Refrigeration station energy-saving control method and system based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based refrigeration station energy-saving control method and system, and relates to the technical field of energy-saving control, and the method comprises the steps: collecting multi-modal data, carrying out the optimization through a PINN neural network, generating a global thermal field graph and an airflow field through a QVC circuit and a quantum variation optimization algorithm, mapping the global thermal field graph to a quantum Hilbert space, and carrying out the optimization of the global thermal field graph and the airflow field. The method comprises the following steps: predicting a cold load demand through NQ-GRU, generating a chaos sequence by using Sine chaos mapping, extracting local thermal field characteristics, generating actions of equipment parameters through DQN, disturbing actions of optimal equipment parameters by using the chaos sequence, converting into a PLC instruction, and generating an equipment operation instruction set. According to the invention, optimization is carried out by combining an OFT circuit with a PINN neural network, a quantum variational optimization algorithm is introduced, the precision and efficiency of energy-saving control of the refrigeration station are improved, the output of a model is disturbed through a chaotic mapping technology, and the flexibility of equipment operation and the adaptive ability of the system are improved.

Description

Technical Field

[0001] The present invention relates to the field of energy-saving control technology, and in particular to an energy-saving control method and system for a refrigeration station based on artificial intelligence. Background Art

[0002] With the acceleration of industrialization and global warming, the energy consumption of refrigeration systems has become a major factor restricting the sustainable development of the industry. Traditional energy-saving control methods for refrigeration systems usually rely on thermodynamic modeling, rule-based control strategies, or optimization methods driven by empirical formulas. Artificial intelligence, especially deep learning and reinforcement learning, has shown great potential in complex system modeling, prediction, and control. Methods such as PINN neural networks and GRU are gradually being applied to thermal system modeling and control, providing refrigeration systems with more accurate prediction capabilities and better control strategies. At the same time, quantum computing and chaos optimization technologies are also being gradually introduced to solve the bottleneck problems of difficult high-dimensional multivariable coupling modeling and local optimal traps.

[0003] Existing energy-saving optimization methods for water-cooled air conditioners still have shortcomings. Existing PINN methods are mostly based on pure data-driven, lack the embedding of physical laws, and lack the coordinated application of quantum variational optimization and NQ-GRU. It is difficult to fully capture local thermal field characteristics and nonlinear dynamics, resulting in limited cooling load prediction accuracy. Equipment parameter optimization mostly uses static action space, lacks deep integration with chaotic perturbation mechanisms, is prone to falling into local optimality, and has difficulty responding to dynamic environmental changes in real time. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an artificial intelligence-based energy-saving control method and system for refrigeration stations, which solves the problems that the existing PINN methods are mostly based on pure data drive, lack of physical law embedding, and insufficient collaborative application of quantum variational optimization and NQ-GRU, making it difficult to fully capture local thermal field characteristics and nonlinear dynamics, resulting in limited cooling load prediction accuracy, equipment parameter optimization mostly using static action space, lacking deep integration with chaotic perturbation mechanism, easily falling into local optimality, and difficult to respond to dynamic environmental changes in real time.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an energy-saving control method for a refrigeration station based on artificial intelligence, which includes collecting multimodal data, generating an initial pressure field and temperature field through an OFT circuit and linear mapping, optimizing using a PINN neural network, generating a global thermal field map and an airflow field through a QVC circuit and a quantum variational optimization algorithm, mapping the global thermal field map to a quantum Hilbright space, predicting cooling load demand through an NQ-GRU, generating a chaotic sequence using a Sine chaotic map, and extracting local thermal field features, generating device parameter actions through a DQN, perturbing the actions of the optimal device parameters using a chaotic sequence, and converting them into PLC instructions, generating a device operation instruction set, allocating energy storage tasks through a VMD decomposition method, generating an energy storage allocation instruction set, executing the instruction set and setting an alarm mechanism, and storing, collecting, and analyzing the generated multimodal data.

[0008] As a preferred solution of the artificial intelligence-based refrigeration station energy-saving control method of the present invention, wherein: the multimodal data is collected, the initial pressure field and temperature field are generated by OFT circuit and linear mapping, the PINN neural network is used for optimization, the global thermal field map and airflow field are generated by QVC circuit and quantum variational optimization algorithm, the global thermal field map is mapped to quantum Hilbourn space, and the cooling load demand is predicted by NQ-GRU, including:

[0009] Collect multimodal data, including temperature and humidity, wind speed, cooling load, chilled water flow, supply and return water temperature difference, chiller power, cooling tower fan speed, water pump frequency, valve opening, and equipment power consumption data;

[0010] Initialize the virtual CFD calculation domain, extract temperature, humidity, wind speed and cooling load data from the multimodal data, define the virtual grid space, and calculate the number of grid points;

[0011] Extract temperature, humidity and wind speed data from multimodal data and generate grid point coordinates through K-medoids;

[0012] Based on the grid point coordinates and multimodal data, the initial velocity field of each grid point is generated by RBF interpolation method, and the normalized velocity field is generated by mean normalization;

[0013] Using IBM Qiskit, the normalized velocity field is mapped to the quantum state through amplitude encoding. Using the QFT circuit, the frequency domain velocity components are measured. The initial pressure field and initial temperature field are generated by linear mapping and normalized using Min-Max normalization.

[0014] Construct a PINN neural network to calculate mass conservation loss, momentum conservation loss, and energy conservation loss;

[0015] Extract cooling load and wind speed data from multimodal data, and calculate cooling load fluctuation and wind speed fluctuation respectively through wavelet transformation;

[0016] The mass conservation loss, momentum conservation loss, and energy conservation loss are weighted and summed to generate the PINN total loss;

[0017] Initialize the Adam optimizer, calculate the total PINN loss through forward propagation iteration, use automatic differentiation to calculate the gradient of the total PINN loss with respect to the PINN parameters, use the Adam optimizer based on the gradient to update the PINN weights and biases, and monitor the relative change of the gradient norm through the relative change method until the relative change stops decreasing and stops the optimization;

[0018] Input the normalized velocity field, normalized initial pressure field, normalized initial temperature field, multimodal data, and grid point coordinates into the optimized PINN neural network, and output the optimized velocity field, pressure field, and temperature field;

[0019] Based on multimodal data, the turbulent viscosity parameters and heat transfer coefficients are calculated using the RANS method;

[0020] Based on the thermal conductivity coefficient and turbulent viscosity coefficient, a QVC circuit is constructed to encode the optimized velocity field and temperature field into the initial quantum state;

[0021] Through the quantum variational optimization algorithm, CFD conservation loss is embedded to update the QVC parameters. When the maximum number of iterations is reached, the update is stopped and the local thermal field and airflow field are output.

[0022] Use the kd-tree algorithm to stitch the local thermal field and airflow field respectively, smooth them through Gaussian filtering, and generate the global thermal field map and airflow field;

[0023] The global thermal field map is constructed into a time series through a sliding window, converted into a quantum state through angle encoding, mapped to the quantum Hilbert space, and entangled state features are generated through QVFE. The dimension is reduced through quantum principal component analysis to generate an enhanced feature subset.

[0024] An NQ-GRU model is constructed. Based on the enhanced feature subset, the update gate and reset gate are optimized through the quantum gating mechanism. The hidden state at the current moment is calculated, mapped through the fully connected layer, and the cooling load demand forecast for the future time is output.

[0025] As a preferred solution of the energy-saving control method for a refrigeration station based on artificial intelligence described in the present invention, the method uses Sine chaotic mapping to generate a chaotic sequence, extracts local thermal field features, generates device parameter actions through DQN, uses the chaotic sequence to perturb the optimal device parameter actions, and converts them into PLC instructions to generate a device operation instruction set, including:

[0026] Based on the chaotic sequence characteristic analysis experiment, the initial value is set, and the Sine chaotic mapping formula is used for iterative calculation until the maximum number of calculations is reached. All the calculation results are spliced into a chaotic sequence.

[0027] Based on the calculation results of the variance of the global thermal field map and the fluctuation amplitude of the cooling load demand, the number of sub-areas N is determined;

[0028] Based on the chaotic sequence, the center point coordinates of N sub-regions are generated by polynomial interpolation;

[0029] Extract the local thermal field features of N sub-regions, use Sine chaotic mapping to perturb them, splice the perturbed local thermal field features into state vectors, and calculate the entropy value of each state vector using the entropy calculation method;

[0030] Use the biased sample variance method to calculate the local variance of each state vector, generate a dynamic exploration boundary, and splice it into a dynamic exploration boundary set;

[0031] Use k-means clustering to group the exploration boundary set and set the initial priority exploration area;

[0032] Calculate the coverage of the initial priority exploration area, set the coverage threshold based on historical operating data, filter areas with coverage greater than the coverage threshold, and output the priority exploration area and state vector;

[0033] Build an MPC model and define the optimization objective function;

[0034] Based on the priority exploration area and state vector, the predicted equipment parameters in the future are output and optimized by minimizing the optimization objective function. The disturbance is applied through Sine chaos mapping, and the disturbance results are integrated into a short-term control sequence.

[0035] Build a DQN model and set the reward function;

[0036] By maximizing the reward function to maximize the Q value, the global thermal field map and state vector are input into the DQN model, the action of the optimal device parameters is output, and the chaotic sequence is used to perturb the action of the optimal device parameters;

[0037] Convert the post-disturbance actions and short-term control sequences into PLC instructions to generate the equipment operation instruction set.

[0038] As a preferred solution of the artificial intelligence-based refrigeration station energy-saving control method of the present invention, wherein: the energy storage task allocation is performed by the VMD decomposition method to generate the energy storage allocation instruction set, including:

[0039] Based on the difference between the target power and cooling load demand, U modal IMF components are output through VMD decomposition method;

[0040] The center frequency of each IMF component is extracted through FFT. The low-frequency threshold and high-frequency threshold are set based on the time-scale mapping method. The high-frequency modal task group is generated by screening the components with a center frequency greater than the high-frequency threshold, and the low-frequency modal task group is generated by screening the components with a center frequency less than the low-frequency threshold.

[0041] The high-frequency modal tasks are assigned to the supercapacitor, and the low-frequency modal task group is assigned to the battery. An energy storage task allocation plan is generated, converted into PLC instructions, and an energy storage allocation instruction set is generated.

[0042] As a preferred solution of the artificial intelligence-based refrigeration station energy-saving control method of the present invention, the execution of the instruction set and setting of the alarm mechanism include:

[0043] The equipment operation instruction set and energy storage distribution instruction set are executed through the PLC controller, the multimodal data after execution is collected, and the alarm mechanism is set based on the multimodal data after execution.

[0044] As a preferred solution of the artificial intelligence-based refrigeration station energy-saving control method of the present invention, the multimodal data is pre-processed, including:

[0045] The multimodal data are decomposed by EMD decomposition, and the decomposition results are smoothed and denoised using Kalman filtering. The denoised multimodal data are normalized by Z-score and timestamps are added.

[0046] As a preferred solution of the artificial intelligence-based refrigeration station energy-saving control method of the present invention, the multimodal data generated by storage, collection and analysis includes:

[0047] The collected multimodal data is combined with the prediction results generated by the analysis, the equipment operation instruction set, and the energy storage allocation instruction set to generate a complete evaluation report through data visualization and store it in the central database.

[0048] In a second aspect, the present invention provides an artificial intelligence-based refrigeration station energy-saving control system, comprising:

[0049] The collection and optimization module is used to collect multimodal data, generate initial pressure and temperature fields through OFT circuits and linear mapping, optimize them using the PINN neural network, generate global thermal field maps and airflow fields through QVC circuits and quantum variational optimization algorithms, map the global thermal field maps to quantum Hilbourn space, and predict cooling load demand through NQ-GRU;

[0050] The equipment prediction module is used to generate chaotic sequences using Sine chaotic mapping, extract local thermal field features, generate equipment parameter actions through DQN, use chaotic sequences to perturb the optimal equipment parameter actions, and convert them into PLC instructions to generate equipment operation instruction sets;

[0051] The energy storage allocation module is used to allocate energy storage tasks using the VMD decomposition method and generate an energy storage allocation instruction set;

[0052] The execution storage module is used to execute the instruction set and set the alarm mechanism, and store the multimodal data generated by collection and analysis.

[0053] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the artificial intelligence-based refrigeration station energy-saving control method as described in the first aspect of the present invention is implemented.

[0054] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the artificial intelligence-based refrigeration station energy-saving control method as described in the first aspect of the present invention.

[0055] The beneficial effects of the present invention are as follows: the present invention collects multimodal data, uses a PINN neural network for optimization, generates a global thermal field map and an airflow field through a QVC circuit and a quantum variational optimization algorithm, maps the global thermal field map to a quantum Hilbright space, predicts cooling load demand through an NQ-GRU, uses a Sine chaotic map to generate a chaotic sequence, extracts local thermal field features, generates device parameter actions through a DQN, uses the chaotic sequence to perturb the optimal device parameter actions, and converts them into PLC instructions to generate a device operation instruction set; the accuracy and efficiency of energy-saving control of a refrigeration station are improved, and the flexibility of device operation and the system adaptability are enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] Figure 1 This is a flow chart of the energy-saving control method for a refrigeration station based on artificial intelligence in Example 1.

[0058] Figure 2 This is a schematic diagram of the energy-saving control system for a refrigeration station based on artificial intelligence in Example 1.

[0059] Figure 3 Schematic diagram of the parameters and actions of the chaotic perturbation device in Example 1.

[0060] Figure 4 Schematic diagram of the quantum optimization module in Example 1. DETAILED DESCRIPTION

[0061] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0063] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0064] Example 1, reference Figures 1 to 4 , which is the first embodiment of the present invention, provides an energy-saving control method for a refrigeration station based on artificial intelligence, comprising the following steps:

[0065] S1. Collect multimodal data, generate initial pressure and temperature fields through OFT circuits and linear mapping, optimize using PINN neural network, generate global thermal field maps and airflow fields through QVC circuits and quantum variational optimization algorithm, map the global thermal field map to quantum Hilbourn space, and predict cooling load demand through NQ-GRU.

[0066] Specifically, we collect multimodal data, including:

[0067] Collect multimodal data through smart sensors and perform preprocessing;

[0068] The intelligent sensors include temperature and humidity, wind speed, cooling load, flow, temperature difference, power, speed, frequency, rotary position and power consumption sensors;

[0069] The multimodal data includes temperature and humidity, wind speed, cooling load, chilled water flow, supply and return water temperature difference, chiller power, cooling tower fan speed, water pump frequency, valve opening, and equipment power consumption data;

[0070] The preprocessing includes decomposing the multimodal data by EMD decomposition, smoothing and denoising the decomposition results by using Kalman filtering, normalizing the denoised multimodal data by Z-score and adding a timestamp.

[0071] By using multiple types of sensors such as temperature and humidity, wind speed, flow, and power, combined with EMD decomposition, Kalman filtering, and Z-score normalization, we ensure the temporal consistency, stability, and universal scale expression of data, providing a clean and usable input source for subsequent models and improving computational stability.

[0072] Furthermore, the initial pressure and temperature fields are generated through OFT circuits and linear mapping, optimized using the PINN neural network, and the global thermal field map and airflow field are generated through QVC circuits and quantum variational optimization algorithms. The global thermal field map is mapped to the quantum Hilbourn space, and the cooling load demand is predicted using the NQ-GRU, including:

[0073] Initialize the virtual CFD calculation domain, extract the temperature, humidity, wind speed and cooling load data from the multimodal data, define the virtual grid space, and calculate the number of grid points. The formula is:

[0074] V=α·Q+β·v,

[0075]

[0076] Among them, N grid is the number of grid points, Q is the cooling load, v is the wind speed, α and β are the coefficients of cooling load and wind speed respectively (based on empirical analysis), V is the effective calculation volume, R 3 is the grid resolution (verified by experiments);

[0077] Extract temperature, humidity and wind speed data from multimodal data and generate grid point coordinates through K-medoids;

[0078] Based on the grid point coordinates and multimodal data, the initial velocity field of each grid point is generated by RBF interpolation method, and the normalized velocity field is generated by mean normalization;

[0079] Using IBM Qiskit, the normalized velocity field was mapped to a quantum state via amplitude encoding, and the frequency domain velocity components were measured using a QFT circuit.

[0080] Based on the frequency domain velocity component, the initial pressure field and initial temperature field are generated by linear mapping and normalized using Min-Max normalization;

[0081] Construct a PINN neural network, including input layer, hidden layer and output layer;

[0082] Based on the normalized velocity field, the mass conservation loss is calculated by the difference method. Based on the chiller power and cooling tower fan speed data in the multimodal data, the momentum conservation loss is calculated by the Navier-Stokes equation. Based on the normalized initial temperature field and the cooling load data of the multimodal data, the energy conservation loss is calculated by the heat conduction equation.

[0083] Extract cooling load and wind speed data from multimodal data, and calculate cooling load fluctuation and wind speed fluctuation respectively through wavelet transformation;

[0084] The mass conservation loss, momentum conservation loss, and energy conservation loss are weighted and summed (the weights of mass conservation loss, momentum conservation loss, and energy conservation loss are initialized by uniform initialization method, and the thresholds of cooling load fluctuation and wind speed fluctuation are set respectively by empirical rules. The cooling load fluctuation is greater than the cooling load fluctuation threshold or the wind speed fluctuation is greater than the wind speed fluctuation threshold. The energy conservation weight and momentum conservation weight are adjusted so that the total weight of the three is always 1) to generate the PINN total loss;

[0085] Initialize the Adam optimizer, calculate the total PINN loss through forward propagation iteration, use automatic differentiation to calculate the gradient of the total PINN loss with respect to the PINN parameters, use the Adam optimizer based on the gradient to update the PINN weights and biases, and monitor the relative change of the gradient norm through the relative change method until the relative change stops decreasing and stops the optimization;

[0086] Input the normalized velocity field, normalized initial pressure field, normalized initial temperature field, multimodal data, and grid point coordinates into the optimized PINN neural network, and output the optimized velocity field, pressure field, and temperature field;

[0087] Based on multimodal data, the turbulent viscosity parameters and heat transfer coefficients are calculated using the RANS method;

[0088] Based on the thermal conductivity coefficient and turbulent viscosity coefficient, a QVC circuit is constructed, including quantum bits, RY, RZ, and CNOT gates. The optimized velocity field and temperature field are encoded into the initial quantum state. The formula is:

[0089]

[0090] Where Z is the normalization factor, v' and k are the thermal conductivity coefficient and turbulent viscosity coefficient respectively, |ψ is the initial quantum state, M is the number of grid points, and a i is the quantum state amplitude of the i-th grid point, x i and u i are the coordinates of the i-th grid point and the modulus of the optimized velocity field (calculated based on the L2 norm), T i is the optimized temperature field of the i-th grid point;

[0091] Through the quantum variational optimization algorithm, CFD conservation loss is embedded, the QVC parameters are updated, and the update is stopped when the maximum number of iterations is reached (based on cross-validation settings), and the local thermal field and airflow field are output;

[0092] Use the kd-tree algorithm to stitch the local thermal field and airflow field respectively, smooth them through Gaussian filtering, and generate the global thermal field map and airflow field;

[0093] The global thermal field map is constructed into a time series through a sliding window, converted into a quantum state through angle encoding, mapped to the quantum Hilbert space, and entangled state features are generated through QVFE. The dimension is reduced through quantum principal component analysis to generate an enhanced feature subset.

[0094] Build an NQ-GRU model, including input layer, GRU layer, fully connected layer, and output layer;

[0095] Physical simulation data is collected through the digital twin platform, mapped to the quantum Hilbert space, and a simulation-enhanced feature subset is generated. This feature subset is then input into the NQ-GRU model for training and optimized using the MSE loss function.

[0096] Based on the enhanced feature subset, the update gate and reset gate are optimized through the quantum gating mechanism to calculate the hidden state at the current moment;

[0097] The current hidden state is mapped to the prediction target through the fully connected layer, and the cooling load demand at the predicted future time is output.

[0098] The velocity field is mapped to the quantum frequency domain through amplitude coding and QFT circuit, and the pressure field and temperature field are derived by linear mapping, which greatly enhances the expression ability of frequency domain characteristics in physical modeling and helps to reveal the microscopic mechanism of high-frequency disturbances and energy consumption changes in the flow field. By introducing three types of physical loss functions, namely mass conservation, momentum conservation and energy conservation, and adaptively adjusting the loss weight according to the cooling load and wind speed fluctuations, the physical consistency of the model and its robustness under dynamic boundary conditions are effectively enhanced. The QVC circuit and variational optimization algorithm are used to encode the heat transfer coefficient and turbulent viscosity information, and the adjustable electric Path gate optimization generates airflow fields, achieving scene reconstruction that is highly consistent with physical laws and providing high-precision input data for cooling load prediction. After the global thermal field map is converted into a time series, it is mapped to a quantum state through angle encoding. QVFE is then used to extract entanglement features, and quantum PCA is combined for dimensionality reduction, effectively compressing feature dimensions while retaining key information, improving learning efficiency and generalization capabilities. A quantum gating mechanism is introduced to dynamically adjust the update gate and reset gate weights, giving the model better memory stability and nonlinear modeling capabilities when learning complex time series patterns, further improving prediction accuracy and engineering usability.

[0099] S2. Use Sine chaotic mapping to generate chaotic sequences and extract local thermal field features. Generate device parameter actions through DQN. Use chaotic sequences to perturb the optimal device parameter actions and convert them into PLC instructions to generate device operation instruction sets, including:

[0100] Based on the chaotic sequence characteristic analysis experiment, the initial value is set, and the Sine chaotic mapping formula is used for iterative calculation until the maximum number of calculations is reached (based on the actual application scenario). All the calculation results are spliced into a chaotic sequence. The formula is:

[0101]

[0102] Among them, q c is the initial value of the chaotic mapping formula iteration, c is the iteration index, b is the control parameter (set according to the maximum chaos requirement), and π is the circumference of a circle;

[0103] The number of sub-areas is determined based on the variance of the global thermal field map (calculated using the unbiased sample variance formula) and the fluctuation range of the cooling load demand (the difference between the maximum and minimum values). The formula is:

[0104] N=τ·e+γ·ΔL,

[0105] Where N is the number of sub-regions, e is the variance of the global thermal field map, ΔL is the fluctuation amplitude of the cooling load demand, τ and γ are the weight coefficients of the variance of the global thermal field map and the fluctuation amplitude of the cooling load demand, respectively (initialized based on uniform distribution);

[0106] Based on the chaotic sequence, the center point coordinates of N sub-regions are generated by polynomial interpolation;

[0107] Extract the local thermal field characteristics of N sub-regions, the formula is:

[0108]

[0109] Where F is the local thermal field characteristic, is the average temperature of the sub-region (obtained by summing and averaging the temperatures of the sub-regions), is the temperature gradient (calculated by convolving the temperature field in the global thermal field map of the sub-region using the Sobel operator), and A is the cooling load demand;

[0110] Use Sine chaotic mapping to perturb the local thermal field characteristics, splice the perturbed local thermal field characteristics into state vectors, and calculate the entropy value of each state vector using the entropy calculation method;

[0111] The biased sample variance method is used to calculate the local variance of each state vector, generate a dynamic exploration boundary, and splice it into a dynamic exploration boundary set. The formula is:

[0112]

[0113] Among them, B is the dynamic exploration boundary, is the mean of the state vector, t y,r is the distribution critical value (obtained based on the t-distribution table), y is the degree of freedom (based on the t-distribution mutation strategy, combined with the entropy value of each state vector, set by the Sigmoid function), r is the confidence level (obtained based on historical operating condition analysis), and θ is the standard deviation of the state vector (calculated based on the square root method);

[0114] Use k-means clustering to group the exploration boundary set, sort the entropy value of each state vector after grouping in descending order, and select the top j regions with the highest entropy value in the grouping result (based on the cumulative contribution rate of entropy value) and set them as the initial priority exploration regions;

[0115] Calculate the coverage of the initial priority exploration area (the ratio of the exploration boundary set to the initial priority exploration area), set a coverage threshold based on historical operating condition data, filter areas with coverage greater than the coverage threshold, and output the priority exploration area and state vector;

[0116] Construct an MPC model, including input layer, design layer and output layer;

[0117] The simulation data set is obtained through the COMSOL simulation platform to train the MPC model;

[0118] Based on the priority exploration area and state vector, the MPC model outputs the predicted equipment parameters for the future time, including chiller load, cooling tower fan speed, water pump frequency, and valve opening;

[0119] Based on the average entropy value of the priority exploration area, the optimization objective function is defined as follows:

[0120]

[0121] Among them, J is the optimization objective function value, is the equal weight coefficient, E is the sum of the power consumption of all devices, E max is the maximum value of the sum of the power consumption of all devices, ΔT” and ΔP are the temperature deviation and pressure deviation (the difference between the predicted temperature and the actual temperature and the difference between the optimized pressure field and the actual pressure field), PMV sc To predict the PMV value (based on the PMV model), H avg is the average entropy value of the priority area;

[0122] The prediction results of the MPC model are optimized by minimizing the optimization objective function, and disturbances are applied through Sine chaos mapping, and the disturbance results are integrated into a short-term control sequence;

[0123] Build a DQN model, including input layer, hidden layer, and output layer;

[0124] Use Gazebo to obtain simulation data and train the DQN model;

[0125] Set the reward function, the formula is:

[0126]

[0127] Among them, R is the function reward value, and o is the coverage rate of the priority exploration area;

[0128] By maximizing the reward function to maximize the Q value, the global thermal field map and state vector are input into the DQN model, and the action with the optimal device parameters is output;

[0129] Use chaotic sequences to perturb the action of optimal device parameters;

[0130] Convert the post-disturbance actions and short-term control sequences into PLC instructions to generate the equipment operation instruction set.

[0131] The device parameter actions and thermal field characteristics are disturbed by Sine chaotic mapping, effectively avoiding the local optimal trap in reinforcement learning, helping the system to escape the overfitting area, and improving exploration efficiency and control accuracy. The local thermal field characteristics extracted at the sub-region scale, combined with temperature gradient and load changes, make system regulation more targeted, respond to local abnormal conditions, and improve the system's adaptability to environmental dynamics. Entropy value is combined with local variance to construct a dynamic exploration boundary, which not only considers state complexity but also quantifies the range of uncertainty changes. It can dynamically adjust the tolerance range of the control strategy to environmental disturbances and improve the robustness of system operation. Through entropy-based clustering and coverage screening strategies, not only resources are allocated and equipment is adjusted in high-value areas, avoiding redundant control of the entire space, but the optimization objective function integrates equipment power consumption, temperature and pressure deviations, PMV human comfort index and information entropy indicators, reflecting the joint optimization idea of multiple objectives and multiple constraints, which not only improves energy utilization efficiency but also takes into account operational stability and terminal comfort.

[0132] S3. Allocate energy storage tasks using the VMD decomposition method to generate an energy storage allocation instruction set, including:

[0133] Based on the difference between the target power (obtained by weighted summation of mechanical power and thermal power) and the cooling load demand, the VMD decomposition method is used to output U modal IMF components (set based on actual demand). The formula is:

[0134]

[0135] Among them, p 1 and p 2 are mechanical power and thermal power respectively, ΔP * h is the pressure difference (obtained based on the pressure field), Q' h is the cold water flow rate, η h is the equipment efficiency (obtained based on actual equipment testing), m is the mass flow rate (based on the product of cold water flow rate and water density), ΔT 1 is the supply and return water temperature difference, c p is the specific heat capacity of water, h is the index of the device;

[0136] The center frequency of each IMF component is extracted through FFT. The low-frequency threshold and high-frequency threshold are set based on the time-scale mapping method. The high-frequency modal task group is generated by screening the components with a center frequency greater than the high-frequency threshold, and the low-frequency modal task group is generated by screening the components with a center frequency less than the low-frequency threshold.

[0137] The high-frequency modal tasks are assigned to the supercapacitor, and the low-frequency modal task group is assigned to the battery. An energy storage task allocation plan is generated, converted into PLC instructions, and an energy storage allocation instruction set is generated.

[0138] By inputting the difference between the target power and the cooling load as a non-stationary signal into the VMD framework, the multi-scale fluctuation information of the system load can be effectively extracted. The center frequency of each mode is obtained through FFT, and the frequency threshold is set using the time scale mapping method. This enables the system to divide tasks into high-frequency and low-frequency tasks according to the speed of modal changes, realizing frequency domain feature-driven task allocation. The frequency-based task allocation strategy not only improves the utilization efficiency of the energy storage unit, but also extends the equipment service life and optimizes the energy flow path.

[0139] S4, executing the instruction set and setting the alarm mechanism, storing the multimodal data collected and analyzed;

[0140] Specifically, executing the instruction set and setting the alarm mechanism include:

[0141] Execute the equipment operation instruction set and energy storage distribution instruction set through the PLC controller, collect the multimodal data after execution, set the alarm threshold (based on the historical equipment operation status setting), and if the multimodal data after execution deviates from the alarm threshold When (Y is the floating range, set based on statistical analysis), a low-level alarm is triggered and staff are notified to conduct an inspection;

[0142] If the multimodal data after execution deviates from the alarm threshold Y, an intermediate alarm is triggered and the staff is notified to make adjustments and repairs;

[0143] If the multimodal data after execution deviates from the alarm threshold 2Y, a high-level alarm will be triggered and the machine will be shut down immediately.

[0144] By executing operation and energy storage allocation instructions through PLC, not only standardized control of process flows such as refrigeration and energy supply is achieved, but also data and behavior linkage between different modules are effectively integrated, supporting rapid response to environmental disturbances and predictive maintenance. Statistical analysis methods are used to set floating thresholds, which significantly enhances the system's adaptability and early warning capabilities, dynamically adjusts alarm sensitivity, reduces false alarms and missed alarms, and ensures the accuracy and timeliness of alarms. By dividing alarms into three levels, not only the system's fault tolerance is improved, but also the allocation of operation and maintenance resources is optimized.

[0145] Furthermore, the multimodal data generated by storage, collection and analysis includes:

[0146] The collected multimodal data is combined with the prediction results generated by the analysis, the equipment operation instruction set, and the energy storage allocation instruction set to generate a complete evaluation report through data visualization and store it in the central database.

[0147] Through data visualization and the generation of evaluation reports, multi-dimensional information is presented to users, helping them to quickly understand the system's operating status and optimization results. This provides data-driven decision support for management, helping to improve the scientific nature and accuracy of management. By storing information in a central database, it provides data support for global optimization and improves the efficiency of collaborative work.

[0148] This embodiment also provides an artificial intelligence-based refrigeration station energy-saving control system, including:

[0149] The collection and optimization module is used to collect multimodal data, generate initial pressure and temperature fields through OFT circuits and linear mapping, optimize them using the PINN neural network, generate global thermal field maps and airflow fields through QVC circuits and quantum variational optimization algorithms, map the global thermal field maps to quantum Hilbourn space, and predict cooling load demand through NQ-GRU;

[0150] The equipment prediction module is used to generate chaotic sequences using Sine chaotic mapping, extract local thermal field features, generate equipment parameter actions through DQN, use chaotic sequences to perturb the optimal equipment parameter actions, and convert them into PLC instructions to generate equipment operation instruction sets;

[0151] The energy storage allocation module is used to allocate energy storage tasks using the VMD decomposition method and generate an energy storage allocation instruction set;

[0152] The execution storage module is used to execute the instruction set and set the alarm mechanism, and store the multimodal data generated by collection and analysis.

[0153] This embodiment also provides a computer device, which is suitable for the case of an artificial intelligence-based refrigeration station energy-saving control method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the artificial intelligence-based refrigeration station energy-saving control method proposed in the above embodiment.

[0154] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0155] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the energy-saving control method for a refrigeration station based on artificial intelligence as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0156] In summary, the present invention collects multimodal data, uses a PINN neural network for optimization, generates a global thermal field map and airflow field through a QVC circuit and a quantum variational optimization algorithm, maps the global thermal field map to a quantum Hilbright space, predicts cooling load demand through NQ-GRU, generates a chaotic sequence using a Sine chaotic map, extracts local thermal field features, generates device parameter actions through a DQN, uses the chaotic sequence to perturb the optimal device parameter actions, and converts them into PLC instructions to generate a device operation instruction set; thereby improving the accuracy and efficiency of energy-saving control in refrigeration stations, and enhancing the flexibility of equipment operation and the system's adaptability.

[0157] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based refrigeration station energy-saving control method, characterized by: include, Collect multimodal data, generate initial pressure and temperature fields through OFT circuits and linear mapping, optimize using PINN neural network, generate global thermal field maps and airflow fields through QVC circuits and quantum variational optimization algorithm, map the global thermal field map to quantum Hilbourn space, and predict cooling load demand through NQ-GRU; Use Sine chaotic mapping to generate chaotic sequences and extract local thermal field features. Generate device parameter actions through DQN. Use chaotic sequences to perturb the optimal device parameter actions and convert them into PLC instructions to generate device operation instruction sets. Energy storage tasks are allocated using VMD decomposition method to generate energy storage allocation instruction sets; Execute instruction sets and set alarm mechanisms, store, collect and analyze the multimodal data generated.

2. The artificial intelligence-based refrigeration station energy-saving control method according to claim 1, characterized in that: The method collects multimodal data, generates initial pressure and temperature fields through OFT circuits and linear mapping, optimizes using a PINN neural network, generates a global thermal field map and airflow field through a QVC circuit and a quantum variational optimization algorithm, maps the global thermal field map to a quantum Hilbourt space, and predicts cooling load demand through an NQ-GRU, including: Collect multimodal data, including temperature and humidity, wind speed, cooling load, chilled water flow, supply and return water temperature difference, chiller power, cooling tower fan speed, water pump frequency, valve opening, and equipment power consumption data; Initialize the virtual CFD calculation domain, extract temperature, humidity, wind speed and cooling load data from the multimodal data, define the virtual grid space, and calculate the number of grid points; Extract temperature, humidity and wind speed data from multimodal data and generate grid point coordinates through K-medoids; Based on the grid point coordinates and multimodal data, the initial velocity field of each grid point is generated by RBF interpolation method, and the normalized velocity field is generated by mean normalization; Using IBM Qiskit, the normalized velocity field is mapped to the quantum state through amplitude encoding. Using the QFT circuit, the frequency domain velocity components are measured. The initial pressure field and initial temperature field are generated by linear mapping and normalized using Min-Max normalization. Construct a PINN neural network to calculate mass conservation loss, momentum conservation loss, and energy conservation loss; Extract cooling load and wind speed data from multimodal data, and calculate cooling load fluctuation and wind speed fluctuation respectively through wavelet transformation; The mass conservation loss, momentum conservation loss, and energy conservation loss are weighted and summed to generate the PINN total loss; Initialize the Adam optimizer, calculate the total PINN loss through forward propagation iteration, use automatic differentiation to calculate the gradient of the total PINN loss with respect to the PINN parameters, use the Adam optimizer based on the gradient to update the PINN weights and biases, and monitor the relative change of the gradient norm through the relative change method until the relative change stops decreasing and stops the optimization; Input the normalized velocity field, normalized initial pressure field, normalized initial temperature field, multimodal data, and grid point coordinates into the optimized PINN neural network, and output the optimized velocity field, pressure field, and temperature field; Based on multimodal data, the turbulent viscosity parameters and heat transfer coefficients are calculated using the RANS method; Based on the thermal conductivity coefficient and turbulent viscosity coefficient, a QVC circuit is constructed to encode the optimized velocity field and temperature field into the initial quantum state; Through the quantum variational optimization algorithm, CFD conservation loss is embedded to update the QVC parameters. When the maximum number of iterations is reached, the update is stopped and the local thermal field and airflow field are output. Use the kd-tree algorithm to stitch the local thermal field and airflow field respectively, smooth them through Gaussian filtering, and generate the global thermal field map and airflow field; The global thermal field map is constructed into a time series through a sliding window, converted into a quantum state through angle encoding, mapped to the quantum Hilbert space, and entangled state features are generated through QVFE. The dimension is reduced through quantum principal component analysis to generate an enhanced feature subset. An NQ-GRU model is constructed. Based on the enhanced feature subset, the update gate and reset gate are optimized through the quantum gating mechanism. The hidden state at the current moment is calculated, mapped through the fully connected layer, and the cooling load demand forecast for the future time is output.

3. The artificial intelligence-based refrigeration station energy-saving control method according to claim 2, characterized in that: The method uses Sine chaotic mapping to generate a chaotic sequence, extracts local thermal field features, generates device parameter actions through DQN, uses the chaotic sequence to perturb the optimal device parameter actions, and converts them into PLC instructions to generate a device operation instruction set, including: Based on the chaotic sequence characteristic analysis experiment, the initial value is set, and the Sine chaotic mapping formula is used for iterative calculation until the maximum number of calculations is reached. All the calculation results are spliced into a chaotic sequence. Based on the calculation results of the variance of the global thermal field map and the fluctuation amplitude of the cooling load demand, the number of sub-areas N is determined; Based on the chaotic sequence, the center point coordinates of N sub-regions are generated by polynomial interpolation; Extract the local thermal field features of N sub-regions, use Sine chaotic mapping to perturb them, splice the perturbed local thermal field features into state vectors, and calculate the entropy value of each state vector using the entropy calculation method; Use the biased sample variance method to calculate the local variance of each state vector, generate a dynamic exploration boundary, and splice it into a dynamic exploration boundary set; Use k-means clustering to group the exploration boundary set and set the initial priority exploration area; Calculate the coverage of the initial priority exploration area, set the coverage threshold based on historical operating data, filter areas with coverage greater than the coverage threshold, and output the priority exploration area and state vector; Build an MPC model and define the optimization objective function; Based on the priority exploration area and state vector, the predicted equipment parameters in the future are output and optimized by minimizing the optimization objective function. The disturbance is applied through Sine chaos mapping, and the disturbance results are integrated into a short-term control sequence. Build a DQN model and set the reward function; By maximizing the reward function to maximize the Q value, the global thermal field map and state vector are input into the DQN model, the action of the optimal device parameters is output, and the chaotic sequence is used to perturb the action of the optimal device parameters; Convert the post-disturbance actions and short-term control sequences into PLC instructions to generate the equipment operation instruction set.

4. The artificial intelligence-based refrigeration station energy-saving control method according to claim 3, characterized in that: The energy storage task allocation is performed by the VMD decomposition method to generate an energy storage allocation instruction set, including: Based on the difference between the target power and cooling load demand, U modal IMF components are output through VMD decomposition method; The center frequency of each IMF component is extracted through FFT. The low-frequency threshold and high-frequency threshold are set based on the time-scale mapping method. The high-frequency modal task group is generated by screening the components with a center frequency greater than the high-frequency threshold, and the low-frequency modal task group is generated by screening the components with a center frequency less than the low-frequency threshold. The high-frequency modal tasks are assigned to the supercapacitors, and the low-frequency modal task group is assigned to the battery. An energy storage task allocation plan is generated, converted into PLC instructions, and an energy storage allocation instruction set is generated.

5. The artificial intelligence-based refrigeration station energy-saving control method according to claim 4, characterized in that: The execution of the instruction set and setting of the alarm mechanism include: The equipment operation instruction set and energy storage distribution instruction set are executed through the PLC controller, the multimodal data after execution is collected, and the alarm mechanism is set based on the multimodal data after execution.

6. The artificial intelligence-based refrigeration station energy-saving control method according to claim 2, characterized in that: The multimodal data is preprocessed, including: The multimodal data are decomposed by EMD decomposition, and the decomposition results are smoothed and denoised using Kalman filtering. The denoised multimodal data are normalized by Z-score and timestamps are added.

7. The artificial intelligence-based refrigeration station energy-saving control method according to claim 6, characterized in that: The multimodal data generated by the storage, collection and analysis includes: The collected multimodal data is combined with the prediction results generated by the analysis, the equipment operation instruction set, and the energy storage allocation instruction set to generate a complete evaluation report through data visualization and store it in the central database.

8. An artificial intelligence-based energy-saving control system for a refrigeration station, based on the artificial intelligence-based energy-saving control method for a refrigeration station according to any one of claims 1 to 7, characterized in that: include, The collection and optimization module is used to collect multimodal data, generate initial pressure and temperature fields through OFT circuits and linear mapping, optimize them using the PINN neural network, generate global thermal field maps and airflow fields through QVC circuits and quantum variational optimization algorithms, map the global thermal field maps to quantum Hilbourn space, and predict cooling load demand through NQ-GRU; The equipment prediction module is used to generate chaotic sequences using Sine chaotic mapping, extract local thermal field features, generate equipment parameter actions through DQN, use chaotic sequences to perturb the optimal equipment parameter actions, and convert them into PLC instructions to generate equipment operation instruction sets; The energy storage allocation module is used to allocate energy storage tasks using the VMD decomposition method and generate an energy storage allocation instruction set; The execution storage module is used to execute the instruction set and set the alarm mechanism, and store the multimodal data generated by collection and analysis.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the artificial intelligence-based refrigeration station energy-saving control method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the artificial intelligence-based refrigeration station energy-saving control method according to any one of claims 1 to 7 are implemented.

Images