PLC-based intelligent control system and method for thermal power plant

Abstract

The embodiment of the invention relates to the field of intelligent control, and provides a PLC-based intelligent control system and method for a thermal power plant, and the system comprises the steps: collecting a time queue of flame temperature values and oxygen content values, transmitting the time queue to a PLC-based fan rotating speed controller, and controlling the flame temperature values and the oxygen content values in the PLC-based fan rotating speed controller; and performing time sequence implicit association on the flame temperature value and the oxygen content value by adopting a data analysis and encoding technology based on artificial intelligence, so as to intelligently obtain an ideal oxygen content estimated value according to dynamic memory interaction information between flame temperature time sequence implicit association characteristics and oxygen content time sequence implicit association characteristics. And a fan rotating speed adjusting instruction is automatically generated based on the oxygen content error between the current time point and the current time point. By analyzing the time sequence information of the flame temperature and the oxygen content in real time, the combustion efficiency can be improved, and the optimal state can be kept.

Description

Technical Field

[0001] The present disclosure relates to the field of intelligent control, and more specifically, to an intelligent control system and method for a thermal power plant based on a PLC. Background Art

[0002] In modern thermal power plants, the optimization of the combustion process is crucial for improving efficiency, reducing emissions, and ensuring the safe operation of equipment. As a key parameter, the oxygen content directly affects the combustion effect through precise control. Too low oxygen content may lead to incomplete combustion, generating more pollutants such as carbon monoxide (CO); while too high oxygen content means that excessive excess air is heated and discharged, resulting in energy waste. Therefore, by precisely regulating the oxygen supply, full combustion of fuel can be achieved, the generation of harmful substances can be reduced, and the equipment can be protected, thus ensuring the efficient, clean, and safe operation of the thermal power plant.

[0003] Traditional oxygen content control methods usually rely on fixed set values and simple feedback mechanisms based on oxygen sensors. Specifically, due to changes in power demand, fuel properties, or external environmental conditions, the optimal oxygen concentration during the combustion process is not fixed. Fixed set points cannot adapt to these changes in real time, which may lead to low efficiency or excessive emissions. In addition, there are time delays and errors between sensor signal processing and the response of the control system in some traditional simple feedback regulation methods, which may result in untimely and inaccurate regulation. Especially when the combustion conditions change rapidly, the untimely response of the system may cause instantaneous combustion instability.

[0004] Therefore, an optimized intelligent control scheme for a thermal power plant is desired. Summary of the Invention

[0005] The present disclosure aims to at least solve one of the problems existing in the prior art, and provides an intelligent control system and method for a thermal power plant based on a PLC.

[0006] One aspect of the present disclosure provides an intelligent control method for a thermal power plant based on a PLC, including:

[0007] Collecting a time queue of flame temperature values and a time queue of oxygen content values through a sensor component;

[0008] Transmitting the time queue of the flame temperature values and the time queue of the oxygen content values to a fan speed controller based on a PLC;

[0009] In the fan speed controller based on a PLC, performing sequential implicit correlation encoding on the time queue of the flame temperature values and the time queue of the oxygen content values respectively to obtain a flame temperature sequential implicit correlation encoding vector and an oxygen content sequential implicit correlation encoding vector;

[0010] In the PLC - based fan speed controller, performing dynamic memory attention interaction on the flame temperature - time - series implicit correlation encoding vector and the oxygen content - time - series implicit correlation encoding vector to obtain a flame temperature - oxygen content - time - series fine - grained interaction encoding vector, including: calculating a flame temperature - oxygen content - time - series external knowledge - optimized fine - grained feature interaction matrix between the flame temperature - time - series implicit correlation encoding vector and the oxygen content - time - series implicit correlation encoding vector; based on the flame temperature - oxygen content - time - series external knowledge - optimized fine - grained feature interaction matrix, modulating and fusing the flame temperature - time - series implicit correlation encoding vector and the oxygen content - time - series implicit correlation encoding vector to obtain the flame temperature - oxygen content - time - series fine - grained interaction encoding vector;

[0011] In the PLC - based fan speed controller, based on the flame temperature - oxygen content - time - series fine - grained interaction encoding vector, obtaining an estimated value of the ideal oxygen content;

[0012] In the PLC - based fan speed controller, based on the estimated value of the ideal oxygen content, obtaining a fan speed adjustment instruction.

[0013] Optionally, performing time - series implicit correlation encoding on the time queue of the flame temperature values and the time queue of the oxygen content values respectively to obtain a flame temperature - time - series implicit correlation encoding vector and an oxygen content - time - series implicit correlation encoding vector, including: respectively inputting the time queue of the flame temperature values and the time queue of the oxygen content values into a sequence encoder based on a forward LSTM model to obtain the flame temperature - time - series implicit correlation encoding vector and the oxygen content - time - series implicit correlation encoding vector.

[0014] Optionally, calculating the flame temperature - oxygen content - time - series external knowledge - optimized fine - grained feature interaction matrix between the flame temperature - time - series implicit correlation encoding vector and the oxygen content - time - series implicit correlation encoding vector, including:

[0015] Performing fine - grained interaction on the flame temperature - time - series implicit correlation encoding vector and the oxygen content - time - series implicit correlation encoding vector to obtain a flame temperature - oxygen content - time - series fine - grained feature interaction matrix;

[0016] Performing external knowledge attention interaction on the flame temperature - oxygen content - time - series fine - grained feature interaction matrix to obtain the flame temperature - oxygen content - time - series external knowledge - optimized fine - grained feature interaction matrix.

[0017] Optionally, optimize the fine-grained feature interaction matrix based on the external knowledge of the flame temperature-oxygen content time series, and modulate and fuse the flame temperature time series implicit correlation encoding vector and the oxygen content time series implicit correlation encoding vector to obtain the flame temperature-oxygen content time series fine-grained interaction encoding vector, including:

[0018] Perform time series fine-grained correlation between the flame temperature-oxygen content time series external knowledge optimized fine-grained feature interaction matrix and the flame temperature time series implicit correlation encoding vector and the oxygen content time series implicit correlation encoding vector respectively to obtain the optimized flame temperature time series implicit correlation encoding vector and the optimized oxygen content time series implicit correlation encoding vector;

[0019] Calculate the element-wise division between the optimized flame temperature time series implicit correlation encoding vector and the optimized oxygen content time series implicit correlation encoding vector to obtain the flame temperature-oxygen content time series fine-grained interaction encoding vector.

[0020] Optionally, perform time series fine-grained correlation between the flame temperature-oxygen content time series external knowledge optimized fine-grained feature interaction matrix and the flame temperature time series implicit correlation encoding vector and the oxygen content time series implicit correlation encoding vector respectively to obtain the optimized flame temperature time series implicit correlation encoding vector and the optimized oxygen content time series implicit correlation encoding vector, including:

[0021] Perform a linear transformation on the flame temperature time series implicit correlation encoding vector to obtain a flame temperature query feature vector and a flame temperature value feature vector, and use the flame temperature-oxygen content time series external knowledge optimized fine-grained feature interaction matrix as the key matrix, and perform fine-grained modulation of the converter on the flame temperature query feature vector, the flame temperature value feature vector and the key matrix to obtain the optimized flame temperature time series implicit correlation encoding vector;

[0022] Perform a linear transformation on the oxygen content time series implicit correlation encoding vector to obtain an oxygen content query feature vector and an oxygen content value feature vector, and use the flame temperature-oxygen content time series external knowledge optimized fine-grained feature interaction matrix as the key matrix, and perform the fine-grained modulation of the converter on the oxygen content query feature vector, the oxygen content value feature vector and the key matrix to obtain the optimized oxygen content time series implicit correlation encoding vector.

[0023] Optionally, based on the flame temperature-oxygen content time series fine-grained interaction encoding vector, obtain an ideal oxygen content estimate value, including: inputting the flame temperature-oxygen content time series fine-grained interaction encoding vector into an ideal oxygen content dynamic optimizer based on a decoder to obtain the ideal oxygen content estimate value.

[0024] Optionally, inputting the fine-grained interaction encoding vector of the flame temperature-oxygen content time series into a decoder-based ideal oxygen content dynamic optimizer to obtain the estimated value of the ideal oxygen content includes: multiplying the decoding weight matrix of the decoder by the fine-grained interaction encoding vector of the flame temperature-oxygen content time series to obtain a decoded fine-grained interaction encoding vector of the flame temperature-oxygen content time series, and calculating the position-by-position sum value of all eigenvalues in the decoded fine-grained interaction encoding vector of the flame temperature-oxygen content time series to obtain the estimated value of the ideal oxygen content.

[0025] Optionally, obtaining a fan speed adjustment command based on the estimated value of the ideal oxygen content includes:

[0026] Extracting the oxygen content value at the current time point from the time queue of the oxygen content values, and calculating the difference between the oxygen content value at the current time point and the estimated value of the ideal oxygen content to obtain an oxygen content error;

[0027] Inputting the oxygen content error into a PID controller to obtain the fan speed adjustment command.

[0028] On the other hand, the present disclosure also provides an intelligent control system for a thermal power plant based on a PLC, including:

[0029] A flame temperature oxygen content data acquisition module for collecting a time queue of flame temperature values and a time queue of oxygen content values through a sensor assembly;

[0030] A flame temperature oxygen content data transmission module for transmitting the time queue of the flame temperature values and the time queue of the oxygen content values to a fan speed controller based on a PLC;

[0031] A flame temperature oxygen content data encoding module for respectively performing time series implicit correlation encoding on the time queue of the flame temperature values and the time queue of the oxygen content values in the fan speed controller based on a PLC to obtain a flame temperature time series implicit correlation encoding vector and an oxygen content time series implicit correlation encoding vector;

[0032] The flame temperature - oxygen content data interaction module is used to perform dynamic memory attention interaction on the PLC - based fan speed controller for the temporal implicit correlation encoding vectors of the flame temperature and the oxygen content to obtain a fine - grained interaction encoding vector of the flame temperature - oxygen content. Among them, the flame temperature - oxygen content data interaction module includes: an external knowledge optimization unit for calculating a fine - grained feature interaction matrix of the external knowledge optimization of the flame temperature - oxygen content time series between the temporal implicit correlation encoding vector of the flame temperature and the temporal implicit correlation encoding vector of the oxygen content; a feature modulation and fusion unit for modulating and fusing the temporal implicit correlation encoding vector of the flame temperature and the temporal implicit correlation encoding vector of the oxygen content based on the fine - grained feature interaction matrix of the external knowledge optimization of the flame temperature - oxygen content time series to obtain the fine - grained interaction encoding vector of the flame temperature - oxygen content time series;

[0033] The ideal oxygen content estimation module is used to obtain an ideal oxygen content estimation value in the PLC - based fan speed controller based on the fine - grained interaction encoding vector of the flame temperature - oxygen content time series;

[0034] The fan speed adjustment instruction generation module is used to obtain a fan speed adjustment instruction in the PLC - based fan speed controller based on the ideal oxygen content estimation value.

[0035] Optionally, the flame temperature - oxygen content data encoding module is further used to: input the time queue of the flame temperature value and the time queue of the oxygen content value into a sequence encoder based on the forward LSTM model respectively to obtain the temporal implicit correlation encoding vector of the flame temperature and the temporal implicit correlation encoding vector of the oxygen content.

[0036] Compared with the prior art, the present disclosure collects the time queue of the flame temperature value and the time queue of the oxygen content value through a sensor component and transmits them to a PLC - based fan speed controller. In the PLC - based fan speed controller, artificial - intelligence - based data analysis and encoding technologies are used to perform temporal implicit correlation on the flame temperature value and the oxygen content value, so as to intelligently obtain an ideal oxygen content estimation value according to the dynamic memory interaction information between the temporal implicit correlation features of the flame temperature and the temporal implicit correlation features of the oxygen content, and automatically generate a fan speed adjustment instruction based on the oxygen content error from the current time point. By analyzing the temporal information of the flame temperature and the oxygen content in real - time, the present disclosure can quickly and intelligently identify the optimal combustion conditions. Compared with the traditional method, it can dynamically adjust the fan speed according to real - time data, thereby helping to improve the combustion efficiency and maintain its optimal state. Brief Description of the Drawings

[0037] One or more embodiments are exemplarily illustrated by the pictures in the corresponding drawings. These exemplary illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements. Unless otherwise stated, the drawings in the figures do not constitute a scale limitation.

[0038] Figure 1 FIG. is a flowchart of an intelligent control method for a thermal power plant based on a PLC according to an embodiment of the present disclosure;

[0039] Figure 2 FIG. is a flowchart of step S140 in the intelligent control method for a thermal power plant based on a PLC according to an embodiment of the present disclosure;

[0040] Figure 3 FIG. is a flowchart of step S220 in the intelligent control method for a thermal power plant based on a PLC according to an embodiment of the present disclosure;

[0041] Figure 4 FIG. is a block diagram of an intelligent control system for a thermal power plant based on a PLC according to an embodiment of the present disclosure. Detailed Embodiments

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the following will elaborate on each embodiment of the present disclosure in conjunction with the drawings. However, those of ordinary skill in the art can understand that in each embodiment of the present disclosure, many technical details are provided to help readers better understand the present disclosure. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions required to be protected by the present disclosure can still be achieved. The following division of each embodiment is for convenience of description and should not constitute any limitation on the specific implementation manner of the present disclosure. Each embodiment can be combined and cross-referenced with each other without conflict.

[0043] In modern thermal power plants, the optimization of the combustion process is crucial for improving efficiency, reducing emissions, and ensuring the safe operation of equipment. As a key parameter, the oxygen content directly affects the combustion effect through precise control. Too low oxygen content may lead to incomplete combustion, generating more pollutants such as carbon monoxide (CO); while too high oxygen content means that too much excess air is heated and discharged, resulting in energy waste. Therefore, by precisely regulating the oxygen supply, full combustion of fuel can be achieved, reducing the generation of harmful substances, protecting the equipment, and thus ensuring the efficient, clean, and safe operation of the thermal power plant.

[0044] Traditional oxygen content control methods usually rely on fixed setpoints and simple feedback mechanisms based on oxygen sensors. Specifically, due to changes in power demand, fuel properties, or external environmental conditions, the optimal oxygen concentration during the combustion process is not fixed. Fixed setpoints cannot adapt to these changes in real time, which may lead to low efficiency or excessive emissions. In addition, there are time delays and errors between sensor signal processing and control system response in some traditional simple feedback regulation methods, which may result in untimely and inaccurate regulation. Especially when combustion conditions change rapidly, untimely system response may cause instantaneous combustion instability.

[0045] Therefore, to address the above technical problems, the technical concept of the present disclosure is to collect the time queue of flame temperature values and the time queue of oxygen content values through a sensor component and transmit them to a PLC-based fan speed controller. In the PLC-based fan speed controller, artificial intelligence-based data analysis and coding techniques are used to perform temporal implicit association on the flame temperature values and the oxygen content values, so as to intelligently obtain an ideal oxygen content estimated value based on the dynamic memory interaction information between the temporal implicit association features of the flame temperature and the temporal implicit association features of the oxygen content, and automatically generate a fan speed adjustment instruction based on the oxygen content error with the current time point. By analyzing the temporal information of flame temperature and oxygen content in real time, the present disclosure can quickly and intelligently identify the optimal combustion conditions. Compared with traditional methods, it can dynamically adjust the fan speed according to real-time data, thus helping to improve combustion efficiency and maintain its optimal state.

[0046] Figure 1 The flowchart of the PLC-based intelligent control method for a thermal power plant according to an embodiment of the present disclosure is as Figure 1As shown, the intelligent control method for a thermal power plant based on PLC according to an embodiment of the present disclosure includes: S110, collecting a time queue of flame temperature values and a time queue of oxygen content values through a sensor assembly; S120, transmitting the time queue of the flame temperature values and the time queue of the oxygen content values to a PLC-based fan speed controller; S130, in the PLC-based fan speed controller, respectively performing sequential implicit association coding on the time queue of the flame temperature values and the time queue of the oxygen content values to obtain a flame temperature sequential implicit association coding vector and an oxygen content sequential implicit association coding vector; S140, in the PLC-based fan speed controller, performing dynamic memory attention interaction on the flame temperature sequential implicit association coding vector and the oxygen content sequential implicit association coding vector to obtain a flame temperature-oxygen content sequential fine-grained interaction coding vector; S150, in the PLC-based fan speed controller, obtaining an ideal oxygen content estimated value based on the flame temperature-oxygen content sequential fine-grained interaction coding vector; S160, in the PLC-based fan speed controller, obtaining a fan speed adjustment instruction based on the ideal oxygen content estimated value.

[0047] In step S110, a time queue of flame temperature values and a time queue of oxygen content values are collected through a sensor assembly. It should be understood that in a thermal power plant, the flame temperature value and the oxygen content value are two crucial parameters, which are of great significance for optimizing the combustion process, improving efficiency, and reducing emissions. Specifically, the flame temperature value refers to the actual temperature of the flame during the combustion process, which reflects the intensity and efficiency of the combustion after the fuel is mixed with air. By monitoring the flame temperature, it can be ensured that the combustion occurs within the optimal temperature range, thereby improving the energy conversion efficiency. The oxygen content value refers to the remaining oxygen concentration in the combustion chamber that has not participated in the reaction, usually expressed as a volume percentage, which reflects whether the actual amount of air supplied during the combustion process is sufficient or excessive. Appropriately adjusting the oxygen content can ensure complete combustion of the fuel and avoid incomplete combustion caused by insufficient oxygen or energy waste caused by excessive oxygen. Therefore, in the technical solution of the present disclosure, collecting the time queue of the flame temperature values and the time queue of the oxygen content values through the sensor assembly and analyzing and processing them can guide the automatic adjustment of the fan speed and the like to achieve more precise combustion control.

[0048] As the eyes and ears of the entire system, the performance of sensors directly determines the quality of the acquired data. For the measurement of flame temperature, high-temperature sensors such as thermocouples or infrared thermometers are usually used. These devices can withstand the extreme environment inside the boiler and have fast response characteristics, enabling them to capture temperature changes within a short period. For oxygen content, oxygen sensors are generally used for monitoring, such as electrochemical oxygen sensors or zirconia oxygen sensors. They can accurately detect the remaining oxygen concentration in the flue gas, reflecting whether the air supply in the combustion chamber is sufficient. To ensure the integrity and continuity of the data, these sensors need to be distributed at different positions in the combustion chamber according to specific rules, including but not limited to key parts such as near the burner, the flame centerline, and the flue gas outlet, so as to comprehensively cover the entire combustion area and thus more accurately reflect the actual working conditions.

[0049] Next, how to effectively organize and manage the data from each sensor has become an important issue. For this purpose, an efficient data acquisition system has been constructed. This system is not only responsible for converting the analog signals scattered everywhere into digital signals, but also for performing preliminary processing on the raw data, such as operations like filtering and calibration, to remove noise interference and ensure numerical accuracy. In addition, considering the strong electromagnetic interference in the environment of thermal power plants, a series of anti-interference measures have also been taken in the hardware design. For example, shielded cables are selected to connect the sensors and the acquisition terminal, or EMI / RFI filter circuits are built-in to suppress the influence of external electromagnetic radiation. At the same time, to improve the stability and reliability of the system, redundant modules are usually equipped. Even if a certain component fails, it can ensure that other parts work normally without affecting the coherence of the overall data stream.

[0050] When the above preparations are completed, it enters the core stage - the generation of time-series data. The so-called "time queue" refers to a series of observed values arranged in sequence at a certain time interval (sampling frequency), which reflects the trend of the target variable evolving over time. For the flame temperature, this means recording the current temperature reading every fixed period (such as once per second); for the oxygen content, it is to synchronously record the corresponding oxygen concentration level under the same time basis. Such an approach can not only observe the state snapshot at a single time point, but also uncover potential change rules and development patterns, laying a solid foundation for further in-depth analysis.

[0051] It's worth noting that in actual applications, due to the inherent complexity and variability of the combustion process, such as fuel type switching and load fluctuations, even with the same sensor installed in the same location, data obtained over different time periods can vary significantly. Therefore, in addition to relying solely on hardware, software algorithms are also needed for auxiliary adjustments. For example, machine learning models can be trained on historical data to build predictive models to anticipate future scenarios. Alternatively, adaptive filtering technology can be introduced to dynamically correct sensor outputs to make them more realistic. These methods help improve data quality, enabling subsequent PLC-based fan speed controllers to make more accurate and effective decisions. In summary, the time series acquisition of flame temperature and oxygen content values using sensor components is a comprehensive project that integrates multiple advanced technologies. From carefully selecting high-performance sensors suitable for the site environment, to building a stable and efficient data acquisition platform, to meticulously designing the time series construction process, every detail has been carefully considered and implemented to provide the most robust data support for intelligent control in thermal power plants.

[0052] In step S120, the time series of the flame temperature values and the time series of the oxygen content values are transmitted to the PLC-based fan speed controller. Specifically, transmitting the time series of the flame temperature values and the oxygen content values to the PLC establishes a closed-loop control system, ensuring that an appropriate amount of oxygen is always present during the combustion process. This prevents incomplete combustion due to insufficient oxygen and energy waste due to excessive air. This not only enables immediate response based on current conditions but also predicts future needs through data analysis, thereby achieving a more efficient combustion process.

[0053] Specifically, before data transmission, all data from the sensor must undergo preliminary processing to ensure that its quality and format are suitable for subsequent use. This includes but is not limited to signal conditioning (such as amplification and filtering), analog signal to digital signal conversion (ADC), time stamping, and other operations. These steps are usually completed by a front-end data acquisition device located near the sensor. This device can be an integrated intelligent unit that is responsible for aggregating data from multiple sensors and performing necessary preprocessing tasks. In this way, not only can the workload of the main control system be reduced, but the response speed of the overall system can also be improved.

[0054] Next, when the preprocessed data is ready, it is necessary to consider how to efficiently transmit it to the PLC-based fan speed controller. Given the complex internal space layout and strong electromagnetic interference sources in thermal power plants, choosing the appropriate communication medium and technology is particularly important. For short-distance data exchange, wired connection methods such as shielded twisted pair or coaxial cable can be used. They have good noise resistance performance and can ensure the stability and accuracy of data transmission. However, with the expansion of the power plant scale and the increase in wiring difficulty, wireless communication has gradually become a more flexible option. For example, using low-power wide-area network technologies such as ZigBee, Wi-Fi, or LoRaWAN designed specifically for the industrial Internet of Things, a data transmission network covering the whole plant can be quickly established without damaging the existing structure. Such wireless solutions are not only easy to install but also can flexibly adjust the node positions according to actual needs, greatly facilitating the later expansion and maintenance work.

[0055] It is worth noting that regardless of the physical layer technology adopted, a series of software-level measures are also needed to ensure the quality of data transmission. First is the choice of protocol. An efficient communication protocol should have the following characteristics: support two-way authentication to enhance security; have a powerful error detection and correction mechanism; provide sufficient bandwidth to meet the needs of a large number of concurrent data streams; and simplify the configuration process as much as possible for the convenience of on-site engineers. Currently, industrial standard protocols such as Modbus RTU / ASCII, Profibus-DP, and EtherCAT are widely used in PLC control systems. They are optimized for different application scenarios, and users can make the best choice according to specific requirements. In addition, for possible packet loss or delay problems, buffer settings and retransmission strategies are usually also set to ensure that important information will not be lost due to temporary network fluctuations.

[0056] Once the data successfully arrives at the PLC port, the next step is how to effectively manage and parse this data. Since a PLC is essentially a real-time control system that needs to continuously read input variables and calculate output instructions according to a predetermined algorithm, the newly arrived data must be processed as soon as possible to avoid backlog. For this reason, modern PLCs often have built-in hardware accelerators dedicated to data processing or adopt a multi-core processor architecture, enabling smooth operation even in the face of massive amounts of data. On the other hand, to enable the PLC to better understand the received information, the data packet itself also needs to follow a certain structural specification. For example, it is packed in a specific format before sending, including fields such as identifiers, lengths, and content bodies, so that the receiving party can quickly identify the physical meaning corresponding to each data item, thereby accelerating the decoding speed. Through the effective implementation of the above measures, thermal power plants can manage the combustion process on a more refined and scientific basis, thereby promoting the green transformation and development of the entire industry.

[0057] In the practical application of intelligent control in thermal power plants, the design of the above transmission mechanism is not only for simply moving data, but more importantly for building an efficient, reliable and secure data circulation pipeline, enabling the PLC to timely obtain accurate information on flame temperature and oxygen content, thereby providing a solid foundation for the intelligent regulation of the fan speed. By combining advanced communication technologies and data processing methods, seamless connection of data from the collection point to the control center is ensured, enabling the fan speed controller based on PLC to dynamically adjust according to the latest combustion state, maintaining the best efficiency and stability of the combustion process. The combined effect of these measures ultimately realizes the refined management of the combustion process, providing a strong guarantee for the efficient, clean and safe operation of thermal power plants.

[0058] In step S130, in the PLC-based fan speed controller, temporal implicit correlation encoding is respectively performed on the time queue of the flame temperature value and the time queue of the oxygen content value to obtain a flame temperature temporal implicit correlation encoding vector and an oxygen content temporal implicit correlation encoding vector. Specifically, in the embodiments of the present disclosure, performing temporal implicit correlation encoding on the time queue of the flame temperature value and the time queue of the oxygen content value respectively to obtain a flame temperature temporal implicit correlation encoding vector and an oxygen content temporal implicit correlation encoding vector includes: respectively inputting the time queue of the flame temperature value and the time queue of the oxygen content value into a sequence encoder based on a forward LSTM model to obtain the flame temperature temporal implicit correlation encoding vector and the oxygen content temporal implicit correlation encoding vector. It should be understood that considering that the time queue data of the flame temperature and the oxygen content exhibit specific temporal characteristics as time changes, and within different local time periods, these data show different implicit patterns and interconnections. Therefore, in order to effectively capture the dependencies existing between the flame temperature and the oxygen content at different time scales, so as to better understand the temporal dynamic characteristics of the data, in the technical solution of the present disclosure, the time queue of the flame temperature value and the time queue of the oxygen content value are respectively input into a sequence encoder based on a forward LSTM model to capture the complex implicit patterns and interrelationships in the time series, and obtain a flame temperature temporal implicit correlation encoding vector and an oxygen content temporal implicit correlation encoding vector. It is worth mentioning that the forward LSTM model is a variant of the long short-term memory network, which is suitable for dealing with and predicting long-term dependence problems in time series data. In particular, the forward LSTM model specifically refers to an LSTM model that processes data in chronological order from front to back in time series data. This means that it starts from the earliest time point and gradually processes to the most recent time point, which is very useful for time series prediction tasks with clear causal relationships, because it can utilize all previous historical information to predict the state at future moments. In this way, the sequential dependencies and temporal trend changes of the flame temperature and the oxygen content can be captured and refined to better characterize the state changes at future moments, so as to achieve precise control of the fan speed.

[0059] Specifically, here, the time queue of the flame temperature value is taken as an example for specific illustration:

[0060] First of all, considering that the time series of the flame temperature value has significant temporal characteristics, that is, the data at each time point is not only affected by the current factors, but may also be affected by multiple previous time points. Therefore, a multi-layer stacked forward LSTM network is constructed, aiming to extract deep feature representations from the original data and convert them into information helpful for subsequent decision-making. The design goal of this network is to ensure that the model can make full use of historical information to predict future state changes, and thus provide a basis for fan speed adjustment.

[0061] At the input end of the network, a time series of flame temperature values collected from the sensor component is received. To improve the quality and representational ability of the data, these raw values first pass through an Embedding Layer. Although the flame temperature itself is a continuous variable, by mapping it to a higher-dimensional space, the richness of the data can be increased, enabling it to contain more implicit information. This step is crucial for subsequent feature learning as it allows the model to operate on the data at a higher level of abstraction, rather than just based on the raw temperature readings.

[0062] Next, the high-dimensional vector after embedding is fed into the first layer of LSTM units. Each LSTM unit internally consists of four main components: a forget gate, an input gate, a cell state update mechanism, and an output gate. When data for a time step enters this layer, the LSTM unit first uses the forget gate to determine which previous memories should be retained or discarded; then, the input gate determines which information needs to be added to the cell state based on the new input; the cell state update mechanism combines the results of these two gates to form the latest memory state; finally, the output gate calculates the output result for this time step based on the updated cell state. In this process, the gating mechanism inside the LSTM unit plays a key role, enabling the model to effectively and selectively remember or forget certain information, thus better adapting to the characteristics of time series data.

[0063] As the data flows along the time axis, the output processed by the first layer of LSTM units is passed to the next layer, forming a multi-layer stacked network structure. Such a design allows the model to learn more abstract and complex feature representations because each layer can further refine and generalize the information based on the previous layer. For example, in the second layer of LSTM, the model can not only focus on the local changes at a single time point but also identify the global trends spanning multiple time steps. Additionally, to prevent overfitting, a Dropout layer can be added after some LSTM layers, randomly discarding a portion of the neuron connections to enhance the generalization ability of the model. The role of the Dropout layer is to simulate the effects of different neuron combinations, improving the model's adaptability to unknown data. In summary, the network structure based on the forward LSTM model can not only effectively capture the long-term dependencies in the time series of flame temperature values but also possess strong representational ability and flexibility, thus helping to achieve more efficient combustion control strategies.

[0064] In step S140, the PLC-based fan speed controller performs a dynamic memory attention interaction on the flame temperature time series implicit correlation encoding vector and the oxygen content time series implicit correlation encoding vector to obtain a flame temperature-oxygen content time series fine-grained interaction encoding vector. Furthermore, considering that the flame temperature time series implicit correlation features and the oxygen content time series implicit correlation features have temporal interactions and influence relationships at different time points, and that there are deeper and implicit relationships between the two, in order to explore the deep, fine-grained interactions hidden in the time series and better capture the complex dynamic relationship between the two, in the technical solution disclosed herein, the flame temperature time series implicit correlation encoding vector and the oxygen content time series implicit correlation encoding vector are subjected to a dynamic memory attention interaction to obtain a flame temperature-oxygen content time series fine-grained interaction encoding vector. In particular, by introducing external knowledge, the dynamic memory attention interaction mechanism can guide the model to better understand and process complex combustion processes, enabling the model to further refine more expressive feature representations based on the original encoding. These representations not only contain the original time series information but also incorporate the complex interactions between them.

[0065] Figure 2 FIG is a flow chart of step S140 in the PLC-based intelligent control method for a thermal power plant according to an embodiment of the present disclosure. Specifically, in the embodiment of the present disclosure, as Figure 2 As shown, step S140: in the PLC-based fan speed controller, the flame temperature time series implicit association coding vector and the oxygen content time series implicit association coding vector are subjected to dynamic memory attention interaction to obtain a flame temperature-oxygen content time series fine-grained interaction coding vector, including: S210, calculating the flame temperature-oxygen content time series external knowledge optimized fine-grained feature interaction matrix between the flame temperature time series implicit association coding vector and the oxygen content time series implicit association coding vector; S220, based on the flame temperature-oxygen content time series external knowledge optimized fine-grained feature interaction matrix, the flame temperature time series implicit association coding vector and the oxygen content time series implicit association coding vector are modulated and fused to obtain the flame temperature-oxygen content time series fine-grained interaction coding vector.

[0066] Specifically, in step S210, the flame temperature-oxygen content time series external knowledge optimized fine-grained feature interaction matrix between the flame temperature time series implicit association coding vector and the oxygen content time series implicit association coding vector is calculated, including:

[0067] Fine-grained interaction is performed on the flame temperature time series implicit correlation coding vector and the oxygen content time series implicit correlation coding vector to obtain a flame temperature-oxygen content time series fine-grained feature interaction matrix. This process can be expressed as follows:

[0068]

[0069] Among them, V1 is the time-series implicit correlation coding vector of the flame temperature, V2 is the time-series implicit correlation coding vector of the oxygen content, and V2 T is the transposed vector of V2, is matrix multiplication, and M p is the time-series fine-grained feature interaction matrix of flame temperature-oxygen content;

[0070] Performing external knowledge attention interaction on the time-series fine-grained feature interaction matrix of flame temperature-oxygen content to obtain the time-series external knowledge optimized fine-grained feature interaction matrix of flame temperature-oxygen content, and this process can be expressed by the formula:

[0071]

[0072] Among them, M p is the time-series fine-grained feature interaction matrix of flame temperature-oxygen content, M k and M v respectively represent learnable memory parameter matrices, M k T is the transposed matrix of M k , norm(·) represents the normalization function, and M y is the time-series external knowledge optimized fine-grained feature interaction matrix of flame temperature-oxygen content.

[0073] It should be understood that in order to capture and mine the fine-grained interaction information between the time-series implicit correlation coding vector of the flame temperature and the time-series implicit correlation coding vector of the oxygen content, the fine-grained interaction is performed on the time-series implicit correlation coding vector of the flame temperature and the time-series implicit correlation coding vector of the oxygen content to reveal the possible deep and non-linear dynamic interaction patterns between the flame temperature and the oxygen content, and obtain the time-series fine-grained feature interaction matrix of flame temperature-oxygen content.

[0074] Correspondingly, performing external knowledge attention interaction on the time-series fine-grained feature interaction matrix of flame temperature-oxygen content to utilize external knowledge (such as physical and chemical principles, expert rules, historical data, etc.) to embed the professional knowledge in these fields into the feature interaction, so that the model can better understand the complex physical and chemical reactions during the combustion process, and obtain a more expressive and informative time-series external knowledge optimized fine-grained feature interaction matrix of flame temperature-oxygen content.

[0075] Figure 3 is the flowchart of step S220 in the intelligent control method of a thermal power plant based on a PLC according to an embodiment of the present disclosure. Specifically, in the embodiment of the present disclosure, as Figure 3As shown, step S220: based on the flame temperature-oxygen content time series external knowledge optimization fine-grained feature interaction matrix, the flame temperature time series implicit association coding vector and the oxygen content time series implicit association coding vector are modulated and fused to obtain the flame temperature-oxygen content time series fine-grained interaction coding vector, including: S310, the flame temperature-oxygen content time series external knowledge optimization fine-grained feature interaction matrix is respectively associated with the flame temperature time series implicit association coding vector and the oxygen content time series implicit association coding vector to obtain the optimized flame temperature time series implicit association coding vector and the optimized oxygen content time series implicit association coding vector; S320, the position point division between the optimized flame temperature time series implicit association coding vector and the optimized oxygen content time series implicit association coding vector is calculated to obtain the flame temperature-oxygen content time series fine-grained interaction coding vector.

[0076] More specifically, in the embodiment of the present disclosure, in step S310, the flame temperature-oxygen content time series external knowledge optimized fine-grained feature interaction matrix is respectively associated with the flame temperature time series implicit association coding vector and the oxygen content time series implicit association coding vector to obtain an optimized flame temperature time series implicit association coding vector and an optimized oxygen content time series implicit association coding vector, including:

[0077] The flame temperature time series implicit association coding vector is linearly transformed to obtain a flame temperature query feature vector and a flame temperature value feature vector, and the flame temperature-oxygen content time series external knowledge is used to optimize the fine-grained feature interaction matrix as a key matrix. The flame temperature query feature vector, the flame temperature value feature vector, and the key matrix are fine-grained modulated by a converter to obtain the optimized flame temperature time series implicit association coding vector. This process can be expressed as follows:

[0078]

[0079] Among them, V1 is the flame temperature time series implicit correlation coding vector, M y is the flame temperature-oxygen content time series external knowledge optimization fine-grained feature interaction matrix, W 1q and b 1q are the flame temperature query embedding matrix and the flame temperature first bias vector, V 1q is the flame temperature query feature vector, W 1v and b 1v are the flame temperature embedding matrix and the second bias vector of flame temperature, V 1v is the flame temperature characteristic vector, M y T M yThe transposed matrix of, d1 is V 1q The length of, softmax(·) is the softmax function, and V1′ is the optimized implicit correlation encoding vector of the flame temperature time series;

[0080] Perform a linear transformation on the implicit correlation encoding vector of the oxygen content time series to obtain an oxygen content query feature vector and an oxygen content value feature vector, and use the fine-grained feature interaction matrix optimized by the flame temperature-oxygen content time series external knowledge as the key matrix. Perform fine-grained modulation of the converter on the oxygen content query feature vector, the oxygen content value feature vector, and the key matrix to obtain the optimized implicit correlation encoding vector of the oxygen content time series. This process can be expressed by the formula:

[0081]

[0082] Among them, V2 is the implicit correlation encoding vector of the oxygen content time series, and M y is the fine-grained feature interaction matrix optimized by the flame temperature-oxygen content time series external knowledge, softmax(·) is the softmax function, W 2q and b 2q are the oxygen content query embedding matrix and the first oxygen content bias vector respectively, V 2q is the oxygen content query feature vector, W 2v and b 2v are the oxygen content value embedding matrix and the second oxygen content bias vector respectively, V 2v is the oxygen content value feature vector, d2 is the length of V 2q and V2′ is the optimized implicit correlation encoding vector of the oxygen content time series.

[0083] It should be understood that a linear transformation is performed on the implicit correlation encoding vector of the flame temperature time series to obtain a flame temperature query feature vector and a flame temperature value feature vector, and the fine-grained feature interaction matrix optimized by the external knowledge integrated with the external knowledge is used as the key matrix, and fine-grained modulation of the converter is performed on it. It can be understood that through the self-attention layer in the converter, the interaction between the flame temperature query feature vector and the key matrix can achieve a very fine match, realizing the information interaction and integration between the internal and external knowledge of the flame temperature feature, ensuring that the features at each time point are fully considered, and ensuring that the flame temperature feature can make full use of the external knowledge to obtain advantages, and obtaining the optimized implicit correlation encoding vector of the flame temperature time series.

[0084] Similarly, a linear transformation is performed on the time-series implicit correlation encoding vector of the oxygen content to obtain an oxygen content query feature vector and an oxygen content value feature vector, and the same processing flow is performed on them with the external knowledge optimized fine-grained feature interaction matrix, so as to utilize the features obtained from the external knowledge of the oxygen content to enhance the representation ability of the oxygen content feature and obtain an optimized time-series implicit correlation encoding vector of the oxygen content.

[0085] In step S320, the element-wise division is calculated between the optimized time-series implicit correlation encoding vector of the flame temperature and the optimized time-series implicit correlation encoding vector of the oxygen content to obtain the time-series fine-grained interaction encoding vector of the flame temperature - oxygen content.

[0086] The processing process of the above step S320 can be expressed by the following formula:

[0087]

[0088] where V1′ is the optimized time-series implicit correlation encoding vector of the flame temperature, V2′ is the optimized time-series implicit correlation encoding vector of the oxygen content, and V 12 is the time-series fine-grained interaction encoding vector of the flame temperature - oxygen content.

[0089] That is, calculate the element-wise division between the optimized time-series implicit correlation encoding vector of the flame temperature and the optimized time-series implicit correlation encoding vector of the oxygen content to highlight the relative changes of the flame temperature and the oxygen content at each time point, so as to more sensitively respond to the subtle changes during the combustion process, and obtain the final time-series fine-grained interaction encoding vector of the flame temperature - oxygen content, thereby making a more refined adjustment.

[0090] In step S150, in the PLC-based fan speed controller, based on the time-series fine-grained interaction encoding vector of the flame temperature - oxygen content, an ideal oxygen content estimation value is obtained. Specifically, in the embodiment of the present disclosure, obtaining an ideal oxygen content estimation value based on the time-series fine-grained interaction encoding vector of the flame temperature - oxygen content includes: inputting the time-series fine-grained interaction encoding vector of the flame temperature - oxygen content into an ideal oxygen content dynamic optimizer based on a decoder to obtain the ideal oxygen content estimation value. That is, decoding processing is performed on the time-series fine-grained interaction encoding vector of the flame temperature - oxygen content obtained by performing dynamic fine-grained interaction using the time-series implicit correlation encoding vector of the flame temperature and the time-series implicit correlation encoding vector of the oxygen content, so as to intelligently obtain the ideal oxygen content estimation value.

[0091] More specifically, in the embodiments of the present disclosure, inputting the fine-grained interactive coding vector of the flame temperature-oxygen content time series into the decoder-based ideal oxygen content dynamic optimizer to obtain the ideal oxygen content estimation value includes: multiplying the decoding weight matrix of the decoder by the fine-grained interactive coding vector of the flame temperature-oxygen content time series to obtain a decoded fine-grained interactive coding vector of the flame temperature-oxygen content time series, and calculating the position-by-position sum value of all eigenvalues in the decoded fine-grained interactive coding vector of the flame temperature-oxygen content time series to obtain the ideal oxygen content estimation value.

[0092] In step S160, in the PLC-based fan speed controller, based on the ideal oxygen content estimation value, a fan speed adjustment instruction is obtained. Specifically, in the embodiments of the present disclosure, obtaining a fan speed adjustment instruction based on the ideal oxygen content estimation value includes: extracting the oxygen content value at the current time point from the time queue of the oxygen content values, and calculating the difference between the oxygen content value at the current time point and the ideal oxygen content estimation value to obtain an oxygen content error; inputting the oxygen content error into a PID controller to obtain the fan speed adjustment instruction. Correspondingly, in order to be able to understand in real time the deviation between the current oxygen concentration in the combustion chamber and the ideal oxygen content estimation value, so as to better control the air (oxygen) supply during the combustion process, thereby optimizing the combustion efficiency, in the technical solution of the present disclosure, the oxygen content value at the current time point is extracted from the time queue of the oxygen content values, and the difference between the oxygen content value at the current time point and the ideal oxygen content estimation value is calculated to obtain an oxygen content error. In this way, the oxygen content can be kept within the ideal range, and it can be ensured that the fuel burns fully, maximizing the energy conversion efficiency and reducing unnecessary energy waste. It should be understood that the adjustment of the fan speed is an important link in the combustion control system, which directly affects the ratio of fuel to air in the combustion chamber, that is, the air-fuel ratio. By precisely adjusting the fan speed, it can be ensured that the amount of air entering the combustion chamber is appropriate, especially the amount of oxygen, so that the fuel can burn fully and maximize the energy conversion efficiency. It can be understood that the PID controller (Proportional-Integral-Derivative controller) is a control algorithm widely used in industrial automation, which can dynamically adjust the output according to the error and quickly respond to any deviation, so as to ensure that the system is always close to the ideal set value.

[0093] Particularly, in a specific embodiment of the present disclosure, the input of the oxygen content error into the PID controller to obtain the fan speed adjustment instruction can be implemented in the following manner:

[0094] The PID controller is a control algorithm widely used in the field of industrial automation. It can dynamically adjust the output according to the error, quickly respond to any deviation, and thus ensure that the system always approaches the ideal set value. Specifically, the PID controller consists of three main parts - Proportional (P), Integral (I), and Derivative (D). They each handle different aspects of the error and work together to achieve the optimal regulation of the fan speed.

[0095] In the Proportional (P) part, the controller directly multiplies the current error value to generate an output signal proportional to the magnitude of the error. The advantage of this method is its rapid response, being able to immediately respond to large errors. However, the drawback is that it cannot completely eliminate the steady-state error, that is, the small deviation that still exists when the system reaches a stable state. To solve this problem, the Integral (I) part is introduced. The Integral part calculates the result of the cumulative error over time and adds the result to the output signal. This enables even small persistent errors to gradually accumulate until the system returns to the desired state. However, the integral action may also lead to overshoot, that is, the system responds excessively and exhibits oscillations. Therefore, in practical applications, it is necessary to carefully set the integral coefficient to balance the relationship between the response speed and stability.

[0096] Finally, the Derivative (D) part focuses on the rate of change of the error. It calculates the first derivative of the error, predicts the possible future trends, and takes measures in advance to suppress them. The derivative action helps to reduce the overshoot and smooth the transition process, but it may amplify the effects in the face of noise interference or high-frequency fluctuating data. Therefore, when designing a PID controller, it is usually necessary to weigh the roles of proportional, integral, and derivative in combination with the actual situation to find the parameter combination most suitable for the current application scenario.

[0097] When the oxygen content error is input into the PID controller, the above three links work together, comprehensively considering the current error, historical cumulative error, and error change trend, and calculate an appropriate adjustment amount. This adjustment amount is then converted into a specific fan speed adjustment instruction to change the air flow rate into the combustion chamber. In actual operation, the fan speed adjustment instruction usually appears as a specific value or range, which can directly correspond to the speed set value of the fan motor or a proportional factor for increasing or decreasing the existing speed. For example, if the system detects that the oxygen content at the current time point is lower than the estimated value of the ideal oxygen content, it means that there is insufficient oxygen in the combustion chamber, which may lead to incomplete combustion. At this time, the generated fan speed adjustment instruction may indicate an increase in the fan speed to supply more air to the combustion chamber, thereby increasing the oxygen concentration. Conversely, if the oxygen content is too high, it may indicate a decrease in the fan speed to reduce the supply of excess air and avoid unnecessary energy waste. To ensure the effectiveness and safety of the instruction, a series of rules and limiting conditions, such as the maximum allowable speed and the minimum safety interval, are generally predefined inside the PLC to prevent equipment damage or safety accidents caused by misoperations in extreme situations.

[0098] To achieve this goal, the fan speed adjustment instruction needs to be transmitted to the variable frequency drive (VFD) of the fan through an appropriate communication protocol. The VFD is an electronic device that can receive instructions from the PLC and adjust the AC frequency and voltage output to the fan motor accordingly, thereby controlling the motor speed. This adjustment method can not only precisely control the operating speed of the fan but also effectively protect the motor from overload or other abnormal conditions.

[0099] Through the effective implementation of the above measures, thermal power plants can manage the combustion process on a more refined and scientific basis, thereby promoting the green transformation and development of the entire industry.

[0100] In summary, the intelligent control method for a thermal power plant based on PLC according to the embodiments of the present disclosure is elucidated. It collects the time queues of the flame temperature values and the oxygen content values through the sensor assembly and transmits them to the PLC-based fan speed controller. In the PLC-based fan speed controller, data analysis and coding techniques based on artificial intelligence are used to perform temporal implicit association on the flame temperature values and the oxygen content values. Based on this, the ideal oxygen content estimation value is intelligently obtained according to the dynamic memory interaction information between the temporal implicit association features of the flame temperature and the temporal implicit association features of the oxygen content, and the fan speed adjustment instruction is automatically generated based on the oxygen content error from the current time point. By analyzing the temporal information of the flame temperature and the oxygen content in real time, the present disclosure can quickly and intelligently identify the optimal combustion conditions. Compared with the traditional method, it can dynamically adjust the fan speed according to real-time data, thereby helping to improve the combustion efficiency and maintain its optimal state.

[0101] Figure 4 FIG. is a block diagram of an intelligent control system for a thermal power plant based on PLC according to an embodiment of the present disclosure. As Figure 4As shown, the intelligent control system 100 of a thermal power plant based on PLC according to an embodiment of the present disclosure includes: a flame temperature and oxygen content data acquisition module 110, configured to collect a time queue of flame temperature values and a time queue of oxygen content values through a sensor assembly; a flame temperature and oxygen content data transmission module 120, configured to transmit the time queue of the flame temperature values and the time queue of the oxygen content values to a PLC-based fan speed controller; a flame temperature and oxygen content data encoding module 130, configured to perform sequential implicit correlation encoding on the time queue of the flame temperature values and the time queue of the oxygen content values respectively in the PLC-based fan speed controller to obtain a flame temperature sequential implicit correlation encoding vector and an oxygen content sequential implicit correlation encoding vector; a flame temperature and oxygen content data interaction module 140, configured to perform dynamic memory attention interaction on the PLC-based fan speed controller for the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector to obtain a flame temperature-oxygen content sequential fine-grained interaction encoding vector, wherein the flame temperature and oxygen content data interaction module includes: an external knowledge optimization unit, configured to calculate a flame temperature-oxygen content sequential external knowledge optimization fine-grained feature interaction matrix between the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector; a feature modulation and fusion unit, configured to perform modulation and fusion on the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector based on the flame temperature-oxygen content sequential external knowledge optimization fine-grained feature interaction matrix to obtain the flame temperature-oxygen content sequential fine-grained interaction encoding vector; an ideal oxygen content estimation module 150, configured to obtain an ideal oxygen content estimation value in the PLC-based fan speed controller based on the flame temperature-oxygen content sequential fine-grained interaction encoding vector; and a fan speed adjustment instruction generation module 160, configured to obtain a fan speed adjustment instruction in the PLC-based fan speed controller based on the ideal oxygen content estimation value.

[0102] Here, those skilled in the art can understand that the specific operations of each step in the above intelligent control system of a thermal power plant based on PLC have been introduced in detail in the description of the Figures 1 to 3 intelligent control method of a thermal power plant based on PLC above, and thus, the repeated description thereof will be omitted.

[0103] As described above, the intelligent control system 100 of a thermal power plant based on PLC according to an embodiment of the present disclosure can be implemented in various wireless terminals, such as a server having an intelligent control algorithm for a thermal power plant based on PLC. In a possible implementation manner, the intelligent control system 100 of a thermal power plant based on PLC according to an embodiment of the present disclosure can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the intelligent control system 100 of a thermal power plant based on PLC can be a software module in the operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the intelligent control system 100 of a thermal power plant based on PLC can also be one of the numerous hardware modules of the wireless terminal.

[0104] Alternatively, in another example, the intelligent control system 100 of a thermal power plant based on PLC and the wireless terminal can also be separate devices, and the intelligent control system 100 of a thermal power plant based on PLC can be connected to the wireless terminal through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0105] Those of ordinary skill in the art can understand that the above embodiments are specific implementation manners for implementing the present disclosure, and in actual applications, various changes can be made to them in form and details without departing from the spirit and scope of the present disclosure.

Claims

1. An intelligent control method for a thermal power plant based on PLC, characterized in that, Including: A time queue of flame temperature values and a time queue of oxygen content values collected by a sensor component; Transmitting the time queue of the flame temperature values and the time queue of the oxygen content values to a PLC-based fan speed controller; In the PLC-based fan speed controller, performing sequential implicit correlation encoding on the time queue of the flame temperature values and the time queue of the oxygen content values respectively to obtain a flame temperature sequential implicit correlation encoding vector and an oxygen content sequential implicit correlation encoding vector; In the PLC-based fan speed controller, performing dynamic memory attention interaction on the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector to obtain a flame temperature-oxygen content sequential fine-grained interaction encoding vector, including: calculating a flame temperature-oxygen content sequential external knowledge optimized fine-grained feature interaction matrix between the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector; Based on the flame temperature-oxygen content sequential external knowledge optimized fine-grained feature interaction matrix, modulating and fusing the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector to obtain the flame temperature-oxygen content sequential fine-grained interaction encoding vector; In the PLC-based fan speed controller, obtaining an ideal oxygen content estimated value based on the flame temperature-oxygen content sequential fine-grained interaction encoding vector; In the PLC-based fan speed controller, obtaining a fan speed adjustment instruction based on the ideal oxygen content estimated value.

2. The intelligent control method for a thermal power plant based on PLC according to claim 1, wherein Performing sequential implicit correlation encoding on the time queue of the flame temperature values and the time queue of the oxygen content values respectively to obtain a flame temperature sequential implicit correlation encoding vector and an oxygen content sequential implicit correlation encoding vector, including: respectively inputting the time queue of the flame temperature values and the time queue of the oxygen content values into a sequence encoder based on a forward LSTM model to obtain the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector.

3. The intelligent control method for a thermal power plant based on PLC according to claim 2, wherein, Calculating a flame temperature-oxygen content sequential external knowledge optimized fine-grained feature interaction matrix between the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector, including: Performing fine-grained interaction on the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector to obtain a flame temperature-oxygen content sequential fine-grained feature interaction matrix; Performing external knowledge attention interaction on the flame temperature-oxygen content sequential fine-grained feature interaction matrix to obtain the flame temperature-oxygen content sequential external knowledge optimized fine-grained feature interaction matrix.

4. The intelligent control method for a thermal power plant based on PLC according to claim 3, wherein Based on the flame temperature-oxygen content sequential external knowledge optimized fine-grained feature interaction matrix, modulating and fusing the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector to obtain the flame temperature-oxygen content sequential fine-grained interaction encoding vector, including: Perform temporal fine-grained association of the optimized fine-grained feature interaction matrix of the flame temperature-oxygen content time series external knowledge with the hidden correlation encoding vector of the flame temperature time series and the hidden correlation encoding vector of the oxygen content time series respectively to obtain an optimized hidden correlation encoding vector of the flame temperature time series and an optimized hidden correlation encoding vector of the oxygen content time series; Calculate the element-wise division between the optimized hidden correlation encoding vector of the flame temperature time series and the optimized hidden correlation encoding vector of the oxygen content time series to obtain the fine-grained interaction encoding vector of the flame temperature-oxygen content time series.

5. The intelligent control method for a thermal power plant based on PLC according to claim 4, characterized in that, Performing temporal fine-grained association of the optimized fine-grained feature interaction matrix of the flame temperature-oxygen content time series external knowledge with the hidden correlation encoding vector of the flame temperature time series and the hidden correlation encoding vector of the oxygen content time series respectively to obtain an optimized hidden correlation encoding vector of the flame temperature time series and an optimized hidden correlation encoding vector of the oxygen content time series, includes: Perform a linear transformation on the hidden correlation encoding vector of the flame temperature time series to obtain a flame temperature query feature vector and a flame temperature value feature vector, and use the optimized fine-grained feature interaction matrix of the flame temperature-oxygen content time series external knowledge as the key matrix. Perform fine-grained modulation of the converter on the flame temperature query feature vector, the flame temperature value feature vector, and the key matrix to obtain the optimized hidden correlation encoding vector of the flame temperature time series; Perform a linear transformation on the hidden correlation encoding vector of the oxygen content time series to obtain an oxygen content query feature vector and an oxygen content value feature vector, and use the optimized fine-grained feature interaction matrix of the flame temperature-oxygen content time series external knowledge as the key matrix. Perform the fine-grained modulation of the converter on the oxygen content query feature vector, the oxygen content value feature vector, and the key matrix to obtain the optimized hidden correlation encoding vector of the oxygen content time series.

6. The intelligent control method for a thermal power plant based on PLC according to claim 5, characterized in that, Based on the fine-grained interaction encoding vector of the flame temperature-oxygen content time series, obtain an estimated value of the ideal oxygen content, including: input the fine-grained interaction encoding vector of the flame temperature-oxygen content time series into an ideal oxygen content dynamic optimizer based on a decoder to obtain the estimated value of the ideal oxygen content.

7. The intelligent control method for a thermal power plant based on PLC according to claim 6, characterized in that, Input the fine-grained interaction encoding vector of the flame temperature-oxygen content time series into an ideal oxygen content dynamic optimizer based on a decoder to obtain the estimated value of the ideal oxygen content, including: multiply the decoding weight matrix of the decoder by the fine-grained interaction encoding vector of the flame temperature-oxygen content time series to obtain a decoded fine-grained interaction encoding vector of the flame temperature-oxygen content time series, and calculate the element-wise sum of all eigenvalues in the decoded fine-grained interaction encoding vector of the flame temperature-oxygen content time series to obtain the estimated value of the ideal oxygen content.

8. The intelligent control method for a thermal power plant based on PLC according to claim 7, characterized in that Based on the estimated value of the ideal oxygen content, obtain a fan speed adjustment instruction, including: Extract the oxygen content value at the current time point from the time queue of the oxygen content value, and calculate the difference between the oxygen content value at the current time point and the estimated value of the ideal oxygen content to obtain an oxygen content error; Input the oxygen content error into the PID controller to obtain the fan speed adjustment command.

9. An intelligent control system for a thermal power plant based on PLC, characterized in that, It includes: A flame temperature and oxygen content data acquisition module, which is used to collect the time queue of the flame temperature value and the time queue of the oxygen content value through the sensor assembly; A flame temperature and oxygen content data transmission module, which is used to transmit the time queue of the flame temperature value and the time queue of the oxygen content value to the fan speed controller based on PLC; A flame temperature and oxygen content data encoding module, which is used to perform sequential implicit correlation encoding on the time queue of the flame temperature value and the time queue of the oxygen content value respectively in the fan speed controller based on PLC to obtain the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector; A flame temperature and oxygen content data interaction module, which is used to perform dynamic memory attention interaction on the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector in the fan speed controller based on PLC to obtain the flame temperature-oxygen content sequential fine-grained interaction encoding vector. Among them, the flame temperature and oxygen content data interaction module includes: an external knowledge optimization unit, which is used to calculate the flame temperature-oxygen content sequential external knowledge optimization fine-grained feature interaction matrix between the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector; a feature modulation and fusion unit, which is used to perform modulation and fusion on the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector based on the flame temperature-oxygen content sequential external knowledge optimization fine-grained feature interaction matrix to obtain the flame temperature-oxygen content sequential fine-grained interaction encoding vector; An ideal oxygen content estimation module, which is used to obtain the ideal oxygen content estimation value in the fan speed controller based on PLC based on the flame temperature-oxygen content sequential fine-grained interaction encoding vector; A fan speed adjustment command generation module, which is used to obtain the fan speed adjustment command in the fan speed controller based on PLC based on the ideal oxygen content estimation value.

10. The intelligent control system for a thermal power plant based on a PLC according to claim 9, characterized in that, The flame temperature and oxygen content data encoding module is also used to: input the time queue of the flame temperature value and the time queue of the oxygen content value into the sequence encoder based on the forward LSTM model respectively to obtain the flame temperature sequential implicit correlation encoding vector and the oxygen content sequential implicit correlation encoding vector.

Images