

The virtual classification error unit 211 may calculate a classification error value through the air conditioner virtual data estimated from the virtual sensor 210 .

That is, the virtual classification error unit 211 may calculate a classification error value by comparing values of the air conditioner virtual data estimated by the virtual sensor using the actual failure information.

The virtual failure diagnosis unit 212 may generate diagnostic information by determining whether there is a failure based on the classification error value.

For example, when an air conditioner having a compressor failure is predicted by a virtual sensor using the present system, the result shown in FIG. 6 may be obtained through the virtual classification error unit 211 .

Referring to Figure 5, as a result of comparing the values estimated from the energy balance virtual sensor, the inflation device virtual sensor, and the compression map virtual sensor, respectively, the classification error value of the energy balance virtual sensor is 4% or less, that of the inflation device virtual sensor Although the classification error value was less than 3%, it can be seen that the classification error value of the compressor map virtual sensor was up to 19%. Here, it was confirmed that the classification error value of the compressor map virtual sensor increased as the degree of compressor failure increased.

Accordingly, the virtual failure diagnosis unit 212 may generate diagnostic information by determining a failure when the classification error value is greater than the error threshold value based on the error threshold value after obtaining the above result. Conversely, when the classification error value is smaller than the error threshold value, it is determined that there is no failure and diagnostic information can be generated.

Also, it is possible to generate diagnostic information by deriving which part of the air conditioner is defective according to which sensor has estimated a value having a classification error value exceeding the error threshold.

For example, it is possible to generate diagnostic information due to a compressor failure, an expansion device failure, and the like.

The virtual diagnosis output unit 213 generates a first fault diagnosis result including the air conditioner virtual data estimated from the virtual sensor 210 and the diagnosis information generated from the virtual fault diagnosis unit 212, and the database 20; It can be transmitted to the deep learning unit 22 and the fault diagnosis integration unit 23 to be stored or used.

The deep learning unit 22 implements a failure diagnosis and analysis algorithm having the concept of deep learning, and learns the building temperature data and the first failure diagnosis result of the virtual sensor unit 21 in the deep learning model, A second fault diagnosis result can be generated through simulation verification using a deep learning model.

Referring to FIG. 6 , the deep learning unit 22 may include a model selection unit 220 , a deep learning verification unit 221 , a deep learning failure diagnosis unit 222 , and a deep learning diagnosis output unit 223 . .

The model selection unit 220 may learn the building temperature data stored in the database in the deep learning model and the first failure diagnosis result of the virtual sensor unit 21 and select the learned deep learning model.

In this case, the building temperature data and the failure diagnosis results can be used for learning the deep learning model through a normalization process, which can prevent the learning results from divergence. Here, the normalization process may be performed using a GPU parallel processing technique, but is not limited thereto.

In addition, the model selection unit 220 may determine the prediction accuracy of the selected deep learning model, and select the deep learning model again when out of the confidence range. That is, only when the prediction accuracy falls within the confidence range, the deep learning model can be finally selected so that the next process can be performed.

This can improve the accuracy of the final diagnostic result in the system.

The deep learning verification unit 221 may obtain a verification result by performing energy saving simulation verification through the deep learning model selected by the model selection unit 220 .

Energy saving simulation verification predicts the peak power of a building through a deep learning model, and the predicted building power amount can be obtained as a verification result. can also be obtained

The deep learning failure diagnosis unit 222 may compare the verification result obtained from the deep learning verification unit 221 with the actual building energy information as described above to determine whether there is a failure and generate diagnostic information.

The deep learning diagnosis output unit 223 may generate a second fault diagnosis result through the diagnosis information, store it in the database 20 , and transmit it to the fault diagnosis integration unit 23 for use.

The fault diagnosis integration unit 23 may generate a final diagnosis result by integrating the first fault diagnosis result generated from the virtual sensor unit 21 and the second fault diagnosis result generated from the deep learning unit 22 .

In this case, the fault diagnosis integration unit 23 may determine one of a fault state, a warning state, and a no fault state to generate a final diagnosis result, but is not limited thereto.

In the above case, the fault diagnosis integration unit 23 may determine a fault state when both the first fault diagnosis result and the second fault diagnosis result are fault determinations.

In addition, when only one of the first and second failure diagnosis results is a failure determination, the failure diagnosis integration unit 23 may determine a warning state.

Finally, the fault diagnosis integration unit 23 may determine that the first and second fault diagnosis results are no faults, and determine the fault-free state.

After generating the final diagnosis result in this way, the fault diagnosis integrator 23 may store the final diagnosis result in the database 20 and transmit it to the manager terminal 3 .

The failure recognition unit 24 may check whether repair or replacement is performed for a certain period of time when repair or replacement is required according to the final diagnosis result generated by the failure diagnosis integration unit 23 .

That is, when the final diagnosis result is a failure state, repair and replacement are required. Accordingly, it is possible to check whether the manager has actually performed the repair and replacement in response.

If the failure is not resolved for a certain period of time, a notification may be transmitted to the manager terminal 3 to re-recognize the failure to take action.

The manager terminal 3 may be a terminal of a manager who is an air conditioning device user, a building manager, a resident, or the like, and may include a PC, a tablet, a personal digital assistant (PDA), etc. in addition to a mobile terminal.

The manager terminal 3 can receive the final diagnosis result from the management server 2 and provide it to the manager, and can receive all kinds of information in real time, so that the manager can monitor it in real time.

In addition, it allows the manager to control the air conditioner through the management server (2) using the manager terminal (3).

7 is a flowchart showing a building energy failure diagnosis and analysis method using a building energy failure diagnosis and analysis system integrated with a virtual sensor and deep learning according to an embodiment of the present invention, and FIG. 8 is a step S1 of FIG. 7 sequentially It is a flowchart shown, and FIG. 9 is a flowchart sequentially showing step S2 of FIG. 7 .

Referring to FIG. 7 , the method for diagnosing and analyzing a building energy failure using a building energy failure diagnosis and analysis system integrating a virtual sensor and deep learning according to an embodiment of the present invention includes a virtual sensor diagnosis step (S1), a deep learning diagnosis step (S2) and a fault diagnosis integration step (S3) may be included.

Specifically, in the virtual sensor diagnosis step (S1), the virtual sensor unit 21 of the management server 2 estimates the air conditioner virtual data using the building temperature data measured from the temperature sensor unit 1, and then A first failure diagnosis result may be generated through the virtual data.

To this end, step S1 may include an estimation step (S10), an error calculation step (S11), a virtual diagnosis step (S12), and a result output step (S13), as shown in FIG. 8 .

In the estimating step ( S10 ), the virtual sensor 210 of the virtual sensor unit 21 may estimate virtual data of the air conditioner using the building temperature data measured from the temperature sensor unit 1 .

At this time, in step S10, the virtual sensor unit 21 includes at least one of a compressor power sensor, a refrigerant mass flow sensor, a gas flow sensor, and a refrigerant charge sensor, and uses the compressor power consumption, refrigerant mass flow rate, and gas volume flow rate as virtual data of the air conditioner. , a refrigerant charge amount, etc. can be estimated, which is the same as the system description, and thus will be omitted.

In the error calculation step S11, the virtual classification error unit 211 of the virtual sensor unit 21 may calculate a classification error value through the air conditioner virtual data. A classification error value can be calculated by comparing data values with each other.

In the virtual diagnosis step ( S12 ), the virtual fault diagnosis unit 212 of the virtual sensor unit 21 may determine whether there is a fault based on the classification error value and generate diagnosis information. This is the same as that described in the above system, so a detailed description will be omitted.

In the result output step (S13), the virtual diagnosis output unit 213 of the virtual sensor unit 21 generates a first failure diagnosis result through the air conditioner virtual data and diagnosis information, stores it in the database 20, and stores it in the management server ( 2) can be transmitted to the deep learning unit 22, the fault diagnosis integration unit 23, and the like.

In the deep learning diagnosis step (S2), the deep learning unit 22 of the management server 2 learns the building temperature data and the first failure diagnosis result in the deep learning model, and uses the learned deep learning model to verify the simulation. A second fault diagnosis result may be generated.

Referring to FIG. 9 , step S2 may include a model selection step (S20), a deep learning verification step (S21), a deep learning diagnosis step (S22), and a result output step (S23).

In the model selection step (S20), the model selection unit 220 of the deep learning unit 22 learns the building temperature data stored in the database 20 in the deep learning model and the first failure diagnosis result of the virtual sensor unit 21, , the trained deep learning model can be selected.

In addition, in step S20, after selecting a deep learning model, the prediction accuracy of the selected deep learning model can be determined. At this time, if the prediction accuracy is out of the confidence range, the deep learning model is selected again, there is. A detailed description will be omitted because it overlaps with the description of the system.

In the deep learning verification step (S21), the deep learning verification unit 221 of the deep learning unit 22 may obtain a verification result by performing energy saving simulation verification through the selected deep learning model. This is the same as that described in the above system, so a detailed description will be omitted.

In the deep learning diagnosis step (S22), the deep learning failure diagnosis unit 222 of the deep learning unit 22 may compare the verification result with the actual building energy information to determine whether there is a failure and generate diagnostic information.

In the result output step (S23), the deep learning diagnosis output unit 223 of the deep learning unit 22 generates a second failure diagnosis result through the diagnosis information, and stores it in the database 20 of the management server 2, It can be transmitted to the fault diagnosis integration unit 23 .

In the fault diagnosis integration step S3 , the fault diagnosis integration unit 23 of the management server 2 may integrate the first and second fault diagnosis results to generate a final diagnosis result. Since the integration method is the same as that described in the above system, a detailed description thereof will be omitted.

In addition, the building energy failure diagnosis and analysis method using the building energy failure diagnosis and analysis system integrated with the virtual sensor and deep learning according to the embodiment of the present invention is performed after the step S3, according to the generated final diagnosis result, the management step (S4) may further include.

That is, if the final diagnosis result generated in step S3 is in a faulty state, repair and replacement are required, and accordingly, it can be checked through the management step (S4) whether maintenance and replacement have been performed for a certain period of time.

In step S4, after confirming whether maintenance and replacement have been implemented for a certain period of time, if the enforcement is not performed, a notification may be transmitted to the manager terminal 3 .

As described above, the building energy failure diagnosis and analysis system that integrates the virtual sensor and deep learning according to an embodiment of the present invention, and the building energy failure diagnosis and analysis method using the same, have the concept of a virtual sensor. By integrating the algorithm and the failure diagnosis and analysis algorithm with the concept of an artificial neural network, it is possible to efficiently maintain/repair the air conditioning system of the building through the optimized energy utilization.

In other words, it collects building temperature data from the temperature sensor unit and uses deep learning technology to automatically diagnose and analyze failures accurately and in a short time based on the data estimated through the virtual sensor referring to thermodynamic and hydrodynamic information. The cause of the failure can be intuitively identified, allowing the administrator to respond quickly.

Accordingly, time wasted in the maintenance/repair process can be saved, manager satisfaction is improved, and building energy can be easily managed under optimal conditions.

The embodiment of the present invention described above is not implemented only through an apparatus and/or method, but may be implemented through a program for realizing a function corresponding to the configuration of the embodiment of the present invention, a recording medium in which the program is recorded, etc. Also, such an implementation can be easily implemented by those skilled in the art to which the present invention pertains from the description of the above-described embodiments.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improved forms of the present invention are also provided by those skilled in the art using the basic concept of the present invention as defined in the following claims. is within the scope of the right.

1: temperature sensor unit
2: Management Server
20: Database
21: virtual sensor unit
210: virtual sensor
211: virtual classification error part
212: virtual fault diagnosis unit
213: virtual diagnostic output unit
22: Deep Learning Department
220: model selection unit
221: deep learning verification unit
222: deep learning fault diagnosis unit
223: deep learning diagnostic output unit
23: fault diagnosis integration unit
24: fault recognition unit
3: Administrator terminal

Claims (12)

a database for collecting and storing building temperature data measured from a temperature sensor unit in a building in which at least one air conditioner is installed;
a virtual sensor unit for estimating virtual data of the air conditioner by using the building temperature data and generating a first failure diagnosis result through the virtual data of the air conditioner;
A deep learning unit for learning the building temperature data and the first failure diagnosis result of the virtual sensor unit in a deep learning model, and generating a second failure diagnosis result through simulation verification using the learned deep learning model;
A building energy failure diagnosis and analysis system that integrates a virtual sensor and deep learning including a management server including a failure diagnosis integrator for generating a final diagnosis result by integrating the first and second failure diagnosis results.
According to claim 1,
The virtual sensor unit,
a virtual sensor for estimating virtual data of an air conditioner using the building temperature data;
a virtual classification error unit for calculating a classification error value through the air conditioner virtual data;
a virtual fault diagnosis unit that determines whether there is a fault based on the classification error value and generates diagnostic information; and
A building energy failure diagnosis and analysis system that integrates a virtual sensor and deep learning including a virtual diagnosis output unit that generates a first failure diagnosis result through the air conditioning device virtual data and diagnosis information.
3. The method of claim 2,
The temperature sensor unit,
_{Evaporating temperature, T eva} , Condensing temperature, T _cond , Suction line temperature, T _suc , Liquid line temperature, T of the air conditioner as the building temperature data _liquid ), Discharging line temperature (T _dis ), Condenser Inlet dry temperature (T _{cond .in} ), Condenser Oulet dry temperature (T _{cond .out} ), Evaporator inlet temperature (Evaporator) Inlet dry temperature, T _eva.in ) and evaporator outlet temperature (Evaporator Oulet dry temperature, T _{eva .out} ) A building energy failure diagnosis and analysis system incorporating a virtual sensor and deep learning, characterized in that it is measured.
4. The method of claim 3,
The virtual sensor is
a compressor power sensor for estimating compressor power usage using the evaporation temperature and the condensation temperature;
a refrigerant mass flow sensor for estimating a refrigerant mass flow rate through the building temperature data; a gas flow sensor for estimating a gas volume flow rate of at least one of a condenser and an evaporator through the building temperature data; and
A building energy failure diagnosis and analysis system integrating deep learning and a virtual sensor including at least one of a refrigerant charge amount sensor for estimating a refrigerant charge amount using the evaporation temperature, condensation temperature, intake line temperature, and liquid line temperature.
According to claim 1,
The deep learning unit,
a model selection unit for learning the building temperature data stored in the database and a first failure diagnosis result of the virtual sensor unit in the deep learning model, and selecting the learned deep learning model;
a deep learning verification unit that performs energy saving simulation verification through the selected deep learning model;
A deep learning failure diagnosis unit that compares the verification result obtained from the deep learning verification unit with the actual building energy information to determine whether there is a failure and generates diagnostic information; and
A building energy failure diagnosis and analysis system that integrates a virtual sensor and deep learning including a deep learning diagnostic output unit that generates a second failure diagnosis result through the diagnostic information.
6. The method of claim 5,
The model selection unit,
A building energy failure diagnosis and analysis system that integrates a virtual sensor and deep learning, characterized in that it determines the prediction accuracy of the selected deep learning model and reselects the deep learning model if it is out of the confidence range.
According to claim 1,
The fault diagnosis integration unit,
If both the first and second failure diagnosis results are failure determination, it is determined as a failure state,
If only one of the first and second failure diagnosis results is a failure determination, it is determined as a warning state,
Building energy failure diagnosis and analysis system integrating virtual sensor and deep learning, characterized in that when the first and second failure diagnosis results are both failure-free determination, the failure-free state is determined and a final diagnostic result is generated.
8. The method of claim 7,
The management server,
When repair and replacement are required according to the final diagnosis result generated from the failure diagnosis integration unit, the building energy failure diagnosis and analysis system.
In a building energy failure diagnosis and analysis method using a building energy failure diagnosis and analysis system that integrates a virtual sensor and deep learning,
(a) a virtual sensor diagnosis step of estimating the air conditioner virtual data using the building temperature data measured from the temperature sensor unit, and then generating a first failure diagnosis result through the air conditioner virtual data;
(b) a deep learning diagnosis step of learning the building temperature data and the first failure diagnosis result in the deep learning model, and generating a second failure diagnosis result through simulation verification using the learned deep learning model; and
(c) a failure diagnosis integration step of generating a final diagnosis result by integrating the first and second failure diagnosis results; and a building energy failure diagnosis and analysis method.
10. The method of claim 9,
The step (a) is,
an estimation step of estimating virtual data of an air conditioner using the building temperature data;
an error calculation step of calculating a classification error value through the air conditioner virtual data;
a virtual diagnosis step of generating diagnostic information by judging whether there is a failure based on the classification error value; and
and a result output step of generating a first failure diagnosis result through the air conditioner virtual data and diagnosis information.
10. The method of claim 9,
Step (b) is,
a model selection step of learning the building temperature data stored in the database in the deep learning model and the first failure diagnosis result of the virtual sensor unit, and selecting the learned deep learning model;
A deep learning verification step of performing energy saving simulation verification through the selected deep learning model;
A deep learning diagnostic step that compares the verification result with the actual building energy information to determine whether there is a failure and generates diagnostic information;
and a result output step of generating a second failure diagnosis result through the diagnosis information.
10. The method of claim 9,
After step (c),
If repair and replacement are required according to the generated final diagnosis result, the building energy failure diagnosis and analysis method further comprising a management step of checking whether repair and replacement have been performed for a certain period of time.

Images