Boiler start-stop control method and device

Abstract

This application relates to the field of boiler technology, and particularly to a boiler start-up and shutdown control method and device. The method includes: extracting equipment feature information corresponding to the boiler equipment and the user's operational intent from the voice information based on the text information corresponding to the user's voice information; generating a feedforward pulse compensation command corresponding to the frequency converter output sequence of the boiler equipment's coal feeder based on the equipment feature information and operational intent; optimizing the frequency converter output parameters of the boiler equipment's coal feeder based on the feedforward pulse compensation command and the boiler equipment's real-time operating data to generate a start-up control command or a stop control command for the boiler equipment, and controlling the boiler equipment to execute the corresponding start-up or stop action according to the start-up control command or stop control command. This application can perform feedforward pulse compensation on the frequency converter output sequence of the coal feeder based on the translated text of the operator's voice alarm, thereby solving the problem of thermal collapse of the pulverizing system caused by the post-positioning of traditional signal threshold alarms.

Description

Technical Field

[0001] This application relates to the field of boiler technology, and in particular to a boiler start-up and shutdown control method and device. Background Technology

[0002] In related technologies, in the field of thermal power generation, the start-up and shutdown monitoring of boilers and their pulverizing systems usually relies on a preset sensor threshold alarm mechanism. The system only triggers protection or adjustment actions when the operating parameters exceed the critical value.

[0003] However, in related technologies, this reactive, passive response mode, which triggers protection or adjustment actions only after operating parameters exceed critical values, is unable to respond promptly to thermal fluctuations caused by sudden anomalies such as coal feeder failure or blockage. It also cannot parse and convert the voice and text provided by maintenance personnel, thus failing to provide pulse-type dynamic compensation for the frequency converter output of the coal feeder before the risk of thermal collapse of the pulverizing system is formed. This makes it difficult to effectively ensure the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown, and urgently needs to be addressed. Summary of the Invention

[0004] This application provides a boiler start-up and shutdown control method and device to solve the problems in related technologies where boiler start-up and shutdown monitoring is mostly a reactive, post-processing mode, which makes it difficult to respond in a timely manner to thermal fluctuations caused by sudden anomalies. It also fails to parse and convert the voice and text of maintenance personnel, thus making it impossible to perform pulse-type dynamic compensation on the frequency converter output of the coal feeder before the risk of thermal collapse of the pulverizing system is formed. This makes it difficult to effectively ensure the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown.

[0005] The first aspect of this application provides a boiler start-up and shutdown control method, comprising the following steps: responding to a user's voice information, extracting equipment feature information corresponding to the boiler equipment and operation intention corresponding to the user from the voice information based on the text information corresponding to the voice information; generating a feedforward pulse compensation command corresponding to the frequency converter output sequence of the boiler equipment's coal feeder based on the equipment feature information and the operation intention; optimizing the frequency converter output parameters of the boiler equipment's coal feeder based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment, so as to generate a start-up control command or a stop control command for the boiler equipment according to the optimized frequency converter output parameters, and controlling the boiler equipment to perform the corresponding start-up action or stop action according to the start-up control command or stop control command.

[0006] Optionally, in one embodiment of this application, the step of extracting equipment feature information corresponding to the boiler equipment and operation intent corresponding to the user from the text information corresponding to the voice information includes: extracting equipment number entity and anomaly type entity from the text information based on pre-set named entities, so as to determine the equipment feature information based on the equipment number entity and the anomaly type entity; and extracting fuzzy text and negative text from the text information based on the contextual features of the equipment feature information, so as to determine the user's operation intent based on the fuzzy text and the negative text.

[0007] Optionally, in one embodiment of this application, generating a feedforward pulse compensation command corresponding to the frequency converter output sequence of the coal feeder of the boiler equipment based on the equipment feature information and the operation intention includes: converting the equipment feature information and the operation intention into a semantic vector; determining the current fault intensity score of the boiler equipment based on the semantic vector; comparing the current fault intensity score with a preset compensation threshold, and generating the feedforward pulse compensation command when the comparison result is that the current fault intensity score is greater than the preset compensation threshold.

[0008] Optionally, in one embodiment of this application, optimizing the variable frequency output parameters of the coal feeder of the boiler equipment based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment includes: acquiring the original sensor data of the boiler and performing timestamp alignment processing on the original sensor data to obtain the real-time operating data; calculating the communication delay of the feedforward pulse compensation command based on the sending time and feedback time of the feedforward pulse compensation command; and optimizing the variable frequency output parameters of the coal feeder of the boiler equipment based on the real-time operating data when the communication delay is less than or equal to a preset allowable delay time threshold.

[0009] Optionally, in one embodiment of this application, optimizing the variable frequency output parameters of the coal feeder of the boiler equipment based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment includes: obtaining the real-time feedback current value of the frequency converter of the boiler equipment, and calculating the current deviation value between the preset current value of the frequency converter and the real-time feedback current value; generating an adjustment factor based on the ratio of the current deviation value to the preset deviation threshold, and determining the feedforward compensation strength of the frequency converter according to the adjustment factor, so as to optimize the variable frequency output parameters of the coal feeder according to the feedforward compensation strength.

[0010] Optionally, in one embodiment of this application, generating the start-up control command or stop-down control command of the boiler equipment based on the optimized variable frequency output parameters includes: acquiring the real-time hot air temperature and real-time pulverized coal concentration of the boiler equipment; calculating the stability prediction value of the boiler equipment when the real-time hot air temperature is less than a preset allowable hot air temperature threshold and the real-time pulverized coal concentration is greater than a preset pulverized coal concentration threshold; and generating the start-up control command or stop-down control command of the boiler equipment based on the stability prediction value and a preset instability threshold.

[0011] A second aspect of this application provides a boiler start-stop control device, comprising: an extraction module, configured to, in response to a user's voice information, extract equipment feature information corresponding to the boiler equipment and an operation intention corresponding to the user from the voice information based on text information corresponding to the voice information; a generation module, configured to, based on the equipment feature information and the operation intention, generate a feedforward pulse compensation command corresponding to the frequency converter output sequence of the boiler equipment's coal feeder; and a control module, configured to, based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment, optimize the frequency converter output parameters of the boiler equipment's coal feeder, so as to generate a start-up control command or a stop-down control command for the boiler equipment according to the optimized frequency converter output parameters, and control the boiler equipment to perform the corresponding start-up or stop-down action according to the start-up control command or the stop-down control command.

[0012] Optionally, in one embodiment of this application, the extraction module includes: a first extraction unit, configured to extract device number entity and anomaly type entity from the text information based on pre-defined named entities, so as to determine the device feature information based on the device number entity and the anomaly type entity; and a second extraction unit, configured to extract ambiguous text and negative text from the text information based on the contextual features of the device feature information, so as to determine the user's operation intention based on the ambiguous text and the negative text.

[0013] Optionally, in one embodiment of this application, the generation module includes: a conversion unit, configured to convert the device feature information and the operation intention into a semantic vector; a determination unit, configured to determine the current fault intensity score of the boiler equipment based on the semantic vector; and a comparison unit, configured to compare the current fault intensity score with a preset compensation threshold, and generate the feedforward pulse compensation command when the comparison result is that the current fault intensity score is greater than the preset compensation threshold.

[0014] Optionally, in one embodiment of this application, the control module includes: a processing unit, configured to acquire the original sensor data of the boiler and perform timestamp alignment processing on the original sensor data to obtain the real-time operating data; a first calculation unit, configured to calculate the communication delay of the feedforward pulse compensation command based on the sending time and feedback time of the feedforward pulse compensation command; and an optimization unit, configured to optimize the frequency conversion output parameters of the coal feeder of the boiler equipment based on the real-time operating data when the communication delay is less than or equal to a preset allowable delay time threshold.

[0015] Optionally, in one embodiment of this application, the control module includes: a first acquisition unit, configured to acquire the real-time feedback current value of the frequency converter of the boiler equipment, and calculate the current deviation value between the preset current value of the frequency converter and the real-time feedback current value; and a first generation unit, configured to generate an adjustment factor based on the ratio of the current deviation value to a preset deviation threshold, and determine the feedforward compensation strength of the frequency converter according to the adjustment factor, so as to optimize the frequency conversion output parameters of the coal feeder according to the feedforward compensation strength.

[0016] Optionally, in one embodiment of this application, the control module includes: a second acquisition unit, configured to acquire the real-time hot air temperature and real-time pulverized coal concentration of the boiler equipment; a second calculation unit, configured to calculate a stability prediction value of the boiler equipment when the real-time hot air temperature is less than a preset allowable hot air temperature threshold and the real-time pulverized coal concentration is greater than a preset pulverized coal concentration threshold; and a second generation unit, configured to generate an start control command or a stop control command for the boiler equipment based on the stability prediction value and a preset instability threshold.

[0017] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the boiler start-up and shutdown control method as described in the above embodiments.

[0018] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the boiler start-up and shutdown control method described above.

[0019] A fifth aspect of this application provides a computer program product, including a computer program that, when executed, is used to implement the boiler start-up and shutdown control method described above.

[0020] This application embodiment can parse the user's voice information, extract equipment abnormality features and operation intentions, generate a feedforward pulse compensation command corresponding to the frequency converter output sequence of the coal feeder, and then fuse the command with real-time operating data to adjust the frequency converter output parameters of the coal feeder. Based on the adjusted frequency converter output parameters, the boiler start-up and shutdown process is dynamically optimized to achieve thermal stability of the pulverizing system. This invention achieves timely response to thermal fluctuations caused by sudden anomalies by converting the voice alarm text of operators into feedforward pulse compensation commands. It provides pulse-based dynamic compensation to the frequency converter output of the coal feeder before the risk of thermal collapse of the pulverizing system materializes, effectively ensuring the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown. This solves the problems of related technologies where boiler start-up and shutdown monitoring is mostly a reactive, after-the-fact response mode, making it difficult to respond promptly to thermal fluctuations caused by sudden anomalies. Furthermore, it cannot parse and convert the voice text of maintenance personnel, thus failing to provide pulse-based dynamic compensation to the frequency converter output of the coal feeder before the risk of thermal collapse of the pulverizing system materializes, and thus failing to effectively ensure the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown.

[0021] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0022] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a boiler start-up and shutdown control method according to an embodiment of this application; Figure 2 This is a schematic diagram of the framework of a boiler start-up and shutdown intelligent monitoring system based on a large language model according to an embodiment of this application; Figure 3 A flowchart illustrating an embodiment of the intelligent monitoring method for boiler start-up and shutdown based on a large language model, according to this application; Figure 4 This is a schematic diagram of the boiler start-up and shutdown control device provided according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application.

[0023] Figure label: 10- Boiler start-stop control device; 100- Extraction module, 200- Generation module and 300- Control module; 501- Memory, 502- Processor and 503- Communication interface. Detailed Implementation

[0024] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0025] The following description, with reference to the accompanying drawings, illustrates a boiler start-up and shutdown control method and apparatus according to embodiments of this application. Addressing the issues mentioned in the background art regarding the difficulty in timely responding to thermal fluctuations caused by sudden anomalies in boiler start-up and shutdown monitoring, and the inability to parse and convert the voice and text of maintenance personnel, thus failing to provide pulse-type dynamic compensation for the feeder frequency converter output before the risk of thermal collapse in the pulverizing system arises, and thus failing to effectively ensure the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown, this application provides a boiler start-up and shutdown control method. In this method, user voice information can be parsed to extract equipment anomaly characteristics and operational intentions to generate a feedforward pulse compensation command corresponding to the feeder frequency converter output sequence. This command is then fused with real-time operating data to adjust the feeder frequency converter output parameters. Based on the adjusted frequency converter output parameters, the boiler start-up and shutdown process is dynamically optimized to achieve thermal stability of the pulverizing system. This technology converts operator voice alarms into feedforward pulse compensation commands, enabling timely responses to thermal fluctuations caused by sudden anomalies. It provides pulse-based dynamic compensation to the feeder frequency converter output before the risk of thermal collapse in the pulverizing system materializes, effectively ensuring the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown. This addresses the problems of existing technologies where boiler start-up and shutdown monitoring is often reactive and reactive, making it difficult to respond promptly to thermal fluctuations caused by sudden anomalies. Furthermore, these technologies cannot parse and convert operator voice alarms, thus hindering the provision of pulse-based dynamic compensation to the feeder frequency converter output before the risk of thermal collapse in the pulverizing system materializes, and ultimately failing to effectively guarantee the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown.

[0026] Specifically, Figure 1 This is a flowchart of a boiler start-up and shutdown control method provided in an embodiment of this application.

[0027] like Figure 1 As shown, the boiler start-up and shutdown control method includes the following steps: In step S101, in response to the user's voice information, based on the text information corresponding to the voice information, the equipment feature information corresponding to the boiler equipment and the operation intention corresponding to the user are extracted from the voice information.

[0028] Understandably, in the actual operation of boiler equipment, relevant maintenance personnel will promptly observe the boiler's operating status. In the actual operation of thermal power plants, when an anomaly occurs in the pulverizing system, experienced maintenance personnel will often issue alarms via voice, such as "#4 coal mill vibration abnormal, suspected coal shortage," etc.

[0029] Traditional boiler start-up and shutdown monitoring methods, which trigger protection or adjustment actions only after operating parameters exceed critical values, are mostly reactive and reactive. This approach struggles to respond promptly to thermal fluctuations caused by sudden anomalies such as coal feeder failure or blockage. Especially when maintenance personnel convey operational experience and vague predictions via voice alarms, traditional monitoring systems cannot parse and convert this context-rich text. Consequently, they cannot provide pulse-like dynamic compensation for the feeder's frequency converter output before the risk of thermal collapse in the pulverizing system materializes, making it difficult to effectively ensure the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown. In some embodiments, the present application embodiments may first set up certain sound pickup devices in the boiler monitoring system to collect the voice signals given by the operation and maintenance personnel, and use speech recognition technology to convert them into text format to obtain the text information corresponding to the voice signal.

[0030] Subsequently, this embodiment of the application can invoke a pre-trained large language model to perform deep semantic analysis on the translated text, so as to perform semantic analysis and fault classification on the translated text of the voice alarm from the maintenance personnel, and extract the equipment feature information corresponding to the boiler equipment and the operation intent corresponding to the user. Here, the user refers to the maintenance personnel of the boiler equipment.

[0031] Pre-training is a strategy for training deep learning models. Its core lies in using large-scale datasets to initially train the model, enabling it to learn general feature representations. This process is similar to the basic learning stage humans undergo before learning new knowledge, accumulating experience through extensive reading and observation.

[0032] A pre-trained large language model generally refers to a language model training task designed based on a large-scale corpus (including language training materials such as sentences and paragraphs), and a large-scale neural network algorithm structure trained to learn and implement it. The final large-scale neural network algorithm structure and parameters are the pre-trained language model. Subsequent tasks can be carried out on the basis of this model to extract features or fine-tune the task to achieve specific task objectives.

[0033] The idea behind pre-training is to first train a set of model parameters on a single task, then use these parameters to initialize the network model parameters, and finally use the initialized network model to train on other tasks, resulting in models adapted for those tasks. By pre-training on large-scale corpora, neural language representation models can learn powerful language representation capabilities, extracting rich syntactic and semantic information from text.

[0034] Pre-trained large language models can provide tokens containing rich semantic information and sentence-level features for downstream tasks. They can also be fine-tuned directly on the pre-trained model for downstream tasks, making it convenient and quick to obtain downstream-specific models.

[0035] The neural network algorithm structure used to train the pre-trained language model can be CNN, RNN, LSTM, etc., or it can be a model built with attention networks, such as Transformer, BERT, GPT, Clip, etc. This application does not limit it. An attention network is a network model that uses an attention mechanism for training. This model extracts more important feature information from the input sequence by assigning different weights to each part of the input sequence, so that the model can finally obtain a more accurate output.

[0036] Optionally, in one embodiment of this application, based on the text information corresponding to the voice information, extracting equipment feature information corresponding to the boiler equipment and operation intention corresponding to the user from the voice information includes: extracting equipment number entity and anomaly type entity from the text information based on pre-set named entities, so as to determine equipment feature information based on equipment number entity and anomaly type entity; and extracting fuzzy text and negative text from the text information based on the contextual features of the equipment feature information, so as to determine the user's operation intention based on fuzzy text and negative text.

[0037] In some embodiments, during the semantic parsing and fault classification process, this application further refines the deep analysis method of speech-to-text, and achieves accurate understanding of the speech information of operation and maintenance personnel through a triple mechanism of named entity recognition, context feature extraction and sentiment analysis.

[0038] Specifically, the semantic parsing process in this application embodiment includes, but is not limited to, the following sub-steps: First, the translated text is segmented and part-of-speech tagged to identify key information units. Second, named entity recognition technology is used to mark entity equipment feature information such as equipment number (e.g., "#4 coal mill"), abnormal phenomena (e.g., "abnormal vibration"), and fault prediction (e.g., "suspected coal shortage") from the text.

[0039] Meanwhile, embodiments of this application can construct contextual feature vectors to analyze the operational intent contained in the text information corresponding to the voice information. For example, when words such as "abnormal vibration" or "coal shortage" appear, it can be determined that the maintenance personnel intend to indicate insufficient coal supply and that the coal feeding rate needs to be increased to prevent the coal mill from grinding dry or the boiler from burning unstablely.

[0040] For example, when extracting equipment feature information corresponding to boiler equipment from voice information based on text information corresponding to voice information, this application can first extract equipment number entity and anomaly type entity from text information based on pre-set named entities, so as to determine equipment feature information related to boiler equipment based on equipment number entity and anomaly type entity.

[0041] In practical implementation, the embodiments of this application can first perform named entity recognition on the speech-translated text. For example, using a variety of pre-defined named entities, a certain named entity recognition model (such as a BiLSTM-CRF model or a Transformer-based model) can be trained to obtain a pre-trained named entity recognition model, which can then be integrated into a pre-trained large language model.

[0042] Therefore, in practical applications, the embodiments of this application can utilize the pre-trained named entity recognition model to perform sequence annotation on the text, marking the equipment number entity and the anomaly type entity in the voice information of maintenance personnel, so as to determine the equipment feature information of the boiler equipment.

[0043] For example, given the text message "#4 coal mill vibration abnormal, suspected coal shortage", the pre-trained named entity recognition model can identify "#4 coal mill" as the equipment number entity, and "vibration abnormal" and "coal shortage" as the abnormality type entities. This entity information is stored in a structured manner for subsequent fault localization.

[0044] Furthermore, based on the contextual features of these device characteristic information, embodiments of this application can extract ambiguous and negative text from the text information to determine the user's operational intent based on the ambiguous and negative text.

[0045] For example, this application can extract the contextual features of these equipment characteristic information (equipment number entity and anomaly type entity in voice information), and pay particular attention to statements containing ambiguous expressions. This is because power plant maintenance personnel often use uncertain words such as "suspected," "possibly," and "probably" in emergency situations, and these words contain important predictive information.

[0046] Therefore, embodiments of this application can construct a fuzzy lexicon and dependency parsing in a pre-trained large language model to determine whether the text contains vague textual expressions such as "suspected coal shortage" or "possible blockage". When such vague textual expressions are detected, embodiments of this application can mark them as "pending confirmation" and appropriately lower the action threshold or add a manual confirmation step in subsequent control decisions to maintain vigilance while avoiding misjudgment.

[0047] For example, before practical application, this application can pre-construct a fuzzy lexicon. This lexicon can pre-include words with uncertain semantics such as "suspected," "possible," "probably," "maybe," "seems like," and "perhaps," and label each fuzzy word with its part of speech and semantic category. Secondly, dependency parsing is used to parse the text after speech recognition and translation, generating a dependency parsing tree with words as nodes and syntactic relations as edges. By identifying the dependency relations between fuzzy words and other words (for example, "suspected" as an adverb forms an "advmod" modifying relationship with "coal shortage," and "possible" as...), the application can further analyze the fuzzy lexicon. To determine the target entity or event description modified by the fuzzy word, the auxiliary verb is used to form an "aux" or "xcomp" relationship with "blockage". Furthermore, this application can combine the attention mechanism and context embedding representation in the large language model to jointly extract the fuzzy word and its dependent objects. When the word in the fuzzy word library matches the dependency relationship successfully and the target entity belongs to the equipment abnormality type (such as "coal shortage", "blockage", "abnormal vibration", etc.), it is determined that the text contains a fuzzy text expression, and the abnormal entity corresponding to the expression and its degree of fuzziness confidence are marked for subsequent extraction of operational intent.

[0048] To further quantify the urgency of the fault, this embodiment also introduces a sentiment analysis algorithm to calculate a negative sentiment index E in the text information based on negative text. This algorithm can be designed, but is not limited to, based on the following principle: the more urgent the alarm tone and the more negative words used by the maintenance personnel, the more severe the on-site situation. The specific formula for calculating the negative sentiment index is as follows: E = (Number of angry words × Weight of negative words) / Total number of sentences.

[0049] Among them, negative text refers to words in the text that express strong emotions such as urgency, anger, and surprise; the number of angry words refers to the number of words in the text that express strong emotions such as urgency, anger, and surprise (such as "hurry up", "quick", "terrible"); the weight of negative words can be pre-assigned according to the degree of negativity of the words, for example, "abnormal" has a weight of 1 and "collapse" has a weight of 2; "total number of sentences" is used for normalization processing to avoid deviations caused by different text lengths.

[0050] In practical implementation, when negative words such as "abnormal", "coal shortage", and "blockage" appear in the alarm text, the embodiments of this application can calculate the E value based on the frequency of these negative words (number of angry words) and word weight.

[0051] If the calculated E value exceeds a preset threshold (e.g., 6) in actual application, the embodiment of this application can determine that the maintenance personnel are indicating that the current alarm is a serious problem, that is, the current voice information given by the maintenance personnel indicates that there is a serious problem with the boiler equipment, and the operation intention is to trigger the emergency handling mechanism immediately.

[0052] The emergency response mechanism includes, but is not limited to: automatically increasing the monitoring frequency of relevant equipment, sending high-priority alarm pop-ups to the shift supervisor's workstation, suspending automatic control and waiting for manual confirmation, or directly activating the preset emergency protection program.

[0053] This application embodiment can transform unstructured voice alarms into structured equipment anomaly characteristic information (such as equipment ID, fault type, severity) and operation intentions (such as "increase coal feed" and "early compensation") through semantic parsing and fault classification, laying the foundation for subsequent control decisions.

[0054] Furthermore, by converting subjective tone into quantitative indicators through sentiment analysis, the embodiments of this application can enable the system to more accurately identify truly urgent alarms, avoid wasting response resources on alarms with bland statements, and thus achieve optimized allocation of limited monitoring resources.

[0055] Step S102: Based on the equipment feature information and operation intention, generate a feedforward pulse compensation command corresponding to the frequency conversion output sequence of the coal feeder of the boiler equipment.

[0056] In some embodiments, after obtaining the equipment anomaly characteristics and operational intentions, this application embodiment needs to map this high-level semantic information into low-level control commands. Therefore, this application embodiment can generate feedforward pulse compensation commands for the corresponding coal feeder frequency converter output sequence based on these equipment anomaly characteristics and operational intentions.

[0057] The process will now be explained in more detail.

[0058] Optionally, in one embodiment of this application, a feedforward pulse compensation instruction corresponding to the frequency converter output sequence of the boiler equipment's coal feeder is generated based on equipment feature information and operation intention. This includes: converting the equipment feature information and operation intention into a semantic vector; determining the current fault intensity score of the boiler equipment based on the semantic vector; comparing the current fault intensity score with a preset compensation threshold, and generating a feedforward pulse compensation instruction when the comparison result shows that the current fault intensity score is greater than the preset compensation threshold.

[0059] In some embodiments, during the process of generating feedforward instructions based on the equipment characteristic information of the boiler equipment and the operation intentions of the maintenance personnel, this embodiment further refines the specific implementation method from speech-to-text to fault intensity quantification, ensuring that the output of the large language model can be transformed into a quantifiable basis for control decisions.

[0060] The core of transforming the output of a large language model into a quantifiable basis for control decisions lies in constructing a mapping model from the semantic space to the control parameter space.

[0061] Based on this, the embodiments of this application can first construct a semantic vector S={w1,w2,...,wn} from the extracted device feature information and operation intention, where wi represents the semantic word embedding of the i-th word. That is, each word wi is mapped to a dense vector of fixed dimensions through a pre-trained word embedding model (such as Word2Vec, BERT, etc.), and is therefore a semantic word embedding.

[0062] By combining all semantic word embedding vectors in text order, a semantic vector S can be formed. For example, for the alarm text "#4 coal mill vibration abnormal, suspected coal shortage", after word segmentation and word embedding, a multi-dimensional semantic matrix is ​​obtained by concatenating multiple word vectors, including equipment identification words, status description words, and fault inference words.

[0063] Furthermore, in this embodiment, the semantic vector can be input into a multi-layer neural network model, and after nonlinear transformation, the current fault intensity score F_score of the boiler equipment is output. This score reflects the urgency of the current abnormal state requiring compensation. When the current fault intensity score F_score is greater than a preset compensation threshold, this embodiment can generate a feedforward pulse compensation command.

[0064] Specifically, in this embodiment, the semantic vector S can be input into a pre-trained multi-layer neural network model. The structure of this neural network typically includes an input layer, several hidden layers (such as fully connected layers, convolutional layers, or recurrent neural network layers), and an output layer. The input layer receives the semantic vector S, which undergoes multiple non-linear transformations in the hidden layers (including linear transformations and activation function processing such as ReLU and Sigmoid) to extract and abstract semantic features from the text layer by layer. Finally, the output layer outputs a scalar value F_score, which is the current fault intensity score. This score quantifies the severity of the device anomaly reflected in the current alarm text, and its value range can be normalized to [0,1] or [0,100].

[0065] For example, the design and implementation process of the multilayer neural network model in this application can be represented, but is not limited to, as follows: Algorithm Structure and Training Process: This neural network adopts a fully connected feedforward architecture, consisting of one input layer, two hidden layers, and one output layer. The input layer has the same dimension as the semantic vector S (e.g., using a 768-dimensional BERT embedding). The first hidden layer can have 128 neurons, the second hidden layer can have 32 neurons, and the output layer has one neuron that directly outputs the fault intensity score F_score (ranging from 0 to 100).

[0066] Inter-layer connections can be fully connected, with a batch normalization layer sequentially connected after each hidden layer to accelerate convergence and alleviate internal covariate bias, and a dropout layer (with a dropout rate of 0.3) to prevent overfitting.

[0067] The activation function can be, but is not limited to, ReLU (Modified Linear Unit) to introduce nonlinearity and avoid gradient vanishing. The training data can be the historical operation records of a 600MW supercritical unit over the past three years, with a total of 12,000 maintenance voice alarm texts and their corresponding fault intensity scores manually labeled (the average score was taken by three senior shift supervisors after back-to-back scoring as the label).

[0068] Data preprocessing: After speech-to-text conversion, modal particles and repetitive words were removed, equipment numbering formats were standardized, and semantic vectors were extracted using a pre-trained BERT model for the power sector. Mean squared error (MSE) was used as the loss function, and Adam was selected as the optimization algorithm (learning rate 0.001, β1=0.9, β2=0.999). The batch size was set to 32, the training epochs were 200, and an early stopping strategy was adopted (the training terminated if the validation set loss did not decrease for 10 consecutive epochs).

[0069] Furthermore, key hyperparameters can be selected, but are not limited to, according to the following criteria: the number of hidden layer neurons (128, 32) is determined through grid search, with the search range covering [64, 128, 256] and [16, 32, 64]. On the validation set, the combination of 128+32 achieves the lowest MSE and has a moderate number of parameters; the dropout rate is 0.3, which is derived from monitoring the degree of overfitting (generalization is considered good when the difference between the training set and the validation set loss is less than 5%); the learning rate is 0.001, which is a commonly used starting value for the Adam optimizer in similar text regression tasks, and has been confirmed to be optimal through learning rate decay experiments (0.0005~0.005).

[0070] The parameter adjustment range can be, but is not limited to, the following: the hidden layer width can be adjusted to 64~256, the Dropout rate can be adjusted to 0.2~0.5, and the learning rate can be adjusted to 0.0001~0.005. In actual deployment, it can be dynamically adjusted according to the amount of data in the new power plant.

[0071] Simulation results show that the model's mean absolute error (MAE) is 4.37 (range 0-100) on 2000 independent test samples, which is a 65.9% improvement compared to the baseline model using only TF-IDF features and linear regression (MAE=12.85); it is also 44.8% lower than the same structured network without pre-trained semantic vectors (MAE=7.92).

[0072] More importantly, during the actual boiler start-up and shutdown process, when the F_score output by the model exceeds the preset threshold (80 points), feedforward pulse compensation is triggered. Compared with the traditional threshold alarm method, the early warning time for the risk of thermal collapse of the pulverizing system is shortened from an average of 6.2 seconds to 1.8 seconds (based on statistical analysis of 21 coal interruption events on site), which verifies that the neural network model can accurately quantify the degree of urgency of the fault reflected by the voice alarm of the operation and maintenance personnel.

[0073] To determine whether to trigger a compensation action and the intensity of the compensation, this application embodiment can pre-set a compensation threshold, i.e., a preset compensation threshold. Here, the preset compensation threshold can be understood as the maximum boundary value of the current fault intensity score. In this application embodiment, the preset compensation threshold can be, but is not limited to, set to Th = 8 × max(F_score_m), that is, the preset compensation threshold in this application embodiment can be set to 80% of the historical maximum fault intensity (or the preset benchmark maximum value) F_score_m.

[0074] The purpose of this design is to determine that a serious event requiring active intervention is only identified when the current fault intensity reaches more than 80% of the historical peak, thereby avoiding overreaction to minor anomalies or false alarms.

[0075] In the specific judgment logic, the processor compares the current fault intensity score F_score with the preset compensation threshold Th: when the current fault intensity score F_score ≥ Th, it can be determined that a compensation instruction needs to be generated, and the value of F_score is mapped to specific compensation parameters (such as pulse amplitude, duration, etc.) to call the feedforward compensation generation module to output the instruction.

[0076] For example, the specific form of the compensation instruction in this application is a feedforward pulse sequence, which includes, but is not limited to, parameters such as compensation amplitude, pulse width, and pulse interval. For example, for the alarm "#4 coal mill coal cut-off", the compensation instruction generated by the system may be: within the next 30 seconds, increase the frequency of the inverter output corresponding to the #4 coal feeder by 5Hz based on the current set value, continue for 5 seconds, then decrease by 2Hz, and continue for another 5 seconds to form a pulse compensation waveform.

[0077] This feedforward pulse compensation can quickly replenish the instantaneous shortage of pulverized coal caused by coal interruption, prevent the sudden drop of the boiler heat load, and avoid system oscillation caused by long-term overcompensation.

[0078] When F_score < Th, it can be determined that the current abnormal degree of the boiler equipment is still within the tolerable range. Only alarm information is recorded without triggering automatic compensation, waiting for manual confirmation or further observation.

[0079] Through this thresholding process, the embodiments of the present application can effectively achieve the deterministic mapping from the semantic understanding result to the control action, which can not only effectively ensure the timely response to serious faults but also avoid the impact of frequent intervention on the system stability.

[0080] Step S103: Based on the feedforward pulse compensation instruction and the real-time working condition data of the boiler equipment, optimize the frequency conversion output parameters of the coal feeder of the boiler equipment, and generate an opening control instruction or a stop control instruction for the boiler equipment according to the optimized frequency conversion output parameters, and control the boiler equipment to perform the corresponding opening action or stop action according to the opening control instruction or stop control instruction.

[0081] As a possible implementation manner, the embodiments of the present application can, after generating the feedforward pulse compensation instruction, perform fusion control on the feedforward pulse compensation instruction and the real-time working condition data to adjust and optimize the frequency conversion output parameters of the coal feeder.

[0082] Specifically, after generating the feedforward pulse compensation instruction, the embodiments of the present application do not simply issue the instruction directly to the frequency converter, but fuse it with the real-time working condition data of the boiler equipment to ensure the safety and effectiveness of the compensation action. It is equivalent to dynamically correcting and limiting the compensation instruction by collecting the real-time feedback of the on-site sensors of the boiler equipment.

[0083] The ultimate goal of feedforward compensation is to ensure the thermal stability of the boiler during the start-up and shut-down processes. Therefore, according to the adjusted and optimized frequency conversion output parameters, the embodiments of the present application can generate an opening control instruction or a stop control instruction for the boiler equipment, so as to control the boiler equipment to perform the opening action according to the generated opening control instruction, or control the boiler equipment to perform the corresponding stop action according to the stop control instruction, thereby completing the dynamic optimization of the boiler start-up and shut-down processes, especially paying attention to the smooth transition of the whole process from ignition to load carrying and from load carrying to boiler shutdown, so as to achieve the thermal stability of the coal pulverizing system.

[0084] Among them, in the process of fusion control of the feedforward pulse compensation instruction and the real-time working condition data, the embodiments of the present application further introduce a compensation stability period judgment mechanism based on the pulverized coal flow velocity and the primary air flow rate to ensure that the feedforward pulse compensation instruction is only executed when the system has the stable regulation ability, and avoid blind compensation leading to pulverized coal deposition or powder feeding interruption under the condition that the flow conditions are not met.

[0085] For example, in this embodiment of the application, a pulverized coal flow rate sensor and a primary air flow rate measuring device can be installed on the pulverized coal feeding pipeline of the pulverized coal feeding system to collect the pulverized coal flow rate v_flow and the primary air flow rate q_air in real time, respectively. These two parameters are key indicators for judging the pulverized coal conveying status: if the pulverized coal flow rate is too low, it may cause pulverized coal to deposit and block in the pipeline; if the primary air flow rate is insufficient, it will not be able to effectively carry the pulverized coal into the furnace.

[0086] The controller continuously monitors the two parameters mentioned above and compares them with the preset minimum air supply threshold and minimum pulverized coal velocity threshold, respectively. The preset minimum air supply threshold refers to the pre-set minimum air supply boundary value of the pulverizing system, which the system must reach at least to operate normally. The preset minimum pulverized coal velocity threshold refers to the pre-set minimum pulverized coal velocity boundary value of the pulverizing system, which the system must reach at least to operate normally.

[0087] By comparing the results, the embodiments of this application can determine whether the boiler's pulverized coal feeding system meets the following conditions for entering the compensation stabilization period: q_air > q_min and v_flow > v_min. Here, q_min is the minimum air supply threshold, which can be preset according to pipeline design parameters and pulverized coal conveying requirements, for example, it can be taken as 60%-70% of the design air volume; v_min is the minimum pulverized coal velocity threshold, which can usually be determined based on the pulverized coal particle size and pipeline layout, generally not less than 15-18 m / s, to ensure that the pulverized coal is in a suspended conveying state.

[0088] It should be noted that the minimum air supply threshold and the minimum coal powder velocity threshold can be determined by those skilled in the art based on the actual situation. The embodiments in this application are only illustrative and do not impose specific limitations.

[0089] When the conditions q_air > q_min and v_flow > v_min are met, it indicates that the current coal feeding system is in a stable operating range—there is sufficient airflow to carry coal powder and sufficient flow velocity to prevent deposition. At this time, performing feedforward pulse compensation is safe and effective. In this stable state, embodiments of this application can enable a proportional-integral-derivative (PID) controller to finely adjust the frequency converter output parameters.

[0090] The PID controller's setpoint is derived from the feedforward pulse compensation command after fusing with the real-time operating data of the boiler equipment. The feedback value is the actual frequency converter output parameters (such as frequency and current). The control output is calculated through the PID algorithm to achieve closed-loop stable control of the coal feeder speed.

[0091] If the above conditions are not met, such as q_air≤q_min or v_flow≤v_min, the powder feeding system is determined to be in an unstable flow state. In this case, even if there is a need for feedforward compensation, the embodiment of this application will not enable PID control for the time being. Instead, the flow conditions will be improved by adjusting the wind baffle or starting the purging program. PID control will be put into operation after the conditions are met.

[0092] This hierarchical control strategy ensures that compensation actions are always performed under safe flow conditions, effectively preventing control failures or safety hazards caused by deteriorating flow conditions.

[0093] Furthermore, when generating multiple start-up and shutdown strategies based on the compensation instructions after fusion control, this embodiment of the application can also evaluate the start-up and shutdown strategies through a decision tree model during the normal start-up and shutdown process of the boiler, combining runtime sequence data T_seq and equipment health status index H, to obtain the evaluation score of the current start-up and shutdown strategy, and select the start-up and shutdown strategy with the smallest evaluation score to be executed.

[0094] For example, in this application embodiment, two key decision input parameters can be constructed first. The first parameter is runtime sequence data T_seq, which records key timing information during boiler start-up and shutdown, such as the estimated time from ignition to grid connection and the remaining time from load reduction to shutdown, reflecting the urgency of the time dimension. The second parameter is the equipment health status index H, which is obtained by fusing and analyzing monitoring data such as vibration, temperature, and current of key equipment (such as coal mills, coal feeders, and fans), quantifying the current health level of the equipment. The value range can be normalized to [0, 100], with higher values ​​indicating better health status.

[0095] In this embodiment, these two parameters can be input into a pre-built decision tree model. The decision tree model is trained based on historical start-up and shutdown data and expert experience. Its internal nodes contain judgment conditions for T_seq and H, and the leaf nodes correspond to specific start-up and shutdown strategy options (such as "fast grid connection strategy", "smooth grid connection strategy", "delayed start-up strategy", etc.).

[0096] During the decision tree traversal, the strategy corresponding to each leaf node is quantified and scored using an evaluation formula. The scoring formula can be, but is not limited to, expressed as follows: Score = λ1 × T_seq + λ2 × H.

[0097] Here, λ1 and λ2 are the weighting factors for time priority and safety factor, respectively. These two weights can be pre-assigned by domain experts based on power plant operation procedures and safety requirements. For example, in scenarios where the grid load is tight and a rapid response is required, λ1 can be set higher (e.g., 0.7) and λ2 lower (e.g., 0.3) to prioritize time; in stages of equipment aging or frequent failures, λ1 can be set lower (e.g., 0.3) and λ2 higher (e.g., 0.7) to prioritize safety.

[0098] By traversing all feasible start-up and shutdown strategies, this embodiment of the application can calculate the Score value corresponding to each strategy and select the strategy with the smallest Score as the final execution plan. The reason for selecting the minimum value is that a smaller T_seq indicates a more urgent time, and a smaller H indicates a worse health status. Both of these increase the operational risk. Therefore, a lower score means that the strategy has the least comprehensive risk under the current constraints, which is most conducive to achieving a smooth start-up and shutdown.

[0099] Through this decision tree evaluation mechanism based on runtime sequence data and equipment health status, the embodiments of this application can quantify the evaluation and optimization of start-up and shutdown strategies, so as to dynamically adjust the start-up and shutdown strategies according to real-time operating conditions, and achieve the best balance between time and safety. That is, to ensure that the start-up and shutdown time is shortened as much as possible while meeting safety requirements, improve the unit's economy, and effectively avoid operational problems caused by blindly pursuing speed or being overly conservative.

[0100] Additionally, during the dynamic optimization of the boiler start-up and shutdown process, this application embodiment further introduces an intelligent optimization mechanism based on combustion efficiency calculation and genetic algorithm optimization to achieve the global optimal configuration of the coal feeder frequency conversion output sequence, thereby improving the unit's economy while ensuring thermal stability.

[0101] For example, this application can first calculate the fuel combustion efficiency η of the boiler. This efficiency reflects the completeness of pulverized coal combustion in the furnace and is a key indicator for evaluating combustion economy. The formula for calculating fuel combustion efficiency is as follows: η = (CO content × air volume) / (pulverized coal consumption × combustion time).

[0102] Among them, CO content can be measured in real time by flue gas analyzer, reflecting the magnitude of incomplete combustion loss; air supply volume can be obtained by primary air and secondary air flow measurement devices; pulverized coal consumption can be indirectly calculated by coal feeder speed and pulverized coal concentration; combustion time can be determined according to boiler load and furnace residence time.

[0103] It should be noted that the fuel combustion efficiency formula in this embodiment is a simplified efficiency characterization model. In practical applications, the calculation method of fuel combustion efficiency can be adjusted by those skilled in the art according to the boiler type and measuring point configuration. For example, parameters such as flue gas oxygen content and fly ash carbon content can be introduced for correction. This embodiment is only for illustrative purposes and does not impose specific limitations.

[0104] While calculating the real-time efficiency η, this embodiment of the application can also monitor the pressure change ΔP_in in the furnace in real time. Pressure fluctuation is an important indicator of combustion stability; excessive pressure fluctuation may indicate combustion oscillation or the risk of flameout. By acquiring pressure values ​​in real time through a furnace pressure transmitter, this embodiment of the application can calculate the pressure change amplitude ΔP_in in the furnace per unit time.

[0105] To achieve comprehensive optimization of combustion economy and stability, this application embodiment can construct a multi-objective optimization function: Obj = ω1 × η + ω2 × ΔP_in. Here, ω1 and ω2 are the trade-off coefficients between efficiency and pressure variation, respectively, and their sum is typically 1. The values ​​of ω1 and ω2 can be dynamically adjusted according to the unit's operating phase: during normal steady-state operation, ω1 can be appropriately increased to pursue economy; during start-up and shutdown processes or low-load operation, ω2 can be increased to prioritize stability. By adjusting the trade-off coefficients, the optimization objective can be flexibly switched between different operating modes.

[0106] The goal of the optimization problem is to find the optimal frequency converter output sequence that minimizes Obj. Since the frequency converter output of the coal feeder involves a complex optimization space with multiple variables, nonlinearity, and multiple constraints, this embodiment of the application can use a genetic algorithm to solve it. The specific implementation steps of the genetic algorithm include, but are not limited to, the following: (1) Encoding and Population Initialization: The frequency conversion output sequence to be optimized (such as the set value sequence of the coal feeder frequency in the future) is encoded into a chromosome, with each gene position corresponding to the frequency value at a certain time. An initial population containing several candidate solutions is randomly generated.

[0107] (2) Fitness function design: The reciprocal of the optimization target Obj is used as the fitness function. The smaller Obj is, the higher the fitness.

[0108] (3) Selection operation: Using roulette or tournament selection methods, individuals with higher fitness are selected from the current population to enter the next generation.

[0109] (4) Crossover operation: Randomly select two parent chromosomes and exchange gene segments at random positions to generate offspring individuals.

[0110] (5) Mutation operation: Randomly perturb certain gene loci of an individual with a certain probability to introduce new search directions and avoid getting trapped in local optima.

[0111] (6) Iteration and Termination: Repeat the selection, crossover, and mutation operations until the preset number of iterations or fitness convergence is reached. Finally, the individual with the highest fitness is decoded as the optimal frequency conversion output sequence.

[0112] In actual boiler start-up and shutdown control, PID control, genetic optimization, and feedforward pulse compensation follow a hierarchical collaborative logic of "feedforward pulse priority triggering, PID real-time closed-loop regulation, and genetic algorithm offline / slow cycle global optimization".

[0113] Specifically, when the voice alarm from the maintenance personnel generates a feedforward pulse compensation command through semantic parsing, the command has the highest real-time priority. The controller immediately outputs a pulse waveform (such as a frequency step) according to the command to quickly respond to sudden anomalies. At the same time, the PID controller remains online but does not interfere with the pulse execution phase. After the pulse compensation is completed (or during the interval of the compensation period), the PID controller automatically takes over and uses the fused target value as the setpoint to perform closed-loop fine-tuning of the frequency converter output parameters to eliminate steady-state errors.

[0114] Genetic algorithms operate on a slower timescale (e.g., every 5 minutes or after each start-stop process). Based on multi-objective optimization functions such as historical combustion efficiency and pressure fluctuations, they solve for the globally optimal variable frequency output sequence offline or online in the background. The optimization results are then injected into the PID controller in the form of a "reference trajectory" or "setpoint bias" to gradually correct the baseline parameters of the PID controller, but do not directly overwrite the instantaneous commands of the feedforward pulse.

[0115] The switching logic is managed by a state machine: the system monitors the pulverized coal flow rate, primary air flow rate, and operating condition stability indicators in real time. When the stability condition of "q_air>q_min and v_flow>v_min" is met, the PID and genetic optimization outputs are allowed to take effect; once a voice alarm triggers the feedforward pulse compensation condition (F_score≥Th), the state machine immediately switches control to the feedforward pulse generator, temporarily disables the PID integral term (to prevent integral saturation), resumes PID control after the pulse sequence is completed, and records the compensation effect for the genetic algorithm to update the fitness function in the next iteration.

[0116] It should be noted that the genetic algorithm itself does not directly intervene in real-time pulses. Instead, it updates the controller's feedforward gain table or PID weight coefficients periodically to make the pulse amplitude of the next feedforward compensation more accurate, thereby achieving a synergy between fast response and global optimization.

[0117] Leveraging the global optimization capabilities of genetic algorithms, this embodiment can automatically search for the optimal coal feeder control strategy that combines combustion efficiency and pressure stability while satisfying thermal stability constraints, thus achieving a leap from experience-based control to intelligent optimization. In practical applications, this optimization process can be performed offline and then downloaded to the controller, or it can be executed online periodically to dynamically adapt to changes in operating conditions.

[0118] Through the closed-loop control of the above four steps, this application can quickly convert the voice alarms of operation and maintenance personnel into feedforward pulse compensation commands and deeply integrate them with real-time operating conditions to achieve fine-grained control of the frequency converter output of the coal feeder. This effectively prevents the thermal collapse of the pulverizing system caused by the lag of traditional threshold alarms and significantly improves the safety and economy of the boiler start-up and shutdown process.

[0119] Optionally, in one embodiment of this application, optimizing the frequency conversion output parameters of the boiler feeder based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment includes: acquiring the original sensor data of the boiler and performing timestamp alignment processing on the original sensor data to obtain real-time operating data; calculating the communication delay of the feedforward pulse compensation command based on the sending time and feedback time of the feedforward pulse compensation command; and optimizing the frequency conversion output parameters of the boiler feeder based on the real-time operating data when the communication delay is less than or equal to a preset allowable delay time threshold.

[0120] In other embodiments, this application introduces a timestamp alignment mechanism for multi-sensor data and a device delay compensation mechanism to ensure that the timing of the execution of the feedforward pulse compensation command is accurate and to avoid misalignment between the compensation action and the actual system state due to data transmission delay.

[0121] Specifically, to ensure the spatiotemporal consistency of multi-source data, this application can perform timestamp alignment processing on the raw sensor data obtained from different sensors (such as current sensors, coal flow sensors, and temperature sensors) to obtain effective real-time operating condition data.

[0122] Then, in this embodiment of the application, the device communication delay Δt = t_set - t_actual of the feedforward pulse compensation command can be calculated based on the sending time and feedback time of the feedforward pulse compensation command. The feedforward compensation action is only executed when the communication delay Δt is within the allowable range (Δt ≤ Δt_max), so as to avoid misalignment of the compensation timing due to communication delay.

[0123] For example, a boiler start-up and shutdown monitoring system involves a large number of field sensors and actuators, including temperature sensors, pressure sensors, flow sensors, frequency converters, and electric valves. These devices are distributed in different locations on-site and connected to the control system via fieldbus (such as Modbus, PROFIBUS) or industrial Ethernet. The data sampling frequency and communication latency of each device vary. If unaligned data is used directly for control decisions, control commands may be issued based on outdated operating information, significantly reducing the compensation effect or even causing adverse effects.

[0124] Therefore, embodiments of this application can establish a unified time synchronization mechanism at the system level. The control system periodically synchronizes with an NTP time server or GPS clock source to ensure that all modules within the system have a unified time reference. All data collected by sensors is timestamped before entering the controller, recording the precise moment of data acquisition.

[0125] In the data fusion processing stage, the embodiments of this application can first perform timestamp alignment on the data from different sensors. Specifically, when it is necessary to fuse multi-source data such as current, temperature, and flow rate for control decisions, the system finds the most recently sampled data from each sensor and checks the difference between its timestamp and the current time. Only data with timestamps within the allowed time window (e.g., within 500ms) is used for fusion calculation; data outside the window is considered expired data and awaits the next refresh.

[0126] Furthermore, for the execution of feedforward compensation commands, this application embodiment introduces a delay compensation mechanism. That is, this application embodiment can record the sending time t_set of the compensation command after it is generated. After receiving the command, the frequency converter or other actuator feeds back the receiving time t_actual of the compensation command (i.e., the feedback time of the compensation command). The formula for calculating the communication delay can be, but is not limited to, expressed as: Δt = t_set - t_actual.

[0127] This value reflects the network transmission time and device processing time that the compensation command takes from being sent to being received.

[0128] Next, the controller determines whether Δt meets the execution condition: feedforward compensation is only allowed when Δt ≤ Δt_max. Here, Δt_max is a preset allowable delay time threshold, i.e., the maximum allowable delay time, which is preset according to the real-time requirements of the control system and the response speed of the process. For example, for frequency converter control of a coal feeder, it can be set to 200ms-500ms.

[0129] If Δt exceeds this threshold, it indicates abnormal congestion in the communication link or slow device response. In this case, the opportunity to execute compensation may have passed, and the system will abandon the compensation command, record a communication delay alarm, and wait for reassessment in the next control cycle. This time alignment and delay compensation mechanism ensures that feedforward control commands always act on field devices within the effective time window, improving the reliability and accuracy of the control system.

[0130] Optionally, in one embodiment of this application, optimizing the variable frequency output parameters of the coal feeder of the boiler equipment based on feedforward pulse compensation commands and real-time operating data of the boiler equipment includes: obtaining the real-time feedback current value of the frequency converter of the boiler equipment, and calculating the current deviation value between the preset current value of the frequency converter and the real-time feedback current value; generating an adjustment factor based on the ratio of the current deviation value to the preset deviation threshold, and determining the feedforward compensation strength of the frequency converter according to the adjustment factor, so as to optimize the variable frequency output parameters of the coal feeder according to the feedforward compensation strength.

[0131] In some embodiments, during the process of fusing and controlling the feedforward pulse compensation command and the real-time operating data of the boiler equipment, this application further introduces an adaptive adjustment mechanism based on the inverter feedback current to ensure that the execution intensity of the feedforward pulse compensation command matches the actual operating conditions and avoids system oscillation caused by over-compensation.

[0132] Specifically, in this embodiment, the real-time feedback current value I_real can be read from the inverter communication interface at millisecond intervals (e.g., 100ms). This value directly reflects the real-time load status of the coal feeder drive motor. Simultaneously, this embodiment can obtain the current set current I_set of the current coal feeder inverter output from the DCS, i.e., the inverter's preset current value.

[0133] Then, the embodiments of this application can calculate the current deviation between the real-time feedback current value and the current set current value I_set: ΔI = I_set - I_real. This deviation reflects the difference between the actual load and the set load of the coal feeder, indirectly indicating the change in coal flow: when ΔI is positive, it indicates that the actual current is lower than the set value, which usually means that the coal feed is insufficient, the motor load is reduced, and there may be coal shortage or blockage; when ΔI is negative, it indicates that the actual current is higher than the set value, which may mean that the coal flow is too large or mechanical jamming.

[0134] If ΔI is large (e.g., the absolute value exceeds the allowable threshold I_threshold), it indicates that the system is already in a state of large fluctuation. In this case, compensation should be applied with caution to avoid exacerbating the oscillation.

[0135] To prevent the system from overshooting due to the full application of compensation commands when ΔI is large, this application embodiment designs an adaptive adjustment factor (adaptive weighting factor) based on the ratio of the current deviation value to the preset deviation threshold I_threshold: w=min(1,(ΔI / I_threshold)). β ), where β is the adjustment coefficient (usually between 1 and 2), with a value range of 1.0 to 2.0, used to control the sensitivity of the weighting factor to deviation.

[0136] The role of this weighting factor is as follows: when the actual deviation ΔI is small, w approaches 1, and the compensation command is executed in full; when ΔI approaches or exceeds the threshold, w decreases accordingly, and the compensation intensity is weakened. Ultimately, the actual compensation intensity applied to the inverter is ΔF × w, where ΔF is the theoretical compensation amount.

[0137] Wherein, the preset deviation threshold I_threshold is the maximum allowable current deviation value of the system. This threshold can be preset by those skilled in the art based on historical operating data and equipment rated parameters, for example, it can be 5%-10% of the rated current. This embodiment is only illustrative and does not impose specific limitations. β is the adjustment coefficient. In actual control processes, the calculation of the theoretical compensation amount can be, but is not limited to, expressed as follows: First, calculate the absolute value of ΔI, then substitute it into the formula to calculate w. When |ΔI| is much smaller than I_threshold, (ΔI / I_threshold) β The value of w approaches 0, but due to the constraint of the min function, the actual value of w is 1, which means that the compensation instruction is fully executed; when |ΔI| is close to I_threshold, w is close to 1; when |ΔI| exceeds I_threshold, w is less than 1 and decreases as the deviation increases, which means that the compensation strength is weakened.

[0138] Subsequently, in this embodiment of the application, the obtained current fault intensity score F_score can be mapped to the theoretical compensation amount ΔF according to a preset mapping rule. The mapping rule between the current fault intensity score F_score and the theoretical compensation amount ΔF can be, but is not limited to, a linear function (e.g., ΔF = ΔF_max × (F_score - Th) / (F_score_max - Th)), a piecewise function, or a lookup table method based on expert experience, where ΔF_max is the maximum allowable compensation amplitude of the system (e.g., a frequency increment of 5Hz), and F_score_max is the upper limit of the fault intensity score (e.g., 100).

[0139] In addition, other compensation parameters such as pulse width and pulse interval can also be similarly mapped based on F_score or in combination with the fault type, and finally form a feedforward pulse compensation instruction including parameters such as ΔF, pulse width, and interval.

[0140] Finally, multiply the theoretical compensation amount ΔF by the weight factor w to obtain the actual applied feedforward compensation intensity. Then, the embodiments of this application can superimpose this feedforward compensation intensity on the current set value of the frequency converter to complete the adjustment and optimization of the frequency conversion output parameters of the coal feeder. Through this adaptive adjustment, this application can ensure the compensation effect while effectively suppressing mechanical shock and thermal fluctuations caused by forced compensation under large deviation conditions.

[0141] Optionally, in an embodiment of this application, an opening control instruction or a stop control instruction for the boiler equipment is generated according to the optimized frequency conversion output parameters, including: obtaining the real-time hot air temperature and real-time pulverized coal concentration of the boiler equipment; when the real-time hot air temperature is less than the preset allowable hot air temperature threshold and the real-time pulverized coal concentration is greater than the preset pulverized coal concentration threshold, calculating the stability prediction value of the boiler equipment; generating an opening control instruction or a stop control instruction for the boiler equipment according to the stability prediction value and the preset instability threshold.

[0142] In some embodiments, during the dynamic optimization of the boiler start-stop process, this application also introduces a critical state discrimination and predictive control mechanism based on hot air temperature and pulverized coal concentration to cope with the instability risk most likely to occur in the start-stop stage of the coal pulverizing system.

[0143] Specifically, the embodiments of this application can respectively and real-time collect the hot air temperature T_hot and pulverized coal concentration C_coal of the boiler equipment through temperature sensors and concentration sensors arranged on the pulverized coal pipes at the outlet of the coal mill.

[0144] The controller continuously monitors these two key parameters and thereby determines whether the boiler equipment meets the unstable critical state condition: T_hot < T_min and C_coal > C_th.

[0145] Among them, T_min is the preset allowable hot air temperature threshold, which can be understood here as the lowest allowable hot air temperature of the boiler and can generally be set according to the coal type characteristics and burner design requirements. For example, for bituminous coal, it can be set to 150°C; C_th is the preset pulverized coal concentration threshold, which can be understood here as the critical pulverized coal concentration of the boiler and can generally be set according to the requirements of preventing deposition in the powder delivery pipeline and combustion stability requirements. For example, a reasonable range of 0.3 - 0.5 kg / kg can be taken, etc.

[0146] The specific preset allowable hot air temperature threshold and preset coal powder concentration threshold can be set or adjusted by those skilled in the art according to the actual situation. The embodiments in this application are only illustrative and do not impose specific limitations.

[0147] When T_hot is detected to be lower than the minimum allowable temperature T_min and C_coal is higher than the critical concentration C_th, the embodiments of this application can determine that the pulverizing system is in an unstable critical state. At this time, the coal powder concentration is too high but the hot air temperature is insufficient, the hot air carrying capacity is insufficient, the coal powder is easy to deposit and block, and it is difficult to ignite stably after being injected into the furnace, which can easily lead to risks such as incomplete combustion and re-burning of the tail flue.

[0148] At this point, traditional control logic may continue to increase the coal feed according to the established procedure, which may exacerbate the risk. To avoid such miscontrol, the embodiments of this application adopt a predictive control strategy, that is, control is carried out by predicting the evolution trend of the coal feeding system.

[0149] To predict the system's evolution trend, this application's embodiments construct a prediction model: Y_pred=k×(T_hot+ln(C_coal)), where k is an empirical coefficient reflecting the system's sensitivity to changes in hot air temperature and pulverized coal concentration, which can be obtained through field experiments or fitting historical data. The model is designed based on the fact that the effect of hot air temperature on combustion stability is approximately linear, while the effect of pulverized coal concentration is more significant on a logarithmic scale—the higher the concentration, the faster the stability deteriorates.

[0150] In practical applications, this model can predict the next step of the system's combustion stability index, i.e. the stability prediction value at the next moment, based on the real-time hot air temperature and real-time pulverized coal concentration, combined with the logarithmic relationship of hot air temperature and the linear relationship of pulverized coal concentration.

[0151] Based on the prediction result Y_pred and the preset instability threshold, the controller can determine the evolution trend of the powder feeding system. Here, the preset instability threshold can be understood as a pre-set upper limit boundary of the predicted value.

[0152] If the predicted value Y_pred exceeds the preset instability threshold, it indicates that the coal feeding system is at risk of instability. In this case, the start of the coal feeding program can be delayed, that is, the start-up time of the coal mill or coal feeder can be appropriately postponed, and coal powder can be added after the hot air temperature rises, so as to avoid sudden combustion fluctuations caused by injecting a large amount of coal powder under low hot air conditions.

[0153] Specifically, when the original start-stop sequence requires starting the coal feeder or opening the pulverized coal valve, the controller issues a hold command, delaying the execution of this action and waiting for the hot air temperature to rise. During the delay, the system may increase the hot air temperature by adjusting the hot air damper or prematurely engaging the oil gun. Once T_hot rises to a safe range and the predictive model output returns to normal, the pulverized coal feeding program is resumed. Through this feedforward predictive control, the embodiments of this application can effectively avoid the risk of sudden fluctuations caused by blindly feeding pulverized coal under critical conditions.

[0154] The following is a specific embodiment to explain the boiler start-up and shutdown control method in this application.

[0155] Figure 2 This is a schematic diagram illustrating the framework of a boiler start-up and shutdown intelligent monitoring system based on a large language model, according to one embodiment of this application. Figure 2 As shown, the boiler start-up and shutdown intelligent monitoring system 200 based on a large language model includes, but is not limited to, the following four modules: The extraction module 201 is the front-end intelligent processing unit of the system, responsible for semantic analysis and fault classification of the translated text of voice alarms from maintenance personnel, extracting equipment anomaly features and operational intentions. This module integrates a speech recognition engine and a large language model inference engine. The speech recognition engine collects the speech signals of maintenance personnel through a microphone array and uses an end-to-end deep learning model (such as Conformer or Transformer-Transducer) to convert speech into text in real time. The large language model inference engine loads a pre-trained language model (such as BERT, GPT series, or domain-adjusted models) to perform named entity recognition, sentiment analysis, and intent understanding on the translated text, outputting structured anomaly feature vectors and operational intent labels. The output of the extraction module is transmitted to the feedforward impulse compensation module via a high-speed data bus.

[0156] The feedforward pulse compensation module 202 is the decision-making core of the system. Based on the abnormal features extracted from the module's output and the operational intent, it generates feedforward pulse compensation commands for the corresponding frequency converter output sequence of the coal feeder. This module integrates a semantic-to-parameter mapping model, including a multi-layer neural network and a fault intensity calculation unit. The neural network maps semantic vectors to a fault intensity score F_score and compares it with a preset threshold Th. When the triggering condition is met, the compensation command generation unit generates a feedforward command sequence containing parameters such as pulse amplitude, width, and interval, based on the fault type and intensity, combined with a pre-set compensation rule base or expert system. This module also has self-learning capabilities, allowing it to optimize the mapping model parameters based on historical compensation feedback. The generated compensation commands are sent to the control module via the industrial bus.

[0157] Control module 203, the system's execution and coordination unit, is responsible for fusing feedforward pulse compensation commands with real-time operating data to adjust the frequency converter output parameters of the coal feeder. This module reads real-time data from multiple sensors, including the frequency converter feedback current, pulverized coal flow rate, primary air flow rate, hot air temperature, and pulverized coal concentration, via I / O acquisition cards and a fieldbus interface. Internally, it integrates an adaptive weight calculation unit, a time alignment unit, a PID control algorithm, and stabilization period judgment logic. The control module first performs timestamp alignment and validity verification on the multi-source data. Then, based on the above methods, it calculates adaptive weights, determines the compensation stabilization period, handles communication delays, and finally generates the fused frequency converter output setpoint. This setpoint is then written to the coal feeder frequency converter via an analog output channel or fieldbus, achieving closed-loop control.

[0158] Output module 204 is the system's optimization and display unit, responsible for dynamically optimizing the boiler start-up and shutdown process based on the adjusted variable frequency output parameters to achieve thermal stability of the pulverizing system. This module receives real-time parameters from the control module, combines runtime sequence data and equipment health status indicators, and runs a decision tree model, a prediction model, and a genetic algorithm optimization engine. The internally integrated prediction unit predicts system stability, the decision unit evaluates start-up and shutdown strategies, and the optimization unit solves for the optimal variable frequency sequence. The optimization results are displayed to maintenance personnel through a human-machine interface (HMI) in the form of trend charts, alarm lists, and operation suggestions. Furthermore, automatically executed optimization strategies can be written into the DCS logic or directly sent to the underlying control system via interfaces such as OPC.

[0159] The four modules are connected via high-speed industrial Ethernet or an internal data bus. Communication between modules uses standardized data formats and protocols to ensure real-time and reliable data transmission. The extraction module, feedforward pulse compensation module, control module, and output module can be deployed within the same industrial control computer, or they can be distributed across multiple devices according to functional and reliability requirements. For example, the extraction module can be deployed in an intelligent voice workstation, and the control module in a DCS controller. The modules collaborate via network communication. This modular design maintains a clear functional division while offering flexible deployment options and good scalability, enabling it to adapt to the technical upgrade needs of different power plants.

[0160] This system can be deployed in the unit control room of a power plant and seamlessly integrated with existing DCS systems, PLC control systems and field instruments to form a complete intelligent monitoring closed loop. Figure 3 This is a flowchart illustrating an embodiment of the intelligent monitoring method for boiler start-up and shutdown based on a large language model, according to this application. Figure 3 As shown, based on this system, the boiler start-up and shutdown control process can be represented, but is not limited to, as follows: S1: Perform semantic analysis and fault classification on the translated text of voice alarms from operators, and extract equipment abnormality features and operational intentions; S2: Generate feedforward pulse compensation commands for the corresponding coal feeder frequency converter output sequence based on equipment abnormality characteristics and operational intentions; S3: Integrate feedforward pulse compensation commands with real-time operating data to control and adjust the frequency conversion output parameters of the coal feeder; S4: Dynamically optimize the boiler start-up and shutdown process based on the adjusted frequency converter output parameters to achieve thermal stability of the pulverizing system.

[0161] According to the boiler start-up and shutdown control method proposed in the embodiments of this application, the user's voice information can be parsed, the abnormal characteristics of the equipment and the operation intention can be extracted, and a feedforward pulse compensation command corresponding to the frequency converter output sequence of the coal feeder can be generated. Then, the command is fused with real-time operating data to adjust the frequency converter output parameters of the coal feeder. Based on the adjusted frequency converter output parameters, the boiler start-up and shutdown process is dynamically optimized to achieve thermal stability of the pulverizing system. This invention achieves timely response to thermal fluctuations caused by sudden anomalies by converting the voice alarm text of operators into feedforward pulse compensation commands. It provides pulse-based dynamic compensation to the frequency converter output of the coal feeder before the risk of thermal collapse of the pulverizing system materializes, effectively ensuring the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown. This solves the problems in related technologies, such as the difficulty in timely response to thermal fluctuations caused by sudden anomalies during boiler start-up and shutdown monitoring, the inability to parse and convert the voice text of maintenance personnel, and the inability to provide pulse-based dynamic compensation to the frequency converter output of the coal feeder before the risk of thermal collapse of the pulverizing system materializes, thus failing to effectively ensure the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown.

[0162] Next, the boiler start-up and shutdown control device according to the embodiments of this application is described with reference to the accompanying drawings.

[0163] Figure 4 This is a schematic diagram of the boiler start-stop control device according to an embodiment of this application.

[0164] like Figure 4 As shown, the boiler start-stop control device 10 includes: an extraction module 100, a generation module 200, and a control module 300.

[0165] The extraction module 100 is used to respond to the user's voice information and extract the equipment feature information corresponding to the boiler equipment and the operation intention corresponding to the user from the voice information based on the corresponding text information; the generation module 200 is used to generate a feedforward pulse compensation command corresponding to the frequency conversion output sequence of the boiler equipment's coal feeder based on the equipment feature information and operation intention; the control module 300 is used to optimize the frequency conversion output parameters of the boiler equipment's coal feeder based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment, so as to generate a start control command or a stop control command for the boiler equipment according to the optimized frequency conversion output parameters, and control the boiler equipment to perform the corresponding start action or stop action according to the start control command or stop control command.

[0166] Optionally, in one embodiment of this application, the extraction module 100 includes: a first extraction unit, configured to extract device number entity and anomaly type entity from text information based on pre-defined named entities, so as to determine device feature information based on device number entity and anomaly type entity; and a second extraction unit, configured to extract fuzzy text and negative text from text information based on contextual features of device feature information, so as to determine the user's operation intention based on fuzzy text and negative text.

[0167] Optionally, in one embodiment of this application, the generation module 200 includes: a conversion unit for converting equipment feature information and operation intention into semantic vectors; a determination unit for determining the current fault intensity score of the boiler equipment based on the semantic vectors; and a comparison unit for comparing the current fault intensity score with a preset compensation threshold, and generating a feedforward pulse compensation command when the comparison result is that the current fault intensity score is greater than the preset compensation threshold.

[0168] Optionally, in one embodiment of this application, the control module 300 includes: a processing unit, configured to acquire raw sensor data of the boiler and perform timestamp alignment processing on the raw sensor data to obtain real-time operating condition data; a first calculation unit, configured to calculate the communication delay of the feedforward pulse compensation command based on the sending time and feedback time of the feedforward pulse compensation command; and an optimization unit, configured to optimize the frequency conversion output parameters of the boiler equipment's coal feeder based on the real-time operating condition data when the communication delay is less than or equal to a preset allowable delay time threshold.

[0169] Optionally, in one embodiment of this application, the control module 300 includes: a first acquisition unit, configured to acquire the real-time feedback current value of the frequency converter of the boiler equipment and calculate the current deviation value between the preset current value of the frequency converter and the real-time feedback current value; and a first generation unit, configured to generate an adjustment factor based on the ratio of the current deviation value to a preset deviation threshold, determine the feedforward compensation strength of the frequency converter according to the adjustment factor, and optimize the frequency conversion output parameters of the coal feeder according to the feedforward compensation strength.

[0170] Optionally, in one embodiment of this application, the control module 300 includes: a second acquisition unit, used to acquire the real-time hot air temperature and real-time pulverized coal concentration of the boiler equipment; a second calculation unit, used to calculate the stability prediction value of the boiler equipment when the real-time hot air temperature is less than a preset allowable hot air temperature threshold and the real-time pulverized coal concentration is greater than a preset pulverized coal concentration threshold; and a second generation unit, used to generate an start control command or a stop control command for the boiler equipment based on the stability prediction value and a preset instability threshold.

[0171] It should be noted that the explanation of the above-mentioned boiler start-up and shutdown control method embodiment also applies to the boiler start-up and shutdown control device of this embodiment, and will not be repeated here.

[0172] According to the boiler start-up and shutdown control device proposed in the embodiments of this application, the user's voice information can be parsed, equipment abnormality characteristics and operation intentions can be extracted, and a feedforward pulse compensation command corresponding to the frequency converter output sequence of the coal feeder can be generated. The command is then fused with real-time operating data to adjust the frequency converter output parameters of the coal feeder. Based on the adjusted frequency converter output parameters, the boiler start-up and shutdown process is dynamically optimized to achieve thermal stability of the pulverizing system. This invention achieves timely response to thermal fluctuations caused by sudden anomalies by converting the voice alarm text of operators into feedforward pulse compensation commands. It provides pulse-based dynamic compensation to the frequency converter output of the coal feeder before the risk of thermal collapse of the pulverizing system materializes, effectively ensuring the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown. This solves the problems in related technologies, such as the difficulty in timely response to thermal fluctuations caused by sudden anomalies during boiler start-up and shutdown monitoring, the inability to parse and convert the voice text of maintenance personnel, and the inability to provide pulse-based dynamic compensation to the frequency converter output of the coal feeder before the risk of thermal collapse of the pulverizing system materializes, thus failing to effectively ensure the thermal stability and timely response of the pulverizing system during boiler start-up and shutdown.

[0173] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0174] When the processor 502 executes the program, it implements the boiler start-up and shutdown control method provided in the above embodiments.

[0175] Furthermore, electronic devices also include: Communication interface 503 is used for communication between memory 501 and processor 502.

[0176] The memory 501 is used to store computer programs that can run on the processor 502.

[0177] Memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0178] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0179] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0180] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0181] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the boiler start-up and shutdown control method described above.

[0182] This application also provides a computer program product, including a computer program that can run computer instructions. When the computer instructions are executed by a processor, they implement the boiler start-up and shutdown control method provided in this application.

[0183] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0184] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0185] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0186] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0187] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0188] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0189] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0190] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A boiler start-up and shutdown control method, characterized in that, Includes the following steps: In response to the user's voice information, based on the text information corresponding to the voice information, the device feature information corresponding to the boiler equipment and the operation intention corresponding to the user are extracted from the voice information; Based on the equipment feature information and the operation intention, a feedforward pulse compensation command corresponding to the frequency conversion output sequence of the coal feeder of the boiler equipment is generated. Based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment, the frequency conversion output parameters of the coal feeder of the boiler equipment are optimized. According to the optimized frequency conversion output parameters, the start control command or stop control command of the boiler equipment is generated, and the boiler equipment is controlled to perform the corresponding start action or stop action according to the start control command or stop control command.

2. The method according to claim 1, characterized in that, The step of extracting equipment feature information corresponding to the boiler equipment and operation intent corresponding to the user from the voice information based on the text information corresponding to the voice information includes: Based on pre-defined named entities, the device number entity and the anomaly type entity are extracted from the text information to determine the device feature information according to the device number entity and the anomaly type entity. Based on the contextual features of the device feature information, fuzzy text and negative text are extracted from the text information to determine the user's operational intent.

3. The method according to claim 1, characterized in that, The step of generating a feedforward pulse compensation command corresponding to the frequency converter output sequence of the coal feeder of the boiler equipment based on the equipment feature information and the operation intention includes: The device feature information and the operation intention are converted into semantic vectors; Based on the semantic vector, the current fault intensity score of the boiler equipment is determined; The current fault intensity score is compared with the preset compensation threshold, and the feedforward pulse compensation command is generated when the comparison result shows that the current fault intensity score is greater than the preset compensation threshold.

4. The method according to claim 1, characterized in that, The optimization of the frequency conversion output parameters of the coal feeder of the boiler equipment based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment includes: The original sensor data of the boiler is acquired, and the original sensor data is timestamped to obtain the real-time operating data. Based on the transmission and feedback times of the feedforward pulse compensation command, the communication delay of the feedforward pulse compensation command is calculated. When the communication delay is less than or equal to a preset allowable delay time threshold, the frequency conversion output parameters of the coal feeder of the boiler equipment are optimized based on the real-time operating data.

5. The method according to claim 1, characterized in that, The optimization of the frequency conversion output parameters of the coal feeder of the boiler equipment based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment includes: Obtain the real-time feedback current value of the frequency converter of the boiler equipment, and calculate the current deviation value between the preset current value of the frequency converter and the real-time feedback current value; An adjustment factor is generated based on the ratio of the current deviation value to the preset deviation threshold. The feedforward compensation strength of the frequency converter is determined according to the adjustment factor, so as to optimize the frequency conversion output parameters of the coal feeder according to the feedforward compensation strength.

6. The method according to claim 1, characterized in that, The step of generating start-up or stop control commands for the boiler equipment based on the optimized variable frequency output parameters includes: Obtain the real-time hot air temperature and real-time pulverized coal concentration of the boiler equipment; When the real-time hot air temperature is less than the preset allowable hot air temperature threshold and the real-time pulverized coal concentration is greater than the preset pulverized coal concentration threshold, the stability prediction value of the boiler equipment is calculated. Based on the stability prediction value and the preset instability threshold, the start-up control command or stop control command of the boiler equipment is generated.

7. A boiler start-up and shutdown control device, characterized in that, include: The extraction module is used to respond to the user's voice information and extract the equipment feature information corresponding to the boiler equipment and the operation intention corresponding to the user from the voice information based on the text information corresponding to the voice information; The generation module is used to generate a feedforward pulse compensation command corresponding to the frequency conversion output sequence of the coal feeder of the boiler equipment based on the equipment feature information and the operation intention. The control module is used to optimize the frequency conversion output parameters of the coal feeder of the boiler equipment based on the feedforward pulse compensation command and the real-time operating data of the boiler equipment, so as to generate the start control command or stop control command of the boiler equipment according to the optimized frequency conversion output parameters, and control the boiler equipment to perform the corresponding start action or stop action according to the start control command or stop control command.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the boiler start-up and shutdown control method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the boiler start-up and shutdown control method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed, it is used to implement the boiler start-up and shutdown control method as described in any one of claims 1-6.

Images