An intelligent refrigeration system based on cold load historical data to predict and adjust fresh air volume

Abstract

The application provides an intelligent refrigeration system for predicting and adjusting fresh air volume based on cold load historical data, which comprises an evaporator, a heat exchanger and a building unit, water in the heat exchanger enters the evaporator, exchanges heat with refrigerant in the evaporator to become cold water, then enters the heat exchanger, exchanges heat with air in the heat exchanger, and then returns to the evaporator to form a heat exchange cycle; air enters the heat exchanger, exchanges heat with cold water in the heat exchanger to form cold air, and then enters the building unit for refrigeration; the air entering the heat exchanger comprises mixed air of fresh air and return air, the return air is indoor air discharged after refrigeration of the building unit, the fresh air comprises a fresh air pipeline and a fan arranged on the fresh air pipeline, the fan is connected with a control system in data mode, and is used for controlling the frequency of the fan so as to control the fresh air volume entering the heat exchanger. The application adjusts the refrigeration capacity by controlling the flow of fresh air, determines the frequency of the fan, and adjusts the fresh air volume of the fan according to the frequency.

Description

Technical Field

[0001] This invention relates to the field of refrigeration technology, and more specifically to a compression refrigeration air conditioning system. Background Technology

[0002] Large buildings, especially large public buildings, generally suffer from high energy consumption and low energy efficiency. Especially during the cooling season, due to the unpredictability and fluctuation of the building's cooling load, air conditioning units often operate in a traditional fixed mode, resulting in a large amount of energy waste.

[0003] Existing air conditioning units are air handling equipment assembled from various air handling functional sections. They are suitable for air conditioning systems with resistance greater than 100Pa. The air handling functional sections of the unit include: air mixing, flow equalization, filtration, cooling, primary and secondary heating, dehumidification, humidification, supply fan, return fan, water spray, noise reduction, and heat recovery units. While units using variable frequency compressors can change cooling output by altering the compressor's operating frequency, the frequency change is still adjusted according to a fixed frequency range set in the program, meaning there are only a few fixed frequencies. Furthermore, the control of variable frequency compressors is too complex, and the manufacturing of compressors is significantly more expensive and complex than that of fixed frequency compressors. Variable frequency systems also require higher pressure resistance from the compressor and system valves, resulting in higher costs for the compressor, system valves, and variable frequency controller. Moreover, the variable frequency controller technology is more complex and difficult to master. Current central air conditioning cooling capacity distribution, i.e., the cooling capacity in cooling mode, is generally adjusted by temperature regulators located at the air conditioning terminals. However, this adjustment method suffers from low adjustment accuracy and is prone to resource waste.

[0004] The existing cooling capacity allocation technology still has some shortcomings. Cooling capacity allocation relies too much on meteorological and time variables. Due to changes in external conditions or the difficulty in obtaining data, there is still a large degree of uncertainty in the input of time and meteorological variables. In addition, the traditional heat transfer calculation method has high requirements for building physics model parameters, is highly targeted to projects, has many prediction input variables, and the system calculation is complex.

[0005] Therefore, in order to reduce reliance on external information such as meteorological and time variables and ensure the universality of the conclusions, this invention aims to propose an intelligent cooling system that predicts and adjusts the fresh air volume based on historical cooling load data, so as to maximize energy saving while meeting user needs. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent refrigeration system that predicts and adjusts the fresh air volume based on historical cooling load data, so as to maximize energy saving while meeting user needs.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] A smart refrigeration system for predicting and adjusting fresh air volume based on historical cooling load data includes an evaporator, a heat exchanger, and building units. Water enters the evaporator and exchanges heat with the refrigerant to become chilled water, which then enters the heat exchanger again to exchange heat with air entering the heat exchanger before returning to the evaporator, forming a heat exchange cycle. Air enters the heat exchanger and exchanges heat with the chilled water to form chilled air, which then enters the building units for cooling. The air entering the heat exchanger includes a mixture of fresh air and return air, where the return air is the indoor air exhausted after cooling from the building units. The system includes a fresh air duct and a fan installed on the duct. The fan is connected to a control system to control its frequency, thereby controlling the amount of fresh air entering the heat exchanger.

[0009] As an improvement, the control system automatically controls the frequency of the fans according to the cooling capacity required by the building unit, thereby controlling the amount of fresh air entering the heat exchanger.

[0010] As an improvement, when the cooling capacity required by a building unit increases, the frequency of the automatically controlled fan increases; when the cooling capacity required by a building unit decreases, the frequency of the automatically controlled fan decreases.

[0011] As an improvement, the cooling capacity required by the building unit is predicted based on the personnel modality acquisition module, building modality, meteorological modality, and equipment modality.

[0012] As an improvement, personnel modality includes the number of personnel, the type of personnel activity, and the distribution of personnel.

[0013] As an improvement, building modalities include building material properties, building location, and leakage levels of doors and windows.

[0014] As an improvement, the meteorological modality includes outdoor temperature and humidity, solar irradiance, and wind direction.

[0015] As an improvement, equipment modes include the operating power of equipment such as lighting equipment and large electrical appliances.

[0016] As an improvement, the air entering the heat exchanger consists of 50-80% return air and 20-50% fresh air, which are mixed before entering the heat exchanger.

[0017] As an improvement, a control method for the aforementioned refrigeration system includes the following steps:

[0018] The first step is to collect data from each module through embedded external modules, which mainly include personnel modal acquisition modules, building modal acquisition modules, meteorological modal acquisition modules, and equipment modal acquisition modules;

[0019] The second step is to preprocess the data for each modality separately;

[0020] The third step is to standardize the data features;

[0021] The fourth step is to extract features from the data of each modality. The image data of different modules in the image modality are set up with different RNNs according to their respective needs.

[0022] The fifth step is to fuse the data from each modality.

[0023] Compared with the prior art, the present invention has the following advantages:

[0024] 1. The cooling system adjusts the cooling capacity by regulating the flow rate of fresh air. The fan on the fresh air duct is connected to the control system for data control of the fan frequency. Based on the predicted load, the required air volume is calculated, the proportion of fresh air is determined, and the fan frequency is adjusted accordingly.

[0025] 2. This invention uses a multimodal fusion approach to predict cooling capacity and introduces a convolutional neural network to identify and predict personnel modalities. This increases the accuracy of prediction while focusing on the uneven distribution of personnel in large spaces of large buildings, making the personnel load distribution more reasonable and the prediction data more comprehensive and accurate. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall structure of the refrigeration system of the present invention.

[0027] Figure 2 This is a schematic diagram of the mixed return air and fresh air structure of the present invention.

[0028] Figure 3 This is a schematic diagram of the refrigeration system module allocation of the present invention.

[0029] Figure 4 This is a schematic diagram illustrating the working principle of the energy-saving control module of the refrigeration system of the present invention.

[0030] Figure 5 This is a schematic diagram of the working principle of the energy-saving module of the present invention. Detailed Implementation

[0031] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0032] Unless otherwise specified, in this article, " / " represents division, and "×" and "*" represent multiplication.

[0033] Figure 1A refrigeration system for controlling the cooling capacity of different building units is demonstrated. The system includes an evaporator 1, a condenser 2, a compressor 3, and an expansion valve 4, which are connected in sequence to form a refrigeration cycle system.

[0034] The system also includes a heat exchanger 5 and a building unit 6. Water in the heat exchanger 5 enters the evaporator 1, exchanges heat with the refrigerant in the evaporator 1 to become cold water, and then enters the heat exchanger 5 to exchange heat with the air entering the heat exchanger 5, and then returns to the evaporator 1 to form a heat exchange cycle. Air enters the heat exchanger to exchange heat with the cold water in the heat exchanger to form cold air, which then enters the building unit 6 for cooling. The air entering the heat exchanger includes a mixture of fresh air and return air. The return air is the indoor air discharged after cooling by the building unit. The system includes a fresh air duct 7 and a fan 8 installed on the fresh air duct 7. The fan 8 is connected to the control system for data control of the fan frequency, thereby controlling the amount of fresh air entering the heat exchanger.

[0035] This invention controls the cooling capacity of different building units by adjusting the frequency of the fans in the fresh air ducts. Controlling cooling capacity through a single fan frequency results in a simple and reliable control structure. Furthermore, because the return air temperature is lower than the outdoor fresh air temperature, by returning the exhaust air from the building units to the heat exchanger, the cooling capacity of the return air can be fully utilized, saving energy.

[0036] As an improvement, the air entering the heat exchanger consists of 50-80% return air and 20-50% fresh air, which are mixed before entering the heat exchanger. By controlling the ratio of fresh air to return air, sufficient fresh air input can be ensured while fully utilizing the cooling capacity of the return air, thus guaranteeing energy savings.

[0037] As an improvement, the control system automatically controls the frequency of the fan 8 based on the cooling capacity required by building unit 6, thereby controlling the amount of fresh air entering the heat exchanger. Intelligent control is achieved by intelligently controlling the fan frequency.

[0038] As an improvement, when the cooling capacity required by a building unit increases, the frequency of the automatically controlled fan increases; when the cooling capacity required by a building unit decreases, the frequency of the automatically controlled fan decreases. Through this intelligent control, accurate regulation of the cooling capacity is achieved.

[0039] Figure 2 A schematic diagram of the ductwork for mixing return air and fresh air is shown. (For example...) Figure 2As shown, the return air duct 9 is arranged perpendicularly to the fresh air duct 7. A mixing component 10 is installed in the return air duct 9. The mixing component 10 is located upstream of the connection between the fresh air duct and the return air duct. The mixing component 10 is a blade structure, which includes multiple blades. The extension direction of the gap between adjacent blades is parallel to the central axis of the fresh air duct.

[0040] A fresh air mixing device 11 is installed in the fresh air duct 7. The fresh air mixing device 11 has a blade structure and is located upstream of the connection point of the fresh air duct 7. The blade structure has multiple blades, and the extension direction of the flow channel between adjacent blades is parallel to the central axis of the return air duct. Through the above arrangement, the gaps between the distribution components of the return air duct and the fresh air duct are parallel to each other, which can reduce the impact and noise during fluid mixing.

[0041] As an improvement, the return air layer containing the return air distribution component and the fresh air layer containing the fresh air distribution component are staggered. This staggered arrangement ensures that during mixing, the fresh air flowing from the fresh air distribution component enters precisely between the two staggered return air layers, while the return air flowing from the return air distribution layer enters precisely between the adjacent fresh air layers. This staggered arrangement of the return air layer and the fresh air layer results in a more uniform distribution, and also reduces noise and impact.

[0042] The spacing between the return air channels formed by adjacent blades of the homogenizing component 10 is different. As the distance from the fresh air duct increases, the spacing between the return air channels formed by adjacent blades becomes smaller and smaller.

[0043] Near the fresh air outlet, the fresh air flow is at its maximum, resulting in the best mixing of fresh and return air. However, this leads to uneven mixing overall. As the distance from the outlet increases, the amount of fresh air decreases, and the mixing deteriorates, resulting in poor overall mixing performance. This invention addresses this by improving the blade spacing, causing the return air flow to decrease gradually along the direction of the fresh air entering the return air duct, thus achieving balanced mixing throughout the entire area and improving overall mixing uniformity.

[0044] As an improvement, the spacing of the return air duct can be set by adjusting the size of the blades. For example, the wider the blade is, the further away it is from the fresh air duct, and the spacing between adjacent blades can be reduced by increasing the width.

[0045] Based on existing air conditioners, this invention creatively proposes a mixed regulation system for fresh air and return air, which includes two opposing regulating devices at the upstream and downstream ends to achieve dual control and dual flow equalization, resulting in a significant equalization effect between return air and fresh air and high economic benefits.

[0046] Preferably, the spacing between adjacent blades forming the return airflow channel increases with increasing distance from the fresh air duct. This variation in spacing improves the mixing ratio and achieves better overall uniformity.

[0047] Preferred, such as Figure 2 As shown, the return air duct is equipped with a return air flow equalization device 12, which is located downstream of the connection between the fresh air duct and the return air duct. The flow equalization device 12 has a multi-blade structure. As an improvement, the flow channel spacing between adjacent blades farther away from the fresh air duct is larger. With this arrangement, the mixed air flows as far away from the fresh air duct as possible, thereby improving the situation where the mixed air is concentrated near the fresh air duct after mixing, and making the mixed air evenly distributed throughout the return air duct.

[0048] As an improvement, the spacing between adjacent blades of the return air equalization device 12 increases with increasing distance from the fresh air duct. This configuration further ensures even distribution of mixed air throughout the return air duct.

[0049] As an improvement, the flow equalization device 12 channels and the homogenizing component 10 channels are arranged perpendicularly to each other. Because the two sets of regulating devices are perpendicular to each other, the flow can be diffused in all four directions, both horizontally and vertically, resulting in a higher degree of homogenization.

[0050] This invention includes two devices, upstream and downstream, that enable dual control and dual flow equalization, resulting in a significant effect on the equalization of return and fresh air ratios. Furthermore, the flow equalization device ensures uniform distribution throughout the entire duct. The cooperation of these two devices achieves flow control, uniform proportional distribution, and uniform fuel distribution throughout the entire duct.

[0051] Preferably, multiple mixing devices 12 are provided, and the gaps between adjacent blades in the mixing devices become smaller and smaller along the flow direction of the fluid in the return air duct. By varying the amplitude, it is possible to further ensure uniform flow distribution while maintaining low flow resistance.

[0052] Existing methods for predicting cooling load still have some shortcomings:

[0053] 1. It is difficult to predict the cooling load of public places by determining the flow of people, which makes it difficult to determine the personnel load. It can only be estimated by experience. If the estimated cooling load is too high, it will waste electricity and reduce energy efficiency. If the estimated cooling load is too low, it will cause the indoor temperature and humidity to be too high, failing to achieve the cooling effect and affecting the comfort of indoor personnel.

[0054] 2. If the air conditioning load forecasting method cannot take into account the distribution of people in a large space. In a large space, the density of people in different areas varies, resulting in large temperature differences between different areas. The temperature and humidity are high in densely populated areas, resulting in low human comfort, while the cooling capacity overflows in sparsely populated areas, causing energy waste.

[0055] 3. The different physical properties of different buildings can significantly affect the cooling load. The heat transfer coefficient of the building envelope directly affects the heat gain and loss of the building, while the heat transfer coefficient of exterior windows and the solar heat gain coefficient also affect the indoor temperature. Therefore, physical modeling can provide a more intuitive way to observe and summarize the physical properties of buildings.

[0056] 4. Weather changes have a significant impact on cooling load. The building envelope exchanges heat directly with the outdoor environment. When the outdoor temperature rises, the amount of heat transferred into the room through the building envelope increases, and the load changes accordingly. At the same time, in order to meet the requirements of indoor air quality, it is necessary to introduce fresh outdoor air. The rise in outdoor temperature will inevitably lead to the rise in the temperature of the fresh air, which will increase the fresh air load of processing the fresh air to a state that can be supplied. Therefore, weather is also one of the factors that cannot be ignored when affecting cooling load.

[0057] As an improvement, the cooling capacity required by the building unit is predicted based on the personnel modality acquisition module, building modality, meteorological modality, and equipment modality.

[0058] As an improvement, personnel modality includes the number of personnel, the type of personnel activity, and the distribution of personnel.

[0059] As an improvement, building modalities include building material properties, building location, and leakage levels of doors and windows.

[0060] As an improvement, the meteorological modality includes outdoor temperature and humidity, solar irradiance, and wind direction.

[0061] As an improvement, equipment modes include the operating power of equipment such as lighting equipment and large electrical appliances.

[0062] The above predictions take into account numerous factors, making the predictions more accurate.

[0063] As an improvement, a building cooling load prediction method based on a multimodal data model includes the following steps:

[0064] The first step is to collect data from each module through embedded external modules. The embedded external modules mainly include personnel modality acquisition module, building modality acquisition module, meteorological modality acquisition module, and equipment modality acquisition module. The specific data collected by each acquisition modality is shown below.

[0065] The personnel modality mainly collects image data. The image information, such as monitoring screens in different areas, is input into a pre-trained multimodal fusion recognition network to extract three modal features, including the number of people, the type of people's activities, and the distribution of people.

[0066] Building modality analysis involves acquiring image data, including building drawings and surveillance footage, and extracting necessary data such as building structure and interior / exterior wall area; numerical data includes building material properties, building location, and door / window leakage rate.

[0067] Meteorological modal data collection primarily consists of numerical data, including outdoor temperature and humidity, solar irradiance, and wind direction.

[0068] Equipment modality: Collected numerical data includes the operating power of lighting equipment, large electrical appliances, etc., and image data includes monitoring screens. Two modal features can be extracted from these: equipment start-up and shutdown status, and the number of equipment in operation.

[0069] The second step involves preprocessing each modality of data. For numerical data, the preprocessing method is the same as for traditional time series data: handling missing and outlier values. Missing values ​​are replaced by the average of the previous and next time series data. The lower quartile, median, and upper quartile are identified, and the upper and lower bounds are calculated using the formulas. Then, all data for that feature are iterated again; values ​​outside these bounds are considered outliers. If the current value is smaller than the lower bound, it is replaced with the lower bound; if the current value is greater than the upper bound, it is replaced with the upper bound. For image modality data preprocessing, image data is cropped to resize the image to match the model's input dimensions. Unnecessary background is removed, reducing interference and improving image processing speed.

[0070] The third step is to standardize the data features by using a standardization formula to narrow down the data range, ensuring that all data are between 0 and 1. Where x′ is the result after processing by the Min-Max normalization transformation function, x is the population sample data, and x max It is the maximum value in the sample data, while x min It is the minimum value in the sample data.

[0071] The fourth step involves feature extraction for each modality of data. In the image modality, different RNNs are used for different modules of image data according to their specific needs. For the personnel quantity modality, RGB modal features are extracted from surveillance footage. Each 15-minute surveillance video from each area is treated as a tensor, which is then divided into x blocks. The feature extraction layer extracts features related to various personnel in the surveillance images to obtain a feature tensor. The extracted feature codes are used as input to the convolutional layer for convolution operations, and finally, the feature tensor is output through a global averaging pooling layer. For the personnel activity type modality, in addition to collecting dynamic video and converting it to RGB data for storage in the image set, thermal imagery should also be collected, converted to RGB data, and stored in the image set. A convolutional neural network system is established, including five feature extraction layers corresponding to five typical levels of heat output. After global pooling of the five feature codes, feature tensors corresponding to the five typical heat outputs are output, ultimately achieving the goal of identifying personnel activity types. For architectural drawing images, a convolutional neural network system is also used for feature extraction, setting up feature extraction layers for exterior and interior walls, etc., and finally outputting feature tensors corresponding to interior and exterior walls, interior and exterior windows, etc. The data modes use extreme values ​​and average values ​​as feature extraction values. For the building mode, the physical properties of building materials, building location, and leakage of doors and windows can be directly used. For the meteorological mode, outdoor temperature and humidity, solar radiation, wind direction, and other data should be used as the final feature data because the air conditioning design should meet the cooling load requirements under the most extreme conditions. For the equipment mode, parameters such as operating power are used because the equipment parameters usually change linearly and have strong regularity when the equipment is adjusted. Therefore, the average value within each hour is usually used as the feature parameter when extracting features.

[0072] The fifth step involves fusing data from different modalities. A shared embedding layer is created as a common target for features from different modalities, enabling joint learning. Features from different modalities are mapped into the shared embedding layer and learn towards a virtual mean vector. Regardless of whether the input is text or image data, their feature representations will be optimized to approximate this virtual mean vector. Temporal correlation learning is performed using an LSTM module. The feature data from the previous step, which changed over time, is used to build a database based on hourly data, such as the number of people, outdoor temperature and humidity, and the number of machine workstations. Other parameters remain constant, creating a data matrix with the same time step. A training model is then built. First, the multi-feature information from different modalities in the feature extraction network is used as samples input into the LSTM. A forgetting gate determines which information should be forgotten or retained from the memory unit at each time step, using the following formula.

[0073] f t =δ(W f ·[h t-1 ,x t ]+b f )

[0074] Where δ is the logistic activation function, i.e., the sigmoid function, whose output range is (0,1). This function is used to compress the weighted sum to the (0,1) interval, thereby determining the output ratio of cell state information. t-1 ,x t ] indicates concatenating the hidden state from the previous time step with the current input; b f It is the bias vector of the forget gate, and its dimension is the same as f. t The dimensions are the same, used to adjust the activation value of the forget gate; f t It is a value between 0 and 1, which determines how much information from the cell's state at the previous moment has been "forgotten".

[0075] At the same time, the data is input into the input gate.

[0076] The input gate consists of two parts: a sigmoid layer that determines which values ​​will be updated, and a tanh layer that creates a new vector of candidate values, which will be added to the state. The formula for the input gate is as follows:

[0077] i t =δ(W i ·[h t-1 x t ]+b i )

[0078] C t =tanh(W c ·[h t-1 x t ]+b c )

[0079] wherein, i t It is the output of the input gate, C t It is the state of the candidate memory cell, W i W c and b i b c These are the relevant weights and biases, respectively. Finally, the output gate determines which part of the memory cell's state will be output to the hidden state, calculated using the following formula:

[0080] O t =δ(W0·[h t-1 x t ]+b O )

[0081] h t =O t ×tanh(C t )

[0082] Among them, O t δ is the output of the sigmoid function of the output gate, and its output range is (0, 1). This function is used to compress the weighted sum to the (0, 1) interval, thereby determining the output ratio of cell state information; W0 is the weight matrix of the output gate, and its shape depends on the hidden state h at the previous time step. t-1 The dimension and the current input x t The dimension; [h t-1 x t [] indicates that the previous state h is hidden. t-1 and the current input x t Concatenate the vectors to form a longer vector; b O It is the bias vector of the output gate, and its dimension is O. t The dimensions are the same, used to adjust the activation value of the output gate. Calculate the hidden state h at the current time step. t When, where tanh is the hyperbolic tangent function, its output range is (-1, 1); C t This represents the current state of the cell. (Through O) t and tanh(C t The product of ) allows the output gate to selectively output information from the cell state as the hidden state h at the current time step. t .

[0083] By using LSTM to predict the future trend of characteristic data within a four-month air-conditioning season, the mean squared error f between the predicted values ​​of the number of people, outdoor temperature and humidity, and number of machine workstations and the historical sample values ​​of the same parameters is calculated. t , Among them, y i1 Let y1 be the predicted value of the number of personnel, outdoor temperature and humidity, and number of machine workstations, and y2 be the historical true value of the number of personnel, outdoor temperature and humidity, and number of machine workstations. Multiple errors are summed, and through continuous iteration, the minimum error is found based on the negative gradient direction. The predicted data at this point is then used as the final feature data for each mode and input into the load calculation model. The model includes: using the formula Q = KF·(t wq -t n Calculate the load on the building envelope, where Q is the hourly cooling load of the building envelope, F is the area of ​​the building envelope, K is the heat transfer coefficient of the building envelope, and t wq The temperature for calculating the hourly cooling load of the building envelope is t. n Indoor calculated temperature, using the formula CL sb =C CLsb C sb Q sb Calculate the equipment load, where CL sb The hourly cooling load C of the equipment CLsbThe equipment cooling load factor, C, is obtained according to the specifications. sb Q is the equipment correction factor. sb This refers to the heat dissipation of the equipment. The formula Q is used. r =n·Φ·Q rt Calculate personnel load, where Q r The cooling load caused by heat dissipation from the human body, Φ is the clustering coefficient, taken as 0.93, Q rt Let n be the heat dissipation from the human body and n be the number of people in the air-conditioned room. The loads calculated using the above formulas are summed to obtain the final predicted cooling load. This application also discloses a multimodal cooling load prediction and energy-saving control system.

[0084] It comprises three modules: a multimodal data acquisition module for collecting historical multimodal data; a cooling load prediction module for acquiring multimodal data for the prediction period, extracting data features, and establishing a prediction model; and an energy-saving control module for intelligent energy-saving control based on the predicted cooling load value for that period, using a genetic algorithm to deduce the optimal operating parameters of the refrigeration equipment to adapt to the current cooling load and avoid unnecessary energy waste.

[0085] In the energy-saving control module, a genetic algorithm is established for parameter optimization. First, the required adjustment parameters are identified, focusing on regulating the most energy-consuming components of the entire air conditioning system. These parameters should include chilled water pump flow rate, and preferably further include chiller outlet water temperature, cooling water pump flow rate, chilled water pump flow rate, and fan air volume. Second, the range of values ​​for the adjustment parameters and the fitness evaluation function, i.e., the energy consumption calculation formula, are determined. Then, random values ​​are generated within the parameter range, encoded, and used to form individuals, i.e., chromosomes. N chromosomes are randomly generated, N = 2, 3, 4..., forming population 1. The energy consumption of each chromosome in population 1 is then calculated using a fitness evaluation function. The top 10% of chromosomes in population 1 with the lowest energy consumption (i.e., the highest fitness evaluation value) are retained, and the remaining chromosomes undergo crossover and mutation operations to obtain a new population 2. It is then determined whether population 2 satisfies the maximum number of iterations or whether the optimal individuals of several adjacent populations show no significant change. If so, the optimal solution is output. If not, population 2 is used to replace population 1, and iterative operations are performed to finally obtain the optimal combination parameters that minimize overall energy consumption. The air conditioning components are then adjusted in a timely manner to achieve energy saving.

[0086] The refrigeration system includes a cooling water circulation system and a chilled water circulation system, which together form a water circulation system. The water circulation system is equipped with two screw chiller units, one in operation and one on standby, along with two chilled water pumps, two cooling water pumps, and two cooling towers. Chilled water is cooled by cooling water in the chiller units and then enters a manifold, where it exchanges heat with the air in the combined air conditioning units. After being heated, it is pressurized by the chilled water pumps and returned to the chiller units, completing the chilled water circulation. Cooling water, after acquiring heat from the chilled water in the chiller units, enters the cooling towers, is cooled, and then pressurized by the cooling water pumps before returning to the chiller units, completing the cooling water circulation. The air circulation system uses an all-air system. Chilled water is transported to the corresponding floors via the water system and then enters the air conditioning units on those floors for heat exchange. It collects 70% return air and 30% fresh air, mixes them, and then exchanges heat with the chilled water to cool them down. A single-pass return air dew point supply method is used to determine the air supply state point. Once the air supply state is reached, the air is distributed to the offices through air ducts and diffusers.

[0087] This refrigeration system regulates cooling capacity by adjusting the chilled water outlet temperature and the air supply flow rate. The air supply flow rate of each diffuser in the office building is controlled by at least one valve, which is controlled for continuous time intervals of equal length. Based on the predicted load, the percentage of the predicted output load relative to the total load of the air conditioning system is calculated, and the valve opening is adjusted accordingly. The valve opening percentage is determined over several time intervals, which should be the time intervals for the predicted output load. The adjustment threshold is between 60% and 100%, i.e., the fan frequency is between the minimum opening of 60% and the maximum opening of 100%. For example, the preset first range can be 60-80%, and the preset first adjustment threshold is 80% of the maximum fan frequency opening. If the predicted load rate is lower than 60%, the adjustment is made within the preset first range; when the predicted load rate is greater than 60%, the fan frequency should be adjusted within a second preset threshold, which is 80-100%.

[0088] While the present invention has been disclosed above with reference to preferred embodiments, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. An intelligent refrigeration system for predicting and adjusting fresh air volume based on historical cooling load data, the system comprising an evaporator, a heat exchanger, and building units. Water enters the evaporator, exchanges heat with the refrigerant in the evaporator to become chilled water, then enters the heat exchanger again to exchange heat with air entering the heat exchanger, and then returns to the evaporator, forming a heat exchange cycle. Air enters the heat exchanger and exchanges heat with the chilled water in the heat exchanger to form chilled air, which then enters the building units for cooling. The air entering the heat exchanger includes a mixture of fresh air and return air. The return air is the indoor air exhausted after cooling by the building units. The fresh air includes a fresh air duct and a fan installed on the fresh air duct. The fan is data-connected to a control system for controlling the fan frequency, thereby controlling the fresh air volume entering the heat exchanger. The system includes return air... The return air duct and the fresh air duct are mixed before entering the heat exchanger. The return air duct is perpendicular to the fresh air duct. A homogenizing component is installed in the return air duct, located upstream of the connection point between the fresh air duct and the return air duct. The homogenizing component has a blade structure, consisting of multiple blades, with the extension direction of the gap between adjacent blades parallel to the central axis of the fresh air duct. A fresh air mixing device is installed in the fresh air duct, also with a blade structure, located upstream of the connection point. This device has multiple blades, with the extension direction of the flow channel between adjacent blades parallel to the central axis of the return air duct. The spacing between the return air flow channels formed by adjacent blades of the homogenizing component varies, decreasing with increasing distance from the fresh air duct.

2. The refrigeration system as described in claim 1, characterized in that, The control system automatically controls the frequency of the fans according to the cooling capacity required by the building unit, thereby controlling the amount of fresh air entering the heat exchanger.

3. The refrigeration system as described in claim 2, characterized in that, When the cooling capacity required by a building unit increases, the frequency of the automatically controlled fan increases; when the cooling capacity required by a building unit decreases, the frequency of the automatically controlled fan decreases.

4. The refrigeration system as described in claim 2, characterized in that, The cooling capacity required by the building unit is predicted based on the human modality acquisition module, building modality, meteorological modality, and equipment modality.

5. The refrigeration system as described in claim 4, characterized in that, Personnel modalities include the number of personnel, the types of personnel activities, and the distribution of personnel.

6. The refrigeration system as described in claim 4, characterized in that, Building modalities include building material properties, building location, and leakage levels of doors and windows.

7. The refrigeration system as described in claim 4, characterized in that, Meteorological modes include outdoor temperature and humidity, solar irradiance, and wind direction.

8. The refrigeration system as described in claim 4, characterized in that, Equipment modes include the operating power of equipment such as lighting equipment and large electrical appliances.

9. The refrigeration system as described in claim 1, characterized in that, The air entering the heat exchanger consists of 50-80% return air and 20-50% fresh air. The return air and fresh air are mixed before entering the heat exchanger.

10. A control method for a refrigeration system as described in any one of claims 1-9, comprising the following steps: The first step is to collect data from each module through embedded external modules, which mainly include personnel modal acquisition modules, building modal acquisition modules, meteorological modal acquisition modules, and equipment modal acquisition modules; The second step is to preprocess the data for each modality separately; The third step is to standardize the data features; The fourth step is to extract features from the data of each modality. The image data of different modules in the image modality are set up with different RNNs according to their respective needs. The fifth step is to fuse the data from each modality.

Images