TWI470217B - On - line performance evaluation method of cooling tower - Google Patents

Description

Cooling tower line performance evaluation method

The present invention relates to a method for evaluating the performance of a cooling device, and more particularly to an online performance evaluation method for a cooling water tower.

Cooling towers are a typical heat exchange unit that is widely used in factories, commercial buildings, and other buildings that have large air conditioning or cooling load equipment requirements. Since the accuracy of the cooling tower performance calculation will seriously affect the efficiency of the system operation, the performance evaluation of the cooling tower must be extra cautious.

The Cooling Tower established by the Cooling Tower Institute (CTI) uses the cooling tower performance evaluation method specified in the standard specification. It is the most commonly used evaluation method, but its experimental conditions are set so rigorously that it is generally used in the factory. The cooling tower cannot obtain data that meets the experimental conditions. The biggest reason is that it is impossible for the factory to change the normal operation of the cooling tower in order to obtain the data that meets the test conditions. This will affect the profit of the plant. Therefore, the CTI cooling tower performance evaluation method is only suitable for testing the cooling tower in the laboratory. The performance is not able to perform an on-line performance assessment of the cooling towers in the actual plant.

Therefore, it is necessary to provide an innovative and progressive cooling tower performance evaluation method to solve the above problems.

The invention provides an online performance evaluation method for a cooling water tower, comprising the following steps: (a) calculating a fan outlet air volume; (b) using a multi-model local model network (Local Model Network, LMN) method to establish a cooling water tower outlet water temperature model; (c) establish a predictive characteristic curve according to the cooling water tower outlet water temperature model: and (d) calculate the cooling tower efficiency by using the predicted characteristic curve and the original cooling tower characteristic curve .

The invention establishes a cooling water tower outlet water temperature model by using historical data and using a multi-model Local Model Network (LMN) method, and the cooling water tower outlet water temperature model can be performed without affecting the normal operation of the cooling tower. Make a correct and objective performance assessment of the cooling tower. Moreover, the implementation of the present invention does not increase the operating cost of the enterprise and, therefore, is well suited for conducting on-line performance evaluation of cooling towers in actual plants.

The embodiments of the present invention can be more clearly understood, and the objects, features, and advantages of the present invention will become more apparent. The details are as follows.

(L/G) _design ‧ ‧ liquid to gas ratio under design conditions

(L/G) _test ‧ ‧ liquid to gas ratio under test conditions

G‧‧‧ air volume

1 is a flow chart showing the method for evaluating the performance of the cooling tower of the present invention; FIG. 2 is a flow chart showing the modeling of the multi-model local model network method of the present invention; and FIG. 3 is a graph showing the predicted value and calculated value of the characteristic value of the cooling tower of the present invention. The comparison results; and Figure 4 shows the characteristic curve of the cooling tower of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a flow chart showing the method for evaluating the performance of the cooling water tower of the present invention. Referring to step S11 of Fig. 1, the fan outlet air volume is calculated. In this embodiment, the parameters for calculating the fan outlet air volume include the cooling water flow rate, the cooling water inlet and outlet temperature difference, the cooling water inlet and outlet heat enthalpy difference, and the cooling water inlet and outlet humidity difference, because the air volume change is affected by the parameters.

In addition, since the wet bulb thermometer or hygrometer is not installed at the fan outlet, the outlet humidity cannot be known. The invention adopts the data of taking the relative humidity of the inlet air by more than 85%, and performs the calculation on the basis that the relative humidity of the outlet has reached saturation of 100%. In addition, in order to avoid the difference between the high and low speeds of the fan and affect each other's results, the present invention adds the conditions that the six fans are both high or low. After calculation, the average air volume of the fan high and low speeds in each month is shown in Table 1, and the unit is Mm ³ .

From the calculation results in Table 1, it can be seen that the air volume of the high and low speed fans is almost the same, but the actual air volume has some slight errors in the calculation results of the high and low speed air volume. It is speculated that there may be two reasons: one is the relative humidity of the outlet air 100%. The assumption is unreasonable, because the inlet air temperature is higher when the relative humidity of the inlet air is 85%, the amount of water that can reach the relative humidity of 100% is more, so the relative humidity of the outlet may not reach 100%; The change of the fan's power consumption is reflected in the change of the air volume is not rapid, the fan power consumption change can be measured immediately, but the air volume change can not be immediately expressed, and the air volume involves many variables in the calculation, which may be caused by the disturbance. error. Therefore, in the present embodiment, the average air volume of three months is defined as the air volume of the fan at a high and low speed.

2 shows a modeling flow chart of the multi-model partial model network method of the present invention. Referring to step S12 and FIG. 2 of FIG. 1 , a cooling water tower outlet water temperature model is established by a multi-model local model network (LMN) method. In this embodiment, the multi-model partial model network method includes the following steps: Step S21: According to the inlet water temperature, the air temperature, the air humidity, and the fan power consumption. Fuzzy c-Mean (FCM) grouping; step S22: establishing satisfactory fuzzy c-means grouping according to inlet water temperature, air temperature, air humidity and fan power consumption; step S23: establishing linearity in each group a model; step S24: using the distance of the data from each of the cluster centers to give a weight to the global model; and step S25: establishing a multi-model weighted global cooling water tower outlet water temperature prediction function. The following is a detailed description of the flow of the cooling water tower outlet water temperature model for steps S21 to S25.

[Multi-model local model network (LMN)]

For multivariable systems with R-dimensional input M-dimensional output, the LMN-based multi-model is described as:

among them, Represents the system output vector at time k, σ _i is the scheduling function, which is the scheduling variable The function, Is the input variable The local model, the number of scheduling functions is n .

Although the selection of scheduling functions is diverse, the more common are Gaussian bells, ASMOD, MARS, and Kernel functions, among which the Gaussian bells function is the most common: Where c _i is the central variable of the Gaussian function and s _i is the width of the Gaussian function.

In order to ensure the uniformity of the input space division, each scheduling function is standardized:

The local model f _i in equation (1) can be in any form: non-linear or linear form, state space or input-output form, and discrete or continuous form, and the like. The invention uses a linear model:

It can be known from equation (1) that the multi-model modeling problem is the identification problem of the parameters n , c _i , s _i and θ _i . In order to avoid the nonlinear optimization problem of parameter identification, the above parameters are usually divided into two parts for identification, namely the scheduling function parameters n , c _i , s _i and the local model parameters θ _i .

[fuzzy c-means grouping]

Compared with other subgroups, the fuzzy c-means (FCM) grouping method is a well-established and simplest method of grouping, so it has been widely used in industrial processes.

Given data sample set The goal of the FCM clustering algorithm is to obtain the membership matrix U = [ μ _{i, k} ] _{c × N} and the cluster center V = [ v ₁ , v ₂ , ..., v _c ] by obtaining the objective function. When the number of clusters c is constant, the objective function of the FCM clustering algorithm can be expressed as follows: Satisfy: Where w is the index weight that affects the degree of fuzzification of the membership matrix, w (1, ∞). To solve the minimization problem of (5), you can get:

[satisfactory fuzzy c-mean grouping]

The satisfactory fuzzy c-means grouping algorithm specifies the initial number of clusters c = 2, and decides whether to add new clusters according to the modeling accuracy index that the user cares about ( c [2, c ^* ]). If the clustering result is not satisfactory, find a sample that is the most similar to the clustering center v _i ~ v _c in the data sample set as the new clustering center v _{c +1} , and use v _i ~ v _{c +1} as the initial grouping. calculating new center nonrandom membership matrix U _0, and the system re-partitioning c +1 by FCM algorithm. Repeat the above steps as required by the performance indicators until satisfactory results are obtained. Since the number of clustering is significantly reduced, and there is no need to re-initialize the random clustering center and membership matrix during the calculation process, the calculation amount will be greatly reduced, and the convergence speed of the algorithm will be significantly accelerated. When the grouping is over, the parameters to be identified n , c _i and the grouping result have the following relationship: n = c (9)

c _i = v _i , i =1,..., n (10) and s _i can be determined by the nearest neighbor heuristic algorithm: Where p is the number of the nearest neighbors of the i- th group; c _l ( l =1,..., p ) is the clustering center of the most neighbors of c _i .

Define z = [1 1...1] ^T , Where d is the number of elements in c _i , then the applicable domain of the i- th local model can be described as:

[Local Model Parameter Identification]

After determining the parameters n , c _i , s _i in the scheduling function, the scheduling function values of each data sample in the data set Φ can be calculated, so the identification problem of each local model parameter θ _i is simplified to a linear optimization problem. It can be solved by the least square method or the Kalman filter inverse algorithm. Its specific form is: Where i =1,..., n , m =1,..., M , J _i represent the performance index of the i- th local model, N _i is the number of data samples in the i- th neighborhood, y _im ( k ) and Represents the expected output value and the actual output value of the i- th local model, respectively.

By arranging the θ _{im of} each local model in a vector, the local performance index can be rewritten to the form of Θ _m : J _mL (Θ _m )=( Y _m -Ψ'Θ _m ) ^T Q ( Y _m -Ψ'Θ _m ) (14) where, , θ _im is the mth column vector of θ _i ; Y _m =[ y _m (1) y _m (2)... y _m ( N )] ^T is the actual output vector of the mth channel of the system;

Q = diag (max _i ρ _i (1) max _i ρ _i (2)...max _i ρ _i ( N )).

Minimizing J _mL gives an estimate of Θ _m :

Fig. 3 is a graph showing the comparison between the predicted value and the calculated value of the characteristic value of the cooling water tower of the present invention. Referring to step S13 and FIG. 3 of FIG. 1 , a prediction characteristic curve is established according to the water temperature model of the cooling water tower outlet. In this embodiment, the air temperature is set to 33 ° C, 35 ° C and 37 ° C, the corresponding relative humidity is 81.5%, 71.0% and 61.7%, so that the wet bulb temperature is 30 ° C design conditions, and Under the condition that the inlet water temperature is 45.6 ° C and the cooling water flow rate is 2.1×10 ⁷ kg/hr, the fan water consumption power is substituted into the outlet water temperature model of the cooling water tower, and the outlet water temperature under the condition can be obtained, and The air volume of the fan at the power consumption can be calculated to obtain the liquid-gas ratio and calculate the characteristic value (KaV/L) under the condition. As shown in Fig. 3, the cooling tower characteristic curves under three different inlet air conditions can be established, and it can be found that the curves have similar results.

In order to check whether the predicted characteristic curve has accuracy, in the present embodiment, the actual data within ±5% of the design value error is taken to calculate the KaV/L value and the liquid-gas ratio, and the result is the same as the cooling water tower outlet water temperature model. The established predictive characteristic curve (dotted line of Figure 3) coincides, indicating that the predicted characteristic curve does have accuracy.

Figure 4 is a graph showing the characteristics of the cooling tower of the present invention. Referring to step S14 and FIG. 4 of FIG. 1 , the cooling tower efficiency is calculated by using the predicted characteristic curve and the original cooling tower characteristic curve. In this embodiment, the predicted characteristic curve (dotted line of FIG. 4) intersects with a state curve (dashed line of FIG. 4), and the intersection point is a test point of the cooling water tower, and the test point corresponds to a liquid under a test condition. Gas ratio (L/G) _test . In addition, the original cooling tower characteristic curve (solid line in Fig. 4) intersects the state curve, and the intersection point is the design point of the cooling water tower, and the design point corresponds to the liquid-gas ratio (L/G) _design under a design condition. Thereafter, the liquid-gas ratio (L / G) _design calculation of the cooling tower efficiency (L / G) _test conditions and the design of the liquid-gas ratio under the test conditions. In the present embodiment, the cooling tower efficiency is equal to the ratio of liquid to gas test conditions (L / G) _test liquid to gas ratio with the design conditions of (L / G) _design ratio × 100%, the following formula .

Cooling tower efficiency (%) = [(L / G) _test / (L / G) _design ] × 100%

The efficiency of the cooling tower after calculation in this embodiment is about 46.7%, which is obviously different from the calculation result of the American Cooling Water Tower Association (CTI) performance evaluation method (76.3%), and the working state is judged according to the damage of the on-site cooling water tower. It is far from the original design state, so the cooling tower efficiency cannot exceed 70%. In other words, the CTI performance evaluation method overestimates the efficiency of the cooling tower, and the online performance evaluation method of the present invention is objective and practical.

The invention can make a correct and objective performance evaluation of the cooling water tower without affecting the normal operation of the cooling water tower, and the implementation of the invention does not increase the operating cost of the enterprise, and therefore is very suitable for use in an actual factory. The cooling tower is evaluated for online performance.

The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the scope of the present invention. The scope of the invention should be as set forth in the appended claims.

Claims (6)

An online performance evaluation method for a cooling tower comprises the steps of: (a) calculating a fan outlet air volume; and (b) establishing a cooling water tower outlet water temperature model by a multi-model Local Model Network (LMN) method; a predictive characteristic curve is established according to the cooling water tower outlet water temperature model; and (d) the cooling water tower efficiency is calculated by using the predicted characteristic curve and the original cooling tower characteristic curve, and the predicted characteristic curve intersects with a state curve, and the intersection point is The test point of the cooling tower corresponding to the ratio of liquid to gas under a test condition. The online performance evaluation method of the cooling tower of claim 1, wherein the parameter (a) calculating the fan outlet air volume comprises a cooling water flow rate, a cooling water inlet and outlet temperature difference, a cooling water inlet and outlet thermal enthalpy difference, and a cooling water inlet and outlet humidity difference. The online performance evaluation method of the cooling tower of claim 1, wherein the step (a) defines the average air volume of the three months as the air volume of the fan at a high and a low speed. The online performance evaluation method for the cooling tower of claim 1, wherein the multi-model partial model network method of step (b) comprises the following steps: (b1) establishing blur according to inlet water temperature, air temperature, air humidity, and fan power consumption The c-mean (FCM) grouping; (b2) establish a satisfactory fuzzy c-mean grouping according to the inlet water temperature, air temperature, air humidity and fan power consumption; (b3) establish a linear model in each group; (b4) using the distance of the data from each cluster center to give the proportion of the global model; and (b5) Establish a multi-model weighted global cooling water tower outlet water temperature prediction function. The online performance evaluation method of the cooling tower of claim 1, wherein the original cooling tower characteristic curve of the step (d) intersects the state curve, and the intersection point is a design point of the cooling water tower, and the design point corresponds to a design condition. Liquid to gas ratio. The method for evaluating the performance of the cooling tower of claim 5, wherein the step (d) further comprises calculating the cooling tower efficiency by using the ratio of liquid to gas under the test condition and the ratio of liquid to gas under the design condition, the cooling tower efficiency is equal to the test The ratio of the liquid-gas ratio under the conditions to the liquid-gas ratio under the design conditions is ×100%.