Automatic temperature control heating method and system

Abstract

The invention discloses an automatic temperature control heating method and system. The method comprises the steps that heating state data are obtained in real time and preprocessed; a deep neural network is adopted to predict the heating state data to obtain heating control parameters; predicting the historical heating state sequence data by adopting a long short-term memory network to obtain a future parameter change value; a reinforcement learning algorithm is adopted for reinforcement feedback, and an optimization control signal for the heating control parameters is obtained; performing fuzzy reasoning by adopting an adaptive neural fuzzy reasoning system, and adaptively adjusting rules and weights to obtain a fuzzy control signal; and integrating the parameters and the signals, and generating a heating control signal for the heating equipment through weighted fusion. According to the technical scheme, intelligent automatic control over the heating equipment is achieved, the monitoring capacity and the process management and control capacity of the heating equipment are greatly improved, the heating quality of products is improved, the rejection rate of the products is reduced, and the automation and intelligence level is improved.

Description

Technical Field

[0001] The present invention relates to the field of heating technology, and in particular to an automatic temperature control heating method and an automatic temperature control heating system. Background Art

[0002] At present, the existing technology for the heating process of products basically relies on technicians or operators to manually monitor and adjust heating parameters, which not only consumes a lot of man-hours, but also varies from person to person in terms of the ability to cope with external changing conditions, such as temperature changes in the morning, noon, afternoon, and throughout the year, as well as the aging process of the heating block. Moreover, manual monitoring through spot inspections and other means cannot be timely and cannot effectively identify changes in conditions in a timely manner. It often has a serious lag, and also brings about large fluctuations in product quality, which can easily cause quality and safety hazards. Summary of the invention

[0003] In response to the above problems, the present invention provides an automatic temperature control heating method and system, which monitors and optimizes the material heating process through an artificial intelligence algorithm, predicts and infers the heating state data through a deep neural network, a long short-term memory network, a reinforcement learning algorithm and an adaptive neural fuzzy inference system, and obtains a heating control signal for the heating equipment, thereby realizing intelligent automatic control of the heating equipment, greatly improving the monitoring capability and process control capability of the heating equipment, improving the heating quality of the product, reducing the product scrap rate, and improving the level of automation and intelligence.

[0004] To achieve the above object, the present invention provides an automatic temperature control heating method, comprising: Acquire heating status data of the heating equipment for the material heating process in real time, and pre-process the heating status data; Using a deep neural network to predict the heating state data to obtain heating control parameters; The long short-term memory network is used to predict the historical heating state sequence data with time series to obtain the future parameter change value; Using a reinforcement learning algorithm based on the reinforcement feedback of the heating state data, an optimized control signal for the heating control parameter is obtained; Adopting an adaptive neural fuzzy inference system to perform fuzzy inference on the heating state data and the heating control parameters, adaptively adjusting rules and weights, and obtaining a fuzzy control signal; The heating control parameter, the future parameter change value, the optimization control signal and the fuzzy control signal are integrated to generate a heating control signal for the heating device through weighted fusion.

[0005] In the above technical solution, preferably, the real-time acquisition of heating state data of the heating device for the material heating process and preprocessing of the heating state data include: Acquire in real time the temperature data, humidity data, pressure data, light data, wind speed data and material property data of the heating device for the material heating area as heating state data; Performing standardization and normalization processing on the heating state data; Using Kalman filtering, median filtering and / or sliding average filtering methods to filter the heating state data; Performing abnormal value detection on the heating state data and removing abnormal data; Performing missing data detection on the heating state data, and numerically filling the missing data by interpolation, forward / backward filling and / or mean filling methods; Extracting features from the time features, sliding windows and statistical features of the heating state data; The temperature data, humidity data, pressure data, light data, wind speed data and material property data in the heating state data are fused as comprehensive heating environment model data.

[0006] In the above technical solution, preferably, the use of a deep neural network to predict the heating state data to obtain the heating control parameters specifically includes: A deep neural network is constructed using an input layer, a preset number of hidden layers and an output layer, a mean square error is used as a loss function of the deep neural network, and an Adam optimizer is used to optimize the deep neural network; Using preset material heating data to train a deep neural network, so that the deep neural network can measure the error between the predicted parameters and the actual parameters; The heating state data is input into the deep neural network for prediction to obtain heating control parameters adjusted for the heating device.

[0007] In the above technical solution, preferably, the long short-term memory network is used to predict the historical heating state sequence data with a time series to obtain the future parameter change value, and the specific process includes: The long short-term memory network is constructed by using an input layer, an LSTM unit and an output layer, and the LSTM unit updates weights through a back propagation algorithm and BPTT; The long short-term memory network is trained using preset material heating data, so that the long short-term memory network can predict the parameter change trend of the heating process; Constructing the heating state data with a time series in the historical heating process to obtain the historical heating state series data; The historical heating state sequence data is input into the long short-term memory network for prediction to obtain future parameter change values.

[0008] In the above technical solution, preferably, the reinforcement learning algorithm is used to obtain the optimized control signal for the heating control parameter based on the reinforcement feedback of the heating state data, and the specific process includes: The heating state data is input into the reinforcement learning algorithm Q-learning, and the reinforcement learning algorithm calculates a reward function according to the difference between the actual temperature obtained by executing the current heating action and the target temperature, updates the Q value according to the feedback reward, and obtains an optimized control signal for adjusting the heating control parameters according to the Q value.

[0009] In the above technical solution, preferably, the adaptive neural fuzzy inference system is used to perform fuzzy inference on the heating state data and the heating control parameters, and adaptively adjusts the rules and weights to obtain the fuzzy control signal. The specific process includes: Using the fuzzy logic of the adaptive neural fuzzy inference system ANFIS to process uncertainty, the heating state data is converted into a fuzzy value using a fuzzification method; The fuzzy membership function is automatically adjusted in combination with the preset fuzzy inference rules, and the fuzzy control signal for the heating equipment is derived.

[0010] The present invention further proposes an automatic temperature control heating system, which uses the automatic temperature control heating method disclosed in any one of the above technical solutions, including: A heating data acquisition module is used to acquire the heating state data of the heating equipment for the material heating process in real time and pre-process the heating state data; A parameter deep prediction module, used to predict the heating state data using a deep neural network to obtain heating control parameters; A parameter change prediction module is used to use a long short-term memory network to predict the historical heating state sequence data with a time series to obtain future parameter change values; A parameter reinforcement learning module, used for obtaining an optimized control signal for the heating control parameter based on the reinforcement feedback of the heating state data using a reinforcement learning algorithm; A fuzzy reasoning control module, used for performing fuzzy reasoning on the heating state data and the heating control parameters by using an adaptive neural fuzzy reasoning system, adaptively adjusting rules and weights, and obtaining a fuzzy control signal; The control signal fusion module is used to integrate the heating control parameter, the future parameter change value, the optimization control signal and the fuzzy control signal, and generate a heating control signal for the heating device through weighted fusion.

[0011] In the above technical solution, preferably, the heating data acquisition module is specifically used for: Acquire in real time the temperature data, humidity data, pressure data, light data, wind speed data and material property data of the heating device for the material heating area as heating state data; Performing standardization and normalization processing on the heating state data; Using Kalman filtering, median filtering and / or sliding average filtering methods to filter the heating state data; Performing abnormal value detection on the heating state data and removing abnormal data; Performing missing data detection on the heating state data, and numerically filling the missing data by interpolation, forward / backward filling and / or mean filling methods; Extracting features from the time features, sliding windows and statistical features of the heating state data; The temperature data, humidity data, pressure data, light data, wind speed data and material property data in the heating state data are fused as comprehensive heating environment model data.

[0012] In the above technical solution, preferably, the parameter depth prediction module is specifically used for: A deep neural network is constructed using an input layer, a preset number of hidden layers and an output layer, a mean square error is used as a loss function of the deep neural network, and an Adam optimizer is used to optimize the deep neural network; Using preset material heating data to train a deep neural network, so that the deep neural network can measure the error between the predicted parameters and the actual parameters; Inputting the heating state data into the deep neural network for prediction to obtain heating control parameters adjusted for the heating device; The parameter change prediction module is specifically used for: The long short-term memory network is constructed by using an input layer, an LSTM unit and an output layer, and the LSTM unit updates weights through a back propagation algorithm and BPTT; The long short-term memory network is trained using preset material heating data, so that the long short-term memory network can predict the parameter change trend of the heating process; Constructing the heating state data with a time series in the historical heating process to obtain the historical heating state series data; The historical heating state sequence data is input into the long short-term memory network for prediction to obtain future parameter change values.

[0013] In the above technical solution, preferably, the parameter reinforcement learning module is specifically used for: The heating state data is input into the reinforcement learning algorithm Q-learning, the reinforcement learning algorithm calculates a reward function according to the difference between the actual temperature obtained by executing the current heating action and the target temperature, updates the Q value according to the feedback reward, and obtains an optimized control signal for adjusting the heating control parameter according to the Q value; The fuzzy reasoning control module is specifically used for: Using the fuzzy logic of the adaptive neural fuzzy inference system ANFIS to process uncertainty, the heating state data is converted into a fuzzy value using a fuzzification method; The fuzzy membership function is automatically adjusted in combination with the preset fuzzy inference rules, and the fuzzy control signal for the heating equipment is derived.

[0014] Compared with the prior art, the beneficial effects of the present invention are: monitoring and optimizing the material heating process through artificial intelligence algorithms, predicting and reasoning the heating state data through deep neural networks, long short-term memory networks, reinforcement learning algorithms and adaptive neural fuzzy inference systems, obtaining heating control signals for the heating equipment, and realizing intelligent automatic control of the heating equipment, which greatly improves the monitoring capability and process control capability of the heating equipment, improves the heating quality of the product, reduces the product scrap rate, and improves the level of automation and intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a schematic flow chart of an automatic temperature control heating method disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] The present invention is further described in detail below in conjunction with the accompanying drawings: like Figure 1 As shown, an automatic temperature control heating method according to the present invention includes: Acquire the heating status data of the heating equipment during the material heating process in real time, and pre-process the heating status data; A deep neural network is used to predict the heating state data to obtain the heating control parameters; The long short-term memory network is used to predict the historical heating state sequence data with time series to obtain the future parameter change value; The reinforcement learning algorithm is used to obtain the optimized control signal for the heating control parameters based on the reinforcement feedback of the heating state data; Adopting adaptive neural fuzzy inference system to perform fuzzy inference on heating state data and heating control parameters, adaptively adjusting rules and weights, and obtaining fuzzy control signals; The heating control parameters, future parameter change values, optimized control signals and fuzzy control signals are integrated to generate a heating control signal for the heating equipment through weighted fusion.

[0018] In this embodiment, an artificial intelligence algorithm is used to monitor and optimize the material heating process, and the heating state data is predicted and inferred through deep neural networks, long short-term memory networks, reinforcement learning algorithms, and adaptive neural fuzzy inference systems, respectively, to obtain a heating control signal for the heating equipment, thereby realizing intelligent automatic control of the heating equipment, greatly improving the monitoring capability and process control capability of the heating equipment, improving the heating quality of the product, reducing the product scrap rate, and improving the level of automation and intelligence.

[0019] Specifically, the AI ​​algorithm constructed through deep neural networks, long short-term memory networks, reinforcement learning algorithms and adaptive neural fuzzy inference systems is used to monitor and optimize the heating process of the material. The purpose is to automatically adjust the heating parameters according to factors such as ambient temperature, equipment conditions, and material properties, thereby achieving intelligent optimization of the temperature control process.

[0020] The method is divided into the following key parts: Data acquisition part: including temperature sensor, humidity sensor, material property sensor, etc.

[0021] Intelligent analysis and optimization part: Use machine learning and deep learning algorithms to process data in real time, predict temperature changes during the heating process, and optimize heating strategies.

[0022] Control part: Automatically adjust the power, current or temperature setting of the heating equipment based on the output of the AI ​​algorithm.

[0023] In the above embodiment, preferably, the heating state data of the heating device for the material heating process is obtained in real time, and the heating state data is preprocessed. The specific process includes: Real-time acquisition of temperature data, humidity data, pressure data, light data, wind speed data and material property data of the heating device for the material heating area as heating status data; Standardize and normalize the heating status data; Using Kalman filtering, median filtering and / or sliding average filtering methods to filter the heating state data; Perform outlier detection on heating status data and remove abnormal data; Perform missing data detection on heating status data and numerically fill the missing data by interpolation, forward / backward filling and / or mean filling methods; Extract the time characteristics, sliding window and statistical characteristics of the heating state data; The temperature data, humidity data, pressure data, light data, wind speed data and material property data in the heating state data are fused as comprehensive heating environment model data.

[0024] In this embodiment, data collection is mainly used to obtain real-time information such as temperature, humidity, material type, heating equipment status, etc. inside and outside the oven. These sensor data need to be preprocessed as follows: (1) Standardization and normalization: Standardize the raw data (such as temperature, humidity, etc.) collected by different sensors to ensure that they are in the same dimension to facilitate algorithm processing.

[0025] (2) Data denoising and interpolation: Since sensor data may be noisy or missing, data smoothing, interpolation or filtering methods (Kalman filtering) are used to process the data to ensure data accuracy.

[0026] Specifically, different types of sensors are used to collect different information. The sensors used in this embodiment and the data collected include: (1) Temperature sensor: used to collect real-time temperature data inside and outside the oven (such as thermocouple, RTD temperature sensor, etc.).

[0027] (2) Humidity sensor: used to measure air humidity, which has a certain impact on the thermal efficiency of the drying process.

[0028] (3) Pressure sensor: In some applications, pressure data may be used to evaluate the operating status of the heating equipment.

[0029] (4) Environmental sensors: such as light sensors and wind speed sensors, used to measure the impact of environmental conditions on the heating process.

[0030] (5) Material property sensor: If the heating material is different, different sensors may be required to monitor the material properties, such as the material's heat absorption, thermal conductivity, etc.

[0031] The raw data collected by the above sensors are usually continuous time series data.

[0032] Furthermore, since the data collected by different types of sensors have different dimensions (for example, temperature is in degrees Celsius and humidity is in percentage), directly inputting these data into the AI ​​algorithm may lead to poor model training results. In order to eliminate the differences between dimensions, the data needs to be standardized and normalized.

[0033] The purpose of standardization is to transform the data into a distribution with a mean of 0 and a standard deviation of 1, so that the range of each data point is not too large or too small, and to avoid a certain feature from having too much impact on model training. The standardization formula is: ; in, X is the original data, m is the mean of the data, s is the standard deviation of the data.

[0034] Normalization scales the data to a fixed range (such as 0 to 1), which is sometimes more beneficial for some algorithms (such as neural networks). The normalization formula is: ; in, X min and X max are the minimum and maximum values ​​in the data.

[0035] Furthermore, sensor data may contain noise, which may cause inaccurate model training results. Therefore, noise removal is necessary before inputting data into the AI ​​model.

[0036] Among them, the filtering methods include: Kalman Filter: Kalman filter is a recursive estimation algorithm based on linear system, which can effectively remove noise and estimate the system state. It is particularly suitable for denoising temperature data.

[0037] Median filtering: used to remove instantaneous outliers by sorting neighborhood data and taking the middle value to smooth the data.

[0038] Sliding average filtering: It removes short-term fluctuations by calculating the sliding average of the data series, and is often used to smooth time series data.

[0039] Among them, for sensor data, if there are abnormal values ​​(for example, temperature data that obviously deviates from the normal range), it is necessary to eliminate them by setting thresholds or statistical methods. For example, the standard deviation detection method can be used. When the difference between a data point and the mean exceeds a certain multiple of the standard deviation, it is considered abnormal data and eliminated.

[0040] Furthermore, in practical applications, data may be missing due to sensor failure, external environmental interference or other reasons. Common methods for processing missing data include interpolation and filling.

[0041] Among them, the interpolation method includes: (1) Linear interpolation: Linear interpolation is performed to fill in missing values ​​based on the adjacent data points before and after the missing data point.

[0042] (2) Polynomial interpolation: Use the polynomial relationship between previous and subsequent data points to interpolate missing values.

[0043] (3) Spline interpolation: Smoother than polynomial interpolation, suitable for interpolation when the data fluctuates greatly.

[0044] Filling methods include: (1) Forward filling or backward filling: For time series data, missing values ​​can be filled with the previous value or the next value. This method is suitable for situations where there are fewer missing values ​​and the data changes are relatively stable.

[0045] (2) Mean filling: Fill missing values ​​with the mean of the data set. This method is suitable for situations where the data distribution is relatively uniform.

[0046] Furthermore, in AI models, raw sensor data usually needs to be extracted through features to enhance the model's expressiveness. These features can help the model better capture the underlying patterns in the data.

[0047] Feature extraction methods include: (1) Temporal feature extraction: For time series data, time-related features can be extracted from it. For example, information such as the hour, day of the week, and season of the timestamp can be extracted so that the model can identify temperature change trends within a day or throughout the year.

[0048] (2) Sliding window: In order to capture the short-term trend of time series data, the data is often divided into windows of fixed length for processing. For example, the temperature data of the past 10 minutes can be used to predict the temperature change in the next 5 minutes.

[0049] (3) Statistical characteristics: Calculate the statistical characteristics of the original data, such as mean, variance, maximum value, minimum value, etc., to extract key trend information of the data.

[0050] Furthermore, due to the use of multiple different types of sensors (such as temperature sensors, humidity sensors, etc.), data fusion can synthesize a more accurate environmental model through weighted averaging, Kalman filtering, etc. For example, the data from temperature and humidity sensors can be combined to derive a comprehensive heating environment model.

[0051] In the above embodiment, preferably, a deep neural network is used to predict the heating state data to obtain the heating control parameters, and the specific process includes: A deep neural network is constructed using an input layer, a preset number of hidden layers, and an output layer. The mean square error is used as the loss function of the deep neural network, and the Adam optimizer is used to optimize the deep neural network. The deep neural network is trained using preset material heating data, so that the deep neural network can measure the error between the predicted parameters and the actual parameters; The heating status data is input into the deep neural network for prediction, and the heating control parameters adjusted for the heating equipment are obtained.

[0052] In this embodiment, a deep neural network (DNN) is used to accurately predict the temperature change during the heating process. DNN can capture nonlinear relationships and is suitable for such a complex heating process.

[0053] In this deep neural network, the input layer contains multiple input nodes, which receive data from sensors, including temperature, humidity, material type, etc. The hidden layer contains multiple hidden layers, each of which uses the ReLU (Rectified Linear Unit) activation function to learn complex features in the input data. The output layer is used to output the control parameters of the heating equipment, such as temperature setting value or power.

[0054] The training objective and loss function use mean square error (MSE) to measure the error between the predicted temperature and the actual temperature. The loss function is: ; in, T pred is the temperature predicted by the network, T actual is the actual temperature collected by the sensor, n is the number of data samples.

[0055] Among them, the Adam optimizer is used to adjust the network weights to minimize the loss function, perform backpropagation and update the model.

[0056] In the above implementation, preferably, a long short-term memory network is used to predict the historical heating state sequence data with a time series to obtain the future parameter change value, and the specific process includes: The input layer, LSTM unit and output layer are used to construct a long short-term memory network. The LSTM unit updates the weight through the back propagation algorithm and BPTT. The long short-term memory network is trained using preset material heating data, so that the long short-term memory network can predict the parameter change trend of the heating process; The heating state data with time series in the historical heating process is constructed to obtain the historical heating state sequence data; The historical heating state sequence data is input into the long short-term memory network for prediction to obtain the future parameter change values.

[0057] In this embodiment, since the temperature has the characteristics of a time series during the heating process, the long short-term memory (LSTM) network can better capture the time dependency and predict the temperature change trend.

[0058] In the structure of the LSTM network, the input layer is used to receive historical temperature sequence data and related environmental parameters (time, season, etc.). The LSTM unit is used to memorize the pattern of temperature changes over time, so as to predict future temperatures. LSTM can capture time series dependencies and predict future temperature trends by memorizing past temperature changes. This can help the system adjust the heating strategy in advance to avoid excessive or insufficient heating. The output layer is used to predict future temperature values ​​or control signals (power adjustment) as a supplement to DNN predictions and to adjust the heating strategy.

[0059] During the training process of the LSTM network, the weights are updated through the backpropagation algorithm and BPTT (Backpropagation Through Time) to reduce the error of temperature prediction. The mean square error is used as the loss function during the training process.

[0060] In the above embodiment, preferably, a reinforcement learning algorithm is used to obtain an optimized control signal for the heating control parameters based on the reinforcement feedback of the heating state data. The specific process includes: The heating state data is input into the reinforcement learning algorithm Q-learning, and a heating action (such as adjusting the temperature setting or changing the power) is selected based on the state. The reinforcement learning algorithm calculates the reward function based on the difference between the actual temperature and the target temperature obtained by executing the current heating action, updates the Q value based on the feedback reward, adjusts the future heating strategy, and obtains the optimized control signal for adjusting the heating control parameters based on the Q value.

[0061] Specifically, reinforcement learning (Q-learning) can help the system dynamically adjust the heating strategy in actual operation to cope with different environmental changes (such as weather changes, equipment aging, etc.).

[0062] In the Q-learning algorithm, the state space (s) includes temperature, humidity, time, season, equipment aging, etc. The action space (a) is to adjust the power, current, heating time, etc. of the heating equipment. The reward function (R) is: ; in,T set is the target temperature, T actual is the actual temperature, E For energy consumption, l is the weight of energy consumption.

[0063] Among them, the update formula of Q value is: ; Among them, α is the learning rate, γ is the discount factor, R ( s , a ) is the reward value, s' is the next state, and a' is the next action.

[0064] In the above embodiment, preferably, an adaptive neural fuzzy inference system is used to perform fuzzy inference on the heating state data and the heating control parameters, and adaptively adjust the rules and weights to obtain a fuzzy control signal. The specific process includes: The fuzzy logic of adaptive neuro-fuzzy inference system ANFIS is used to deal with uncertainty, and the heating state data is converted into fuzzy values ​​using fuzzification method; The fuzzy membership function is automatically adjusted in combination with the preset fuzzy inference rules, and the fuzzy control signal for the heating equipment is derived.

[0065] In this implementation, ANFIS combines the advantages of neural networks and fuzzy logic control to handle the uncertainty in the temperature control system. The algorithm obtains a fuzzy rule base by fuzzifying the input variables and training them through a neural network, ultimately achieving precise control.

[0066] Among them, the fuzzy rules are based on the input parameters in the heating process (such as temperature error, heating rate, etc.), and several fuzzy rules are set, such as: If the temperature error is large, increase the heating power; if the temperature error is small, reduce the heating power.

[0067] During the training process, the membership function of the fuzzy rules is adjusted through neural network training to optimize the control effect. The least squares method is used to optimize the parameters during the training process to ensure the accuracy of the temperature control process.

[0068] During the implementation process, ANFIS receives fuzzy inputs from the system (such as temperature error, heating rate, etc.), uses fuzzification methods to convert input data into fuzzy values ​​(such as fuzzy states such as "high temperature" and "low humidity"), and combines neural network training to handle the fuzziness and uncertainty of the input. Through the fuzzy reasoning mechanism, ANFIS derives the temperature control strategy based on the temperature error and other fuzzy parameters in the heating process, and adjusts the control strategy. For example, if the temperature gap is large, the power output should be increased. It can adaptively adjust rules and weights to adapt to different heating situations. The control signal generated by fuzzy reasoning is used to further accurately control the behavior of the heating equipment and ultimately optimize the temperature control process.

[0069] The present invention further provides an automatic temperature control heating system, which uses the automatic temperature control heating method disclosed in any one of the above embodiments, including: A heating data acquisition module is used to acquire the heating state data of the heating equipment for the material heating process in real time and pre-process the heating state data; A parameter deep prediction module is used to predict the heating state data using a deep neural network to obtain heating control parameters; A parameter change prediction module is used to use a long short-term memory network to predict the historical heating state sequence data with a time series to obtain future parameter change values; A parameter reinforcement learning module is used to obtain an optimized control signal for a heating control parameter based on reinforcement feedback of heating state data using a reinforcement learning algorithm; A fuzzy reasoning control module is used to perform fuzzy reasoning on heating state data and heating control parameters using an adaptive neural fuzzy reasoning system, adaptively adjust rules and weights, and obtain a fuzzy control signal; The control signal fusion module is used to integrate heating control parameters, future parameter change values, optimized control signals and fuzzy control signals, and generate heating control signals for heating equipment through weighted fusion.

[0070] In the above implementation, preferably, the heating data acquisition module is specifically used for: Real-time acquisition of temperature data, humidity data, pressure data, light data, wind speed data and material property data of the heating device for the material heating area as heating status data; Standardize and normalize the heating status data; Using Kalman filtering, median filtering and / or sliding average filtering methods to filter the heating state data; Perform outlier detection on heating status data and remove abnormal data; Perform missing data detection on heating status data and numerically fill the missing data by interpolation, forward / backward filling and / or mean filling methods; Extract the time characteristics, sliding window and statistical characteristics of the heating state data; The temperature data, humidity data, pressure data, light data, wind speed data and material property data in the heating state data are fused as comprehensive heating environment model data.

[0071] In the above implementation, preferably, the parameter depth prediction module is specifically used for: A deep neural network is constructed using an input layer, a preset number of hidden layers, and an output layer. The mean square error is used as the loss function of the deep neural network, and the Adam optimizer is used to optimize the deep neural network. The deep neural network is trained using preset material heating data, so that the deep neural network can measure the error between the predicted parameters and the actual parameters; The heating state data is input into a deep neural network for prediction, and the heating control parameters adjusted for the heating equipment are obtained; The parameter change prediction module is specifically used for: The input layer, LSTM unit and output layer are used to construct a long short-term memory network. The LSTM unit updates the weight through the back propagation algorithm and BPTT. The long short-term memory network is trained using preset material heating data, so that the long short-term memory network can predict the parameter change trend of the heating process; The heating state data with time series in the historical heating process is constructed to obtain the historical heating state sequence data; The historical heating state sequence data is input into the long short-term memory network for prediction to obtain the future parameter change values.

[0072] In the above implementation, preferably, the parameter reinforcement learning module is specifically used for: The heating state data is input into the reinforcement learning algorithm Q-learning. The reinforcement learning algorithm calculates the reward function according to the difference between the actual temperature and the target temperature obtained by executing the current heating action, updates the Q value according to the feedback reward, and obtains the optimized control signal for adjusting the heating control parameters according to the Q value; The fuzzy reasoning control module is specifically used for: The fuzzy logic of adaptive neuro-fuzzy inference system ANFIS is used to deal with uncertainty, and the heating state data is converted into fuzzy values ​​using fuzzification method; The fuzzy membership function is automatically adjusted in combination with the preset fuzzy inference rules, and the fuzzy control signal for the heating equipment is derived.

[0073] According to the automatic temperature control heating system disclosed in the above embodiment, the functions to be implemented by each module thereof correspond to the steps of the automatic temperature control heating method disclosed in the above embodiment. During the implementation process, the operations are performed with reference to the above embodiment, which will not be repeated here.

[0074] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An automatic temperature control heating method, characterized in that: include: Acquire heating status data of the heating equipment for the material heating process in real time, and pre-process the heating status data; Using a deep neural network to predict the heating state data to obtain heating control parameters; The long short-term memory network is used to predict the historical heating state sequence data with time series to obtain the future parameter change value; Using a reinforcement learning algorithm based on the reinforcement feedback of the heating state data, an optimized control signal for the heating control parameter is obtained; Adopting an adaptive neural fuzzy inference system to perform fuzzy inference on the heating state data and the heating control parameters, adaptively adjusting rules and weights, and obtaining a fuzzy control signal; The heating control parameter, the future parameter change value, the optimization control signal and the fuzzy control signal are integrated to generate a heating control signal for the heating device through weighted fusion.

2. The automatic temperature control heating method according to claim 1, characterized in that: The real-time acquisition of heating status data of the heating device for the material heating process and preprocessing of the heating status data specifically includes: Acquire in real time the temperature data, humidity data, pressure data, light data, wind speed data and material property data of the heating device for the material heating area as heating state data; Performing standardization and normalization processing on the heating state data; Using Kalman filtering, median filtering and / or sliding average filtering methods to filter the heating state data; Performing abnormal value detection on the heating state data and removing abnormal data; Performing missing data detection on the heating state data, and numerically filling the missing data by interpolation, forward / backward filling and / or mean filling methods; Extracting features from the time features, sliding windows and statistical features of the heating state data; The temperature data, humidity data, pressure data, light data, wind speed data and material property data in the heating state data are fused as comprehensive heating environment model data.

3. The automatic temperature control heating method according to claim 2, characterized in that: The use of a deep neural network to predict the heating state data to obtain the heating control parameters specifically includes: A deep neural network is constructed using an input layer, a preset number of hidden layers and an output layer, a mean square error is used as a loss function of the deep neural network, and an Adam optimizer is used to optimize the deep neural network; Using preset material heating data to train a deep neural network, so that the deep neural network can measure the error between the predicted parameters and the actual parameters; The heating state data is input into the deep neural network for prediction to obtain heating control parameters adjusted for the heating device.

4. The automatic temperature control heating method according to claim 3, characterized in that: The long short-term memory network is used to predict the historical heating state sequence data with a time series to obtain the future parameter change value, and the specific process includes: The long short-term memory network is constructed by using an input layer, an LSTM unit and an output layer, and the LSTM unit updates weights through a back propagation algorithm and BPTT; The long short-term memory network is trained using preset material heating data, so that the long short-term memory network can predict the parameter change trend of the heating process; Constructing the heating state data with a time series in the historical heating process to obtain the historical heating state series data; The historical heating state sequence data is input into the long short-term memory network for prediction to obtain future parameter change values.

5. The automatic temperature control heating method according to claim 4, characterized in that: The reinforcement learning algorithm is used to obtain an optimized control signal for the heating control parameter based on the reinforcement feedback of the heating state data. The specific process includes: The heating state data is input into the reinforcement learning algorithm Q-learning, and the reinforcement learning algorithm calculates a reward function according to the difference between the actual temperature obtained by executing the current heating action and the target temperature, updates the Q value according to the feedback reward, and obtains an optimized control signal for adjusting the heating control parameters according to the Q value.

6. The automatic temperature control heating method according to claim 5, characterized in that: The adaptive neural fuzzy inference system is used to perform fuzzy inference on the heating state data and the heating control parameters, and adaptively adjusts the rules and weights to obtain a fuzzy control signal. The specific process includes: Using the fuzzy logic of the adaptive neural fuzzy inference system ANFIS to process uncertainty, the heating state data is converted into a fuzzy value using a fuzzification method; The fuzzy membership function is automatically adjusted in combination with the preset fuzzy inference rules, and the fuzzy control signal for the heating equipment is derived.

7. An automatic temperature control heating system, characterized in that: The automatic temperature control heating method according to any one of claims 1 to 6 comprises: A heating data acquisition module is used to acquire the heating state data of the heating equipment for the material heating process in real time and pre-process the heating state data; A parameter deep prediction module, used to predict the heating state data using a deep neural network to obtain heating control parameters; A parameter change prediction module is used to use a long short-term memory network to predict the historical heating state sequence data with a time series to obtain future parameter change values; A parameter reinforcement learning module, used for obtaining an optimized control signal for the heating control parameter based on the reinforcement feedback of the heating state data using a reinforcement learning algorithm; A fuzzy reasoning control module, used for performing fuzzy reasoning on the heating state data and the heating control parameters by using an adaptive neural fuzzy reasoning system, adaptively adjusting rules and weights, and obtaining a fuzzy control signal; The control signal fusion module is used to integrate the heating control parameter, the future parameter change value, the optimization control signal and the fuzzy control signal, and generate a heating control signal for the heating device through weighted fusion.

8. The automatic temperature control heating system according to claim 7, characterized in that: The heating data acquisition module is specifically used for: Acquire in real time the temperature data, humidity data, pressure data, light data, wind speed data and material property data of the heating device for the material heating area as heating state data; Performing standardization and normalization processing on the heating state data; Using Kalman filtering, median filtering and / or sliding average filtering methods to filter the heating state data; Performing abnormal value detection on the heating state data and removing abnormal data; Performing missing data detection on the heating state data, and numerically filling the missing data by interpolation, forward / backward filling and / or mean filling methods; Extracting features from the time features, sliding windows and statistical features of the heating state data; The temperature data, humidity data, pressure data, light data, wind speed data and material property data in the heating state data are fused as comprehensive heating environment model data.

9. The automatic temperature control heating system according to claim 8, characterized in that: The parameter depth prediction module is specifically used for: A deep neural network is constructed using an input layer, a preset number of hidden layers and an output layer, a mean square error is used as a loss function of the deep neural network, and an Adam optimizer is used to optimize the deep neural network; Using preset material heating data to train a deep neural network, so that the deep neural network can measure the error between the predicted parameters and the actual parameters; Inputting the heating state data into the deep neural network for prediction to obtain heating control parameters adjusted for the heating device; The parameter change prediction module is specifically used for: The long short-term memory network is constructed by using an input layer, an LSTM unit and an output layer, and the LSTM unit updates weights through a back propagation algorithm and BPTT; The long short-term memory network is trained using preset material heating data, so that the long short-term memory network can predict the parameter change trend of the heating process; Constructing the heating state data with a time series in the historical heating process to obtain the historical heating state series data; The historical heating state sequence data is input into the long short-term memory network for prediction to obtain future parameter change values.

10. The automatic temperature control heating system according to claim 9, characterized in that: The parameter reinforcement learning module is specifically used for: The heating state data is input into the reinforcement learning algorithm Q-learning, the reinforcement learning algorithm calculates a reward function according to the difference between the actual temperature obtained by executing the current heating action and the target temperature, updates the Q value according to the feedback reward, and obtains an optimized control signal for adjusting the heating control parameter according to the Q value; The fuzzy reasoning control module is specifically used for: Using the fuzzy logic of the adaptive neural fuzzy inference system ANFIS to process uncertainty, the heating state data is converted into a fuzzy value using a fuzzification method; The fuzzy membership function is automatically adjusted in combination with the preset fuzzy inference rules, and the fuzzy control signal for the heating equipment is derived.

Images