An AI collaborative control-based smelting furnace gas and electricity consumption optimization method

Abstract

The present application belongs to the technical field of industrial smelting energy consumption control, and discloses a smelting furnace gas and electricity double consumption optimization method based on AI collaborative control, which is suitable for a smelting furnace-dust collector coupled energy-saving system. The method collects multi-dimensional operation parameters through a sensor, inputs the double consumption collaborative deep learning model after Z-score standardization preprocessing, extracts coupled features by CNN, predicts future parameters by LSTM, and then outputs optimization instructions through reinforcement learning, so as to collaboratively control each unit and maintain the air-fuel ratio in the optimal interval of 1.05-1.15. At the same time, the flue gas waste heat is recovered to preheat combustion-supporting air, and the iterative model is fed back every hour in a closed loop. The method breaks through the barrier of independent control of equipment, saves electricity in the dust collector by more than 30%, reduces the gas consumption of the smelting furnace by 18%-25% per unit of product, and meets the emission standard of nitrogen oxides, taking into account energy saving, environmental protection and production safety, and can be widely applied to various industrial smelting scenes such as copper alloy, aluminum alloy and steel.

Description

Technical Field

[0001] This invention relates to the field of energy consumption control technology in industrial smelting, specifically to an AI-based collaborative control method for optimizing the dual consumption of gas and electricity in smelting furnaces, which is particularly suitable for smelting furnace-dust collector coupled energy-saving systems. Background Technology

[0002] In industries such as metallurgy and machinery manufacturing, smelting furnaces are core production equipment. During operation, they consume large amounts of fuel gas for melting the furnace charge, and the associated dust collectors, which handle smelting flue gas and control dust emissions, also require continuous electricity to drive the fans and perform dust removal operations. Combined fuel gas and electricity consumption accounts for over 65% of the total production cost. Currently, smelting furnaces and dust collectors mostly employ independent control modes. Furnace combustion control focuses only on its own fuel gas consumption and furnace temperature stability, while dust collector operation control adjusts fan frequency solely based on real-time flue gas volume. This lack of coordination and linkage between the two systems results in numerous technical shortcomings.

[0003] On the one hand, the amount of flue gas and dust concentration generated by the smelting furnace combustion dynamically changes with the amount of furnace charge and the fuel gas ratio. However, the dust collector's control is lagging and cannot adapt to changes in flue gas parameters in advance. This can easily lead to problems such as excessively high fan frequency resulting in wasted electricity, or unreasonable dust removal intervals causing filter bag blockage and increased resistance. On the other hand, the combustion control of the smelting furnace neglects the recovery and utilization of waste heat from the flue gas. Combustion air enters the furnace at room temperature, requiring more fuel gas to raise the furnace temperature. At the same time, the air-fuel ratio adjustment lacks precise control, which can easily lead to incomplete combustion or excessive nitrogen oxide emissions. In addition, existing optimization methods mostly use a single model to control a single device, which cannot capture the parameter coupling relationship between the smelting furnace and the dust collector, making it difficult to achieve global optimization of both gas and electricity consumption, resulting in limited energy-saving effects.

[0004] Therefore, there is an urgent need for a method that can achieve coordinated control of smelting furnaces and dust collectors, accurately optimize gas and electricity consumption based on AI technology, and balance energy saving and environmental protection, in order to overcome the shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention To address the problems of existing smelting furnaces and dust collectors operating independently, insufficient gas-electricity synergy optimization, limited energy-saving effects, and easy neglect of environmental protection requirements, this invention provides a gas-electricity dual consumption optimization method for smelting furnaces based on AI collaborative control. By constructing a coupled energy-saving system and a multi-model fusion AI architecture, multi-unit collaborative control and closed-loop iterative optimization are achieved. This significantly reduces gas and electricity consumption while ensuring that nitrogen oxide emissions meet standards, thereby improving production stability and economy.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following solution: A method for optimizing the gas and electricity consumption of a smelting furnace based on AI collaborative control is applied to a coupled energy-saving system of a smelting furnace and a dust collector. The system includes a smelting furnace combustion control unit, a dust collector adaptive operation unit, an AI intelligent decision-making unit, and a flue gas waste heat recovery unit. The method includes the following steps: S1. Multi-dimensional parameter acquisition: Through the sensors that are matched with the combustion control unit of the smelting furnace and the adaptive operation unit of the dust collector, multi-dimensional operating parameters of the smelting furnace and the dust collector are collected in real time, and the multi-dimensional operating parameters are transmitted to the AI ​​intelligent decision unit. S2. Data Preprocessing: The Z-score normalization method is used to normalize the collected multi-dimensional operating parameters. The formula is as follows: Where X is the multi-dimensional operating parameter, μ is the mean of the multi-dimensional operating parameter, and σ is the standard deviation of the multi-dimensional operating parameter. The 3σ criterion is used to remove outliers caused by instantaneous fluctuations of the sensor and construct a continuous time series data sample set. S3. AI Model Prediction and Optimization Decision: The time series data sample set is input into the dual-consumption collaborative deep learning model built into the AI ​​intelligent decision unit. The dual-consumption collaborative deep learning model extracts multi-parameter coupled features through the CNN feature extraction module, and predicts the flue gas volume, dust concentration and gas combustion efficiency in the next 5-10 minutes through the LSTM time series prediction module. Then, the reinforcement learning optimization module outputs collaborative optimization instructions for gas flow rate, combustion air ratio, fan frequency and ash cleaning interval with the goal of "lowest gas and electricity consumption", so that the air-fuel ratio is maintained in the optimal range of 1.05-1.15. S4. Multi-unit collaborative control: Based on the optimization instructions, the system synchronously regulates each functional unit: adjusts the gas ratio regulating valve of the smelting furnace combustion control unit to accurately control the ratio of gas to combustion air; adjusts the fan frequency converter and ash removal controller of the dust collector adaptive operation unit to synchronously drive the AI-adjustable flow guide device to optimize the angle of the flow guide blades; and preheats the combustion air to 80-120℃ through the flue gas waste heat recovery unit, and starts auxiliary electric heating compensation when the temperature is insufficient. S5. Closed-loop feedback iteration: Real-time collection of unit product gas consumption, dust collector power consumption and nitrogen oxide emission data to obtain raw data. Based on the above raw data, the energy-saving effect is calculated. The above raw data and the energy-saving effect calculation results are fed back to the reinforcement learning optimization module. The model parameters are iterated and optimized once per hour to ensure that the overall power saving rate of the dust collector is ≥30% and the unit product gas consumption of the smelting furnace is reduced by 18%-25%.

[0007] Preferably, in step S1, the multi-dimensional operating parameters include the furnace charge input, real-time furnace temperature, furnace pressure, gas flow rate, combustion air flow rate, gas pressure, and gas combustion efficiency on the furnace side, as well as the flue gas volume, dust concentration, dust collector resistance, fan current, dust removal interval time, and filter bag surface temperature on the dust collector side.

[0008] Preferably, in step S2, when constructing a continuous time series data sample set, linear interpolation is used to supplement missing data. The proportion of missing data does not exceed 3% of the total data volume. The time span of the continuous time series data sample set is set to 24 hours, and it is divided into a sub-sample set according to each hour. Each sub-sample set contains 120 data points, which makes it easier for the model to capture the parameter change patterns under different working conditions, improve the model's generalization ability, and adapt to the working condition fluctuations of different production batches.

[0009] Preferably, in step S3, the training process of the dual-consumption collaborative deep learning model includes: using the preprocessed time-series data sample set as the training set, dividing the dataset according to the ratio of 70% of the data for model training, 20% for validation, and 10% for testing; the CNN feature extraction module adopts an alternating structure of 3 convolutional layers and 2 pooling layers, with the convolutional kernel size set to 3×3 and the stride to 1, and the pooling layer adopts max pooling to extract multi-parameter coupled features; the LSTM time-series prediction module sets 4 hidden layers, each with 64 neurons, and uses dropout... The threshold value was set to 0.2 to avoid model overfitting, and the ReLU function was used as the activation function. The reinforcement learning optimization module adopted the PPO algorithm, taking the flue gas volume, dust concentration, gas combustion efficiency, and dust collector resistance as state inputs, and the total gas and electricity consumption per unit product, nitrogen oxide emissions, and filter bag wear rate as reward and penalty factors. Positive rewards were given when the total gas and electricity consumption decreased, nitrogen oxide emissions met the standards, and filter bag wear slowed down, while negative penalties were given. The control strategy was iteratively optimized until the model converged, and the convergence accuracy was controlled within ±2% to ensure that the deviation between the predicted value and the actual value did not exceed the set range.

[0010] Preferably, in step S3, the precise control of the air-fuel ratio is achieved through the linkage adjustment of the gas flow rate and the combustion air flow rate. The AI ​​intelligent decision-making unit collects the flow data of the two in real time, calculates the actual air-fuel ratio, and when it deviates from the range of 1.05-1.15, it prioritizes adjusting the combustion air flow rate to avoid large fluctuations in the gas flow rate affecting the stability of the furnace temperature.

[0011] Preferably, in step S4, the AI-adjustable flow guiding device includes flow guiding blades, a stepper motor, and a position sensor. The position sensor provides real-time feedback of the flow guiding blade angle signal. Based on the fan frequency and the predicted flue gas volume, the blade angle is dynamically adjusted to 15-60° to achieve uniform distribution of flue gas inside the dust collector, reduce local resistance of the dust collector, and reduce ineffective power consumption of the fan. The flue gas waste heat recovery unit adopts a tubular heat exchanger made of high-temperature resistant stainless steel, and the heat exchange area is matched according to the rated flue gas volume of the smelting furnace. The auxiliary electric heating adopts a high-frequency induction heating method, and the heating power can be steplessly adjusted within the range of 0-50kW. It is linked with the flue gas waste heat recovery unit to form a closed-loop temperature control. It automatically starts when the preheated combustion air temperature is below 80°C and automatically shuts down when the temperature reaches 120°C. At the same time, the start-stop frequency and heating duration are recorded as reference parameters for model iteration to avoid energy waste.

[0012] Preferably, in step S5, the energy-saving effect calculation formula is: η=[(Q0-Q1)×k+(W0-W1)] / (Q0×k+W0)×100%, where η is the comprehensive energy-saving rate, Q0 is the unit product gas consumption before optimization, Q1 is the unit product gas consumption after optimization, W0 is the unit time power consumption of the dust collector before optimization, W1 is the unit time power consumption of the dust collector after optimization, and k is the gas-electricity energy consumption conversion coefficient; at the same time, the nitrogen oxide emission concentration needs to be controlled below 50mg / m³. When it exceeds this threshold, the model prioritizes adjusting the air-fuel ratio to the range of 1.08-1.12 and increasing the preheating temperature of the combustion air to 100-120℃. By optimizing the combustion conditions, the generation of nitrogen oxides is reduced, taking into account both energy saving and environmental protection goals.

[0013] Preferably, the combustion control unit of the smelting furnace also includes a flame detector and a gas leak alarm. The flame detector monitors the combustion status in real time. When it detects that the flame is unstable or goes out, the AI ​​intelligent decision unit immediately outputs an emergency control command to cut off the gas supply and increase the combustion air flow to 1.2 times the rated value. At the same time, it links the dust collector adaptive operation unit to increase the fan frequency to 80% of the rated frequency to accelerate the discharge of residual gas in the furnace. When the gas leak alarm detects a leak, it immediately triggers an audible and visual alarm and shuts down the entire gas pipeline to ensure production safety.

[0014] The beneficial effects of this invention are as follows: 1. Strong collaborative optimization capability: Breaking down the barriers of independent control of smelting furnace and dust collector, a coupled energy-saving system is constructed. Through AI model, the coupling characteristics of parameters of multiple devices are accurately captured, and the global optimal control of gas and electricity consumption is achieved. Compared with single device optimization, the overall energy saving effect is improved by 10%-15%, which solves the problem of high global energy consumption caused by local optimization in traditional technology. 2. Precise Predictive Control: Employing a CNN-LSTM fusion model, flue gas parameters and combustion efficiency are predicted 5-10 minutes in advance, completely avoiding the lag of traditional control methods. Simultaneously, the air-fuel ratio is precisely controlled within the optimal range of 1.05-1.15, balancing complete combustion with low emissions, and nitrogen oxide emission concentration is stably controlled at 50 mg / m³. 3 the following; 3. Significant energy consumption reduction: Through multiple synergies of flue gas waste heat recovery to preheat combustion air, AI-adjusted guide vanes to reduce dust collector resistance, and closed-loop iterative optimization control strategies, the overall power saving rate of the dust collector is ensured to be ≥30%, the gas consumption per unit product of the smelting furnace is reduced by 18%-25%, the total production cost is significantly reduced, and the economic benefits of the enterprise are improved. 4. High stability and safety: Through data preprocessing, model regularization design and multi-condition adaptation training, the model's ability to cope with load fluctuations is improved. At the same time, it is equipped with flame monitoring, gas leak alarm and emergency control mechanism to ensure safe and stable production operation in all aspects. It can be widely used in supporting systems for various industrial smelting furnaces such as copper alloy, aluminum alloy and steel, and has strong generalization. Attached Figure Description

[0015] Figure 1 This is a flowchart of a method for optimizing gas and electricity consumption in a smelting furnace based on AI collaborative control, according to the present invention. Figure 2 This is a schematic diagram of the energy-saving system of the smelting furnace-dust collector coupling of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto. Example

[0017] like Figure 1 As shown, Figure 1 This is a flowchart illustrating an AI-based collaborative control method for optimizing the gas and electricity consumption of a smelting furnace. The method is applied to a coupled energy-saving system of a smelting furnace and dust collector (e.g.,...). Figure 2 As shown in the figure, the system includes a smelting furnace combustion control unit, a dust collector adaptive operation unit, an AI intelligent decision-making unit, and a flue gas waste heat recovery unit. The method includes the following steps: S1. Multi-dimensional parameter acquisition: Through the sensors that are matched with the combustion control unit of the smelting furnace and the adaptive operation unit of the dust collector, multi-dimensional operating parameters of the smelting furnace and the dust collector are collected in real time, and the multi-dimensional operating parameters are transmitted to the AI ​​intelligent decision unit. S2. Data Preprocessing: The Z-score normalization method is used to normalize the collected multi-dimensional operating parameters. The formula is as follows: Where X is the multi-dimensional operating parameter, μ is the mean of the multi-dimensional operating parameter, and σ is the standard deviation of the multi-dimensional operating parameter. The 3σ criterion is used to remove outliers caused by instantaneous fluctuations of the sensor and construct a continuous time series data sample set. S3. AI Model Prediction and Optimization Decision: The time series data sample set is input into the dual-consumption collaborative deep learning model built into the AI ​​intelligent decision unit. The dual-consumption collaborative deep learning model extracts multi-parameter coupled features through the CNN feature extraction module, and predicts the flue gas volume, dust concentration and gas combustion efficiency in the next 5-10 minutes through the LSTM time series prediction module. Then, the reinforcement learning optimization module outputs collaborative optimization instructions for gas flow rate, combustion air ratio, fan frequency and ash cleaning interval with the goal of "lowest gas and electricity consumption", so that the air-fuel ratio is maintained in the optimal range of 1.05-1.15. S4. Multi-unit collaborative control: Based on the optimization instructions, the system synchronously regulates each functional unit: adjusts the gas ratio regulating valve of the smelting furnace combustion control unit to accurately control the ratio of gas to combustion air; adjusts the fan frequency converter and ash removal controller of the dust collector adaptive operation unit to synchronously drive the AI-adjustable flow guide device to optimize the angle of the flow guide blades; and preheats the combustion air to 80-120℃ through the flue gas waste heat recovery unit, and starts auxiliary electric heating compensation when the temperature is insufficient. S5. Closed-loop feedback iteration: Real-time collection of unit product gas consumption, dust collector power consumption and nitrogen oxide emission data to obtain raw data. Based on the above raw data, the energy-saving effect is calculated. The above raw data and the energy-saving effect calculation results are fed back to the reinforcement learning optimization module. The model parameters are iterated and optimized once per hour to ensure that the overall power saving rate of the dust collector is ≥30% and the unit product gas consumption of the smelting furnace is reduced by 18%-25%.

[0018] Preferably, in step S1, the multi-dimensional operating parameters include the furnace charge input, real-time furnace temperature, furnace pressure, gas flow rate, combustion air flow rate, gas pressure, and gas combustion efficiency on the furnace side, as well as the flue gas volume, dust concentration, dust collector resistance, fan current, dust removal interval time, and filter bag surface temperature on the dust collector side.

[0019] Preferably, in step S2, when constructing a continuous time series data sample set, linear interpolation is used to supplement missing data. The proportion of missing data does not exceed 3% of the total data volume. The time span of the continuous time series data sample set is set to 24 hours, and it is divided into a sub-sample set according to each hour. Each sub-sample set contains 120 data points, which makes it easier for the model to capture the parameter change patterns under different working conditions, improve the model's generalization ability, and adapt to the working condition fluctuations of different production batches.

[0020] Preferably, in step S3, the training process of the dual-consumption collaborative deep learning model includes: using the preprocessed time-series data sample set as the training set, dividing the dataset according to the ratio of 70% of the data for model training, 20% for validation, and 10% for testing; the CNN feature extraction module adopts an alternating structure of 3 convolutional layers and 2 pooling layers, with the convolutional kernel size set to 3×3 and the stride to 1, and the pooling layer adopts max pooling to extract multi-parameter coupled features; the LSTM time-series prediction module sets 4 hidden layers, each with 64 neurons, and uses dropout... The threshold value was set to 0.2 to avoid model overfitting, and the ReLU function was used as the activation function. The reinforcement learning optimization module adopted the PPO algorithm, taking the flue gas volume, dust concentration, gas combustion efficiency, and dust collector resistance as state inputs, and the total gas and electricity consumption per unit product, nitrogen oxide emissions, and filter bag wear rate as reward and penalty factors. Positive rewards were given when the total gas and electricity consumption decreased, nitrogen oxide emissions met the standards, and filter bag wear slowed down, while negative penalties were given. The control strategy was iteratively optimized until the model converged, and the convergence accuracy was controlled within ±2% to ensure that the deviation between the predicted value and the actual value did not exceed the set range.

[0021] Preferably, in step S3, the precise control of the air-fuel ratio is achieved through the linkage adjustment of the gas flow rate and the combustion air flow rate. The AI ​​intelligent decision-making unit collects the flow data of the two in real time, calculates the actual air-fuel ratio, and when it deviates from the range of 1.05-1.15, it prioritizes adjusting the combustion air flow rate to avoid large fluctuations in the gas flow rate affecting the stability of the furnace temperature.

[0022] Preferably, in step S4, the AI-adjustable flow guiding device includes flow guiding blades, a stepper motor, and a position sensor. The position sensor provides real-time feedback of the flow guiding blade angle signal. Based on the fan frequency and the predicted flue gas volume, the blade angle is dynamically adjusted to 15-60° to achieve uniform distribution of flue gas inside the dust collector, reduce local resistance of the dust collector, and reduce ineffective power consumption of the fan. The flue gas waste heat recovery unit adopts a tubular heat exchanger made of high-temperature resistant stainless steel, and the heat exchange area is matched according to the rated flue gas volume of the smelting furnace. The auxiliary electric heating adopts a high-frequency induction heating method, and the heating power can be steplessly adjusted within the range of 0-50kW. It is linked with the flue gas waste heat recovery unit to form a closed-loop temperature control. It automatically starts when the preheated combustion air temperature is below 80°C and automatically shuts down when the temperature reaches 120°C. At the same time, the start-stop frequency and heating duration are recorded as reference parameters for model iteration to avoid energy waste.

[0023] Preferably, in step S5, the energy-saving effect calculation formula is: η=[(Q0-Q1)×k+(W0-W1)] / (Q0×k+W0)×100%, where η is the comprehensive energy-saving rate, Q0 is the unit product gas consumption before optimization, Q1 is the unit product gas consumption after optimization, W0 is the dust collector's unit time power consumption before optimization, W1 is the dust collector's unit time power consumption after optimization, and k is the gas-electricity energy consumption conversion coefficient; at the same time, the nitrogen oxide emission concentration needs to be controlled at 50mg / m³. 3 Below this threshold, when the air-fuel ratio exceeds 1.08-1.12, the model will prioritize adjusting the air-fuel ratio to the range of 1.08-1.12 and increasing the preheating temperature of the combustion air to 100-120℃. By optimizing the combustion conditions, the generation of nitrogen oxides will be reduced, thus balancing energy conservation and environmental protection goals.

[0024] Preferably, the combustion control unit of the smelting furnace also includes a flame detector and a gas leak alarm. The flame detector monitors the combustion status in real time. When it detects that the flame is unstable or goes out, the AI ​​intelligent decision unit immediately outputs an emergency control command to cut off the gas supply and increase the combustion air flow to 1.2 times the rated value. At the same time, it links the dust collector adaptive operation unit to increase the fan frequency to 80% of the rated frequency to accelerate the discharge of residual gas in the furnace. When the gas leak alarm detects a leak, it immediately triggers an audible and visual alarm and shuts down the entire gas pipeline to ensure production safety.

[0025] The present invention will be further described in detail below with reference to specific embodiments. This embodiment takes a copper alloy smelting furnace-bag dust collector coupling system as an example. The rated capacity of the smelting furnace is 1000 kg / batch, and the rated flue gas treatment capacity of the dust collector is 3000 m³ / h.

[0026] An AI-based collaborative control method for optimizing the gas and electricity consumption of a smelting furnace is applied to a coupled energy-saving system of a copper alloy smelting furnace and a bag filter. This system includes a smelting furnace combustion control unit, a bag filter adaptive operation unit, an AI intelligent decision-making unit, and a tubular flue gas waste heat recovery unit. The specific implementation steps are as follows: S1. Multi-dimensional parameter acquisition: Parameters are acquired at a frequency of 30 seconds / time. For the smelting furnace side: furnace charge input rate 600-800 kg / batch, real-time furnace temperature 1180-1250℃, furnace pressure -50 to +50 Pa, and gas flow rate 9-12 m³ / h. 3 / h, combustion air flow rate 130-180m 3 / h, gas pressure 0.35-0.5MPa; dust collector side: flue gas volume 1800-2600m³ / h, dust concentration 80-200mg / m³, dust collector resistance 900-1500Pa, fan current 16-25A, dust cleaning interval 4-8 minutes, filter bag surface temperature 90-120℃. The parameters are transmitted to the AI ​​intelligent decision unit through a dual 5G industrial module and industrial Ethernet, with the transmission delay stable within 80ms.

[0027] S2. Data Preprocessing: The above parameters are processed using the Z-score standardization formula Z=(X-μ) / σ, and the mean μ and standard deviation σ of each parameter are calculated. For example, the mean gas flow rate μ=10.5m³ / h 3 / h, standard deviation σ=0.8m 3 / h, remove outliers exceeding the range [10.5-3×0.8, 10.5+3×0.8], i.e., [8.1, 12.9] (e.g., a certain instantaneous gas flow rate of 14.2m³ / h). 3 / h fluctuation data); by supplementing the three missing furnace pressure data points through linear interpolation, a 24-hour time series data sample set was constructed, which was divided into 24 sub-sample sets, each containing 120 data points.

[0028] S3. AI Model Prediction and Optimization Decision: After the dual-consumption collaborative deep learning model is trained, a time-series data sample set is input. The CNN feature extraction module extracts the coupling features of furnace charge input and flue gas volume, and gas flow rate and dust concentration. The LSTM time-series prediction module predicts the flue gas volume of 2200 m³ / s within the next 8 minutes. 3 / h, dust concentration 130mg / m 3 The gas combustion efficiency is 91.5%. The reinforcement learning module, aiming to minimize both gas and electricity consumption, and considering nitrogen oxide emissions and dust collector resistance constraints, outputs an optimization command: gas flow rate 10.8 m³ / h. 3 The parameters are set as follows: air-fuel ratio 1:16.2, fan frequency 36Hz, ash cleaning interval 5.5 minutes, and air-fuel ratio deviation adjustment threshold set at ±0.02 to ensure stability within the optimal range.

[0029] S4. Multi-unit collaborative control: Adjust the gas ratio regulating valve of the smelting furnace to maintain the gas-to-combustion air ratio of 1:16.2; the flame detector monitors the combustion status in real time; the gas pressure is stabilized at 0.4MPa; the dust collector fan frequency is adjusted to 36Hz, the ash cleaning interval is set to 5.5 minutes, the AI ​​adjustable guide vane angle is adjusted to 32°, and the position sensor feedback angle error is 0.5°; the flue gas waste heat recovery unit preheats the combustion air to 108°C through a tubular heat exchanger, without the need to start the auxiliary electric heating, and the heat exchange efficiency is detected to be 86%.

[0030] S5, Closed-loop feedback iteration: Real-time data collection, optimization of unit product gas consumption from 8.6m³ / h. 3 / kg, dust collector power consumption 125kW·h / h, optimized unit product air consumption 6.88m³ 3 / kg, dust collector power consumption 82kW·h / h, comprehensive energy saving rate calculated according to the formula: η=[(8.6-6.88)×1.2+(125-82)] / (8.6×1.2+125)×100%≈33.3%; nitrogen oxide emission concentration 43mg / m³ 3 This meets environmental protection requirements. The data was fed back to the model, and parameter iteration was completed after 1 hour. After optimization, the fan current dropped to 18.2A, the filter bag resistance stabilized at around 1050Pa, and the operation was in good condition.

[0031] This embodiment operated continuously for 30 days, and the statistics showed that the overall power saving rate of the dust collector was 34.2%, the unit product gas consumption of the smelting furnace was reduced by 20%, and the nitrogen oxide emission concentration was controlled at 40-48 mg / m³. 3 During this period, there were no issues such as filter bag blockage, gas leakage, or abnormal combustion, resulting in a significant improvement in production stability and economy.

[0032] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Based on the technical essence of the present invention, any simple modifications, equivalent substitutions, and improvements made to the above embodiments within the spirit and principles of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for optimizing the gas and electricity consumption of a smelting furnace based on AI collaborative control, characterized in that, An energy-saving system coupled with a smelting furnace and a dust collector is applied. The system includes a smelting furnace combustion control unit, a dust collector adaptive operation unit, an AI intelligent decision-making unit, and a flue gas waste heat recovery unit. The method includes the following steps: S1. Multi-dimensional parameter acquisition: Through the sensors that are matched with the combustion control unit of the smelting furnace and the adaptive operation unit of the dust collector, multi-dimensional operating parameters of the smelting furnace and the dust collector are acquired in real time and transmitted to the AI ​​intelligent decision unit. S2. Data Preprocessing: The Z-score normalization method is used to normalize the collected multi-dimensional operating parameters. The formula is as follows: Where X is the multi-dimensional operating parameter, μ is the mean of the multi-dimensional operating parameter, and σ is the standard deviation of the multi-dimensional operating parameter. The 3σ criterion is used to remove outliers caused by instantaneous fluctuations of the sensor and construct a continuous time series data sample set. S3. AI Model Prediction and Optimization Decision: The time series data sample set is input into the dual-consumption collaborative deep learning model built into the AI ​​intelligent decision unit. The dual-consumption collaborative deep learning model extracts multi-parameter coupled features through the CNN feature extraction module, and predicts the flue gas volume, dust concentration and gas combustion efficiency in the next 5-10 minutes through the LSTM time series prediction module. Then, the reinforcement learning optimization module outputs collaborative optimization instructions for gas flow rate, combustion air ratio, fan frequency and ash cleaning interval with the goal of "lowest gas and electricity consumption", so that the air-fuel ratio is maintained in the optimal range of 1.05-1.

15. S4. Multi-unit collaborative control: Based on the optimization instructions, the system synchronously regulates each functional unit; adjusts the gas ratio regulating valve of the smelting furnace combustion control unit to precisely control the ratio of gas to combustion air; adjusts the fan frequency converter and dust removal controller of the dust collector adaptive operation unit to synchronously drive the AI-adjustable flow guide device to optimize the angle of the flow guide blades; and preheats the combustion air to 80-120℃ through the flue gas waste heat recovery unit, and starts auxiliary electric heating compensation when the temperature is insufficient. S5. Closed-loop feedback iteration: Real-time collection of unit product gas consumption, dust collector power consumption and nitrogen oxide emission data to obtain raw data. Based on the above raw data, the energy-saving effect is calculated. The above raw data and the energy-saving effect calculation results are fed back to the reinforcement learning optimization module. The model parameters are iterated and optimized once per hour to ensure that the overall power saving rate of the dust collector is ≥30% and the unit product gas consumption of the smelting furnace is reduced by 18%-25%.

2. The method for optimizing gas and electricity consumption in a smelting furnace based on AI collaborative control according to claim 1, characterized in that, In step S1, the multi-dimensional operating parameters include the furnace charge input, real-time furnace temperature, furnace pressure, gas flow rate, combustion air flow rate, gas pressure, and gas combustion efficiency on the furnace side, as well as the flue gas volume, dust concentration, dust collector resistance, fan current, dust removal interval time, and filter bag surface temperature on the dust collector side.

3. The method for optimizing gas and electricity consumption in a smelting furnace based on AI collaborative control according to claim 1, characterized in that, In step S2, when constructing the continuous time series data sample set, linear interpolation is used to supplement missing data. The proportion of missing data does not exceed 3% of the total data volume. The time span of the continuous time series data sample set is set to 24 hours, and it is divided into a sub-sample set according to each hour. Each sub-sample set contains 120 data points, which makes it easier for the model to capture the parameter change patterns under different working conditions, improve the model's generalization ability, and adapt to the working condition fluctuations of different production batches.

4. The method for optimizing gas and electricity consumption in a smelting furnace based on AI collaborative control according to claim 1, characterized in that, In step S3, the training process of the dual-consumption collaborative deep learning model includes: using the preprocessed time-series data sample set as the training set, dividing the dataset according to the ratio of 70% of the data for model training, 20% for validation, and 10% for testing; the CNN feature extraction module adopts an alternating structure of 3 convolutional layers and 2 pooling layers, with the convolutional kernel size set to 3×3 and the stride set to 1, and the pooling layer adopts max pooling to extract multi-parameter coupled features; the LSTM time-series prediction module sets 4 hidden layers, each with 64 neurons, and the dropout coefficient is set to... To avoid model overfitting, the activation function is ReLU. The reinforcement learning optimization module uses the PPO algorithm, taking flue gas volume, dust concentration, gas combustion efficiency, and dust collector resistance as state inputs, and total gas and electricity consumption per unit product, nitrogen oxide emissions, and filter bag wear rate as reward and penalty factors. Positive rewards are given when total gas and electricity consumption decreases, nitrogen oxide emissions meet standards, and filter bag wear slows down, while negative penalties are given. The control strategy is iteratively optimized until the model converges, with convergence accuracy controlled within ±2%, ensuring that the deviation between predicted and actual values ​​does not exceed the set range.

5. The method for optimizing gas and electricity consumption in a smelting furnace based on AI collaborative control according to claim 1, characterized in that, In step S3, the precise control of the air-fuel ratio is achieved through the linkage adjustment of the gas flow rate and the combustion air flow rate. The AI ​​intelligent decision-making unit collects the flow data of the two in real time and calculates the actual air-fuel ratio. When it deviates from the range of 1.05-1.15, the combustion air flow rate is adjusted first to avoid large fluctuations in the gas flow rate affecting the stability of the furnace temperature.

6. The method for optimizing gas and electricity consumption in a smelting furnace based on AI collaborative control according to claim 1, characterized in that, In step S4, the AI-adjustable flow guiding device includes flow guiding blades, a stepper motor, and a position sensor. The position sensor provides real-time feedback of the flow guiding blade angle signal. Based on the fan frequency and the predicted flue gas volume, the blade angle is dynamically adjusted to 15-60° to achieve uniform distribution of flue gas inside the dust collector, reduce local resistance of the dust collector, and reduce ineffective power consumption of the fan. The flue gas waste heat recovery unit uses a tubular heat exchanger made of high-temperature resistant stainless steel, and the heat exchange area is matched according to the rated flue gas volume of the smelting furnace. The auxiliary electric heating adopts a high-frequency induction heating method, and the heating power can be steplessly adjusted within the range of 0-50kW. It is linked with the flue gas waste heat recovery unit to form a closed-loop temperature control. It automatically starts when the preheated combustion air temperature is below 80°C and automatically shuts down when the temperature reaches 120°C. At the same time, the start-stop frequency and heating duration are recorded as reference parameters for model iteration to avoid energy waste.

7. The method for optimizing gas and electricity consumption in a smelting furnace based on AI collaborative control according to claim 1, characterized in that, In step S5, the energy-saving effect calculation formula is: η=[(Q0-Q1)×k+(W0-W1)] / (Q0×k+W0)×100%, where η is the comprehensive energy-saving rate, Q0 is the unit product gas consumption before optimization, Q1 is the unit product gas consumption after optimization, W0 is the dust collector's unit time power consumption before optimization, W1 is the dust collector's unit time power consumption after optimization, and k is the gas-electricity energy consumption conversion coefficient; at the same time, the nitrogen oxide emission concentration needs to be controlled at 50mg / m³. 3 Below this threshold, when the air-fuel ratio exceeds 1.08-1.12, the model will prioritize adjusting the air-fuel ratio to the range of 1.08-1.12 and increasing the preheating temperature of the combustion air to 100-120℃. By optimizing the combustion conditions, the generation of nitrogen oxides will be reduced, thus balancing energy conservation and environmental protection goals.

8. The method for optimizing gas and electricity consumption in a smelting furnace based on AI collaborative control according to claim 1, characterized in that, The combustion control unit of the smelting furnace also includes a flame detector and a gas leak alarm. The flame detector monitors the combustion status in real time. When it detects that the flame is unstable or goes out, the AI ​​intelligent decision unit immediately outputs an emergency control command to cut off the gas supply and increase the combustion air flow to 1.2 times the rated value. At the same time, it links the dust collector adaptive operation unit to increase the fan frequency to 80% of the rated frequency to accelerate the discharge of residual gas in the furnace. When the gas leak alarm detects a leak, it immediately triggers an audible and visual alarm and shuts down the entire gas pipeline to ensure production safety.

Images