Automatic air distribution system for cooling tower

Abstract

The invention discloses an automatic air distribution system for a cooling tower, and relates to the technical field of automatic control of cooling towers, and the automatic air distribution system comprises an extreme weather identification module used for collecting environmental data and identifying extreme weather types; the machine body state monitoring module is used for monitoring the physical state of the cooling tower in real time and outputting a machine body state evaluation coefficient; the state correlation evaluation module is used for combining the output of the extreme weather identification module and the output of the machine body state monitoring module, quantifying the threat degree of weather to the cooling tower through an air distribution correlation strategy, and obtaining a weather-health influence coefficient; and the self-adaptive air distribution module is used for dynamically adjusting an air distribution strategy, outputting corresponding control parameters, realizing self-adaptive air distribution control and balancing cooling efficiency and equipment safety. The problems of high energy consumption and low safety of the cooling tower in extreme weather are solved, higher adaptive capacity and automatic control capacity are achieved, and an innovative intelligent solution is provided for the cooling tower industry.

Description

Technical Field

[0001] The invention relates to the technical field of cooling tower automatic control, in particular to an automatic air distribution system for a cooling tower. Background Art

[0002] As an indispensable equipment in the industrial production process, the stable operation of cooling towers is crucial to ensure production safety and energy efficiency. The traditional cooling tower air distribution control system mainly relies on preset fixed control strategies, which are difficult to adapt to changes in environmental conditions, especially in extreme weather or changeable weather conditions, which may cause insufficient cooling effect or energy efficiency waste. Therefore, how to realize intelligent control of cooling towers in complex climate environments has become a key issue in the current engineering technology field.

[0003] The operating status of the cooling tower is closely related to the external environment, especially meteorological factors such as temperature, humidity, wind speed, and the health of the cooling tower itself, which will affect the cooling effect. Traditional control systems usually ignore the interaction of these multiple factors, resulting in poor adaptability and low operating efficiency under different weather conditions. Therefore, it is urgent to design an intelligent air distribution control system that can combine real-time weather changes and equipment status.

[0004] For example, the existing Chinese patent with publication number CN119245191A discloses a cooling tower optimization control method, device, terminal and medium for a central air-conditioning system, and the method includes: obtaining the wet-bulb approximation of the cooling tower and the comprehensive energy efficiency of the central air-conditioning system under the current environment; according to the wet-bulb approximation, determining the control strategy to be optimized of the cooling tower corresponding to the lowest power consumption value of the cooling tower; optimizing the control strategy to be optimized by adjusting the modulation coefficient of the cooling tower model and the interaction information between the cooling tower and the adjacent cooling tower, and executing the optimized control strategy to maintain the comprehensive energy efficiency of the central air-conditioning system at the current moment greater than or equal to the comprehensive energy efficiency at the previous moment. The present invention calculates the wet-bulb approximation by considering the influence of relative air humidity, wind speed synergy and dynamic change of cooling tower load rate on wet-bulb temperature, and determines the control strategy, and further optimizes and executes the control strategy by modulating the interaction information of the cooling tower model and the parallel cooling tower to improve and maintain the comprehensive energy efficiency of the central air-conditioning system.

[0005] The above scheme calculates the control strategy of the cooling tower based on the wet bulb approximation, but relying solely on the wet bulb approximation may result in the control strategy being unable to adapt to other environmental factors that are more complex or extreme, thereby affecting the control effect; and simple interactive information adjustment may not be sufficient to effectively improve the overall efficiency of the system. To this end, the present invention provides an automatic air distribution system for a cooling tower. Summary of the invention

[0006] The object of the present invention is to provide an automatic air distribution system for a cooling tower to solve the existing problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] An automatic air distribution system for a cooling tower, comprising:

[0009] Extreme weather identification module, fuselage state monitoring module, state association evaluation module and adaptive air distribution module; the extreme weather identification module is used to collect environmental data and identify extreme weather types; the fuselage state monitoring module is used to monitor the physical state of the cooling tower in real time and output the fuselage state evaluation coefficient; the state association evaluation module is used to combine the outputs of the extreme weather identification module and the fuselage state monitoring module, and quantify the degree of threat of weather to the cooling tower through the air distribution association strategy to obtain the weather-health impact coefficient; the adaptive air distribution module is used to dynamically adjust the air distribution strategy, output the corresponding control parameters, realize adaptive air distribution control, and balance cooling efficiency and equipment safety;

[0010] The environmental data include temperature, humidity, wind speed, precipitation, air pressure, air quality and lightning activity indicators; a weather training set Dwe is constructed through the environmental data, a critical value of all data in the environmental data is set, and data in the weather training set that is greater than the corresponding weather type critical value is marked as the corresponding extreme weather type, and the extreme weather types include high temperature, low temperature, heavy rain, strong wind and lightning.

[0011] A further improvement of the present invention is that the extreme weather recognition module is equipped with an extreme weather recognition algorithm, and the extreme weather recognition algorithm selects the feature with the largest weather information gain as the decision feature of the current node based on the decision tree method. The weather information gain calculation formula is:

[0012]

[0013] Among them, IG(Dwe,A j ) indicates weather characteristics A j The gain on the training set Dwe, A j represents the jth weather feature in the training set Dwe, j represents any number from 1 to the total number of features in the training set Dwe, Entropy(Dwe) represents the entropy of the training set Dwe, and the calculation formula is where p i is the probability of weather type i in the training set Dwe, Dwe v According to weather characteristics A j The subset divided by the value v, |Dwe v| represents the size of the subset, |Dwe| represents the size of the data set; then the same operation is recursively performed on the subset until convergence, and finally the leaf node of the decision tree represents the specific weather type classification result.

[0014] The present invention is further improved in that the wind distribution association strategy includes a weather threat calculation model, which includes extracting the weather type output of the extreme weather identification module and its corresponding environmental data as the core parameters of weather decision-making, and after standardizing all environmental data, calculating the weather threat index through the dynamic weather weight control principle. The weather threat index calculation formula is: Among them, X n represents the core parameter of weather decision making, X ε represents the εth environmental data parameter, n∈(1,2,…,8), and α represents the weight of the core parameter of weather decision making.

[0015] A further improvement of the present invention is that the dynamic weather weight control principle specifically includes judging the degree of influence of current environmental data on the fuselage state by quantifying the difference between current environmental data and the critical value of weather type, and obtaining the weight of weather decision core parameter. The weather decision core parameter weight calculation formula is α=0.3×(1+(|cow-Scow| / Scow)), wherein cow represents the weather decision core parameter, and Scow represents the weather type critical value corresponding to the weather decision core parameter.

[0016] A further improvement of the present invention is that the air distribution association strategy also includes a weather-health association model, which constructs samples based on time series, uses a bidirectional LSTM basic prediction model, takes the weather threat index and weather type as input, outputs the predicted value of the fuselage's own state change corresponding to the current weather threat index through the dynamic forgetting principle of fuselage health, and judges the LSTM average information gain brought by different weather types through the LSTM information gain to obtain the weather-health impact coefficient.

[0017] A further improvement of the present invention is that the dynamic forgetting principle of the airframe health is realized by regulating the forgetting gate, and the calculation process is expressed as: Among them, c t-1 represents the cell state at time t-1, W f represents the weight matrix of the forget gate, The predicted value of the change in the fuselage state evaluation coefficient output by the LSTM model at time t-1, htv t-1 represents the fuselage status evaluation coefficient at time t-1, x htv Indicates input data; b f represents the bias term, the function sig() represents the Sigmoid activation function, Δhtv tIndicates the change in the fuselage state evaluation coefficient at time t Δhtv t =htv t -htv t-Δt , where Δt represents the set time interval, htv t Represents the fuselage status evaluation coefficient at time t.

[0018] The present invention is further improved in that the LSTM average information gain is sorted from large to small to obtain the weather type importance sequence IMW; the weather type importance imw corresponding to the current weather type is extracted. n , the predicted value of the fuselage state change is standardized to obtain the fuselage health change prediction standard value Sy htv , and obtain the weather-health impact coefficient Wherein, ave(IMW) represents the mean value of the data in the weather type importance sequence.

[0019] The present invention is further improved in that the adaptive air distribution module includes an adaptive air distribution algorithm for cooling towers, and the adaptive air distribution algorithm for cooling towers is based on reinforcement learning and outputs corresponding control parameters. The specific steps include:

[0020] Step 1: Design state space, including weather threat index, fuselage state evaluation coefficient and weather-health impact coefficient;

[0021] Step 2: designing the action space, wherein the action space includes the adjustment parameters of the cooling tower, including the fan speed, the water pump flow, and the opening of the spray system;

[0022] Step 3: Design the reward function RR. Among them, Δtem represents the temperature change at a time interval of Δt, ctc represents the energy consumption at a time interval of Δt, ω1 and ω2 represent weight coefficients; Δhtv no Indicates the change of the fuselage state evaluation coefficient at the current time interval Δt;

[0023] Step 4: The cooling tower adaptive air distribution algorithm is trained by the policy gradient method, and the agent selects an action according to the current weather and the fuselage state evaluation coefficient;

[0024] Step 5: By setting the intelligent sampling principle, the sampling probability is intelligently adjusted to jump out of the local optimal solution during the search process;

[0025] Step 6: When the reward function reaches the set reward function threshold, output the action space parameters at this time.

[0026] A further improvement of the present invention is that the intelligent sampling principle is determined by the weather-health impact coefficient, and the intelligent sampling principle follows the formula where s no represents the state space parameter, a no represents the action space parameter, π(a no |s no ) means in state s no Next, the agent chooses action a no The probability of , μ represents the mean value of the strategy output, represents the initial variance.

[0027] A further improvement of the present invention is that the fuselage state evaluation coefficient is obtained by weightedly averaging the minimum-maximum normalization of the average cooling tower fuselage vibration amplitude, current fluctuation and damper opening and closing flexibility within the monitoring time TT with a time interval of ΔTT to obtain the fuselage state evaluation coefficient.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. The present invention first introduces an extreme weather recognition module, so that the system can recognize and classify different weather types in real time, thereby adjusting the air distribution strategy of the cooling tower according to the specific weather conditions; combined with the fuselage status monitoring module, the system can monitor the health status of the cooling tower in real time to further quantify the impact of environmental changes on the equipment, and ensure the safety of the equipment under extreme weather conditions;

[0030] 2. Secondly, the weather-health impact coefficient is used to quantify the impact of different weather types on the health of the cooling tower body. The system can accurately adjust the cooling tower's air distribution strategy according to real-time weather information and changes in the body state; dynamically adjust the balance between exploration and utilization, and affect the width of the sampling distribution through changes in the weather-health impact coefficient, thereby optimizing the exploration and utilization strategies in the reinforcement learning model, so that the system can flexibly adjust the air distribution plan under changing weather conditions;

[0031] 3. Use the policy gradient method for reinforcement learning training, and avoid lingering in the trap of local optimal solutions through the principle of intelligent sampling. The system dynamically adjusts the exploration degree (i.e., the sampling distribution width) to flexibly adjust the trade-off between exploration and utilization under weather conditions of different sensitivity, avoiding excessive punishment and the trouble of local optimal solutions; through precise sampling strategies, the intelligent agent can conduct more exploration when needed, focus more on optimizing the current strategy in known situations, and improve the system's adaptability to environmental changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a framework diagram of an automatic air distribution system for a cooling tower according to the present invention;

[0033] Figure 2 The present invention is a flow chart of a cooling tower self-adaptive air distribution algorithm for an automatic air distribution system for a cooling tower. DETAILED DESCRIPTION

[0034] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0035] The term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0036] Example 1

[0037] Figure 1 The framework diagram of an automatic air distribution system for a cooling tower disclosed in this embodiment is shown, including:

[0038] Extreme weather identification module, fuselage status monitoring module, status correlation evaluation module and adaptive air distribution module;

[0039] The extreme weather identification module is used to collect environmental data and identify extreme weather types;

[0040] The fuselage state monitoring module is used to monitor the physical state of the cooling tower in real time and output the fuselage state evaluation coefficient;

[0041] The state association evaluation module is used to combine the outputs of the extreme weather identification module and the fuselage state monitoring module, quantify the threat degree of weather to the cooling tower through the wind distribution association strategy, and obtain the weather-health impact coefficient;

[0042] The adaptive air distribution module is used to dynamically adjust the air distribution strategy, output corresponding control parameters, realize adaptive air distribution control, and balance cooling efficiency and equipment safety.

[0043] The extreme weather recognition module is equipped with an extreme weather recognition algorithm. The environmental data includes temperature, humidity, wind speed, precipitation, air pressure, air quality and lightning activity indicators. The weather training set Dwe is constructed through the environmental data, and the critical value of all data in the environmental data is set. The data in the weather training set that is greater than the corresponding weather type critical value is marked as the corresponding extreme weather type. The extreme weather types include high temperature, low temperature, heavy rain, strong wind and lightning.

[0044] Since weather information is relatively discrete, and various weather factors, such as temperature, humidity, wind speed, etc., usually have a nonlinear relationship with the weather type, for example, the combination of high temperature and humidity may be crucial to determine whether a rainstorm occurs. Decision trees can automatically process these complex nonlinear relationships by recursively segmenting data, and screen out the most meaningful features layer by layer. Therefore, the decision tree-based method of the present invention selects the feature with the largest weather information gain as the decision feature of the current node. Since the changes in weather data between different regions and time periods have local dependence, by calculating the weather information gain, the importance of different features to the prediction results can be quantified, which helps optimize the decision tree structure and avoid over-reliance on a certain feature. The weather information gain calculation formula is:

[0045]

[0046] Among them, IG(Dwe,A j ) indicates weather characteristics A j The gain on the training set Dwe, A j represents the jth weather feature in the training set Dwe, j represents any number from 1 to the total number of features in the training set Dwe, Entropy(Dwe) represents the entropy of the training set Dwe, indicating the uncertainty of the data, and the calculation formula is where p i is the probability of weather type i in the training set Dwe, Dwe v According to weather characteristics A j The subset divided by the value v, |Dwe v | represents the size of the subset, |Dwe| represents the size of the data set;

[0047] The feature with the largest weather information gain is selected as the decision feature of the current node, and then the same operation is recursively performed on the subset until the stopping condition is met. Finally, the leaf node of the decision tree represents the specific weather type classification result.

[0048] The fuselage state monitoring module performs minimum-maximum normalization on the average vibration amplitude, current fluctuation and damper opening and closing flexibility of the cooling tower body within the monitoring time TT with a time interval of ΔTT, and then obtains the fuselage state evaluation coefficient by weighted average. The larger the fuselage state evaluation coefficient, the worse the cooling tower body state. The damper opening and closing flexibility is obtained by recording the time from the damper being fully closed to being fully opened.

[0049] The wind distribution association strategy includes a weather threat calculation model, which includes extracting the weather type output of the extreme weather identification module and its corresponding environmental data as the weather decision core parameters, and after standardizing all environmental data, calculating the weather threat index through the dynamic weather weight control principle, and the dynamic weather weight control principle specifically includes:

[0050] By quantifying the difference between the current environmental data and the critical value of the weather type, the influence of the current environmental data on the fuselage state is judged. The greater the difference, the greater the influence of the current environmental data on the fuselage, so the weight of the weather decision core parameter is increased; thereby, the weather decision core parameter weight is obtained, and the weather decision core parameter weight calculation formula is α=0.3×(1+(|cow-Scow| / Scow)), where cow represents the weather decision core parameter, and Scow represents the weather type critical value corresponding to the weather decision core parameter; the weather threat index calculation formula is Among them, X n represents the core parameter of weather decision making, X ε represents the εth environmental data parameter, n∈(1,2,…,8);

[0051] The air distribution association strategy also includes a weather-health association model. First, the change in the fuselage state evaluation coefficient Δhtv at time t is calculated. t =htv t -htv t-Δt , where Δt represents the set time interval, htv t represents the fuselage status evaluation coefficient at time t;

[0052] Since environmental data and airframe health parameters have time series characteristics, not only is the current state change of the airframe greatly affected by the environment, but the previous and next health data of the current state of the airframe are also closely correlated. The unique memory unit and gating mechanism of LSTM can more effectively capture and retain long-term dependencies. Therefore, samples are constructed based on time series, and the bidirectional LSTM basic prediction model can capture comprehensive forward and backward information at the same time. The weather threat index and weather type are used as input, and the predicted value of the airframe's own state change corresponding to the current weather threat index is output;

[0053] The information forgetting ratio determined by the forget gate is directly related to how much output information Δhtv from time t-1 can be obtained at time t t-1And the input information at the current time t is filtered by the forget gate, so it is retained and updated to the unit state at time t; considering that the predicted value and the true value at time t-1 already exist when the prediction is made at time t, the error between the predicted value and the true value at time t-1 is calculated. This error value reflects the accuracy of the LSTM prediction of the fuselage state evaluation coefficient at time t-1. A small error means that the fuselage state at this moment has a positive impact on the calculation of the fuselage state change by the predicted value, while a large error indicates that the historical prediction is of limited help in predicting the fuselage state change at time t.

[0054] Therefore, the dynamic forgetting principle of fuselage health is set. In the process of calculating the weather threat index through the dynamic weather weight control principle, the accuracy of historical predictions is further introduced. The present invention not only considers the standardized environmental data, but also inputs the error between the predicted value and the true value of the fuselage state evaluation coefficient at time t-1 as additional information into the forgetting gate of the LSTM network. In this way, the LSTM network can dynamically adjust the passing rate of the forgetting gate according to the error of the fuselage state evaluation coefficient, intelligently determine how much historical information to forget, and thus optimize the calculation process of the weather threat index;

[0055] The calculation process of the forget gate is expressed as: Among them, c t-1 represents the cell state at time t-1, W f represents the weight matrix of the forget gate, The predicted value of the change in the fuselage state evaluation coefficient output by the LSTM model at time t-1, htv t-1 represents the fuselage status evaluation coefficient value at time t-1, x htv Indicates input data; b f Represents the bias term, and the function sig() represents the Sigmoid activation function;

[0056] The LSTM information gain is used to determine the LSTM average information gain brought by different weather types, and the order is sorted from large to small to obtain the weather type importance sequence IMW; the weather type importance imw corresponding to the current weather type is extracted n , the predicted value of the fuselage state change is standardized to obtain the fuselage health change prediction standard value Sy htv , and obtain the weather-health impact coefficient Wherein, ave(IMW) represents the mean value of the data in the weather type importance sequence. The calculation formula for the importance of the εth weather type in IMW is:

[0057]

[0058] Among them, N trees represents the total number of trees, S εrepresents the split node of weather type ε in tree T, Gain(S ε ) represents the split node S ε LSTM information gain;

[0059] Example 2

[0060] Based on the inventive concept of Example 1, the present invention provides a cooling tower adaptive air distribution algorithm. Figure 2 The present invention shows a flow chart of an adaptive air distribution algorithm for a cooling tower in an automatic air distribution system for a cooling tower. The adaptive air distribution algorithm for a cooling tower is based on reinforcement learning and outputs corresponding control parameters. The specific steps include:

[0061] Step 1: Design state space, including weather threat index, fuselage state evaluation coefficient and weather-health impact coefficient;

[0062] Step 2: designing the action space, wherein the action space includes the adjustment parameters of the cooling tower, including the fan speed, the water pump flow, and the opening of the spray system;

[0063] Step 3: Design the reward function RR. Among them, Δtem represents the temperature change at a time interval of Δt, reflecting the cooling effect of the cooling tower, and ctc represents the energy consumption at a time interval of Δt. It can reflect the energy efficiency of the cooling tower, ω1 and ω2 represent weight coefficients; Δhtv no Indicates the change of the fuselage state evaluation coefficient at the current time interval Δt;

[0064] Step 4: The cooling tower adaptive air distribution algorithm is trained by the policy gradient method. The intelligent agent selects actions based on the current weather and fuselage state evaluation coefficient by minimizing the reward function RR target and optimizing the strategy;

[0065] Step 5: Since the policy gradient method updates the policy by calculating the gradient, the gradient information is local, which means that it only reflects the optimal direction of the current state, resulting in excessive punishment, which in turn causes the agent to be too conservative, causing the cooling tower air distribution parameters to gradually converge to a local optimal solution during the optimization process, without sufficient information to explore other parts of the policy space. Therefore, by setting the intelligent sampling principle, the probability of sampling is intelligently adjusted to jump out of the local optimal solution during the search process; the specific performance is as follows:

[0066] In traditional sampling, the formula π(a no |s no)=exp(RR / τ), τ represents the width of temperature distribution. If this method is large, then even if the reward RR is small, the probability of selecting the action will not be too low, and the distribution will be relatively wide; if τ is small, the action with larger reward will be more likely to be selected, and the distribution will be more concentrated.

[0067] The weather-health impact coefficient reflects the sensitivity of the fuselage health to the current weather type. If the weather-health impact coefficient is large, it means that the fuselage is more sensitive to the current weather type. It can be considered that the system needs more precise control, that is, reduce exploration and improve utilization. In this case, the distribution width can be reduced; if the weather-health impact coefficient is small, it means that the fuselage has a weak response to the current weather type. It can be considered that the current weather has little impact on the fuselage and requires more exploration. In this case, the distribution width can be increased through Gaussian distribution to improve randomness and exploration. Therefore, the sampling probability is determined by the weather-health impact coefficient, and the intelligent sampling principle follows the formula where s no represents the state space parameter, a no represents the action space parameter, π(a no |s no ) means in state s no Next, the agent chooses action a no The probability of , μ represents the mean value of the strategy output, represents the initial variance; during the training process, the dynamic adjustment of the variance helps to better adapt to highly sensitive weather conditions;

[0068] The present invention can enable the intelligent agent to better cope with different types of weather conditions, determine when to perform more frequent sampling, and adaptively adjust the sampling probability by dynamically adjusting the distribution width according to the weather-health impact coefficient; for wind distribution strategies with greater weather impact, they are often prone to continuous trapping in local optimality, so the distribution width is reduced to give them a higher selection probability; for wind distribution strategies with less weather impact, their excellent control strategy characteristics are protected, while avoiding premature end of exploration of them, reducing the probability of selecting them, enhancing their exploration ability, breaking the inherent mode of the algorithm, and prompting the algorithm to explore new solution space; avoiding the algorithm from stagnating at a local optimal solution too early, so as to enhance the overall search range of the algorithm, thereby improving the efficiency and effect of the algorithm as a whole.

[0069] Compared with existing technologies, the system's adaptability to environmental changes is improved; the balance between exploration and utilization is controlled by the weather-health impact coefficient, so that the intelligent agent can conduct more exploration when necessary and focus more on optimizing the current air distribution parameters under deterministic conditions; by accurately controlling the degree of exploration, the cooling tower's air distribution strategy will be more efficient, avoiding meaningless random adjustments, and thus better adapting to cooling needs under different weather conditions.

[0070] Step 6: When the reward function reaches the set reward function threshold, output the action space parameters at this time.

[0071] The threshold and weight may be set by default according to the present invention, or may be set by an operator.

[0072] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0073] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0074] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0075] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0076] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. An automatic air distribution system for a cooling tower, characterized in that: include: Extreme weather identification module, fuselage status monitoring module, status correlation evaluation module and adaptive air distribution module; The extreme weather recognition module is used to collect environmental data and identify extreme weather types; the fuselage state monitoring module is used to monitor the physical state of the cooling tower in real time and output the fuselage state evaluation coefficient; the state association evaluation module is used to combine the outputs of the extreme weather recognition module and the fuselage state monitoring module, and quantify the degree of threat of weather to the cooling tower through the air distribution association strategy to obtain the weather-health impact coefficient; the adaptive air distribution module is used to dynamically adjust the air distribution strategy, output the corresponding control parameters, realize adaptive air distribution control, and balance cooling efficiency and equipment safety; The environmental data include temperature, humidity, wind speed, precipitation, air pressure, air quality and lightning activity indicators; a weather training set Dwe is constructed through the environmental data, a critical value of all data in the environmental data is set, and data in the weather training set that is greater than the corresponding weather type critical value is marked as the corresponding extreme weather type, and the extreme weather types include high temperature, low temperature, heavy rain, strong wind and lightning.

2. The automatic air distribution system for a cooling tower according to claim 1, characterized in that: The extreme weather recognition module is equipped with an extreme weather recognition algorithm. The extreme weather recognition algorithm selects the feature with the largest weather information gain as the decision feature of the current node based on the decision tree method. The weather information gain calculation formula is: Among them, IG(Dwe,A j ) indicates weather characteristics A j The gain on the training set Dwe, A j represents the jth weather feature in the training set Dwe, j represents any number from 1 to the total number of features in the training set Dwe, Entropy(Dwe) represents the entropy of the training set Dwe, and the calculation formula is where p i is the probability of weather type i in the training set Dwe, Dwe v According to weather characteristics A j The subset divided by the value v, |Dwe v | represents the size of the subset, |Dwe| represents the size of the data set; then the same operation is recursively performed on the subset until convergence, and finally the leaf node of the decision tree represents the specific weather type classification result.

3. The automatic air distribution system for a cooling tower according to claim 1, characterized in that: The wind distribution association strategy includes a weather threat calculation model, which includes extracting the weather type output of the extreme weather identification module and its corresponding environmental data as the core parameters of weather decision-making. After standardizing all environmental data, the weather threat index is calculated through the dynamic weather weight control principle. The weather threat index calculation formula is: Among them, X n represents the core parameter of weather decision making, X ε represents the εth environmental data parameter, n∈(1,2,…,8), and α represents the weight of the core parameter of weather decision making.

4. The automatic air distribution system for a cooling tower according to claim 3, characterized in that: The dynamic weather weight control principle specifically includes judging the degree of influence of the current environmental data on the fuselage state by quantifying the difference between the current environmental data and the critical value of the weather type, and obtaining the weight of the weather decision core parameter. The weather decision core parameter weight calculation formula is α=0.3×(1+(|cow-Scow| / Scow)), wherein cow represents the weather decision core parameter, and Scow represents the weather type critical value corresponding to the weather decision core parameter.

5. The automatic air distribution system for a cooling tower according to claim 1, characterized in that: The wind distribution association strategy also includes a weather-health association model. The weather-health association model constructs samples based on time series, uses a bidirectional LSTM basic prediction model, takes the weather threat index and weather type as input, outputs the predicted value of the fuselage's own state change corresponding to the current weather threat index through the dynamic forgetting principle of fuselage health, and judges the LSTM average information gain brought by different weather types through the LSTM information gain to obtain the weather-health impact coefficient.

6. The automatic air distribution system for a cooling tower according to claim 5, characterized in that: The dynamic forgetting principle of airframe health is realized by regulating the forgetting gate. The calculation process is expressed as: Among them, c t-1 represents the cell state at time t-1, W f represents the weight matrix of the forget gate, The predicted value of the change in the fuselage state evaluation coefficient output by the LSTM model at time t-1, htv t-1 represents the fuselage status evaluation coefficient at time t-1, x htv Indicates input data; b f represents the bias term, the function sig() represents the Sigmoid activation function, Δhtv t Indicates the change in the fuselage state evaluation coefficient at time t Δhtv t =htv t -htv t-Δt , where Δt represents the set time interval, htv t Represents the fuselage status evaluation coefficient at time t.

7. The automatic air distribution system for a cooling tower according to claim 5, characterized in that: Sort the LSTM average information gain from large to small to obtain the weather type importance sequence IMW; extract the weather type importance imw corresponding to the current weather type n , the predicted value of the fuselage state change is standardized to obtain the fuselage health change prediction standard value Sy htv , and obtain the weather-health impact coefficient Wherein, ave(IMW) represents the mean value of the data in the weather type importance sequence.

8. The automatic air distribution system for a cooling tower according to claim 1, characterized in that: The adaptive air distribution module includes an adaptive air distribution algorithm for cooling towers. The adaptive air distribution algorithm for cooling towers is based on reinforcement learning and outputs corresponding control parameters. The specific steps include: Step 1: Design state space, including weather threat index, fuselage state evaluation coefficient and weather-health impact coefficient; Step 2: designing the action space, wherein the action space includes the adjustment parameters of the cooling tower, including the fan speed, the water pump flow, and the opening of the spray system; Step 3: Design the reward function RR. Among them, Δtem represents the temperature change at a time interval of Δt, ctc represents the energy consumption at a time interval of Δt, ω1 and ω2 represent weight coefficients; Δhtv no Indicates the change of the fuselage state evaluation coefficient at the current time interval Δt; Step 4: The cooling tower adaptive air distribution algorithm is trained by the policy gradient method, and the agent selects an action according to the current weather and the fuselage state evaluation coefficient; Step 5: By setting the intelligent sampling principle, the sampling probability is intelligently adjusted to jump out of the local optimal solution during the search process; Step 6: When the reward function reaches the set reward function threshold, output the action space parameters at this time.

9. The automatic air distribution system for a cooling tower according to claim 8, characterized in that: The intelligent sampling principle is determined by the weather-health impact coefficient, and the intelligent sampling principle follows the formula where s no represents the state space parameter, a no represents the action space parameter, π(a no |s no ) means in state s no Next, the agent chooses action a no The probability of , μ represents the mean value of the strategy output, represents the initial variance.

10. The automatic air distribution system for a cooling tower according to claim 1, characterized in that: The fuselage state evaluation coefficient is obtained by performing minimum-maximum normalization on the average cooling tower fuselage vibration amplitude, current fluctuation and damper opening and closing flexibility within the monitoring time TT with a time interval of ΔTT, and then taking the weighted average thereof.

Images