A method and system for monitoring the thermal balance state of an aluminum electrolytic cell

Abstract

The present invention provides a method and system for monitoring the thermal balance state of an aluminum electrolytic cell. Compared with the prior art, by establishing a mapping between the aluminum electrolytic cell and the visual virtual cell, the real-time monitoring data of the aluminum electrolytic cell are displayed in real time on the visual virtual cell, and the collected characteristic parameters are deeply mined and deeply learned, which not only realizes the timely and accurate judgment of the thermal balance state of the aluminum electrolytic cell, but also helps the staff to more accurately and timely judge the production safety.

Description

Technical Field

[0001] The present invention relates to the technical field of aluminum electrolysis, and specifically relates to a method and system for monitoring the thermal balance state of an aluminum electrolysis cell. Background Art

[0002] An aluminum electrolysis cell is an important production equipment in a molten salt electrolysis process. During the operation of the aluminum electrolysis cell, the materials in the cell will undergo intense electrochemical reactions under the action of strong direct current and release a large amount of Joule heat. After continuously absorbing the Joule heat, the materials in the cell will gradually form a high-temperature molten salt system. Due to the huge temperature difference between the molten body in the cell and the external environment, the electrolysis cell will continuously dissipate heat to the outside. And the molten salt system used in industrial aluminum electrolysis cells needs to be at a temperature of about 950 °C to carry out the electrolysis process. Therefore, in order to stabilize the electrolysis temperature required for the molten salt system, a reasonable heat insulation structure is provided inside the electrolysis cell. When the internal heat insulation structure is reasonable, the electrolysis cell will be in a thermal balance state during operation, that is, the heat income supplied to the aluminum electrolysis cell by the conversion of electrical energy is equal to the heat expenditure, which is the sum of the heat consumed in the electrolysis process and the heat dissipated from the cell body to the environment. It should be noted that the actual aluminum electrolysis cell cannot always be in an ideal thermal balance state, but from the perspective of long-term stable operation, the thermal field of the electrolysis cell can reach a dynamic thermal balance.

[0003] Due to the complex conditions of the aluminum electrolysis cell and the limitations of the current technical level, in the actual production process of electrolytic aluminum, it is very difficult to control the aluminum electrolysis cell in a dynamic thermal balance, and situations such as red furnaces and leakage of the cell often occur, which will not only cause huge economic losses to the enterprise, but may also trigger serious accidents such as fires and explosions, and even cause casualties to on-site workers. Therefore, in order to avoid such accidents, it is necessary to monitor the thermal balance state of the electrolysis cell.

[0004] Traditional monitoring means for aluminum electrolysis cells, limited by factors such as the complex working conditions and low automation level of aluminum electrolysis cells, mainly rely on the experience of workers for decision-making and cannot judge the thermal balance state of aluminum electrolysis cells in a timely and accurate manner. Summary of the Invention

[0005] The invention object of this application is to provide a method and system for monitoring the thermal balance state of an aluminum electrolysis cell to solve the problem that the existing monitoring means for aluminum electrolysis cells cannot judge the thermal balance state of aluminum electrolysis cells in a timely and accurate manner.

[0006] To achieve the above object, the embodiments of this application adopt the following technical solutions.

[0007] On the one hand, a method for monitoring the thermal balance state of an aluminum electrolysis cell is provided, including the following steps:

[0008] S1. According to the design parameters of the aluminum electrolysis cell, use Unity to construct a visual virtual cell of the aluminum electrolysis cell;

[0009] S2. Register the visualized virtual cell of the aluminum electrolysis cell constructed to HoloLens, and use the depth camera built in HoloLens to obtain the fire-eye image data of the aluminum electrolysis cell; use the temperature sensor to obtain the thermal field data of the aluminum electrolysis cell, and use the current sensor and voltage sensor to obtain the electric field data of the aluminum electrolysis cell;

[0010] S3. Divide the fire-eye image data into a first training set, a first validation set, and a first test set, use the first training set to train the ConvLSTM classifier, use the first validation set to validate the ConvLSTM classifier, and obtain the trained ConvLSTM classifier; divide the thermal field data into a second training set, a second validation set, and a second test set, use the second training set to train the first CNN classifier, use the second validation set to validate the first CNN classifier, and obtain the trained first CNN classifier; divide the thermal field data into a third training set, a third validation set, and a third test set, use the third training set to train the second CNN classifier, use the third validation set to validate the second CNN classifier, and obtain the trained second CNN classifier;

[0011] S4. Use the first test set as the input of the trained ConvLSTM classifier to obtain a first classification result; use the second test set as the input of the trained first CNN classifier to obtain a second classification result; use the third test set as the input of the trained second CNN classifier to obtain a third classification result;

[0012] S5. Use the first classification result, the second classification result, and the third classification result as the input of the Softmax function to obtain the membership degrees of the electrolysis cell belonging to each thermal equilibrium state;

[0013] S6. Process the membership degrees using the evidence theory to obtain the confidence degrees of the electrolysis cell belonging to each thermal equilibrium state, and the thermal equilibrium state corresponding to the highest confidence degree is the final classification result of the thermal equilibrium state;

[0014] S7. Obtain the cell data of the aluminum electrolysis cell; establish a mapping between the aluminum electrolysis cell and the visualized virtual cell, and display the final classification result of the thermal equilibrium state, the thermal field data, the electric field data, and the cell data of the aluminum electrolysis cell on the visualized virtual cell in real time; the cell data includes the cell number of the aluminum electrolysis cell, the cell physical model and structure data, the molecular ratio, the iron-silicon ratio, and the aluminum level; the visualized virtual cell is presented through HoloLens.

[0015] By establishing the mapping between the aluminum electrolysis cell and the visual virtual cell, the real-time monitoring data of the aluminum electrolysis cell are displayed in real time on the visual virtual cell, and the collected characteristic parameters are deeply mined and deeply learned, which not only realizes the timely and accurate judgment of the thermal balance state of the aluminum electrolysis cell, but also helps the staff to judge the production safety more accurately and timely.

[0016] In some embodiments, the definition of the ConvLSTM classifier is as follows:

[0017]

[0018]

[0019]

[0020]

[0021]

[0022] In the formula, σ is the gate, * represents the convolution operator, represents the Hadamard operation, tanh is the hyperbolic tangent function, i t is the input gate at time t, f t is the forget gate at time t, is the cell state at time t, o t is the output gate at time t, H t is the hidden state at time t, is the input value at time t, W xi is the weight of the input gate at of, represents the input, that is, the Huoyan image data, W hi is the weight of the input gate in H t-1 state, W ci is the weight of the input gate at state, b i is the bias value of the input gate, W xf is the weight of the forget gate at of, W hf is the weight of the forget gate in H t-1 state, W cf is the weight of the forget gate at state, b f is the bias value of the forget gate, W xc is the weight of the cell state at time t of, W hc is the weight of the cell state in H t-1 state, b c is the bias value of the cell state at time t, Wxo is the weight of the output gate at , W ho is the weight of the output gate at the H t-1 state, W co is the weight of the output gate at state, b o is the bias value of the output gate.

[0023] In some embodiments, the first CNN classifier and the second CNN classifier are defined as follows:

[0024]

[0025] x l = σ(W l *x l-1 +b l ), l = 1, 2,, L-1,

[0026] In the formula, x0 is the input vector of the first CNN classifier or the second CNN classifier, and x l represents the output of the l-th layer of the first CNN classifier or the second CNN classifier, b l represents the bias of the l-th layer of the first CNN classifier or the second CNN classifier, and W l is the weight of the l-th layer of the first CNN classifier or the second CNN classifier, is the size of the k-th dimension of the input vector, k = 1, 2,..., d; L is the number of layers of the first CNN classifier or the second CNN classifier, and x l-1 is the output of the (l-1)-th layer of the first CNN classifier or the second CNN classifier.

[0027] In some embodiments, the confidence of the electrolytic cell belonging to each thermal equilibrium state is obtained according to the following formula:

[0028]

[0029] In the formula, α is the membership degree of the output of the n-th classifier belonging to the i-th category, N is the total number of classifiers, and Bel(A) i is the confidence of the electrolytic cell A belonging to the i-th category, i = 1, 2, 3, and the operator represents the product.

[0030] In some embodiments, the following steps are included: preprocessing the Huoyan image data to obtain the temporal feature data of the Huoyan image data, and using the temporal feature data of the Huoyan image data as the input of the ConvLSTM classifier for training, validating, and testing the ConvLSTM classifier; the preprocessing includes image denoising processing, image segmentation processing based on the color gamut range, opening and closing operation processing, and feature extraction processing; the temporal feature data includes pixel values, edges, and colors.

[0031] In some embodiments, in S7, the thermal field data and the electric field data are displayed on the visualization virtual cell in the form of a cloud map.

[0032] By presenting the thermal field data and the electric field data in the form of a cloud map to the staff more intuitively, the staff can judge the production safety more accurately and timely.

[0033] On the other hand, a monitoring system for the thermal balance state of an aluminum electrolysis cell is provided, including a camera for acquiring the Huoyan image data of the aluminum electrolysis cell, a temperature sensor for acquiring the thermal field data of the aluminum electrolysis cell, a current sensor and a voltage sensor for acquiring the electric field data of the aluminum electrolysis cell, a HoloLens device, a memory, a processor, and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement the steps in the above-mentioned method for monitoring the thermal balance state of an aluminum electrolysis cell.

[0034] Compared with the prior art, the present invention has at least the following technical effects or advantages: by establishing a mapping between the aluminum electrolysis cell and the visualization virtual cell, the real-time monitoring data of the aluminum electrolysis cell are displayed on the visualization virtual cell in real time, and the collected characteristic parameters are deeply mined and deeply learned, not only realizing the timely and accurate judgment of the thermal balance state of the aluminum electrolysis cell, but also helping the staff to judge the production safety more accurately and timely. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic flow chart of a method for monitoring the thermal balance state of an aluminum electrolysis cell in an embodiment of the present application;

[0036] Figure 2 It is a schematic diagram of the effect of the temperature cloud map and the electric field cloud map of the aluminum electrolysis cell in an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] In order to better understand the above technical solutions, the above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and specific embodiments.

[0038] Embodiment 1

[0039] See Figure 1, a method for monitoring the thermal equilibrium state of an aluminum electrolysis cell, comprising the following steps:

[0040] S1. Based on the design parameters of the aluminum electrolytic cell, Unity is used to build a visual virtual cell of the aluminum electrolytic cell;

[0041] S2. Register the constructed visualized virtual tank of the aluminum electrolysis cell to HoloLens, and use the depth camera of HoloLens to obtain the fire eye image data of the aluminum electrolysis cell; use the temperature sensor to obtain the thermal field data of the aluminum electrolysis cell, and use the current sensor and voltage sensor to obtain the electric field data of the aluminum electrolysis cell;

[0042] S3, dividing the fire eye image data into a first training set, a first verification set and a first test set, using the first training set to train the ConvLSTM classifier, using the first verification set to verify the ConvLSTM classifier, and obtaining a trained ConvLSTM classifier; dividing the thermal field data into a second training set, a second verification set and a second test set, using the second training set to train the first CNN classifier, using the second verification set to verify the first CNN classifier, and obtaining a trained first CNN classifier; dividing the thermal field data into a third training set, a third verification set and a third test set, using the third training set to train the second CNN classifier, using the third verification set to verify the second CNN classifier, and obtaining a trained second CNN classifier;

[0043] S4, using the first test set as the input of the trained ConvLSTM classifier to obtain a first classification result; using the second test set as the input of the trained first CNN classifier to obtain a second classification result; using the third test set as the input of the trained second CNN classifier to obtain a third classification result;

[0044] S5. Using the first classification result, the second classification result and the third classification result as inputs of the Softmax function to obtain the degree of membership of the electrolytic cell to each thermal equilibrium state;

[0045] S6. Processing the membership degree using evidence theory to obtain the confidence that the electrolytic cell belongs to each thermal equilibrium state, and the thermal equilibrium state corresponding to the highest confidence is the final classification result of the thermal equilibrium state;

[0046] S7. Acquire the cell data of the aluminum electrolysis cell; establish a mapping between the aluminum electrolysis cell and the visualized virtual cell, and display the final classification result of the thermal equilibrium state of the aluminum electrolysis cell, thermal field data, electric field data and cell data in real time on the visualized virtual cell; the cell data includes the cell number of the aluminum electrolysis cell, the cell physical model and structural data, molecular ratio, iron-silicon ratio and aluminum level; the visualized virtual cell is presented through HoloLens.

[0047] Among them, the thermal field data and the electric field data are displayed on the visualization virtual slot in the form of a cloud map.

[0048] The Huoyan image data can be preprocessed first (including image denoising, image segmentation based on color gamut range, opening and closing operations, and feature extraction), and then the temporal feature data of the Huoyan image data obtained (including pixel values, edges, and colors) is used as the input of the ConvLSTM classifier.

[0049] The definition of the ConvLSTM classifier is as follows:

[0050]

[0051]

[0052]

[0053]

[0054]

[0055] In the formula, σ is the gate, * represents the convolution operator, represents the Hadamard operation, tanh is the hyperbolic tangent function, i t is the input gate at time t, f t is the forget gate at time t, is the cell state at time t, o t is the output gate at time t, H t is the hidden state at time t, is the input value at time t, W xi is the weight of the input gate at , represents the input, that is, the Huoyan image data, W hi is the weight of the input gate in H t-1 state, W ci is the weight of the input gate at state, b i is the bias value of the input gate, W xf is the weight of the forget gate at , W hf is the weight of the forget gate in H t-1 state, W cf is the weight of the forget gate at state, b f is the bias value of the forget gate, W xc is the weight of the cell state at time t , W hc is the weight of the cell state in Ht-1 Weight of the state, b c Is the bias value of the cell state at time t, W xo Is the weight of the output gate at The weight, W ho Is the weight of the output gate at H t-1 The weight of the state, W co Is the weight of the output gate at The weight of the state, b o Is the bias value of the output gate.

[0056] The definitions of the first CNN classifier and the second CNN classifier are as follows:

[0057]

[0058] x l = σ(W l * x l-1 + b l ), l = 1, 2,, L-1,

[0059] In the formula, x0 is the input vector of the first CNN classifier or the second CNN classifier, and x l Represents the output of the l-th layer of the first CNN classifier or the second CNN classifier, and b l Represents the bias of the l-th layer of the first CNN classifier or the second CNN classifier, and W l Is the weight of the l-th layer of the first CNN classifier or the second CNN classifier, Is the size of the k-th dimension of the input vector, k = 1, 2,..., d; L is the number of layers of the first CNN classifier or the second CNN classifier, and x l-1 Is the output of the (l-1)-th layer of the first CNN classifier or the second CNN classifier.

[0060] The definition of the Softmax function is as follows:

[0061]

[0062] Where z i Is the probability of the i-th class output by the last layer of the classifier (including the ConvLSTM classifier, the first CNN classifier, and the second CNN classifier), I is the number of classes, and α(z i ) is the output membership degree.

[0063] The confidence that the electrolytic cell belongs to each thermal equilibrium state is obtained according to the following formula:

[0064]

[0065] In the formula, α is the membership degree that the output of the n-th classifier belongs to the i-th category, N is the total number of classifiers, and Bel(A)i The confidence level that the electrolytic cell A belongs to category i, where i = 1, 2, 3, and the operation symbol represents the product multiplication.

[0066] In this embodiment, when i = 1, the corresponding electrolytic cell category is a cold cell; when i = 2, the corresponding electrolytic cell category is a normal cell; when i = 3, the corresponding electrolytic cell category is a normal cell.

[0067] Embodiment 2

[0068] The present invention also provides a monitoring system for the thermal balance state of an aluminum electrolytic cell, including a camera (preferably a depth camera) for acquiring the flare image data of the aluminum electrolytic cell, a temperature sensor for acquiring the thermal field data of the aluminum electrolytic cell (i.e., the temperature of the aluminum electrolytic cell), a current sensor and a voltage sensor for acquiring the electric field data of the aluminum electrolytic cell (i.e., the current and voltage of the aluminum electrolytic cell), a HoloLens device, a memory, a processor, and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement the steps in a method for monitoring the thermal balance state of an aluminum electrolytic cell in Embodiment 1.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the thermal balance state of an aluminum electrolysis cell, comprising the following steps: S1. According to the design parameters of the aluminum electrolysis cell, use Unity to construct a visual virtual cell of the aluminum electrolysis cell; S2. Register the constructed visual virtual cell of the aluminum electrolysis cell to HoloLens, and use the depth camera built in HoloLens to obtain the fire-eye image data of the aluminum electrolysis cell; use a temperature sensor to obtain the thermal field data of the aluminum electrolysis cell, and use a current sensor and a voltage sensor to obtain the electric field data of the aluminum electrolysis cell; S3. Divide the Fire Eye image data into a first training set, a first validation set, and a first test set. Use the first training set to train the ConvLSTM classifier, and use the first validation set to validate the ConvLSTM classifier to obtain a trained ConvLSTM classifier. Divide the thermal field data into a second training set, a second validation set, and a second test set. Use the second training set to train the first CNN classifier, and use the second validation set to validate the first CNN classifier to obtain a trained first CNN classifier. Divide the electric field data into a third training set, a third validation set, and a third test set. Use the third training set to train the second CNN classifier, and use the third validation set to validate the second CNN classifier to obtain a trained second CNN classifier. S4. Use the first test set as the input of the trained ConvLSTM classifier to obtain a first classification result. Use the second test set as the input of the trained first CNN classifier to obtain a second classification result. Use the third test set as the input of the trained second CNN classifier to obtain a third classification result. S5. Use the first classification result, the second classification result, and the third classification result as the input of the Softmax function to obtain the membership degrees of the electrolytic cell belonging to each thermal equilibrium state. S6. Use the evidence theory to process the membership degrees to obtain the confidence degrees of the electrolytic cell belonging to each thermal equilibrium state. The thermal equilibrium state corresponding to the highest confidence degree is the final classification result of the thermal equilibrium state. S7. Obtain the cell data of the aluminum electrolytic cell. Establish a mapping between the aluminum electrolytic cell and the visualized virtual cell, and display the final classification result of the thermal equilibrium state, the thermal field data, the electric field data, and the cell data of the aluminum electrolytic cell on the visualized virtual cell in real time. The cell data includes the cell number of the aluminum electrolytic cell, the cell physical model and structure data, the molecular ratio, the iron-silicon ratio, and the aluminum level. The visualized virtual cell is presented through HoloLens.

2. The method for monitoring the thermal balance state of an aluminum electrolysis cell according to claim 1, characterized in that: The definition of the ConvLSTM classifier is as follows: ; ； ； ； ； Wherein, is the gate, represents the convolution operator, represents the Hadamard operation, is the hyperbolic tangent function, is the input gate at time is the forget gate at time is the cell state at time is the output gate at time is the hidden state at time is the input value at time is the weight of the input gate at the weight of the input gate in represents the input, i.e., the Huoyan image data, is the weight of the input gate in the state, is the weight of the input gate at the weight of the input gate in the is the bias value of the input gate, is the weight of the forget gate at the weight of the forget gate in is the weight of the forget gate at the weight of the forget gate in the is the weight of the forget gate at the weight of the forget gate in the is the bias value of the forget gate, is the weight of the cell state at when the weight of the cell state in is the weight of the cell state at the weight of the cell state in the is the weight of the cell state at the bias value of the cell state at the time, is the weight of the output gate at the weight of the output gate in is the weight of the output gate at the weight of the output gate in the is the weight of the output gate at the weight of the output gate in the is the bias value of the output gate.

3. The method for monitoring the thermal balance state of an aluminum electrolysis cell according to claim 1 or 2, characterized in that: The definitions of the first CNN classifier and the second CNN classifier are as follows: ， ， In the formula, is the input vector of the first CNN classifier or the second CNN classifier, represents the output of the th layer of the first CNN classifier or the second CNN classifier, represents the bias of the th layer of the first CNN classifier or the second CNN classifier, is the weight of the th layer of the first CNN classifier or the second CNN classifier, is the size of the k-th dimension of the input vector, k = 1, 2, …, d; L is the number of layers of the first CNN classifier or the second CNN classifier, is the output of the th layer of the first CNN classifier or the second CNN classifier.

4. The method for monitoring the thermal balance state of an aluminum electrolysis cell according to claim 1 or 2, characterized in that: The definition of the Softmax function is as follows: ； Among them is the probability of the output of the last layer of the classifier for the th class, is the number of classes, is the output membership degree.

5. The method for monitoring the thermal balance state of an aluminum electrolysis cell according to claim 1 or 2, characterized in that: The confidence degrees of the electrolytic cell belonging to each thermal equilibrium state are obtained according to the following formula: ； In the formula, is the membership degree that the output of the nth classifier belongs to category, N is the total number of classifiers, is the confidence degree that the electrolytic cell A belongs to category, n = 1, 2, 3, and the operator represents the product.

6. The method for monitoring the thermal balance state of an aluminum electrolysis cell according to claim 1 or 2, characterized in that, It includes the following steps: Preprocess the Fire Eye image data to obtain the time series feature data of the Fire Eye image data, and use the time series feature data of the Fire Eye image data as the input of the ConvLSTM classifier to train, validate, and test the ConvLSTM classifier. The preprocessing includes image denoising processing, image segmentation processing based on the color gamut range, opening and closing operation processing, and feature extraction processing. The time series feature data includes pixel values, edges, and colors.

7. The method for monitoring the thermal balance state of an aluminum electrolytic cell according to claim 1 or 2, characterized in that, In S7, the thermal field data and the electric field data are displayed on the visualized virtual cell in the form of cloud maps.

8. An aluminum electrolytic cell thermal balance state monitoring system, characterized in that: It includes a camera for acquiring the image data of the hot eye of the aluminum electrolytic cell, a temperature sensor for acquiring the thermal field data of the aluminum electrolytic cell, a current sensor and a voltage sensor for acquiring the electric field data of the aluminum electrolytic cell, a HoloLens device, a memory, a processor, and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement the steps in the method according to any one of claims 1-7.

Images