Adaptive intelligent temperature control multi-cooling tower motor linkage integrated system and control method

Abstract

The application discloses a self-adaptive intelligent temperature control multi-cooling tower motor linkage integrated system and a control method, which comprises the following steps: outputting preprocessed multi-cooling tower system operation data sequences; inputting the preprocessed multi-cooling tower system operation data sequences into a disturbance adjustable Transformer model as input, and outputting temperature control trend characteristic sequences; inputting the temperature control trend characteristic sequences into an improved Diff-MoE dynamic scheduling network, and outputting motor unit target control strategies; generating motor unit execution instruction sets; obtaining motor unit real-time execution state feedback data; judging whether a motor open-phase fault, a motor overload fault, a contactor sticking fault or a thermal protection tripping fault occurs in the multi-cooling tower system in real time, and if it is judged that a fault state exists, automatically calling a standby motor switching loop, connecting the standby motor into a corresponding cooling tower loop and updating the motor unit execution instruction sets, so that stable operation is realized. The application realizes intelligent division of labor and load balancing among linkage motor units.

Description

Technical Field

[0001] The present invention relates to the technical field of cooling towers, and particularly to an adaptive intelligent temperature-controlled multi-cooling tower motor linkage integration system and a control method thereof. Background Art

[0002] With the improvement of industrial automation level, multi-cooling tower systems are widely used in large-scale industrial circulating water cooling and heating ventilation and air conditioning projects. The existing multi-cooling tower motor linkage integration control methods mainly rely on traditional PID control, fuzzy control or threshold-based logic decision-making. By collecting inlet and outlet water temperatures, motor states and environmental wind speed parameters in real time, the start-stop linkage and power regulation of cooling tower fans or water pumps are realized.

[0003] The existing technologies show the following defects under actual complex working conditions: The traditional control strategy has a significant lag in response to the change of heat load of the cooling tower group. When the production line load changes suddenly or there are sudden external environmental disturbances, the system often responds slowly, resulting in a decrease in temperature control accuracy, and even overshoot and overcompensation phenomena. Conventional control algorithms are vulnerable to interference in industrial field environments with strong electromagnetic interference, equipment aging or motor loop failures, resulting in the failure of motor start-stop commands and frequent misoperations, and it is difficult to achieve highly reliable linkage control. Existing solutions usually adopt static grouping or simple rotation strategies for motor start-stop allocation, and it is difficult to achieve adaptive optimization allocation of main loads and auxiliary loads according to actual load trends and disturbance characteristics, resulting in high energy consumption and increased equipment wear.

[0004] When the existing system based on a single data stream and fixed control logic processes the environmental disturbances, random noises and thermal inertia effects superimposed in the temperature signal, it lacks in-depth discrimination and prediction of the disturbance intensity, change acceleration and thermal load dynamics, resulting in insufficient modeling and control of the dynamic characteristics of the multi-cooling tower system, and further affecting the energy efficiency and reliability of the entire cooling circuit. Summary of the Invention

[0005] An object of the present invention is to provide an adaptive intelligent temperature-controlled multi-cooling tower motor linkage integration system and a control method thereof, and the present invention realizes intelligent division of labor and load balance among linkage motor groups.

[0006] A control method for an adaptive intelligent temperature-controlled multi-cooling tower motor linkage integration system according to an embodiment of the present invention includes: Collecting and preprocessing the original data set of the operation of the multi-cooling tower system through the adaptive intelligent temperature-controlled multi-cooling tower motor linkage integration system, and outputting a preprocessed multi-cooling tower system operation data sequence; The preprocessed multi-cooling tower system operation data sequence is fed into the perturbation-adjustable Transformer model. The multi-head self-attention mechanism is used to generate temperature control perturbation discrimination weights, complete high-frequency random perturbation denoising and extract temperature control trend feature vectors, and output temperature control trend feature sequence. The temperature control trend feature sequence is input into the load prediction algorithm model running on the host computer or edge computing module. The load prediction algorithm model predicts the heat load change trend for the next control cycle based on historical and real-time data, and outputs the heat load trend index and the expected temperature rise rate. The temperature control trend feature sequence is input into the improved Diff-MoE dynamic scheduling network. Based on the gating routing mechanism, the expert set of motor control strategies is matched to obtain the target power allocation vector and target start-stop combination vector for each cooling tower motor unit, and the target control strategy of the motor unit is output. The PLC controller receives the heat load trend index, expected temperature rise rate and real-time temperature deviation signal. Based on the built-in multi-level fuzzy rules and dynamic priority scheduling logic, it calculates the target power allocation coefficient and target start-stop status of each cooling tower motor and generates a motor group collaborative control strategy. Based on the motor unit collaborative control strategy and the motor unit target control strategy, the multi-cooling tower motors are divided into main load motor units and auxiliary load motor units. The priority of the main load motor units, the load compensation ratio of the auxiliary load motor units and the motor start-stop sequence are determined, and the motor unit execution instruction set is generated. The motor set execution instruction set is sent to the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, which sequentially triggers the PLC controller module and the multi-motor drive module to drive the corresponding motor set to start or stop in the motor start-stop sequence and operate according to the target power allocation, thereby obtaining real-time execution status feedback data of the motor set. Based on the real-time execution status feedback data of the motor set, it can determine in real time whether the multi-cooling tower system has a motor phase loss fault, motor overload fault, contactor sticking fault or thermal protection trip fault. If a fault condition is determined, the backup motor switching circuit is automatically called to connect the backup motor to the corresponding cooling tower circuit and update the motor set execution instruction set to achieve stable operation.

[0007] Optionally, the step of collecting and preprocessing the raw data set of the multi-cooling tower system operation through the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system includes: The system collects raw data on inlet and outlet water temperatures, ambient wind speed, ambient humidity, motor operating status, and historical heat load of a multi-cooling tower system. It performs time synchronization processing and unifies the encoding format to obtain a set of raw data for the operation of the multi-cooling tower system. The system then performs noise filtering, missing data completion, and normalization on the raw data set to output a preprocessed sequence of operating data for the multi-cooling tower system.

[0008] Optionally, the step of the PLC controller generating the motor group cooperative control strategy includes: Establish a multi-level fuzzy rule base with real-time temperature deviation, temperature change rate, and heat load trend index as inputs; Determine the total cooling power level required by the system based on the magnitude and direction of the real-time temperature deviation; The total cooling power level is proactively adjusted by incorporating the heat load trend index.

[0009] Optionally, the perturbation-tunable Transformer model includes: The preprocessed multi-cooling tower system operation data sequence is input into the perturbation-adjustable Transformer model. The preprocessed multi-cooling tower system operation data vector at each time step is transformed by the mapping matrix and time position encoding to obtain the input embedding vector. In the perturbation-tunable Transformer model, based on the inlet and outlet water temperature components in the preprocessed multi-cooling tower system operation data sequence, the temperature change rate sequence and temperature curvature sequence at each moment are calculated. The input embedding vector, temperature change rate sequence and temperature curvature sequence corresponding to the current moment are jointly mapped to construct the perturbation modulation vector. Within each attention head of the perturbation-tunable Transformer model, for each time step, the input embedding vector is linearly transformed using the mapping matrix based on the perturbation modulation vector to generate the query vector, key vector, and value vector, and the perturbation-tunable scaling dot product attention score is calculated. Based on the perturbation-adjustable scaling dot product attention score, a normalization operation is used to obtain the attention weight of each time step to all time steps. The attention weights are used to sum the corresponding value vectors to obtain the attention output vector. The attention output vectors of all attention heads at the same time step are concatenated and linearly projected to obtain the encoded feature vector. Based on the attention weights of all attention heads at each time step, the average attention weight entropy and perturbation modulation energy at the corresponding time step are calculated, and after affine transformation, they are normalized by activation function to obtain the temperature control perturbation discrimination weight. Based on the temperature control disturbance discrimination weight, the inlet and outlet water temperature components at each moment are adaptively denoised and reconstructed to output the denoised inlet and outlet water temperatures. Calculate the rate of change of the noise-reduced inlet and outlet water temperatures based on the noise-reduced inlet and outlet water temperatures. The encoded feature vector, the denoised inlet and outlet water temperature values, the denoised inlet and outlet water temperature change rate, and the temperature control disturbance discrimination weight at each moment are concatenated into a temperature control trend feature vector, and then continuously concatenated to form a temperature control trend feature sequence.

[0010] Optionally, the improved Diff-MoE dynamic scheduling network includes: The temperature control trend feature vectors at all times are summed by time-series weighting to obtain the temperature control state vector at the control cycle level. In the improved Diff-MoE dynamic scheduling network, a conditional diffusion scheduling chain is constructed to generate the intermediate power allocation vector for the current step. For each diffusion step, a conditional denoising predictor is established, and the intermediate power allocation vector is updated by denoising inversion based on the conditional denoising predictor to obtain the intermediate power allocation vector that satisfies the current temperature control state. In the improved Diff-MoE dynamic scheduling network, a set of expert motor control strategies is constructed based on the intermediate power allocation vector that satisfies the current temperature control state, and the expert correction vector is output. In the improved Diff-MoE dynamic scheduling network, a gated routing mechanism is constructed, and the expert routing weights of each motor control strategy expert are generated based on the temperature control state vector at the control cycle level. The expert correction vector is weighted and fused based on the expert routing weight, and the power allocation is projected on a feasibility basis to obtain the target power allocation vector of the generator set. Construct a target start-stop combination vector based on the target power allocation vector of the motor set; The output is the target control strategy for the motor set, which is composed of the target power allocation vector and the target start-stop combination vector.

[0011] Optionally, determining the priority of the main load motor group, the load compensation ratio of the auxiliary load motor group, and the motor start-up and shutdown sequence includes: Read the target power allocation vector of the motor set target control strategy, and construct a multi-cooling tower motor load contribution ranking sequence based on the power allocation ratio of each motor in the target power allocation vector; Based on the load contribution ranking sequence of multiple cooling tower motors, the main load motor group and auxiliary load motor group are defined according to the power distribution ratio from high to low. Based on the power distribution ratio, historical start-stop frequency and cumulative running time of each motor in the main load motor group, the main load motor group is prioritized to obtain the main load motor group priority sequence. Based on the power distribution ratio of each motor in the auxiliary load motor group and the remaining heat dissipation capacity of the main load motor group in the current control cycle, the load compensation ratio of the auxiliary load motor group is calculated. The start-up and shutdown sequence of multiple cooling tower motors is determined based on the priority sequence of the main load motors and the load compensation ratio of the auxiliary load motors. Based on the start-stop sequence of multiple cooling tower motors, the target power allocation ratio, and the division results of main load motors and auxiliary load motors, a set of motor execution instructions is generated.

[0012] Optionally, the main load generator set and the auxiliary load generator set include: The main load motor set is a set of motors with a power distribution ratio of not less than the main load threshold, which is used to bear the basic heat dissipation load of the multi-cooling tower system in the current control cycle. Auxiliary load motor sets are a collection of motors whose power distribution ratio is less than the main load threshold and not less than the auxiliary load threshold. They are used to compensate the main load motor sets under conditions of thermal load fluctuation or sudden load.

[0013] Optionally, the start-stop sequence of the multi-cooling tower motors includes: During startup, the main load motors are started sequentially from high to low according to the priority sequence of the main load motors. When all the main load motors are started and the target power distribution requirements are still not met, the auxiliary load motors are started one by one according to the load compensation ratio of the auxiliary load motors. During shutdown, auxiliary load motors are stopped sequentially from low to high priority. If the system power output still needs to be reduced after all auxiliary load motors have stopped, the main load motors are stopped sequentially from low to high priority.

[0014] An adaptive intelligent temperature control multi-cooling tower motor linkage integrated system includes: The temperature acquisition module contains multiple platinum resistance temperature sensors installed at different locations in the cooling tower to collect raw water temperature data in real time. The sensor acquisition module includes platinum resistance temperature sensors installed in the inlet and outlet water pipes of each cooling tower, and wind speed and humidity sensors installed on site, for real-time acquisition of physical parameters. The signal processing and communication module is used to convert analog signals from sensors into digital signals and transmit them to the PLC controller via fieldbus or industrial Ethernet. The PLC controller module is the core control unit of the system. It has built-in interface program for the load forecasting algorithm model, multi-level fuzzy rules and dynamic scheduling logic, fault diagnosis program and backup switching logic; its digital input points receive fault feedback signals and its digital output points control the actuators. The multi-motor drive module includes five cooling tower drive motors and their independent electrical control circuits. Each motor control circuit sequentially includes a PLC digital output point, an intermediate relay, an AC contactor, and a thermal protection component, forming a three-level isolated drive architecture. The five motors are divided into three groups (main motor group) and two groups (auxiliary motor group) in terms of control logic. The standby motor switching circuit includes a standby motor and its control circuit. Its AC contactor coil is controlled by a PLC. The main circuit is connected in parallel with the main circuit of the working motor. It can be automatically switched by the PLC when any working motor fails. The human-machine interface module connects to the PLC controller and is used for system parameter setting, real-time status display, historical data query, fault alarm, and manual operation intervention.

[0015] The beneficial effects of this invention are: This invention designs a perturbation-adjustable Transformer model that deeply integrates the inlet and outlet water temperature components with the real-time calculated temperature change rate and temperature curvature sequence. It introduces a perturbation modulation vector to dynamically modulate the attention distribution of the embedding vector at each moment. Unlike traditional Transformer or static filtering algorithms, this invention can perform spatiotemporal correlation discrimination of non-real heat load fluctuations based on the actual operating conditions of the cooling tower system, accurately separate environmental disturbances from effective load signals, and significantly improve robustness to high-frequency noise and sudden external disturbances. It can reduce the misjudgment rate of abnormal temperature control response to less than 50% of traditional methods.

[0016] This invention introduces a conditional diffusion model into motor power allocation scheduling. In the Diff-MoE dynamic scheduling network, diverse candidate power allocation patterns are generated through diffusion chains. Combined with temperature control trend feature sequences and expert correction mechanisms, distributed scheduling optimization based on actual heat load and system state is achieved. In each diffusion step, disturbances are adaptively injected according to the temperature control state vector of the cooling tower control cycle, ensuring that the scheduling results are predictive and flexible for future load trends. Through a gated routing mechanism, the system can smoothly switch between multiple expert scheduling strategies such as primary load priority, auxiliary load compensation, and energy consumption balancing, realizing intelligent division of labor and load balancing among linked motor units.

[0017] This invention systematically integrates a mechanism for prioritizing main load motors, allocating load compensation ratios for auxiliary load motors, and adaptively generating start-stop sequences. It correlates power allocation ratios with historical operating data and real-time temperature control trends to dynamically determine the role and response level of each motor. This effectively improves the operational safety redundancy of individual motors and the load balancing capability of multi-machine collaboration in multi-cooling tower systems. It can automatically suppress meaningless start-stops and system oscillations caused by ineffective low-power allocation, ensuring that the main load is activated first and the auxiliary load is smoothly compensated. Furthermore, it prioritizes the shutdown of auxiliary loads during power recovery and shutdown phases, extending the overall service life of the equipment and reducing maintenance costs. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 The flowchart shows an adaptive intelligent temperature control multi-cooling tower motor linkage integrated system and control method proposed in this invention. Figure 2This is a circuit diagram of a PLC controller module for an adaptive intelligent temperature control multi-cooling tower motor linkage integrated system and control method proposed in this invention. Detailed Implementation

[0019] Example 1: Reference Figures 1-2 A control method for an adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, comprising: The original data set of multi-cooling tower system operation is collected by an adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, merged and preprocessed, and output preprocessed multi-cooling tower system operation data sequence. The preprocessed multi-cooling tower system operation data sequence is fed into the perturbation-adjustable Transformer model. The multi-head self-attention mechanism is used to generate temperature control perturbation discrimination weights, complete high-frequency random perturbation denoising and extract temperature control trend feature vectors, and output temperature control trend feature sequence. The temperature control trend feature sequence is input into the load prediction algorithm model running on the host computer or edge computing module. The load prediction algorithm model predicts the heat load change trend for the next control cycle based on historical and real-time data, and outputs the heat load trend index and the expected temperature rise rate. The temperature control trend feature sequence is input into the improved Diff-MoE dynamic scheduling network. Based on the gating routing mechanism, the expert set of motor control strategies is matched to obtain the target power allocation vector and target start-stop combination vector for each cooling tower motor unit, and the target control strategy of the motor unit is output. The PLC controller receives the heat load trend index, expected temperature rise rate and real-time temperature deviation signal. Based on the built-in multi-level fuzzy rules and dynamic priority scheduling logic, it calculates the target power allocation coefficient and target start-stop status of each cooling tower motor and generates a motor group collaborative control strategy. Based on the motor unit collaborative control strategy and the motor unit target control strategy, the multi-cooling tower motors are divided into main load motor units and auxiliary load motor units. The priority of the main load motor units, the load compensation ratio of the auxiliary load motor units and the motor start-stop sequence are determined, and the motor unit execution instruction set is generated. The motor set execution instruction set is sent to the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, which sequentially triggers the PLC controller module and the multi-motor drive module to drive the corresponding motor set to start or stop in the motor start-stop sequence and operate according to the target power allocation, thereby obtaining real-time execution status feedback data of the motor set. Based on the real-time execution status feedback data of the motor set, it can determine in real time whether the multi-cooling tower system has a motor phase loss fault, motor overload fault, contactor sticking fault or thermal protection trip fault. If a fault condition is determined, the backup motor switching circuit is automatically called to connect the backup motor to the corresponding cooling tower circuit and update the motor set execution instruction set to achieve stable operation.

[0020] In this embodiment, the raw data set of the multi-cooling tower system operation is collected and preprocessed by the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, including: The system collects raw data on inlet and outlet water temperatures, ambient wind speed, ambient humidity, motor operating status, and historical heat load of a multi-cooling tower system. It performs time synchronization processing and unifies the encoding format to obtain a set of raw data for the operation of the multi-cooling tower system. The system then performs noise filtering, missing data completion, and normalization on the raw data set to output a preprocessed sequence of operating data for the multi-cooling tower system.

[0021] The preprocessed multi-cooling tower system operation data sequence is constructed as a time series with a length equal to the control cycle length. The preprocessed multi-cooling tower system operation data vector at each moment contains inlet and outlet water temperature components, which are represented by the inlet and outlet water temperature values ​​at the current moment. The sampling time interval between adjacent sampling moments is specified.

[0022] In this embodiment, the steps of the PLC controller generating the motor group cooperative control strategy include: Establish a multi-level fuzzy rule base with real-time temperature deviation, temperature change rate, and heat load trend index as inputs; Determine the total cooling power level required by the system based on the magnitude and direction of the real-time temperature deviation; The total cooling power level is proactively adjusted by incorporating the heat load trend index.

[0023] In this embodiment, the perturbation-adjustable Transformer model includes: The preprocessed multi-cooling tower system operation data sequence is input into the perturbation-adjustable Transformer model. The preprocessed multi-cooling tower system operation data vector at each time step is transformed by the mapping matrix and time position encoding to obtain the input embedding vector. In the perturbation-tunable Transformer model, based on the inlet and outlet water temperature components in the preprocessed multi-cooling tower system operation data sequence, the temperature change rate sequence and temperature curvature sequence at each moment are calculated. The input embedding vector, temperature change rate sequence and temperature curvature sequence corresponding to the current moment are jointly mapped to construct the perturbation modulation vector. In Example 1, the temperature change rate sequence is obtained by subtracting the inlet and outlet water temperature components from the previous moment from the inlet and outlet water temperature components at each moment and dividing by the sampling time interval; the temperature curvature sequence is obtained by subtracting the temperature change rate from the previous moment from the current moment's temperature change rate and dividing by the sampling time interval.

[0024] The joint mapping process includes: performing a linear transformation on the input embedding vector; mapping the temperature change rate sequence to the same feature space as the input embedding vector through the rate of change injection mapping vector; mapping the temperature curvature sequence to the same feature space as the input embedding vector through the curvature injection mapping vector; superimposing the mapped input embedding vector, the mapped temperature change rate feature, and the mapped temperature curvature feature element by element; and generating a perturbation modulation vector corresponding to the current time step through nonlinear activation function processing. By repeating the joint mapping and nonlinear processing process for all time steps, a perturbation modulation vector is constructed. This vector is used to characterize the perturbation intensity, acceleration of change, and thermal inertial response features of the inlet and outlet water temperature signals of the multi-cooling tower system under different load change conditions within the perturbation-tunable Transformer model.

[0025]

[0026] in, Represents the modulation mapping matrix, Indicates the injection vector of the rate of temperature change. This represents the temperature curvature injection vector. This represents the modulation bias vector. Representation of time The corresponding perturbation modulation vector, Represents a sequence of temperature change rates. This represents a temperature curvature sequence.

[0027] Within each attention head of the perturbation-tunable Transformer model, for each time step, the input embedding vector is linearly transformed using the mapping matrix based on the perturbation modulation vector to generate the query vector, key vector, and value vector, and the perturbation-tunable scaling dot product attention score is calculated. For each pair of time points, based on the dot product of the query vector and the key vector, the dynamic coupling between the time delay penalty coefficient set by the control system and the disturbance modulation vector, an adjustable scaling dot product attention score for the disturbance is obtained. This score is used to reflect the dynamic correlation and disturbance pattern similarity between temperature control signals at different time points under the thermal inertia conditions of multiple cooling towers.

[0028]

[0029] in, Indicates time For time The perturbation is adjustable and the scaling of the dot product attention score is adjustable. Represents the dimension of the key vector. Indicates the first The time delay penalty coefficient of the attention head is used to suppress long-distance time miscorrelation under multi-cooling-tower thermal inertia conditions, representing the... The perturbation coupling coefficient of each attention head is used to enhance the correlation between time pairs with similar load variation patterns. This represents the disturbance modulation vector corresponding to time τ. This represents the query vector generated by the m-th attention head in the perturbation-tunable Transformer model, targeting the operating state of the multi-cooling-tower system at time t. In the perturbation-tunable Transformer model, the m-th attention head generates a key vector for the operating state of the multi-cooling tower system at historical time τ.

[0030] Based on the perturbation-adjustable scaling dot product attention score, a normalization operation is used to obtain the attention weight of each time step to all time steps. The attention weights are used to sum the corresponding value vectors to obtain the attention output vector. The attention output vectors of all attention heads at the same time step are concatenated and linearly projected to obtain the encoded feature vector. Based on the attention weights of all attention heads at each time step, the average attention weight entropy and perturbation modulation energy at the corresponding time step are calculated, and after affine transformation, they are normalized by activation function to obtain the temperature control perturbation discrimination weight. In Example 1, the attention weights corresponding to different attention heads at the same time are averaged element by element to obtain the average attention weight distribution at the corresponding time. The average attention weight entropy is obtained by summing the products of each time-related weight in the average attention weight distribution and taking the negative number. This entropy is used to quantify the degree of temperature control time-series correlation concentration of the multi-cooling tower system at the current time. The perturbation modulation energy is obtained by summing the squares of each component in the perturbation modulation vector and normalizing it in combination with the dimension of the perturbation modulation vector. This energy is used to quantify the severity of the load change pattern in the temperature control signal of the multi-cooling tower system at the current time.

[0031] The average attention weight entropy and perturbation modulation energy are linearly weighted and superimposed with a bias term to obtain the intermediate perturbation discriminant. The intermediate perturbation discriminant is normalized by applying a monotonically bounded activation function to the intermediate perturbation discriminant, and the output is the temperature control perturbation discriminant weight between zero and one. The temperature control perturbation discriminant weight is used to represent the proportion of non-real heat load fluctuations caused by environmental disturbances, electromagnetic interference or fluid shock in the inlet and outlet water temperature components of the multi-cooling tower system at the current moment.

[0032] Based on the temperature control disturbance discrimination weight, the inlet and outlet water temperature components at each moment are adaptively denoised and reconstructed to output the denoised inlet and outlet water temperatures. In Example 1, the adaptive denoising reconstruction is performed by calculating the model reconstructed inlet and outlet water temperatures at any given time using temperature reconstruction mapping parameters. The model reconstructed inlet and outlet water temperatures represent the theoretical stable temperature values ​​under the current system operating state and historical time series correlation conditions. The temperature control disturbance discrimination weight is used as the fusion coefficient to perform weighted fusion of the original inlet and outlet water temperature components and the model reconstructed inlet and outlet water temperatures. The temperature control disturbance discrimination weight corresponds to the retention ratio of the original inlet and outlet water temperature components, and the complement of the temperature control disturbance discrimination weight corresponds to the introduction ratio of the model reconstructed inlet and outlet water temperatures, thus obtaining the denoised inlet and outlet water temperature values.

[0033] Calculate the rate of change of the noise-reduced inlet and outlet water temperatures based on the noise-reduced inlet and outlet water temperatures. The rate of change of denoised inlet and outlet water temperature is the difference between the denoised inlet and outlet water temperature values ​​at each moment and the previous moment, divided by the sampling time interval. This is used to quantify the dynamic change trend of cooling tower water temperature after disturbance suppression.

[0034] The encoded feature vector, the denoised inlet and outlet water temperature values, the denoised inlet and outlet water temperature change rate, and the temperature control disturbance discrimination weight at each moment are concatenated into a temperature control trend feature vector, and then continuously concatenated to form a temperature control trend feature sequence.

[0035] In this embodiment, the Diff-MoE dynamic scheduling network is improved, including: The temperature control trend feature vectors at all times are summed by time-series weighting to obtain the temperature control state vector at the control cycle level. In the improved Diff-MoE dynamic scheduling network, a conditional diffusion scheduling chain is constructed to generate the intermediate power allocation vector for the current step. In this embodiment, the entire conditional diffusion scheduling chain is used to generate diverse candidate power allocation patterns when there are environmental disturbances and heat load randomness in a multi-cooling tower system. Each diffusion step has a noise injection coefficient, and each step generates a noise vector with the same dimension as the number of controlled motors. Starting from the initial power allocation vector, each step of the diffusion process generates the power allocation intermediate vector of the current step by multiplying the power allocation intermediate vector and the noise vector of the previous diffusion step by the noise injection coefficient and its complement, and then summing them.

[0036] For each diffusion step, a conditional denoising predictor is established, and the intermediate power allocation vector is updated by denoising inversion based on the conditional denoising predictor to obtain the intermediate power allocation vector that satisfies the current temperature control state. The denoising inversion update is as follows: the conditional denoising predictor takes the power allocation intermediate vector of the current diffusion step and the temperature control state vector of the control cycle as input, outputs the predicted noise, subtracts the product of the complement of the noise injection coefficient and the predicted noise from the power allocation intermediate vector of the diffusion step, divides it by the square root of the noise injection coefficient, and iterates back each diffusion step in reverse order to recover the power allocation intermediate vector that satisfies the current temperature control state.

[0037] In the improved Diff-MoE dynamic scheduling network, a set of expert motor control strategies is constructed based on the intermediate power allocation vector that satisfies the current temperature control state, and the expert correction vector is output. The motor control strategy expert set consists of several motor control strategy experts. Each motor control strategy expert takes the power distribution intermediate vector and the temperature control state vector at the control cycle level as inputs and outputs an expert correction vector through a mapping relationship. Each expert correction vector has the same dimension as the power distribution intermediate vector. The expert correction vector is used to perform expert-based fine-tuning and correction of the power distribution results under different heat load change patterns.

[0038] In the improved Diff-MoE dynamic scheduling network, a gated routing mechanism is constructed, and the expert routing weights of each motor control strategy expert are generated based on the temperature control state vector at the control cycle level. In Example 1, a gated routing mechanism is constructed in the improved Diff-MoE dynamic scheduling network. The gated routing mechanism takes the temperature control state vector at the control cycle level as a unified input. For each motor control strategy expert in the set of motor control strategy experts, the corresponding expert activation score is calculated. The expert activation score is obtained by jointly mapping the temperature trend intensity component, the noise-reduced inlet and outlet water temperature change rate component, and the temperature control disturbance discrimination weight component in the temperature control state vector at the control cycle level. It is used to measure the degree of matching of the motor control strategy expert with the heat load response requirements of the multi-cooling tower system in the current control cycle. Joint mapping includes: linear feature projection of the temperature control state vector at the control cycle level to extract state features related to motor load changes, thermal inertia response, and linkage control intensity; applying nonlinear activation transformation to the projected state features to enhance the ability to distinguish nonlinear thermal load changes and sudden disturbance conditions; and configuring an independent set of gating parameters for each motor control strategy expert to enable different motor control strategy experts to have differentiated response sensitivities under different control modes such as main load dominance, auxiliary load compensation, disturbance suppression priority, or energy consumption balance priority.

[0039] After obtaining the expert activation scores for all motor control strategy experts, the expert activation scores are normalized to ensure that the expert routing weights for all motor control strategy experts are non-negative and sum to one. This yields the expert routing weights for each motor control strategy expert. The expert routing weights represent the degree of participation of each motor control strategy expert in the generation process of the target power allocation vector and target start-stop combination vector of the motor group under the current temperature control state of the control cycle. This enables adaptive expert selection and smooth strategy switching based on the temperature control state of the multi-cooling tower system.

[0040] The expert correction vector is weighted and fused based on the expert routing weight, and the power allocation is projected on a feasibility basis to obtain the target power allocation vector of the generator set. Based on the expert routing weights obtained from the gating routing mechanism, all expert correction vectors are weighted and fused according to the routing weights. The weighted and fused expert correction vectors are added to the intermediate power allocation vector that satisfies the current temperature control state to obtain the weighted and corrected power allocation result. The power allocation result is projected onto the feasible space where all components are greater than or equal to zero and the sum of all components is 1 to obtain the target power allocation vector of the motor group. This allows the target power allocation vector of the motor group to be directly used as the power ratio allocation command in the multi-cooling tower motor linkage control.

[0041]

[0042] in, This represents the target power allocation vector for the generator set. Indicates feasibility projection. Represents the expert correction vector. Indicates the expert routing weight. This represents the intermediate vector for power allocation. This indicates the number of experts in motor control strategies.

[0043] Construct a target start-stop combination vector based on the target power allocation vector of the motor set; In Example 1, the target start-stop combination vector consists of binary state components that correspond one-to-one with the number of controlled motors. Each binary state component is used to represent the target start-up state or target shutdown state of the corresponding motor in the current control cycle.

[0044] For each power allocation ratio component in the target power allocation vector of the motor set, a discretization judgment is performed according to the preset target start-stop judgment rule. The target start-stop judgment rule includes: comparing the corresponding power allocation ratio with the target start-stop combination threshold. When the power allocation ratio is greater than or equal to the target start-stop combination threshold, the target start-stop state of the corresponding motor is set to the enabled state; when the power allocation ratio is less than the target start-stop combination threshold, the target start-stop state of the corresponding motor is set to the disabled state.

[0045] The target start-stop combination threshold is a dimensionless parameter used to establish a mapping relationship between continuous power distribution results and discrete motor start-stop control. This ensures that in the linkage control of a multi-cooling tower system, the motor is only allowed to enter the start state when the target power distribution ratio undertaken by the motor reaches the minimum effective load requirement, thus avoiding frequent start-stop and invalid actions caused by small power distribution.

[0046] The output is the target control strategy for the motor set, which is composed of the target power allocation vector and the target start-stop combination vector.

[0047] In this embodiment, determining the priority of the main load motor set, the load compensation ratio of the auxiliary load motor set, and the motor start-up and shutdown sequence includes: Read the target power allocation vector of the motor set target control strategy, and construct a multi-cooling tower motor load contribution ranking sequence based on the power allocation ratio of each motor in the target power allocation vector; The multi-cooling-tower motor load contribution ranking sequence indicates the degree of contribution of each motor to the total heat dissipation capacity of the cooling tower system within the current control cycle.

[0048] Based on the load contribution ranking sequence of multiple cooling tower motors, the main load motor group and auxiliary load motor group are defined according to the power distribution ratio from high to low. Based on the power distribution ratio, historical start-stop frequency and cumulative running time of each motor in the main load motor group, the main load motor group is prioritized to obtain the main load motor group priority sequence. The priority of the main load motor sets, from high to low, corresponds to the order in which the motors are put into operation first. This is used to reduce the long-term operating stress of a single motor while meeting the heat dissipation requirements.

[0049] Based on the power distribution ratio of each motor in the auxiliary load motor group and the remaining heat dissipation capacity of the main load motor group in the current control cycle, the load compensation ratio of the auxiliary load motor group is calculated. The remaining heat dissipation capacity of the main load motor is determined by the difference between the theoretical maximum heat dissipation capacity of the main load motor and the actual heat dissipation capacity corresponding to its target power allocation vector. The load compensation ratio of the auxiliary load motor is calculated according to the ratio between the sum of the target power allocation ratios of the auxiliary load motors and the remaining heat dissipation capacity of the main load motor. The load compensation ratio of the auxiliary load motor is used to limit the maximum proportion of the total power allocation of the system when the auxiliary load motor participates in heat dissipation, so as to avoid excessive intervention of the auxiliary load motor and increase energy consumption.

[0050] The start-up and shutdown sequence of multiple cooling tower motors is determined based on the priority sequence of the main load motors and the load compensation ratio of the auxiliary load motors. Based on the start-stop sequence of multiple cooling tower motors, the target power allocation ratio, and the division results of main load motors and auxiliary load motors, a set of motor execution instructions is generated.

[0051] In this embodiment, the main load generator set and the auxiliary load generator set include: The main load motor set is a set of motors with a power distribution ratio of not less than the main load threshold, which is used to bear the basic heat dissipation load of the multi-cooling tower system in the current control cycle. Auxiliary load motor sets are a collection of motors whose power distribution ratio is less than the main load threshold and not less than the auxiliary load threshold. They are used to compensate the main load motor sets under conditions of thermal load fluctuation or sudden load.

[0052] In this embodiment, the start-stop sequence of the multi-cooling tower motors includes: During startup, the main load motors are started sequentially from high to low according to the priority sequence of the main load motors. When all the main load motors are started and the target power distribution requirements are still not met, the auxiliary load motors are started one by one according to the load compensation ratio of the auxiliary load motors. During shutdown, auxiliary load motors are stopped sequentially from low to high priority. If the system power output still needs to be reduced after all auxiliary load motors have stopped, the main load motors are stopped sequentially from low to high priority.

[0053] An adaptive intelligent temperature control multi-cooling tower motor linkage integrated system includes: The temperature acquisition module contains multiple platinum resistance temperature sensors installed at different locations in the cooling tower to collect raw water temperature data in real time. The sensor acquisition module includes platinum resistance temperature sensors installed in the inlet and outlet water pipes of each cooling tower, and wind speed and humidity sensors installed on site, for real-time acquisition of physical parameters. The signal processing and communication module is used to convert analog signals from sensors into digital signals and transmit them to the PLC controller via fieldbus or industrial Ethernet. The PLC controller module is the core control unit of the system. It has built-in interface program for the load forecasting algorithm model, multi-level fuzzy rules and dynamic scheduling logic, fault diagnosis program and backup switching logic; its digital input points receive fault feedback signals and its digital output points control the actuators. The multi-motor drive module includes five cooling tower drive motors and their independent electrical control circuits. Each motor control circuit sequentially includes a PLC digital output point, an intermediate relay, an AC contactor, and a thermal protection component, forming a three-level isolated drive architecture. The five motors are divided into three groups (main motor group) and two groups (auxiliary motor group) in terms of control logic. The standby motor switching circuit includes a standby motor and its control circuit. Its AC contactor coil is controlled by a PLC. The main circuit is connected in parallel with the main circuit of the working motor. It can be automatically switched by the PLC when any working motor fails. The human-machine interface module connects to the PLC controller and is used for system parameter setting, real-time status display, historical data query, fault alarm, and manual operation intervention.

[0054] Example 2: In a high-temperature industrial circulating water system with five cooling towers (M1-M5), the system needs to ensure that the outlet water temperature is stable within the target range of 30℃~50℃. The equipment environment is subject to strong winds, electromagnetic disturbances, and diurnal heat load fluctuations. The system adopts the adaptive intelligent temperature control multi-cooling tower motor linkage integrated control method of the present invention, with the following initial settings: Target temperature range: [30℃, 50℃], temperature change rate threshold: rising threshold 1 is 0.3℃ / min, rising threshold 2 is 0.8℃ / min, falling threshold is -0.3℃ / min, motor group A: M1, M2, M3, group B: M4, M5, data acquisition frequency: 10 seconds / time; During system initialization, the operator sets parameters and starts data acquisition. In the first 10 minutes, the water temperature stabilizes at 29.6℃~30.2℃ with small fluctuations and a change rate of approximately -0.07℃ / min. All motors maintain minimum power and rotate in turn.

[0055] The rotation priority is set to M1→M2→M3, and the auxiliary group is in standby mode by default.

[0056] Fifteen minutes later, the production line suddenly increased its load, and the inlet temperature of the cooling tower began to rise significantly.

[0057] Temperature was recorded at 31.2℃ at 10:30:00. At 10:31:00, the temperature was collected at T=32.8℃, and ΔT / Δt=+0.26℃ / min; 10:33:00T=35.6℃, ΔT / Δt=+0.70℃ / min; At this point, the perturbation-adjustable Transformer model received an abnormal change, detecting a short-term increase in the rate of temperature change to 0.8℃ / min, and simultaneously detecting a surge in wind speed sensor readings to 5.2 m / s. The perturbation discrimination weights... The value was increased from 0.28 (normal) to 0.63. The Transformer outputs a temperature control trend feature sequence, and the Diff-MoE dynamic scheduling network makes the following judgments and controls based on the real-time trend: At 10:35:00, T=38.3℃ and ΔT / Δt=0.78℃ / min were detected. The temperature was higher than the lower limit but not higher than the upper limit, indicating a medium load condition. Group A's M1 and M2 automatically switched to full power operation, while Group B remained on standby.

[0058] At 10:37:00, T = 41.5℃, ΔT / Δt = 0.84℃ / min, under the condition of rapid temperature rise, all motors in Group A (M1, M2, M3) and one motor in Group B (M4) start synchronously. The Diff-MoE output power distribution vector... Target start / stop vector .

[0059] At 10:39:00, T = 44.2℃, ΔT / Δt = 0.91℃ / min, the temperature rapidly increases, and all motors (M1-M5) start at full load. , .

[0060] The Transformer makes a judgment on a high-frequency temperature fluctuation caused by a sudden change in wind speed. The value was increased to 0.81, which quickly eliminated false signals and prevented unnecessary system startup.

[0061] The main load threshold is set via Diff-MoE gating output. At this moment, M1, M2, and M3 are the main load motor sets, while M4 and M5 are the auxiliary load motor sets.

[0062] Historical operation analysis: M1 has the longest cumulative runtime, so its priority has been lowered. The priority order of the main load is M2→M3→M1.

[0063] The load compensation ratio for Group B is limited to 30%. The threshold was not exceeded.

[0064] Start-up and shutdown sequence: First, start the main load motor units (M2, M3, M1) in sequence, and then gradually put the B group (M4, M5) into operation; when the temperature drops and the rate of change is lower than -0.3℃ / min, prioritize the sequential shutdown of the B group, and then rotate the main load motor units.

[0065] From 12:00 to 14:00, the temperature remains stable between 30.8℃ and 31.4℃, and ΔT / Δt is below +0.02℃ / min, at which point the motor enters energy-saving mode.

[0066] Based on health parameters and cumulative runtime, Diff-MoE automatically rotates the motors in Group A every 2 hours. After the rotation, the priority of M3 is increased, and the main load sequence becomes M3→M2→M1, while Group B remains in standby mode.

[0067] At 16:00, M4 malfunctioned during operation. Power monitoring revealed a sudden drop in M4 current, which the controller identified as an overload trip.

[0068] At 16:00:10, a fault signal was transmitted, the Diff-MoE gating output was adjusted in real time, the remaining B group was automatically taken over by M5, and the main and auxiliary load distribution was adjusted to... , No water temperature exceeded the limit.

[0069] The entire fault switching process took no more than 30 seconds, the maximum water temperature fluctuation was only 0.4℃, and there were no abnormal starts or stops of the main load. Data comparison—the present invention and the traditional method are shown in Table 1: Table 1. Daily Operational Samples of Heat Load Fluctuations (Partial Data) time T / ℃ ΔT / Δt / (℃ / min) Power allocation of main load group A Assisted Group B power allocation Start-stop strategy Traditional response method (threshold method) The response of this invention (Diff-MoE) 10:30 31.2 +0.26 M1: 0.52, M2: 0.48 M4:0, M5:0 Group A Low Load A1 low speed, stop M1, M2 low speed 10:33 35.6 +0.70 M1: 0.46, M2: 0.38 M4:0.16, M5:0 A2 acceleration, M4 start A1, A2, and A3 are all activated (delayed by 1 minute). M1, M2, and M4 accelerate simultaneously. 10:39 44.2 +0.91 Equal Equal Fully loaded Group B's input was delayed by 2 minutes. Fully loaded and ready to go 11:05 33.8 -0.36 Group A:0.62 Group B: 0.00 Group B is discontinued. A3 fails to stop in time, B4 malfunctions. Group B will exit sequentially, and the main load will be rotated. 16:00 36.9 +0.12 Group A:0.72 M4:0.28 M4 Fault Switching Manual confirmation is required to switch after an alarm is triggered. M4 detection failure, M5 compensation <30 seconds In this embodiment, the Diff-MoE dynamic scheduling network and the disturbance-adjustable Transformer algorithm realize adaptive and robust intelligent scheduling of cooling tower group control under disturbance conditions. Every change in temperature signal is accurately identified, and the response control and main and auxiliary load switching are timely and flexible, which is significantly better than traditional static PID and threshold control methods.

[0070] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A control method for an adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, characterized in that, include: The original data set of multi-cooling tower system operation is collected by an adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, merged and preprocessed, and output preprocessed multi-cooling tower system operation data sequence. The preprocessed multi-cooling tower system operation data sequence is fed into the perturbation-adjustable Transformer model. The multi-head self-attention mechanism is used to generate temperature control perturbation discrimination weights, complete high-frequency random perturbation denoising and extract temperature control trend feature vectors, and output temperature control trend feature sequence. The temperature control trend feature sequence is input into the load prediction algorithm model running on the host computer or edge computing module. The load prediction algorithm model predicts the heat load change trend for the next control cycle based on historical and real-time data, and outputs the heat load trend index and the expected temperature rise rate. The temperature control trend feature sequence is input into the improved Diff-MoE dynamic scheduling network. Based on the gating routing mechanism, the expert set of motor control strategies is matched to obtain the target power allocation vector and target start-stop combination vector for each cooling tower motor unit, and the target control strategy of the motor unit is output. The PLC controller receives the heat load trend index, expected temperature rise rate and real-time temperature deviation signal. Based on the built-in multi-level fuzzy rules and dynamic priority scheduling logic, it calculates the target power allocation coefficient and target start-stop status of each cooling tower motor and generates a motor group collaborative control strategy. Based on the motor unit collaborative control strategy and the motor unit target control strategy, the multi-cooling tower motors are divided into main load motor units and auxiliary load motor units. The priority of the main load motor units, the load compensation ratio of the auxiliary load motor units and the motor start-stop sequence are determined, and the motor unit execution instruction set is generated. The motor set execution instruction set is sent to the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, which sequentially triggers the PLC controller module and the multi-motor drive module to drive the corresponding motor set to start or stop in the motor start-stop sequence and operate according to the target power allocation, thereby obtaining real-time execution status feedback data of the motor set. Based on the real-time execution status feedback data of the motor set, it can determine in real time whether the multi-cooling tower system has a motor phase loss fault, motor overload fault, contactor sticking fault or thermal protection trip fault. If a fault condition is determined, the backup motor switching circuit is automatically called to connect the backup motor to the corresponding cooling tower circuit and update the motor set execution instruction set to achieve stable operation.

2. The control method for an adaptive intelligent temperature control multi-cooling tower motor linkage integrated system according to claim 1, characterized in that, The process of collecting and preprocessing the raw data sets of the multi-cooling tower system operation through the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system includes: The system collects raw data on inlet and outlet water temperatures, ambient wind speed, ambient humidity, motor operating status, and historical heat load of a multi-cooling tower system. It performs time synchronization processing and unifies the encoding format to obtain a set of raw data for the operation of the multi-cooling tower system. The system then performs noise filtering, missing data completion, and normalization on the raw data set to output a preprocessed sequence of operating data for the multi-cooling tower system.

3. The control method of the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system according to claim 1, characterized in that, The steps for the PLC controller to generate a motor group coordinated control strategy include: Establish a multi-level fuzzy rule base with real-time temperature deviation, temperature change rate, and heat load trend index as inputs; Determine the total cooling power required by the system based on the magnitude and direction of the real-time temperature deviation; The total cooling power level is proactively adjusted by incorporating the heat load trend index.

4. The control method of the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system according to claim 1, characterized in that, The perturbation-tunable Transformer model includes: The preprocessed multi-cooling tower system operation data sequence is input into the perturbation-adjustable Transformer model. The preprocessed multi-cooling tower system operation data vector at each time step is transformed by the mapping matrix and time position encoding to obtain the input embedding vector. In the perturbation-tunable Transformer model, based on the inlet and outlet water temperature components in the preprocessed multi-cooling tower system operation data sequence, the temperature change rate sequence and temperature curvature sequence at each moment are calculated. The input embedding vector, temperature change rate sequence and temperature curvature sequence corresponding to the current moment are jointly mapped to construct the perturbation modulation vector. Within each attention head of the perturbation-tunable Transformer model, for each time step, the input embedding vector is linearly transformed using the mapping matrix based on the perturbation modulation vector to generate the query vector, key vector, and value vector, and the perturbation-tunable scaling dot product attention score is calculated. Based on the perturbation-adjustable scaling dot product attention score, a normalization operation is used to obtain the attention weight of each time step to all time steps. The attention weights are used to sum the corresponding value vectors to obtain the attention output vector. The attention output vectors of all attention heads at the same time step are concatenated and linearly projected to obtain the encoded feature vector. Based on the attention weights of all attention heads at each time step, the average attention weight entropy and perturbation modulation energy at the corresponding time step are calculated, and after affine transformation, they are normalized by activation function to obtain the temperature control perturbation discrimination weight. Based on the temperature control disturbance discrimination weight, the inlet and outlet water temperature components at each moment are adaptively denoised and reconstructed to output the denoised inlet and outlet water temperatures. Calculate the rate of change of the noise-reduced inlet and outlet water temperatures based on the noise-reduced inlet and outlet water temperatures. The encoded feature vector, the denoised inlet and outlet water temperature values, the denoised inlet and outlet water temperature change rate, and the temperature control disturbance discrimination weight at each moment are concatenated into a temperature control trend feature vector, and then continuously concatenated to form a temperature control trend feature sequence.

5. The control method of the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system according to claim 1, characterized in that, The improved Diff-MoE dynamic scheduling network includes: The temperature control trend feature vectors at all times are summed by time-series weighting to obtain the temperature control state vector at the control cycle level. In the improved Diff-MoE dynamic scheduling network, a conditional diffusion scheduling chain is constructed to generate the intermediate power allocation vector for the current step. For each diffusion step, a conditional denoising predictor is established, and the intermediate power allocation vector is updated by denoising inversion based on the conditional denoising predictor to obtain the intermediate power allocation vector that satisfies the current temperature control state. In the improved Diff-MoE dynamic scheduling network, a set of expert motor control strategies is constructed based on the intermediate power allocation vector that satisfies the current temperature control state, and the expert correction vector is output. In the improved Diff-MoE dynamic scheduling network, a gated routing mechanism is constructed, and the expert routing weights of each motor control strategy expert are generated based on the temperature control state vector at the control cycle level. The expert correction vector is weighted and fused based on the expert routing weight, and the power allocation is projected on a feasibility basis to obtain the target power allocation vector of the generator set. Construct a target start-stop combination vector based on the target power allocation vector of the motor set; The output is the target control strategy for the motor set, which is composed of the target power allocation vector and the target start-stop combination vector.

6. The control method of the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system according to claim 1, characterized in that, The determination of the priority of the main load motor group, the load compensation ratio of the auxiliary load motor group, and the motor start-up and shutdown sequence includes: Read the target power allocation vector of the motor set target control strategy, and construct a multi-cooling tower motor load contribution ranking sequence based on the power allocation ratio of each motor in the target power allocation vector; Based on the load contribution ranking sequence of multiple cooling tower motors, the main load motor group and auxiliary load motor group are defined according to the power distribution ratio from high to low. Based on the power distribution ratio, historical start-stop frequency and cumulative running time of each motor in the main load motor group, the main load motor group is prioritized to obtain the main load motor group priority sequence. Based on the power distribution ratio of each motor in the auxiliary load motor group and the remaining heat dissipation capacity of the main load motor group in the current control cycle, the load compensation ratio of the auxiliary load motor group is calculated. The start-up and shutdown sequence of multiple cooling tower motors is determined based on the priority sequence of the main load motors and the load compensation ratio of the auxiliary load motors. Based on the start-stop sequence of multiple cooling tower motors, the target power allocation ratio, and the division results of main load motors and auxiliary load motors, a set of motor execution instructions is generated.

7. The control method of the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system according to claim 6, characterized in that, The main load generator set and auxiliary load generator set include: The main load motor set is a set of motors with a power distribution ratio of not less than the main load threshold, which is used to bear the basic heat dissipation load of the multi-cooling tower system in the current control cycle. Auxiliary load motor sets are a collection of motors whose power distribution ratio is less than the main load threshold and not less than the auxiliary load threshold. They are used to compensate the main load motor sets under conditions of thermal load fluctuation or sudden load.

8. The control method of the adaptive intelligent temperature control multi-cooling tower motor linkage integrated system according to claim 6, characterized in that, The start / stop sequence of the multi-cooling tower motors includes: During startup, the main load motors are started sequentially from high to low according to the priority sequence of the main load motors. When all the main load motors are started and the target power distribution requirements are still not met, the auxiliary load motors are started one by one according to the load compensation ratio of the auxiliary load motors. During shutdown, auxiliary load motors are stopped sequentially from low to high priority. If the system power output still needs to be reduced after all auxiliary load motors have stopped, the main load motors are stopped sequentially from low to high priority.

9. An adaptive intelligent temperature control multi-cooling tower motor linkage integrated system, characterized in that, include: The temperature acquisition module contains multiple platinum resistance temperature sensors installed at different locations in the cooling tower to collect raw water temperature data in real time. The sensor acquisition module includes platinum resistance temperature sensors installed in the inlet and outlet water pipes of each cooling tower, and wind speed and humidity sensors installed on site, for real-time acquisition of physical parameters. The signal processing and communication module is used to convert analog signals from sensors into digital signals and transmit them to the PLC controller via fieldbus or industrial Ethernet. The PLC controller module is the core control unit of the system. It has built-in interface program for the load forecasting algorithm model, multi-level fuzzy rules and dynamic scheduling logic, fault diagnosis program and backup switching logic; its digital input points receive fault feedback signals and its digital output points control the actuators. The multi-motor drive module includes five cooling tower drive motors and their independent electrical control circuits. Each motor control circuit sequentially includes a PLC digital output point, an intermediate relay, an AC contactor, and a thermal protection component, forming a three-level isolated drive architecture. The five motors are divided into three groups (main motor group) and two groups (auxiliary motor group) in terms of control logic. The standby motor switching circuit includes a standby motor and its control circuit. Its AC contactor coil is controlled by a PLC. The main circuit is connected in parallel with the main circuit of the working motor. It can be automatically switched by the PLC when any working motor fails. The human-machine interface module is connected to the PLC controller and is used for system parameter setting, real-time status display, historical data query, fault alarm and manual operation intervention.

Images