A prediction method for the number of air cooler fans put into operation based on BP neural network

Abstract

The present invention discloses a method for predicting the number of air cooler fans put into operation based on a BP neural network, which includes data preparation and construction of a data set; dividing the data into a training set and a validation set; using the BP neural network to predict the future inlet valve temperature and outlet valve temperature of the cooling water; and calculating the number of air cooler fans put into operation according to the predicted inlet and outlet valve temperatures of the cooling water in combination with the parameters of the air cooler fans. The present invention solves the problem that when multiple fans fail, the operators cannot determine whether the cooling capacity meets the future heat generation of the valve group, improves the trend perception ability of the operation and maintenance personnel, makes corresponding preparations in time for the future change trends, and has guiding significance for the normal operation and reliable power supply of the power grid. The prediction of the present invention has high credibility, is simple to use, and has a rapid response, making up for the deficiencies of the previous prediction methods.

Description

Technical Field

[0001] The present invention belongs to the field of power systems, and particularly relates to a method for predicting the number of air-cooler fans put into operation. Background Art

[0002] With the improvement of China's economic strength and national living standards, the electricity demand has also increased sharply. Therefore, the UHV DC transmission project has come into being. Among them, the converter valve is the core equipment for realizing the conversion of high-voltage AC and DC electrical energy, and the cooling capacity of the converter valve is an important factor affecting the safe and reliable operation of the HVDC transmission system. During the peak electricity consumption period in summer, the converter valve has a large transmission power and a high outdoor ambient temperature, which is a test for the valve cooling system. At present, the air-cooler is a commonly used outdoor heat exchange device in the valve cooling system of UHV DC transmission, which usually consists of multiple groups of variable-frequency and power-frequency fans. If multiple fans fail, the operators cannot judge whether the cooling capacity meets the future heat dissipation requirements, so they cannot formulate a plan to reduce power or perform artificial auxiliary cooling in advance.

[0003] At present, the ARIMA model (Autoregressive Integrated Moving Average model) and the SVM model (Support Vector Machine) are mainly used to predict the valve cooling operation data, that is, to predict future data through historical data. Such models have strict mathematical theory support and strong interpretability, but they require a large amount of data, a long training time and low prediction accuracy. Summary of the Invention

[0004] The present invention solves the problem that when multiple groups of fans fail under high temperature or high power conditions, the operators' judgment on the future cooling capacity lags behind, reserves time for formulating a treatment plan in advance, and improves the safety performance of the temperature control of the cooling system.

[0005] A method for predicting the number of air-cooler fans put into operation based on the BP neural network includes the following steps:

[0006] (1) Collect the operation parameters of the converter valve, perform normalization processing, and determine the input and output of the BP neural network;

[0007] (2) Select part of the data in the data after the normalization processing obtained in step (1) as the training set, and the remaining data as the verification set;

[0008] (3) Set the structure of the BP neural network, initialize the hyperparameters, train with the data of the training set, and continuously adjust the hyperparameters to make the loss function of the training set meet the requirements; in order to prevent overfitting, substitute the verification set data for verification. When the loss function of the verification set does not meet the requirements, adjust the hyperparameters again and retrain until the errors of both the training set and the verification set meet the requirements;

[0009] (4) Input the planned data, predict the inlet and outlet valve temperatures of the cooling water through a BP neural network; combine the inherent parameters of the valve cooling system to calculate the predicted power loss value of the converter valve; calculate the heat exchange capacity of the variable-frequency fan and the industrial-frequency fan unit according to the operating parameters of the air-cooled cooler fan, and combine the heat balance principle of the converter valve cooling system to predict the total number of fans to be put into operation.

[0010] Further, in step (1), the operating parameters of the converter valve collected include but are not limited to the valve group power, ambient temperature, inlet valve temperature of the cooling water, and outlet valve temperature of the cooling water, where the inlet valve temperature of the cooling water and the outlet valve temperature of the cooling water are used as the output data of the BP neural network, and the rest are used as the input data of the BP neural network.

[0011] Further, in step (1), the normalization methods include linear normalization and / or Z-score standard deviation normalization:

[0012] Linear normalization:

[0013]

[0014] In the above formula, max and min are the maximum and minimum values of the given scaling range, X max , X min are the maximum and minimum values of the data respectively, and X is the value of each piece of data;

[0015] Z-score standard deviation normalization:

[0016]

[0017] In the above formula, μ is the average value, σ is the standard deviation, and X is the value of each piece of data.

[0018] Further, in step (3), the loss function includes one of the mean square error, absolute value variance, mean absolute percentage error, and root mean square error.

[0019] Further, in step (3), the structure of the BP neural network includes the number of input layer units, the number of output layer units, the number of hidden layer neurons, and the number of hidden layers; the hyperparameters include the connection weights and biases between the input layer and the hidden layer, between the hidden layers, and between the hidden layer and the output layer, the learning rate, the activation function, and the number of training times.

[0020] Further, the setting method of the learning rate is a fixed learning rate or an adaptive learning rate; the activation functions include the Sigmoid function, the Relu function, and the Tanh function.

[0021] Further, in step (4), the planned data includes the planned power curve data issued by the dispatching and the temperature prediction curve data of the weather forecasting system.

[0022] Further, in step (4), the operating parameters of the air cooler fan include one or more of the following parameters:

[0023] Specific heat capacity of air; heat transfer area; heat transfer efficiency of the power frequency fan unit; heat transfer air volume of the power frequency fan unit; heat transfer coefficient and air volume of the variable frequency fan unit at the highest operating frequency.

[0024] Further, in step (4), the heat transfer amount of the fan unit is calculated by the effectiveness-heat transfer unit method.

[0025] Further, the air cooler is composed of multiple air coolers with known heat capacities connected in parallel, and variable frequency drives are configured for some of the fans of each air cooler with a known heat capacity;

[0026] The variable frequency fan unit includes: a group of variable frequency fans, composed of two variable frequency fans of each air cooler, and / or a large group of variable frequency fans, composed of multiple groups of the group of variable frequency fans at the same position;

[0027] The power frequency fan unit includes: a group of power frequency fans, composed of two power frequency fans of each air cooler, and / or a large group of power frequency fans, composed of multiple groups of the group of power frequency fans at the same position.

[0028] Further, the heat balance principle of the converter valve cooling system is that the heat generation amount is equal to the heat dissipation amount, that is, the power loss value of the converter valve is approximately equal to the heat dissipation amount of the air cooler fan.

[0029] Further, the method for predicting the total number of fans to be put into operation:

[0030] When the ambient temperature is relatively high, under the principle of preferentially starting the variable frequency fans, it is considered that the variable frequency fans are fully opened and operating at the highest set frequency; in the case where the number of groups N 1max of the variable frequency fan unit has been determined, the number of groups N2 of the power frequency fan unit is determined according to the heat balance principle of the converter valve cooling system, and the predicted number of fans to be put into operation is N 1max + N2.

[0031] After adopting the above solution, the present invention has the following beneficial effects:

[0032] 1. Compared with the ARIMA model and the SVM model that predict future data through a large amount of historical data, the present invention combines the inherent parameters of valve cooling operation with the planned power of the valve group and weather forecast data, has high prediction credibility, is simple to use, has a rapid response, and has high prediction accuracy, which is of great significance for the monitoring of valve cooling temperature;

[0033] 2. By predicting the temperatures of the cooling water inlet and outlet valves and the operating parameters of the air cooler fans, the total number of fans to be put into operation can be calculated, which improves the awareness of the maintenance personnel of the cooling capacity and enables them to make corresponding preparations in a timely manner for the future change trends, and has guiding significance for the normal operation and reliable power supply of the power grid. Description of the Drawings

[0034] Figure 1 It is a flowchart of the method for predicting the number of air cooler fans put into operation according to an embodiment of the present invention;

[0035] Figure 2 It is a schematic diagram of the structural principle of the BP neural network according to an embodiment of the present invention;

[0036] Figure 3 It is a schematic diagram of the operating parameters of the air cooler according to an embodiment of the present invention;

[0037] Figure 4 It is a schematic diagram of the fan unit of the air cooler according to an embodiment of the present invention;

[0038] Figure 5 It is a diagram of the prediction result of the number of air cooler fans put into operation according to an embodiment of the present invention. Detailed Embodiments

[0039] In the following, only some exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the present application. Therefore, the drawings and the description are considered to be exemplary in nature rather than restrictive.

[0040] In the description of the present application, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present application, "a plurality" means two or more unless otherwise specifically defined.

[0041] The following disclosure provides many different embodiments or examples for implementing different structures of the present application. To simplify the disclosure of the present application, the components and settings of specific examples are described below. Of course, they are only examples and are not intended to limit the present application. In addition, the present application may repeat reference numerals and / or reference letters in different examples. Such repetition is for the purpose of simplification and clarity and does not itself indicate the relationship between the various embodiments and / or settings discussed. In addition, the present application provides examples of various specific processes and materials, but those of ordinary skill in the art can be aware of the application of other processes and / or the use of other materials.

[0042] The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, numerous specific details are provided to give a thorough understanding of the embodiments of the present application. However, those skilled in the art will realize that the technical solutions of the present application may be practiced without one or more of these specific details, or other methods, components, materials, devices, or steps, etc. may be adopted. In these cases, well-known structures, methods, devices, implementations, materials, or operations will not be shown or described in detail.

[0043] The block diagrams shown in the drawings do not necessarily have to correspond to physically independent entities. These functional entities or parts of functional entities may be implemented using software, or in one or more hardware modules and / or programmable modules, or in different networks and / or processor devices and / or microcontroller devices.

[0044] Figure 1 It is a flowchart of the control method for the total number of air-cooling fans of the converter valve cooling system. Refer to Figure 1 According to the exemplary embodiment, in S101, the operating parameters of the converter valve are collected. The converter valves to which the method for the total number of air-cooling fans provided in the present application can be applied include line-commutated converters or voltage-source converters. Among them, collecting the operating parameters of the converter valve includes the inlet valve temperature T of the cooling water in , the outlet valve temperature T out , the valve group power P, and the ambient temperature T e , where the valve group power and the ambient temperature are input data, and the inlet valve temperature and the outlet valve temperature of the cooling water are output data.

[0045] According to the exemplary embodiment, in S102, the collected data is linearly normalized, and the calculation formula is as follows:

[0046]

[0047] In the formula, max and min are respectively the maximum and minimum values of the given scaling range. Here, the scaling interval is (-1, 1), X max and X min are respectively the maximum and minimum values of their respective data sets, X is each data value, and a total of 965 groups of data are collected.

[0048] According to the exemplary embodiment, in S103, a suitable BP neural network is constructed. In this example, a shallow neural network is adopted, and the number of hidden layers is 1. Select a suitable number of neurons in the hidden layer, and the empirical calculation formula is as follows:

[0049]

[0050] Wherein, q is the number of neurons in the hidden layer; m is the number of neurons in the input layer, which is 2 in this embodiment; n is the number of neurons in the output layer, which is 2 in this embodiment; the value range of a is 1 to 10.

[0051] According to the exemplary embodiment, in S104, the data is trained. Refer to Figure 2 , the learning process of the training samples can be divided into forward propagation and backward propagation.

[0052] In the forward propagation process, first, the neurons in the input layer receive the original input signal, and then it is propagated forward to the neurons in the hidden layer. The output of the hidden layer is:

[0053]

[0054] W ji represents the connection weight between the j-th neuron in the hidden layer and the i-th neuron in the input layer, X i is the input signal, θ j is the bias of the j-th neuron in the hidden layer, O j is the output of the j-th neuron in the hidden layer, and f represents the activation function of the neurons in the hidden layer. In this example, the relu function is adopted.

[0055] The neurons in the output layer receive the input from the neurons in the hidden layer and calculate the final

[0056] output of the entire network. Similarly, the formula is as follows:

[0057]

[0058] W kj represents the connection weight between the k-th neuron in the output layer and the j-th neuron in the hidden layer, θ k is the bias of the k-th neuron in the output layer, Z k is the final output of the k-th neuron in the output layer, and g is the activation function of the output layer. In this example, the sigmoid function is adopted.

[0059] In backward propagation, an error transfer function is first established. In this example, the MSE is used as the loss function:

[0060]

[0061] In the formula, Y k represents the expected output, that is, the true output value in the training set.

[0062] The error transfer function is differentiated with respect to the weight coefficient to determine the weight value for the next iteration correction:

[0063]

[0064]

[0065] In the formula, α is the learning rate, and ω kj (t) and ω ji (t) are the connection weight coefficients between the hidden layer and the output layer, and between the input layer and the output layer this time. ω kj (t + 1) and ω ji (t + 1) are the weight coefficients updated in the next iteration. In this example, the learning rate α is selected as 0.005, and the maximum number of training times is set to 30,000 times.

[0066] According to the exemplary embodiment, in S105, the obtained weight coefficients and biases are substituted into the validation set for verification, and the error e is calculated. After subsequent adjustment of hyperparameter training and verification, when the number of hidden layer neurons q = 5, the error is the smallest and the prediction effect is the best.

[0067] According to the exemplary embodiment, in S106, the weight coefficients and bias terms when the number of hidden layer neurons q = 5 are used as the final model. In S201, the planned power data and weather forecast data are substituted into the final model, and the predicted values of the cooling water inlet valve temperature and outlet valve temperature are obtained after anti-normalization.

[0068] In high-temperature or high-power working conditions, the variable-frequency fan starts preferentially. It can be approximately considered that all variable-frequency fans are put into operation and operate at the highest frequency. In S202, according to the predicted cooling water inlet and outlet valve temperatures, combined with the inherent parameters, the power loss value Pv of the converter valve is calculated, and the maximum heat transfer amount P 1max of the variable-frequency fan unit and the heat transfer amount P2 of the industrial-frequency fan unit are calculated. The inherent parameters include both the inherent parameters of the converter valve and the inherent parameters of the air cooler. Specifically, the operating parameters of the converter valve include the cooling water flow rate Q when the three-way valve is fully open, the specific heat capacity C water of the cooling water, and the density ρ of the cooling water, which are obtained through the converter valve data manual or through experiments. The inherent parameters of the air cooler include the specific heat capacity C air of the air, the heat transfer area A, the heat transfer efficiency ε of the industrial-frequency fan unit, the heat transfer air volume q of the industrial-frequency fan unit, the heat transfer coefficient k max of the variable-frequency fan unit at the highest operating frequency f max and the air volume q max . The calculation formula for the predicted power loss value Pv of the converter valve is as follows:

[0069] P v = C water ρQ(T out - T in )

[0070] In the formula, Q is the cooling water flow rate when the three-way valve is fully open, C water is the specific heat capacity of the cooling water, ρ is the density of the cooling water, and Tin and T out are the inlet valve temperature and outlet valve temperature of the cooling water respectively.

[0071] According to the exemplary embodiments of the present application, the maximum heat transfer capacity P of the variable-frequency fan unit is calculated by the effectiveness-number of transfer units method (ε-NTU), 1max and the heat transfer capacity P2 of the industrial-frequency fan unit.

[0072] Specifically, the heat transfer capacity is calculated according to the effectiveness-number of transfer units (ε-NTU) of the heat exchanger, and the formula is as follows:

[0073] P 1max = ε max q max c(T w_in - T c_in )

[0074] P2 = εqc(T w_in - T c_in )

[0075] where ε max and q max are the heat transfer effectiveness and air volume of the variable-frequency fan unit at the highest operating frequency respectively, ε and q are the heat transfer effectiveness and air volume of the industrial-frequency fan unit respectively, c is the specific heat capacity of air, T w_in is the inlet air temperature of the air cooler, which is approximately replaced by the predicted ambient temperature in the weather forecast, and T c_in is the inlet water temperature of the air cooler, which is approximately replaced by the predicted cooling water outlet valve temperature T out .

[0076] Referring to Figure 1 , in S203, the number of fan units to be put into operation is determined by using the principle of heat balance. According to the exemplary embodiments of the present application, Figure 3 the air cooler shown in Figure 4 is composed of multiple air coolers with known heat capacities connected in parallel, and among them, some of the fans of each air cooler are configured with frequency converters to form variable-frequency fans. The variable-frequency fans include small-group variable-frequency fans and large-group variable-frequency fans. In the fan unit of the air cooler of the converter valve cooling system, 6 groups of air coolers are connected in parallel, each group of air coolers is configured with 16 fans and 2 of the groups are equipped with frequency converters, as shown in 1max . The small-group variable-frequency fans are composed of two variable-frequency fans in each air cooler, and the large-group variable-frequency fans are composed of multiple small-group variable-frequency fans at the same position. In addition, the small-group industrial-frequency fans are composed of two industrial-frequency fans in each air cooler, and the large-group industrial-frequency fans are composed of multiple small-group industrial-frequency fans at the same position. In this embodiment, there are 2 groups of large-group variable-frequency fans, which are composed of 32 variable-frequency fans (N 1max = 32), and there are 4 groups of large-group industrial-frequency fans, which are composed of 64 industrial-frequency fans (N 2max = 64).

[0077] In S203, the number of variable-frequency fan units and industrial-frequency fan units to be put into operation can be determined, and the calculation formula is as follows:

[0078] N 1max P 1max +N2P2≈P v

[0079] Where N 1max is the total number of variable-frequency fan units, N2 is the number of industrial-frequency fan units, and P 1max is the heat exchange capacity of the variable-frequency fan unit at the highest operating frequency f max . Pv is the predicted power consumption value of the converter valve calculated based on the predicted cooling water inlet and outlet valve temperatures. When it is determined that all variable-frequency fan units are put into operation and the number is N 1max , the number of groups of industrial-frequency fan units N2 can be determined by the above calculation formula; finally, the total number N of variable-frequency fan units and industrial-frequency fan units to be put into operation is determined 1max +N2. For the implementation results of this example, see Figure 5 .

[0080] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of operation combinations. However, those skilled in the art should know that this application is not limited by the described operation sequence. According to this application, some steps can be performed in other sequences or simultaneously.

[0081] In the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0082] Through the above detailed description, those skilled in the art can easily understand that the technical solutions according to the embodiments of this application have one or more of the following advantages.

[0083] By using the prediction method for the total number of air-cooled fans of the converter valve cooling system provided in this application, predicting the cooling water inlet and outlet valve temperatures according to weather forecasts and planned power data, calculating the power loss value of the converter valve, and directly obtaining the total number of fans to be put into operation based on the heat exchange capacity of the fan unit using the heat balance principle, it has high real-time performance and outstanding dynamic performance. It improves the perception of the cooling capacity by the operation and maintenance personnel, makes corresponding preparations in a timely manner for future change trends, and has guiding significance for the normal operation of the power grid and providing reliable power supply.

[0084] The foregoing describes exemplary embodiments of the present application, but does not impose any formal restrictions on the present application. These exemplary embodiments are not intended to be exhaustive or to limit the present application to the precise forms disclosed, and it is obvious that, under the inspiration of the foregoing teachings, those of ordinary skill in the art can make many modifications and variations. Therefore, those skilled in the art should understand that the scope of protection involved in the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) disclosed in the present application that have similar functions.

Claims

1. A method for predicting the number of air-cooler fans put into operation based on a BP neural network, characterized in that, It includes the following steps: (1) Collect the operating parameters of the converter valve, perform normalization processing, and determine the input and output of the BP neural network; Among them, the collected operating parameters of the converter valve include but are not limited to valve group power, ambient temperature, cooling water inlet valve temperature, and cooling water outlet valve temperature. Among them, the cooling water inlet valve temperature and the cooling water outlet valve temperature are used as the output data of the BP neural network, and the rest are used as the input data of the BP neural network; (2) Select part of the data in the normalized data obtained in step (1) as the training set, and the remaining data as the validation set; (3) Set the structure of the BP neural network, initialize the hyperparameters, use the data of the training set for training, and continuously adjust the hyperparameters so that the loss function of the training set meets the requirements; to prevent overfitting, substitute the validation set data for verification. When the loss function of the validation set does not meet the requirements, adjust the hyperparameters again and retrain until the errors of both the training set and the validation set meet the requirements; (4) Input the planned data, predict the cooling water inlet and outlet valve temperatures through the BP neural network; combine the inherent parameters of the valve cooling system to calculate the predicted power loss value of the converter valve; according to the operating parameters of the air cooler fan, calculate the heat transfer capacity of the variable-frequency fan and the industrial-frequency fan unit, and combine the heat balance principle of the converter valve cooling system to predict the total number of fans to be put into operation; The planned data includes the planned power curve data issued by the dispatching and the temperature prediction curve data of the weather forecasting system.

2. The method for predicting the number of air-cooler fans put into operation based on the BP neural network according to claim 1, wherein In step (1), the normalization methods include linear normalization and / or Z-score standard deviation normalization: Linear normalization: In the above formula, max and min are the maximum and minimum values of the given scaling range, and X max , X min are respectively the maximum and minimum values of the data, and X is the value of each piece of data; Z-score standard deviation normalization: In the above formula, μ is the average value, σ is the standard deviation, and X is the value of each piece of data.

3. The air cooler fan input quantity prediction method based on BP neural network according to claim 1, characterized in that In step (3), the loss function includes one of mean square error, absolute value variance, mean absolute percentage error, and root mean square error.

4. The method for predicting the number of air cooler fans put into operation based on the BP neural network according to claim 1, characterized in that, In step (3), the structure of the BP neural network includes the number of input layer units, the number of output layer units, the number of neurons in the hidden layer, and the number of hidden layers; the hyperparameters include the connection weights and biases between the input layer and the hidden layer, between the hidden layers, and between the hidden layer and the output layer, the learning rate, the activation function, and the number of training times.

5. The method for predicting the number of air cooler fans put into operation based on the BP neural network according to claim 4, wherein, The setting method of the learning rate is a fixed learning rate or an adaptive learning rate; the activation functions include the Sigmoid function, the Relu function, and the Tanh function.

6. The method for predicting the number of air cooler fans put into operation based on the BP neural network according to claim 1, wherein, In step (4), the operating parameters of the air cooler fan include one or more of the following parameters: Specific heat capacity of air; heat transfer area; heat transfer efficiency of the industrial-frequency fan unit; heat transfer air volume of the industrial-frequency fan unit; heat transfer coefficient and air volume of the variable-frequency fan unit at the highest operating frequency.

7. The method for predicting the number of air-cooler fans put into operation based on the BP neural network according to claim 1, wherein In step (4), the heat transfer capacity of the fan unit is calculated by using the effectiveness-heat transfer unit method.

8. The method for predicting the number of air-cooler fans put into operation based on the BP neural network according to claim 1, characterized in that, The air cooler is composed of multiple air coolers with known heat capacities connected in parallel, and part of the fans of each air cooler with known heat capacity are configured with frequency converters; The variable-frequency fan unit includes: a group of variable-frequency fans, composed of two variable-frequency fans of each air cooler, and / or a large group of variable-frequency fans, composed of multiple groups of the group of variable-frequency fans at the same position; The industrial-frequency fan unit includes: a small-group industrial-frequency fan, which consists of two industrial-frequency fans of each air cooler, and / or a large-group industrial-frequency fan, which consists of multiple groups of the small-group industrial-frequency fans at the same position.

9. The method for predicting the number of air-cooler fans put into operation based on the BP neural network according to claim 1, wherein, The heat balance principle of the converter valve cooling system is that the heat generation amount is equal to the heat dissipation amount, that is, the power loss value of the converter valve is approximately equal to the heat dissipation amount of the air cooler fan.

10. The method for predicting the number of air cooler fans put into operation based on the BP neural network according to claim 9, wherein, Method for predicting the total number of fans to be put into operation: When the ambient temperature is relatively high, under the principle of preferentially starting the variable-frequency fans, it is considered that the variable-frequency fans are fully opened and operating at the highest set frequency; in the case where the number of variable-frequency fan unit groups N 1max has been determined, the number of industrial-frequency fan unit groups N2 is determined according to the heat balance principle of the converter valve cooling system, and the predicted number of fans to be put into operation is N 1max + N2.

Images