Power station key equipment temperature monitoring and early warning method and system

Abstract

The invention provides a power station key equipment temperature monitoring and early warning method and system, and belongs to the field of temperature monitoring, and the method comprises the following steps: S1, collecting temperature data of power station key equipment, and carrying out the preprocessing of the temperature data; s2, performing feature extraction on the preprocessed one-dimensional time sequence data and the preprocessed two-dimensional image data to obtain a first feature and a second feature, and fusing the first feature and the second feature to obtain a fused feature; s3, based on the fusion features, constructing a deep learning prediction model comprising a short-term memory module and a long-term memory module, and optimizing hyper-parameters of the deep learning prediction model by using a distributed hybrid optimization algorithm to obtain an optimized deep learning prediction model; and S4, acquiring temperature data of the key equipment of the power station in real time through an industrial internet platform, predicting the future temperature of the key equipment of the power station by using the optimized deep learning prediction model to obtain a prediction result, and performing adaptive temperature early warning according to the prediction result and a temperature threshold value of the key equipment of the power station. According to the invention, potential temperature abnormity can be identified in time, the accuracy of temperature prediction is remarkably improved, and the sensitivity and reliability of an early warning system are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature monitoring, and in particular to a temperature monitoring and early warning method and system for key equipment in a power station. Background Art

[0002] The power system is an important infrastructure of the national economy, and the safe and stable operation of key power station equipment is directly related to the reliability and economy of power supply. Taking the Three Gorges Power Station as an example, its operating equipment often has external and internal high temperature abnormalities due to environmental factors, equipment aging, poor contact, current overload or internal faults. If these temperature anomalies are not discovered and handled in time, they may lead to equipment failure or even fire, causing serious economic losses and safety hazards.

[0003] With the development of industrial Internet technology, some temperature monitoring systems based on the Internet of Things and artificial intelligence have been proposed. These systems usually include the perception layer, transmission layer, platform layer and application layer, which can realize real-time monitoring and data analysis of equipment temperature.

[0004] However, in the existing technology, existing systems mostly use simple statistical models or single machine learning algorithms, which are difficult to accurately capture complex temperature change patterns. Faced with massive real-time data streams, the data processing capabilities of existing systems often become a bottleneck, and system parameters are often pre-set, which are difficult to dynamically adjust according to actual operating conditions.

[0005] Therefore, finding a method that can not only accurately capture temperature changes but also dynamically adjust the parameters of the temperature monitoring system according to actual conditions is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the invention

[0006] In view of this, the present invention proposes a temperature monitoring and early warning method and system for key power station equipment, which can timely identify potential temperature anomalies, significantly improve the accuracy of temperature prediction, and greatly enhance the sensitivity and reliability of the early warning system.

[0007] The technical solution of the present invention is implemented as follows: The present invention provides a temperature monitoring and early warning method for key equipment in a power station, comprising the following steps:

[0008] S1. Collect temperature data of key equipment in the power station and pre-process the temperature data; the temperature data includes one-dimensional time series data and two-dimensional image data;

[0009] S2, extracting features from the preprocessed one-dimensional time series data and two-dimensional image data respectively to obtain a first feature and a second feature, and fusing the first feature and the second feature to obtain a fused feature;

[0010] S3. Based on the fusion features, a deep learning prediction model including a short-term memory module and a long-term memory module is constructed, and the hyperparameters of the deep learning prediction model are optimized by using a distributed hybrid optimization algorithm to obtain an optimized deep learning prediction model;

[0011] S4. Collect temperature data of key equipment in power plants in real time through the industrial Internet platform, use the optimized deep learning prediction model to predict the future temperature of key equipment in power plants, obtain the prediction results, and perform adaptive temperature warnings based on the prediction results and the temperature thresholds of key equipment in power plants.

[0012] On the basis of the above technical solutions, preferably, the distributed hybrid optimization algorithm includes a genetic algorithm and an ant colony algorithm.

[0013] On the basis of the above technical solution, preferably, the use of a distributed hybrid optimization algorithm to optimize the hyperparameters of the deep learning prediction model specifically includes:

[0014] Step 1: Initialize multiple computing nodes in a distributed computing environment, and generate initial hyperparameter combinations in parallel on each node to form an initial population;

[0015] Step 2: Execute the genetic algorithm operation on the initial population in parallel to generate a new hyperparameter combination of the genetic algorithm;

[0016] Step 3: Execute the ant colony algorithm-based operations on the initial population in parallel to optimize the hyperparameter combination of the genetic algorithm;

[0017] Step 4: Divide the search space into multiple regions, calculate the fitness and fitness change value of each region respectively, determine the application ratio of genetic algorithm and ant colony algorithm in each region according to the fitness of each region, monitor the fitness change value of each region in real time, and adjust the parameters of the distributed hybrid optimization algorithm according to the fitness change value;

[0018] Step 5: Repeat steps 1 to 4 until the iterative optimization conditions are met, integrate the optimal solutions of each node, and determine the optimal hyperparameter combination of the distributed hybrid optimization algorithm based on the optimal solutions of each node.

[0019] Based on the above technical solution, preferably, in step 4, the application ratio of the genetic algorithm and the ant colony algorithm in each region is determined according to the fitness of each region, specifically including:

[0020] The amount of resources in different regions is determined based on the fitness change value of each region. The calculation formula is as follows:

[0021]

[0022] in, Indicates the amount of resources allocated to the ith region, Base Resources indicates the amount of resources initially allocated to each region, Represents the fitness change value of the ith region, TotalFitness Change includes the sum of fitness changes of all regions, Average Fitness (t) Represents the average fitness change value of the region at the tth iteration, Average Fitness (t-1) Represents the average fitness change value of the region at the t-1th iteration;

[0023] The application ratio of genetic algorithm and ant colony algorithm in each region is determined according to the average value of fitness change of all regions. The calculation formula is as follows:

[0024]

[0025]

[0026] in, represents the proportion of genetic algorithm applied in the ith region, Initial GARatio represents the proportion of genetic algorithm set initially, Global Fitness Change represents the average value of fitness change in all regions, Represents the proportion of ant algorithm applied in the i-th region.

[0027] Based on the above technical solution, preferably, in step 5, the optimal hyperparameter combination of the distributed hybrid optimization algorithm is determined according to the optimal solution of each node, specifically including:

[0028] According to the pheromone concentration of the path, it is judged whether the pheromone concentration is too concentrated. The formula is as follows:

[0029]

[0030] Path Stability Index j represents the stability index of the jth path, τ j represents the pheromone concentration on the jth path, τ k represents the pheromone concentration on the kth path, and n represents the total number of all possible paths;

[0031] If so, the mutation rate and pheromone volatility coefficient are dynamically adjusted to obtain the optimal hyperparameter combination of the distributed hybrid optimization algorithm. The formula for adjusting the mutation rate is:

[0032]

[0033] P mrepresents the mutation rate, Υ represents the amplification factor of the mutation rate, ε represents the fitness change threshold,

[0034] The formula for adjusting the pheromone volatility coefficient is:

[0035]

[0036] Wherein, threshold represents the threshold value, and η represents the amplification factor of the pheromone volatility coefficient.

[0037] Based on the above technical solution, preferably, step 2, executing the operation of the genetic algorithm on the initial population in parallel, specifically includes:

[0038] In each computing node, individual selection is performed based on the fitness value. The calculation formula for individual selection is as follows:

[0039]

[0040] Among them, P(θ ij ) represents the probability of the jth hyperparameter individual on the i-th node being selected, θ ij represents the hyperparameter combination of j individuals on i nodes, N represents the total population size, M represents the number of nodes, Fitness(θ ij ) represents the fitness value of the jth hyperparameter individual on the i-th node;

[0041] Inside each node, the selected individuals are crossovered to generate new offspring. The crossover operation is as follows:

[0042] θ child1 =γ.θ p1 +(1-γ).θ p2 ;

[0043] θ child2 =γ.θ p2 +(1-γ).θ p1 ;

[0044] Among them, γ represents the crossover parameter, θ child1 represents the first offspring individual, θ p1 represents the first parent individual, θ child2 represents the second generation individual, θ p2 represents the second parent individual;

[0045] Perform mutation operations on the newly generated offspring, perform individual selection, crossover operations, and mutation operations in parallel on different nodes, generate multiple new XLSTM hyperparameter combinations, replace the original population with the newly generated XLSTM hyperparameter combinations, and complete an iteration; the mutation operation formula is as follows:

[0046] θ mutated =θ child +δ;

[0047] Among them, θ mutated represents the individual after mutation operation, θ child represents the offspring individual before the mutation operation, and δ represents the random mutation value.

[0048] Based on the above technical solution, preferably, in step three, the operation based on the ant colony algorithm is performed on the initial population in parallel, specifically including:

[0049] The ant colony algorithm performs path selection independently on each computing node. The calculation formula for path selection is as follows:

[0050]

[0051] Among them, P j represents the probability of path j being selected, τ j represents the pheromone concentration on path j, η j represents the heuristic information of path j, τ l represents the pheromone concentration on the optional path l, η l represents the heuristic information of the optional path l, α represents the parameter controlling the influence weight of pheromone, and β represents the parameter controlling the influence weight of heuristic information;

[0052] Through path selection, each ant constructs a complete solution corresponding to a complete combination of hyperparameters. Each node independently updates the pheromone concentration on its path in the following way:

[0053]

[0054] in, represents the pheromone concentration on path j at time t+1, represents the pheromone concentration on path j at time t, ρ represents the pheromone volatility coefficient, Δτ j represents the pheromone increment, Q represents the pheromone increment constant, f(θ best ) represents the current optimal solution θ best The fitness function value of

[0055] Path selection and pheromone update operations are performed in parallel on different nodes to generate multiple new XLSTM hyperparameter combinations.

[0056] More preferably, in step S4, the adaptive temperature warning specifically includes:

[0057] According to the temperature adaptability range of different types of power station equipment, the equipment is classified into multiple temperature range groups;

[0058] Set corresponding temperature thresholds for each temperature range group, including the upper and lower limits of the normal operating temperature range, the first-level warning threshold and the second-level warning threshold. The first-level warning threshold is set to the upper limit of the temperature range minus 5°C, and the second-level warning threshold is set to the upper limit of the temperature range;

[0059] Compare the prediction result with the temperature threshold of the device type and perform corresponding warning actions based on the comparison result:

[0060] When the temperature is within the normal operating temperature range, no warning is triggered;

[0061] When the temperature exceeds the normal operating temperature range but does not reach the first-level warning threshold, monitoring is performed;

[0062] When the temperature reaches or exceeds the first-level warning threshold, a first-level warning is triggered;

[0063] When the temperature reaches or exceeds the secondary warning threshold, a secondary warning is triggered and relevant personnel are notified to conduct equipment inspections.

[0064] In a second aspect, the present invention provides a temperature monitoring and early warning system for key equipment in a power station, which adopts the temperature monitoring and early warning method for key equipment in a power station as described above, and comprises:

[0065] The data acquisition module collects the temperature data of key equipment in the power station and pre-processes the temperature data; the temperature data includes one-dimensional time series data and two-dimensional image data;

[0066] A feature extraction module performs feature extraction on the preprocessed one-dimensional time series data and two-dimensional image data respectively to obtain a first feature and a second feature, and fuses the first feature and the second feature to obtain a fused feature;

[0067] A model optimization module, based on the fusion features, constructs a deep learning prediction model including a short-term memory module and a long-term memory module, and optimizes the hyperparameters of the deep learning prediction model using a distributed hybrid optimization algorithm to obtain an optimized deep learning prediction model;

[0068] The temperature prediction module collects temperature data of key power plant equipment in real time through the industrial Internet platform, uses the optimized deep learning prediction model to predict the future temperature of key power plant equipment, obtains the prediction results, and performs adaptive temperature warnings based on the prediction results and the temperature thresholds of key power plant equipment.

[0069] In a third aspect, the present invention provides an electronic device, characterized in that it includes: at least one processor, at least one memory, a communication interface and a bus; wherein:

[0070] The processor, memory, communication interface and bus complete communication with each other;

[0071] The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement the temperature monitoring and early warning method for key power station equipment as described above.

[0072] The temperature monitoring and early warning method of key power plant equipment of the present invention has the following beneficial effects compared with the prior art:

[0073] (1) By combining the industrial Internet platform, deep learning prediction model and distributed hybrid optimization algorithm, potential temperature anomalies can be identified in a timely manner, achieving high-precision prediction and adaptive early warning of the temperature of key power plant equipment. This not only significantly improves the accuracy of temperature prediction, but also greatly improves the sensitivity and reliability of the early warning system, thereby improving the overall operating efficiency and safety of the power plant.

[0074] (2) By allocating the optimization task to multiple computing nodes, the distributed computing resources are fully utilized, which greatly reduces the optimization time. At the same time, the temperature prediction accuracy is improved by using genetic algorithm and ant colony algorithm, which can better adapt to the temperature prediction tasks of different power station equipment.

[0075] (3) By real-time monitoring of the fitness changes of each search area, intelligently adjusting resource allocation and algorithm weight, and allocating more computing resources to areas with high search potential, it not only accelerates the convergence process of the optimal solution, but also increases the probability of finding the global optimal solution. At the same time, it also enhances the adaptability to different types of power plant equipment and operating environments, making the optimization process more flexible and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0077] Figure 1 It is a flow chart of the temperature monitoring and early warning method of key equipment in a power station of the present invention;

[0078] Figure 2 A block diagram of a temperature monitoring and early warning method for key equipment in a power station according to the present invention;

[0079] Figure 3 This is a block diagram of a distributed hybrid optimization algorithm for the temperature monitoring and early warning method for key power station equipment of the present invention. DETAILED DESCRIPTION

[0080] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0081] like Figure 1 and Figure 2 As shown, the present invention provides a temperature monitoring and early warning method for key equipment in a power station, comprising the following steps:

[0082] S1. Collect temperature data of key equipment in the power station and pre-process the temperature data; the temperature data includes one-dimensional time series data and two-dimensional image data.

[0083] In one embodiment of the present application, preprocessing includes: k-means clustering algorithm, smoothing filter algorithm and CNN algorithm.

[0084] It can be understood that the temperature data is firstly used to remove outliers through the k-means clustering algorithm, then the one-dimensional data noise is removed through the moving smoothing filter algorithm, and finally the two-dimensional data noise is removed through the CNN algorithm.

[0085] S2. Perform feature extraction on the preprocessed one-dimensional time series data and two-dimensional image data respectively to obtain a first feature and a second feature, and fuse the first feature and the second feature to obtain a fused feature.

[0086] In one embodiment of the present application, the preprocessed one-dimensional time series data uses an LSTM model to perform feature extraction to obtain a first feature; the preprocessed two-dimensional image data uses a CNN model to perform feature extraction to obtain a second feature.

[0087] In one embodiment of the present application, a feature splicing method is used to merge the first feature and the second feature.

[0088] S3. Based on the fusion features, a deep learning prediction model including a short-term memory module and a long-term memory module is constructed, and the hyperparameters of the deep learning prediction model are optimized by using a distributed hybrid optimization algorithm to obtain an optimized deep learning prediction model;

[0089] Specifically, the distributed hybrid optimization algorithm includes a genetic algorithm and an ant colony algorithm.

[0090] It can be understood that the present application adopts the XLSTM model to predict the power station temperature, and optimizes the hyperparameters in the deep learning prediction model through the improved hybrid genetic ant colony algorithm to obtain the optimized deep learning prediction model.

[0091] like Figure 3 As shown, in one embodiment of the present application, the hyperparameters of the deep learning prediction model are optimized by using a distributed hybrid optimization algorithm, specifically including:

[0092] Step 1: Initialize multiple computing nodes in a distributed computing environment, and generate initial hyperparameter combinations in parallel on each node to form an initial population;

[0093] Step 2: Execute the genetic algorithm operation on the initial population in parallel to generate a new hyperparameter combination of the genetic algorithm;

[0094] Step 3: Execute the ant colony algorithm-based operations on the initial population in parallel to optimize the hyperparameter combination of the genetic algorithm;

[0095] Step 4: Divide the search space into multiple regions, calculate the fitness and fitness change value of each region respectively, determine the application ratio of genetic algorithm and ant colony algorithm in each region according to the fitness of each region, monitor the fitness change value of each region in real time, and adjust the parameters of the distributed hybrid optimization algorithm according to the fitness change value;

[0096] Step 5: Repeat steps 1 to 4 until the iterative optimization conditions are met, integrate the optimal solutions of each node, and determine the optimal hyperparameter combination of the distributed hybrid optimization algorithm based on the optimal solutions of each node.

[0097] It is understandable that the purpose of executing the genetic algorithm is global exploration, and the purpose of executing the ant colony algorithm is local exploration, that is, first use the genetic algorithm to narrow the search space, and then use the ant colony algorithm to find the optimal solution in a smaller search space. Integrate the optimal solutions of each node and find the best one among the optimal solutions of each node, which is the optimal hyperparameter combination of the distributed hybrid optimization algorithm.

[0098] This application makes full use of distributed computing resources by allocating optimization tasks to multiple computing nodes, greatly reducing the optimization time. At the same time, the combination of genetic algorithm and ant colony algorithm not only ensures the global search capability, but also enhances the local optimization effect, effectively avoids the problem of falling into the local optimum, improves the temperature prediction accuracy, and enables it to better adapt to the temperature prediction tasks of different power station equipment.

[0099] In one embodiment of the present application, the number of nodes and the computing resources (including CPU, memory, etc.) of each node are determined, and the total population is evenly distributed to multiple nodes. The formula for node allocation is:

[0100]

[0101] Among them, Population Nodei represents the population on the i-th node, θ i1represents the hyperparameter of the first individual on the i-th node, θ i2 represents the hyperparameters of the second individual on the i-th node, represents the hyperparameter of the last individual on the i-th node, i represents the node index, N represents the total population size, and M represents the number of nodes;

[0102] Each node independently generates a part of the initial individuals, and generates the initial hyperparameter combination in parallel on each computing node. The formula for generating the initial hyperparameter combination is as follows:

[0103]

[0104] Among them, θ ij represents the hyperparameter combination of j individuals on i nodes, and random() represents the random generation function.

[0105] In one embodiment of the present application, the convergence condition of the distributed hybrid optimization algorithm is set. For example, when the fitness change rate is less than a threshold or reaches the maximum number of iterations, the distributed hybrid optimization algorithm is considered to have converged. The iterative process of the algorithm is controlled according to the convergence condition, and after each round of iteration, it is determined whether it is necessary to continue iterating or terminate the algorithm. After convergence, the optimal solution of each node is found, and then the fitness values ​​of these optimal solutions are evaluated and the value of the mean square error of the objective function is calculated to see which one is the smallest. Finally, the best of these optimal solutions is found, which is the output of the distributed hybrid optimization algorithm, that is, the optimal hyperparameter combination.

[0106] Furthermore, in step 4, the application ratio of the genetic algorithm and the ant colony algorithm in each region is determined according to the fitness of each region, specifically including:

[0107] The amount of resources in different regions is determined based on the fitness change value of each region. The calculation formula is as follows:

[0108]

[0109] in, Indicates the amount of resources allocated to the ith region, Base Resources indicates the amount of resources initially allocated to each region, Represents the fitness change value of the ith region, TotalFitness Change includes the sum of fitness changes of all regions, Average Fitness (t) Represents the average fitness change value of the region at the tth iteration, Average Fitness (t-1) Represents the average fitness change value of the region at the t-1th iteration;

[0110] The application ratio of genetic algorithm and ant colony algorithm in each region is determined according to the average value of fitness change of all regions. The calculation formula is as follows:

[0111]

[0112]

[0113] in, represents the proportion of genetic algorithm applied in the ith region, Initial GARatio represents the proportion of genetic algorithm set initially, Global Fitness Change represents the average value of fitness change in all regions, Represents the proportion of ant algorithm applied in the i-th region.

[0114] Furthermore, in step 5, the optimal hyperparameter combination of the distributed hybrid optimization algorithm is determined according to the optimal solution of each node, specifically including:

[0115] According to the pheromone concentration of the path, it is judged whether the pheromone concentration is too concentrated. The formula is as follows:

[0116]

[0117] Path Stability Index j represents the stability index of the jth path, τ j represents the pheromone concentration on the jth path, τ k represents the pheromone concentration on the kth path, and n represents the total number of all possible paths;

[0118] If so, the mutation rate and pheromone volatility coefficient are dynamically adjusted to obtain the optimal hyperparameter combination of the distributed hybrid optimization algorithm. The formula for adjusting the mutation rate is:

[0119]

[0120] P m represents the mutation rate, Υ represents the amplification factor of the mutation rate, ε represents the fitness change threshold,

[0121] The formula for adjusting the pheromone volatility coefficient is:

[0122]

[0123] Wherein, threshold represents the threshold value, and η represents the amplification factor of the pheromone volatility coefficient.

[0124] It can be understood that whether the pheromone concentration is over-concentrated is used to determine whether the region is trapped in the local optimal solution. If it is trapped in the local optimal solution, the parameters of the genetic algorithm and the ant colony algorithm are changed by dynamically adjusting the mutation rate and the pheromone volatility coefficient to jump out of the local optimal solution, thereby obtaining the optimal hyperparameter combination of the distributed hybrid optimization algorithm.

[0125] This application monitors the fitness changes of each search area in real time, intelligently adjusts resource allocation and algorithm proportion, and allocates more computing resources to areas with high search potential. This not only accelerates the convergence process of the optimal solution, but also increases the probability of finding the global optimal solution. At the same time, it also enhances the adaptability to different types of power station equipment and operating environments, making the optimization process more flexible and efficient.

[0126] Furthermore, in step 2, the genetic algorithm operations are executed in parallel on the initial population, specifically including:

[0127] In each computing node, individual selection is performed based on the fitness value. The calculation formula for individual selection is as follows:

[0128]

[0129] Among them, P(θ ij ) represents the probability of the jth hyperparameter individual on the i-th node being selected, θ ij represents the hyperparameter combination of j individuals on i nodes, N represents the total population size, M represents the number of nodes, Fitness(θ ij ) represents the fitness value of the jth hyperparameter individual on the i-th node;

[0130] Inside each node, the selected individuals are crossovered to generate new offspring. The crossover operation is as follows:

[0131] θ child1 =γ.θ p1 +(1-γ).θ p2 ;

[0132] θ child2 =γ.θ p2 +(1-γ).θ p1 ;

[0133] Among them, γ represents the crossover parameter, θ child1 represents the first offspring individual, θ p1 represents the first parent individual, θ child2 represents the second generation individual, θ p2 represents the second parent individual;

[0134] Perform mutation operations on the newly generated offspring, perform individual selection, crossover operations, and mutation operations in parallel on different nodes, generate multiple new XLSTM hyperparameter combinations, replace the original population with the newly generated XLSTM hyperparameter combinations, and complete an iteration; the mutation operation formula is as follows:

[0135] θ mutated =θ child +δ;

[0136] Among them, θ mutated represents the individual after mutation operation, θ child represents the offspring individual before the mutation operation, and δ represents the random mutation value.

[0137] Furthermore, in step three, the operations based on the ant colony algorithm are performed on the initial population in parallel, specifically including:

[0138] The ant colony algorithm performs path selection independently on each computing node. The calculation formula for path selection is as follows:

[0139]

[0140] Among them, P j represents the probability of path j being selected, τ j represents the pheromone concentration on path j, η j represents the heuristic information of path j, τ l represents the pheromone concentration on the optional path l, η l represents the heuristic information of the optional path l, α represents the parameter controlling the influence weight of pheromone, and β represents the parameter controlling the influence weight of heuristic information;

[0141] Through path selection, each ant constructs a complete solution corresponding to a complete combination of hyperparameters. Each node independently updates the pheromone concentration on its path in the following way:

[0142]

[0143] in, represents the pheromone concentration on path j at time t+1, represents the pheromone concentration on path j at time t, ρ represents the pheromone volatility coefficient, Δτ j represents the pheromone increment, Q represents the pheromone increment constant, f(θ best ) represents the current optimal solution θ best The fitness function value of

[0144] Path selection and pheromone update operations are performed in parallel on different nodes to generate multiple new XLSTM hyperparameter combinations.

[0145] S4. Collect temperature data of key equipment in power plants in real time through the industrial Internet platform, use the optimized deep learning prediction model to predict the future temperature of key equipment in power plants, obtain the prediction results, and perform adaptive temperature warnings based on the prediction results and the temperature thresholds of key equipment in power plants.

[0146] Specifically, in step S4, the adaptive temperature warning specifically includes:

[0147] According to the temperature adaptability range of different types of power station equipment, the equipment is classified into multiple temperature range groups;

[0148] Set corresponding temperature thresholds for each temperature range group, including the upper and lower limits of the normal operating temperature range, the first-level warning threshold and the second-level warning threshold. The first-level warning threshold is set to the upper limit of the temperature range minus 5°C, and the second-level warning threshold is set to the upper limit of the temperature range;

[0149] Compare the prediction result with the temperature threshold of the device type and perform corresponding warning actions based on the comparison result:

[0150] When the temperature is within the normal operating temperature range, no warning is triggered;

[0151] When the temperature exceeds the normal operating temperature range but does not reach the first-level warning threshold, monitoring is performed;

[0152] When the temperature reaches or exceeds the first-level warning threshold, a first-level warning is triggered;

[0153] When the temperature reaches or exceeds the secondary warning threshold, a secondary warning is triggered and relevant personnel are notified to conduct equipment inspections.

[0154] This application can more accurately identify potential temperature anomalies by setting personalized temperature thresholds and multi-level warning strategies for different types of power station equipment, which not only reduces the false alarm rate, but also improves the detection sensitivity of real faults. At the same time, the adaptability of the early warning mechanism enables it to dynamically adjust the early warning parameters according to the operating status of the equipment and environmental changes, ensuring the effectiveness of the early warning system in various complex situations, thereby greatly improving the safe operation level of the power station and the efficiency of equipment management.

[0155] Specifically, a specific embodiment is used to illustrate:

[0156] In the temperature range [60,90], a slowly rising trend value alarm can be set;

[0157] When the temperature continues to rise for a period of time and reaches 85 degrees Celsius, a level 1 alarm will be triggered;

[0158] When the temperature continues to rise to 90 degrees Celsius or above, a level 2 alarm is required. After the level 2 alarm, relevant personnel will be notified immediately and the equipment needs to be shut down for inspection.

[0159] This application combines the industrial Internet platform, deep learning prediction model and distributed hybrid optimization algorithm to achieve high-precision prediction and adaptive early warning of the temperature of key power station equipment, which not only significantly improves the accuracy of temperature prediction, but also greatly improves the sensitivity and reliability of the early warning system. By timely identifying potential temperature anomalies, effectively preventing equipment failures, reducing downtime, and improving the overall operating efficiency and safety of the power station, it helps to realize the intelligent operation, maintenance and management of the power station.

[0160] The present invention provides a temperature monitoring and early warning system for key equipment in a power station, which adopts the temperature monitoring and early warning method for key equipment in a power station as described above, and comprises:

[0161] The data acquisition module collects the temperature data of key equipment in the power station and pre-processes the temperature data; the temperature data includes one-dimensional time series data and two-dimensional image data;

[0162] A feature extraction module performs feature extraction on the preprocessed one-dimensional time series data and two-dimensional image data respectively to obtain a first feature and a second feature, and fuses the first feature and the second feature to obtain a fused feature;

[0163] A model optimization module, based on the fusion features, constructs a deep learning prediction model including a short-term memory module and a long-term memory module, and optimizes the hyperparameters of the deep learning prediction model using a distributed hybrid optimization algorithm to obtain an optimized deep learning prediction model;

[0164] The temperature prediction module collects temperature data of key power plant equipment in real time through the industrial Internet platform, uses the optimized deep learning prediction model to predict the future temperature of key power plant equipment, obtains the prediction results, and performs adaptive temperature warnings based on the prediction results and the temperature thresholds of key power plant equipment.

[0165] The present invention provides an electronic device, characterized in that it comprises: at least one processor, at least one memory, a communication interface and a bus; wherein:

[0166] The processor, memory, communication interface and bus complete communication with each other;

[0167] The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement the temperature monitoring and early warning method for key power station equipment as described above.

[0168] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A temperature monitoring and early warning method for key equipment in a power station, characterized in that: The following steps are involved: S1. Collect temperature data of key equipment in the power station and pre-process the temperature data; The temperature data includes one-dimensional time series data and two-dimensional image data; S2, extracting features from the preprocessed one-dimensional time series data and two-dimensional image data respectively to obtain a first feature and a second feature, and fusing the first feature and the second feature to obtain a fused feature; S3. Based on the fusion features, a deep learning prediction model including a short-term memory module and a long-term memory module is constructed, and the hyperparameters of the deep learning prediction model are optimized by using a distributed hybrid optimization algorithm to obtain an optimized deep learning prediction model; S4. Collect temperature data of key equipment in power plants in real time through the industrial Internet platform, use the optimized deep learning prediction model to predict the future temperature of key equipment in power plants, obtain the prediction results, and perform adaptive temperature warnings based on the prediction results and the temperature thresholds of key equipment in power plants.

2. A temperature monitoring and early warning method for key power station equipment according to claim 1, characterized in that: The distributed hybrid optimization algorithm includes a genetic algorithm and an ant colony algorithm.

3. A temperature monitoring and early warning method for key power station equipment according to claim 2, characterized in that: The method of optimizing the hyperparameters of the deep learning prediction model by using a distributed hybrid optimization algorithm specifically includes: Step 1: Initialize multiple computing nodes in a distributed computing environment, and generate initial hyperparameter combinations in parallel on each node to form an initial population; Step 2: Execute the genetic algorithm operation on the initial population in parallel to generate a new hyperparameter combination of the genetic algorithm; Step 3: Execute the ant colony algorithm-based operations on the initial population in parallel to optimize the hyperparameter combination of the genetic algorithm; Step 4: Divide the search space into multiple regions, calculate the fitness and fitness change value of each region respectively, determine the application ratio of genetic algorithm and ant colony algorithm in each region according to the fitness of each region, monitor the fitness change value of each region in real time, and adjust the parameters of the distributed hybrid optimization algorithm according to the fitness change value; Step 5: Repeat steps 1 to 4 until the iterative optimization conditions are met, integrate the optimal solutions of each node, and determine the optimal hyperparameter combination of the distributed hybrid optimization algorithm based on the optimal solutions of each node.

4. A temperature monitoring and early warning method for key power station equipment as claimed in claim 3, characterized in that: In step 4, the application ratio of genetic algorithm and ant colony algorithm in each area is determined according to the fitness of each area, including: The amount of resources in different regions is determined based on the fitness change value of each region. The calculation formula is as follows: in, Indicates the amount of resources allocated to the ith region, Base Resources indicates the amount of resources initially allocated to each region, Represents the fitness change value of the ith region, Total FitnessChange includes the sum of fitness changes of all regions, Average Fitness (t) Represents the average fitness change value of the region at the tth iteration, Average Fitness (t-1) Represents the average fitness change value of the region at the t-1th iteration; The application ratio of genetic algorithm and ant colony algorithm in each region is determined according to the average value of fitness change of all regions. The calculation formula is as follows: in, Indicates the proportion of genetic algorithm applied in the i-th region, Initial GA Ratio indicates the proportion of genetic algorithm set initially, Global Fitness Change indicates the average value of fitness change in all regions, Represents the proportion of ant algorithm applied in the i-th region.

5. A temperature monitoring and early warning method for key power station equipment according to claim 4, characterized in that: In step 5, the optimal hyperparameter combination of the distributed hybrid optimization algorithm is determined according to the optimal solution of each node, including: According to the pheromone concentration of the path, it is judged whether the pheromone concentration is too concentrated. The formula is as follows: path StabilityIndex j represents the stability index of the jth path, τ j represents the pheromone concentration on the jth path, τ k represents the pheromone concentration on the kth path, and n represents the total number of all possible paths; If so, the mutation rate and pheromone volatility coefficient are dynamically adjusted to obtain the optimal hyperparameter combination of the distributed hybrid optimization algorithm. The formula for adjusting the mutation rate is: P m represents the mutation rate, Υ represents the amplification factor of the mutation rate, ε represents the fitness change threshold, The formula for adjusting the pheromone volatility coefficient is: Wherein, threshold represents the threshold value, and η represents the amplification factor of the pheromone volatility coefficient.

6. A temperature monitoring and early warning method for key equipment in a power station as claimed in claim 5, characterized in that: Step 2: Perform the genetic algorithm operations on the initial population in parallel, including: In each computing node, individual selection is performed based on the fitness value. The calculation formula for individual selection is as follows: Among them, P(θ ij ) represents the probability of the jth hyperparameter individual on the i-th node being selected, θ ij represents the hyperparameter combination of j individuals on i nodes, N represents the total population size, M represents the number of nodes, Fitness(θ ij ) represents the fitness value of the jth hyperparameter individual on the i-th node; Inside each node, the selected individuals are crossovered to generate new offspring. The crossover operation is as follows: i child1 =g.i p1 +(1-γ).θ p2 ; i child2 =g.i p2 +(1-γ).θ p1 ; Among them, γ represents the crossover parameter, θ child1 represents the first offspring individual, θ p1 represents the first parent individual, θ child2 represents the second generation individual, θ p2 represents the second parent individual; Perform mutation operations on the newly generated offspring, perform individual selection, crossover operations, and mutation operations in parallel on different nodes, generate multiple new XLSTM hyperparameter combinations, replace the original population with the newly generated XLSTM hyperparameter combinations, and complete an iteration; the mutation operation formula is as follows: i mutated =θ child +d; Among them, θ mutated represents the individual after mutation operation, θ child represents the offspring individual before the mutation operation, and δ represents the random mutation value.

7. A temperature monitoring and early warning method for key equipment in a power station as claimed in claim 6, characterized in that: Step 3: Perform the ant colony algorithm-based operations on the initial population in parallel, including: The ant colony algorithm performs path selection independently on each computing node. The calculation formula for path selection is as follows: Among them, P j represents the probability of path j being selected, τ j represents the pheromone concentration on path j, η j represents the heuristic information of path j, τ l represents the pheromone concentration on the optional path l, η l represents the heuristic information of the optional path l, α represents the parameter controlling the influence weight of pheromone, and β represents the parameter controlling the influence weight of heuristic information; Through path selection, each ant constructs a complete solution corresponding to a complete combination of hyperparameters. Each node independently updates the pheromone concentration on its path in the following way: in, represents the pheromone concentration on path j at time t+1, represents the pheromone concentration on path j at time t, ρ represents the pheromone volatility coefficient, Δτ j represents the pheromone increment, Q represents the pheromone increment constant, f(θ best ) represents the current optimal solution θ best The fitness function value of Path selection and pheromone update operations are performed in parallel on different nodes to generate multiple new XLSTM hyperparameter combinations.

8. A temperature monitoring and early warning method for key equipment in a power station as claimed in claim 1, characterized in that: In step S4, the adaptive temperature warning specifically includes: According to the temperature adaptability range of different types of power station equipment, the equipment is classified into multiple temperature range groups; Set corresponding temperature thresholds for each temperature range group, including the upper and lower limits of the normal operating temperature range, the first-level warning threshold and the second-level warning threshold. The first-level warning threshold is set to the upper limit of the temperature range minus 5°C, and the second-level warning threshold is set to the upper limit of the temperature range; Compare the prediction result with the temperature threshold of the device type and perform corresponding warning actions based on the comparison result: When the temperature is within the normal operating temperature range, no warning is triggered; When the temperature exceeds the normal operating temperature range but does not reach the first-level warning threshold, monitoring is performed; When the temperature reaches or exceeds the first-level warning threshold, a first-level warning is triggered; When the temperature reaches or exceeds the secondary warning threshold, a secondary warning is triggered and relevant personnel are notified to conduct equipment inspections.

9. A temperature monitoring and early warning system for key equipment in a power station, characterized in that: The method for monitoring and warning the temperature of key equipment in a power station according to any one of claims 1 to 8 comprises: The data acquisition module collects the temperature data of key equipment in the power station and pre-processes the temperature data; the temperature data includes one-dimensional time series data and two-dimensional image data; A feature extraction module performs feature extraction on the preprocessed one-dimensional time series data and two-dimensional image data respectively to obtain a first feature and a second feature, and fuses the first feature and the second feature to obtain a fused feature; A model optimization module, based on the fusion features, constructs a deep learning prediction model including a short-term memory module and a long-term memory module, and optimizes the hyperparameters of the deep learning prediction model using a distributed hybrid optimization algorithm to obtain an optimized deep learning prediction model; The temperature prediction module collects temperature data of key power plant equipment in real time through the industrial Internet platform, uses the optimized deep learning prediction model to predict the future temperature of key power plant equipment, obtains the prediction results, and performs adaptive temperature warnings based on the prediction results and the temperature thresholds of key power plant equipment.

10. An electronic device, characterized in that: include: At least one processor, at least one memory, a communication interface, and a bus; wherein: The processor, memory, communication interface and bus complete communication with each other; The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement the temperature monitoring and early warning method for key power station equipment as described in any one of claims 1-8.

Images