KR100466433B1 - Method for analysis of efficiency in turbo refrigerator and apparatus thereof - Google Patents

Abstract

The present invention provides a refrigeration cycle including a compressor, a condenser, an expander, and an evaporator, a radiant water outlet and an inlet formed to discharge and introduce radiant water between the condenser and the cooling tower, and an endotherm between the evaporator and the freezing target part. And a method for analyzing the performance of a turbo chiller having an endothermic water outlet and an inlet formed to be introduced, the method comprising: measuring temperatures of the endothermic water inlet and outlet; Measuring a temperature of the heat radiation water inlet; Measuring a flow rate of the endothermic water; Obtaining a cooling load of an evaporator from the endothermic water inlet temperature, the endothermic water outlet temperature, and the endothermic water flow rate; The temperature of the endothermic water outlet, the temperature of the radiant water inlet, and the cooling load of the evaporator are set as input variables, two membership functions are set for each of the three input variables, and the data obtained by repeating the measurement process several times is the variable value. Establishing a functional relationship between the variables by a neurofuzzy technique; Analyzing the variation of the performance coefficient according to the variation of each variable by substituting the data measured by the measurement process under arbitrary conditions to the variable of the set function; By presenting a performance analysis system that includes, it is possible to analyze and predict the performance of the freezer under any given conditions for a turbo installed in the field.

Description

METHOD FOR ANALYSIS OF EFFICIENCY IN TURBO REFRIGERATOR AND APPARATUS THEREOF

The present invention relates to a turbo chiller, and more particularly, to a method and apparatus for analyzing and predicting a chiller's performance from measured values under any conditions relating to the temperature of the radiant and endothermic water of the turbo chiller installed in the field and the load of the evaporator. It is about.

Today, the issue of resource depletion and environmental pollution has risen, and interest in energy saving has been increasing all over the world for a long time.

In particular, in recent years, the government has actively invested in energy-saving matters such as establishing an energy service company (ESCO) system and strengthening support for the target companies in order to effectively promote energy-saving projects. And focusing efforts.

In order to operate the energy saving business more efficiently, the process of properly understanding the energy use of the heat source and the air conditioning system currently used must be preceded, and for this purpose, an accurate diagnosis of the refrigeration system operating in the field is first of all necessary.

Refrigerators used in air conditioning systems to handle cooling loads in buildings and factories can be said to have high power consumption and a high potential for improvement.

In particular, the turbo chiller can save a lot of energy by proper design, selection, installation and maintenance. Unlike a small refrigeration system that is sold and installed in one-piece form, it is installed as part of a heat source device in the whole air conditioning system, so the operating temperature of the radiant water and the endothermic water as well as the radiant water, endothermic pump, air conditioner, and piping conditions This is because there are many variables that can affect the performance, such as the flow rate may flow differently from the reference conditions.

In addition, even after installation of the turbo refrigerator, the performance of the air conditioning system can be improved by changing the flow rate of the radiant water or the endothermic water.

In practice, when measuring the performance of a refrigerator installed on site, it is often impossible to obtain performance test data under standard conditions specified by regulation by arbitrarily adjusting the temperature and flow rate of radiant or absorbent water. Because of the control characteristics, the freezer operates abnormally, so the performance analysis in the quasi-steady state should be performed.

Therefore, there is an urgent need to develop a model that can predict the performance of a refrigerator under arbitrary temperature and flow conditions by using performance data of a quasi-steady state of a refrigerator installed and operating in the field.

The present invention solves the above problems, the performance analysis method and apparatus of the turbo-cooler to analyze and predict the performance of the freezer under any given conditions, such as the radiant and endothermic flow rate, temperature of the turbo-cooler installed in the field To provide for that purpose.

1 is a block diagram showing the configuration of the performance analysis apparatus according to the present invention

2 is a graph showing the relationship between the average cooling load and the coefficient of performance measured during operation of the refrigerator;

FIG. 3 is a graph showing only data that satisfies a quasi-normal state among data of FIG. 2;

4 is a configuration diagram showing the structure of the neurofuge model

5 is a graph showing the results of training the neurofuge model using data acquired for four days in early June.

FIG. 6 is a graph comparing performance coefficients predicted using the neurofuge model of FIG. 5 and performance coefficients based on data measured in June to September.

7 is a graph showing the results of training the neurofuge model using data acquired for 10 days in July.

FIG. 8 is a graph comparing performance coefficients predicted using the neurofuge model of FIG. 7 and performance coefficients based on data measured in June to September

9 is a graph showing the results of predicting the performance of the refrigerator according to the load under standard conditions using a trained neurofuge model.

** Description of the symbols for the main parts of the drawings **

10 refrigeration cycle 11: compressor

12 condenser 13 expander

14 evaporator 20 cooling tower

21: heat radiation water inlet temperature measuring means 30: refrigeration target portion

31 endothermic water inlet temperature measuring means 32 endothermic water outlet temperature measuring means

33: endothermic water flow rate measuring means 40: analysis means

The present invention is to provide a model that can be equipped with a measuring means in the minimum portion of the turbo freezer in the field for collecting data necessary for performance analysis of the refrigerator, and predict the performance coefficient of the turbo freezer from the experimental data obtained from the measuring means. will be.

In general, since there are no detailed specifications or performance data of a compressor or evaporator of a turbo refrigerator installed in the field, it is a reality that individual device characteristics of the turbo refrigerator are not predicted or performance degradation due to long-term operation.

In addition, the experimental data necessary for the prediction of the coefficient of performance is not obtained by changing the arbitrary conditions in the laboratory scale experimental apparatus, but is made on the premise of the refrigeration load and the atmospheric condition in the state of being installed and operated in the field. Since it is necessary to obtain, it is very difficult to obtain a wide range of experimental data due to changes in various conditions. Therefore, appropriate modeling is needed to predict the performance of turbo chillers from limited experimental data obtained in the field.

In modeling, it would be ideal to consider individual specifications for each turbocooler. However, due to the lack of specification data and experimental data, it is impossible to obtain satisfactory results because appropriate values of coefficients required for tuning the model cannot be obtained.

Accordingly, the present invention designates the radiant water, the temperature of the endothermic water, and the load of the evaporator, which are the factors that have the most decisive influence on the performance of the refrigeration system, as variables, and employs a neurofuge technique (ANFIS: adaptive neuro-fuzzy inference system). By deriving a functional relationship between the variables, it is proposed a model that can analyze and predict the performance coefficient of the refrigerator from the relationship of the variables represented by the function.

The performance analysis system of a turbo-cooler according to the present invention includes a performance analysis apparatus including measurement means for obtaining data and analysis means for processing the collected data, and a performance analysis method for processing the obtained data by analysis means. Since it can be expressed in two categories, hereinafter, the configuration of the performance analysis device will be described, and then the analysis method performed in the device will be described in detail.

1 is a block diagram showing the configuration of a performance analysis device for a turbo-frigerator according to the present invention.

As shown, performance analysis of a turbo chiller according to the present invention basically comprises a refrigeration cycle (10) comprising a compressor (11), a condenser (12), an expander (13) and an evaporator (14), and a condenser (12). And a radiant water outlet and an inlet formed to flow in and out of the radiant water between the cooling tower 20 and a endothermic water outlet and an inlet formed to flow and enter the endothermic water between the evaporator 14 and the freezing target unit 30. It is used to analyze the performance of a turbocooler.

Temperature measuring means 21 is mounted at the inlet of the radiant water for performance analysis according to the present invention, and temperature measuring means 31 and 32 are also mounted at the inlet and outlet of the endothermic water. The measuring means 33 is mounted.

In addition, an analyzing means 40 for processing the values obtained from the temperature measuring means 21, 31, 32 and the flow rate measuring means 33 is provided and is electrically connected to the respective measuring means.

Here, the temperature measuring means 21 and 31 may be easily implemented by a thermocouple, a platinum resistance temperature sensor, a thermopile, a thermowell into which a platinum resistance temperature sensor is inserted, and the flow rate measurement. The means 23 and 33 may be implemented by an ultrasonic flow meter or the like, and the analyzing means 40 may be implemented by a computer to allow data obtained from the measuring means to be input and processed by a method as described below. have.

Each measurement value is stored in the analysis means 40 at intervals of less than one minute so as to interpret abnormal behavior of the refrigerator.

Hereinafter, a method of analyzing the variation of the performance coefficient of the refrigerator according to the variation of each variable (temperature, flow rate, etc.) from the data obtained from the measuring means will be described.

In order to predict the performance under any conditions of the turbo chiller to be measured, performance measurement in the field must be performed first.

Next, after determining whether enough data for performance prediction under certain conditions has been accumulated or not, the neurofuzzy technique is applied to extract a nonlinear functional relationship between data inputs and outputs. The performance of the turbo chiller is predicted and analyzed by substituting the data measured under the conditions into the function.

On-site performance measurements should be carried out within a range that does not interfere with the operation of the refrigeration machine in operation, so that the temperature measurement means and the flow rate measurement means should be made non-contact.

The neurofuzzy technique, a nonlinear interpolation technique that derives a functional relationship by analyzing the relationship between input and output data, is a system that infers only the correlation between input and output without prior information on the physical characteristics of the object to be analyzed. If there is not enough data available over time, we cannot be sure of the reliability of the trained inference system.

In order to solve this problem, it is necessary to filter the data to be used for training the neurofuge technique by using the prior information on the analysis target.

Many studies on thermodynamic-based performance prediction modeling have shown that the reciprocal of the coefficient of performance (COP) 1 / COP and the reciprocal of the cooling load (Q _e : cooling load) 1 / Q _e under steady-state operating conditions. Is said to have an empirical linear relationship as in equation (1).

The reason why the linear relationship is established is that the irreversibility generated during the heat exchange process between the condenser and the evaporator of the refrigerator under the steady state is proportional to the load on the evaporator side Q _e .

This empirical fact is useful information for extracting quasi-steady state data necessary for performance prediction from the measured data.

Neuropurge training requires extracting as many quasi-steady-state data as possible for all evaporator-side load areas from several days of data, which means that the integral mean time of the data measured at one-minute intervals should be as small as possible.

Since the dynamic time constant of medium and large turbo refrigerators is usually about 2 minutes, the 10-minute average data is applied to minimize the effects of abnormal conditions.

The average value of the result of abnormal state operation such as start, stop, etc. is out of the steady state value. To exclude these abnormal state values, the endothermic water outlet temperature is outside the range of the design steady state temperature ± 0.5 ℃ deviation, In the case of deviation from the linearity of Equation 1, the measurement data is filtered by excluding these values from the application object.

Figure 2 shows the 10-minute average value of the freezer data measured for 4 days, Figure 3 shows only the data that meets the above-mentioned quasi-normal conditions of the measured 10-minute average data.

According to the research results to quantify the functional relationship of Equation 1, it is found that Equation 1 is closely related to the radiant water inlet temperature T _cwi and the endothermic water outlet temperature T _ewo .

Similarly, the radiant water outlet temperature and the endothermic water inlet temperature that may be associated are reflected in the performance of the heat exchanger in conjunction with the cooling treatment load, and thus may be omitted in view of simplifying the number of input variables.

Therefore, the neurofuge model may be assumed to have a functional relationship as shown in Equation 2.

Here, the radiant water inlet temperature T _cwi and the endothermic water outlet temperature T _ewo can be measured by the temperature measuring means 21 and 32, and the cooling load Q _e of the evaporator is measured for the endothermic flow rate and the inlet / outlet of the endothermic water ( It can be calculated | required by Formula (3) from the inlet / outlet temperature difference of an evaporator.

Q _e : Cooling load m _ew : Endothermic flow rate

C _pw : Specific heat of water T _ewi : Evaporator inlet temperature

T _ewo : evaporator outlet temperature

Since the cooling load and the outside air temperature change with time, the temperature of the radiant and endothermic water flowing into the refrigerator is also changed. Accordingly, the partial load operation of the refrigerator and the control of the cooling tower are performed. Power, freezer pressure, etc. will also show periodic fluctuations. Therefore, the data is processed assuming that the time-averaged value by integrating at intervals of 10 minutes or more is assumed as the representative value of the time.

There is no specific rule for setting conditions of neurofuge structures necessary to simulate a given object, but the present invention uses a generalized bell function as a form of membership function for efficient analysis in consideration of the characteristics of the turbo chiller.

As shown in Equation 2, the input variables were set to three variables, and the number of membership functions required for each input variable was set to two.

In the case of setting three membership functions, the characteristics of the refrigerator tend to be over-fitting, because the behavior of the coefficient of performance based on the thermodynamic model shows a nearly linear behavior with respect to the corresponding input variables.

Therefore, setting two membership functions is preferable in terms of reducing the error between the model and the experimental results.

Figure 4 shows the neurofuge structure adopted in the performance analysis method according to the present invention. As shown, the number of fuzzy rules is 2 ³ = 8 and the number of nodes is 34.

The number of nonlinear variables related to the conditional membership function (nonlinear seed function) is 6 × 3 = 18, and the number of linear variables in the conclusion part by Sugeno type inference is 8 × 4 = 32.

The training of neurofuge means to find the optimal values of the 50 variables of the conditional and the conclusion. The hybrid training rule proposed in the neurofuge suggests that the determined linearity is determined after obtaining the conclusion linear variables with the least-squares method for a given training data set. Non-linear variable values are found by gradient descent method by back-propagation of the rate of change of error with respect to the variable values.

This process trains neurofuges until the final error reaches the design range, or by repeating only a fixed number of times.

Ideally, the data needed for neurofuge training should be evenly distributed over the inlet variable space, but in practice this is not possible. Therefore, a reliable neurofuge model should be devised based on short-term freezer performance diagnostics.

It is preferable to diagnose the freezer in the off-peak season rather than the freezer in summer, and take necessary measures.However, in the case of winter when the outside temperature is low, the freezer cannot be tested over the full load range because there is no load on the evaporator side. Diagnosis is impossible.

Therefore, neurofuge training for accurate performance diagnosis is preferably performed by applying the data between June and August.

FIG. 5 shows the results of training the neurofuge model based on data acquired during the first four days of June, and FIG. 6 shows an example of the ability to predict performance measured in June through September using the trained model. .

Since the data from the beginning of June, when the cooling load is not large, cannot obtain the performance data over the full load range of the refrigerator, it can be seen that the performance prediction error of the neurofuge model is large and the reliability is inferior.

On the other hand, the results of training the neurofuge model by acquiring data for 10 days in July are shown in FIG. 7. In the model of FIG. 7, the COP is evenly distributed over all the operable regions, and as shown in FIG. 8, the performance measured in June to September is well predicted within a 5% margin of error.

That is, as a result of applying the performance analysis system of the turbo-frigerator according to the present invention to the actual situation, it is possible to train neurofuge to have an error range within ± 5% if more than 10 days of measurement data is available from June to August. appear.

9 shows the results of predicting the performance using the neurofuge model under the condition that the radiant water inlet temperature and the endothermic water outlet temperature, which are the standard conditions of the freezer performance test specified in Korean Industrial Standard KS B 6270-85, are 32 ° C and 7 ° C, respectively. Indicates.

According to FIG. 9, neurofuge models all predict the same performance except early June and lacking cooling load in September.

A well-trained neurofuge model can be used to determine the coefficient of performance at any refrigeration load, thus predicting power consumption and improving the refrigeration operation control system based on these results, thus contributing to national energy savings.

The present invention provides a method and apparatus for analyzing the performance of a turbo-cooler for analyzing and predicting the performance of the refrigerator under any given conditions, such as the radiant and endothermic flow rates and temperature of a turbo-cooler installed in the field.

Claims (7)

A refrigeration cycle including a compressor, a condenser, an expander, and an evaporator, a radiant water outlet and an inlet configured to discharge and introduce radiant water between the condenser and the cooling tower, and endothermic water between the evaporator and the refrigeration target portion; A method for analyzing the performance of a turbo chiller having an endothermic water outlet and an inlet formed, Measuring the temperature of the endothermic water inlet and outlet; Measuring a temperature of the heat radiation water inlet; Measuring a flow rate of the endothermic water; Obtaining a cooling load of an evaporator by substituting the endothermic water inlet temperature, the endothermic water outlet temperature, and the endothermic water flow rate into Equation 3; The temperature of the endothermic water outlet, the temperature of the radiant water inlet, and the cooling load of the evaporator are set as input variables, two membership functions are set for each of the three input variables, and the data obtained by repeating the measurement process several times is the variable value. Establishing a functional relationship between the variables by a neurofuzzy technique; Analyzing the variation of the performance coefficient according to the variation of each variable by substituting the data measured by the measurement process under arbitrary conditions to the variable of the set function; Performance analysis method of the turbo-chiller comprising. A performance analysis apparatus for implementing the performance analysis method of the turbo chiller of claim 1, Temperature measuring means mounted to said endothermic water inlet and outlet and said radiant water inlet; Flow measurement means mounted to the endothermic water inlet or outlet; Analyzing means for analyzing the variation of the performance coefficient according to the variation of each variable by processing the value obtained from the measuring means by the performance analysis method of claim 1; Performance analysis device of the turbo-chiller comprising. The method of claim 2, The temperature measuring means is a performance analysis device of the turbo-freezer, characterized in that the thermocouple. The method of claim 2, The temperature measuring means is a performance analysis device of a turbo refrigerator, characterized in that the platinum resistance temperature sensor. The method of claim 2, The temperature measuring means is a performance analysis device of the turbo-cooler, characterized in that the thermopile. The method of claim 2, Wherein the temperature measuring means is a performance analysis device of a turbo-cooler, characterized in that the thermowell inserted platinum resistance temperature sensor. The method of claim 2, The flow rate measuring means is a performance analysis device of the turbo-cooler, characterized in that the ultrasonic flow meter.