High-efficiency control method and system for central air-conditioning refrigeration machine room

Abstract

The invention provides a high-efficiency control method and system for a central air conditioner refrigeration machine room, and relates to the technical field of automatic control, and the method comprises the steps: 1, obtaining chilled water supply and return water temperature difference and flow data, and calculating the dynamic cold load demand of a building in combination with the personnel distribution thermal imaging data of each floor; and 2, on the basis of the dynamic cooling load requirement, a dynamic adjustment model for the number of running water chilling units is built, matching degree calculation is conducted on the actual cooling load value and the single refrigerating capacity of the water chilling units through a fuzzy PID control algorithm, and a final unit running combination scheme is generated. By calculating the cooling load requirement, dynamically adjusting the operation parameters of the water chilling unit, the chilled water pump and the cooling tower and balancing the equipment operation time, the system energy efficiency is effectively improved, the energy consumption and the operation and maintenance cost are reduced, and stable operation of the refrigerating system is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of automatic control technology, and in particular to a high-efficiency control method and system for a central air-conditioning refrigeration room. Background Art

[0002] During the operation of a central air conditioning refrigeration room, the coordinated control of the chiller, chilled water pump, and cooling water system directly impacts energy efficiency and equipment life. Traditional control methods, which start and stop units or regulate pump speeds based on fixed thresholds or single parameters (such as the supply and return water temperature difference), are unable to adapt to the dynamic changes in the building's cooling load.

[0003] In the existing technology, the calculation of cooling load demand mostly relies on static models or historical data averages, without real-time integration of personnel distribution thermal imaging data and ambient temperature changes, resulting in large deviations in cooling load prediction.

[0004] For example, when the weather suddenly changes and the ambient temperature rises sharply, the traditional method does not dynamically correct the cooling load forecast value and still allocates the number of operating units according to the historical average, resulting in redundant cooling capacity or temperature out of control in local areas.

[0005] In addition, there is a lack of dynamic linkage between the speed regulation of the chilled water pump and the operating status of the unit, and the pressure difference in the pipeline network fluctuates greatly, further exacerbating energy waste. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a high-efficiency control method and system for a central air-conditioning refrigeration room, which realizes the coordinated optimization of the chilled water pump speed regulation and the unit status.

[0007] In order to solve the above technical problems, the technical solutions of the present invention are as follows: In a first aspect, a high-efficiency control method for a central air-conditioning refrigeration room is provided, the method comprising: Step 1: Obtain the chilled water supply and return temperature difference and flow data, and combine it with the thermal imaging data of the occupant distribution on each floor to calculate the dynamic cooling load demand of the building; Step 2: Based on the dynamic cooling load demand, a dynamic adjustment model for the number of chillers in operation is constructed. The matching degree between the actual cooling load value and the cooling capacity of a single chiller is calculated using a fuzzy PID control algorithm to generate the final unit operation combination plan. Step 3: Based on the final unit operation number combination plan, the real-time value of the chilled water supply and return water pressure difference is used as feedback, combined with the flow demand forecast value, the chilled water pump frequency conversion parameters are optimized through genetic algorithm, and the dynamic control instructions of the chilled water cycle are output; Step 4: Based on the dynamic control instructions for the chilled water cycle, the temperature and flow data of the condenser outlet and the cooling tower inlet are collected in real time. The feedback correction coefficient is extracted by fitting the dynamic temperature-flow relationship curve of the cooling water. The cooling water demand temperature is predicted by combining the neural network algorithm. The cooling tower fan speed and bypass valve opening are dynamically adjusted to generate the cooling water cycle control results. Step 5: Based on the equipment operating parameter optimization results, a model for balancing the equipment operating time is established. Based on the historical operating time data of each chiller, a rotation priority algorithm is used to dynamically allocate unit operating tasks to achieve balanced operating time within the equipment group.

[0008] Furthermore, the temperature difference and flow rate of chilled water supply and return water are obtained, combined with thermal imaging data of occupant distribution on each floor, to calculate the dynamic cooling load demand of the building, including: Based on the infrared radiation intensity distribution of each area in the thermal imaging data, the boundary range of the personnel gathering area is identified, and the dynamic heat source weight coefficient of each floor is calculated based on the heat source density per unit area; The dynamic heat source weight coefficient is integrated with the chilled water supply and return water temperature difference and flow data to calculate the instantaneous cooling load value of each zone layer by layer; Dynamically smooth the instantaneous cooling load value, set a fixed time window length, extract the cooling load value sequence within the window, and generate a smoothed cooling load benchmark value; Based on the smoothed cooling load reference value and the current temperature data collected in real time by the ambient temperature sensor, the historical average temperature of the same period in the past preset time is obtained, the deviation between the current temperature and the historical average temperature is calculated, and the temperature compensation factor is generated proportionally according to the size of the deviation; The temperature compensation factor is superimposed and corrected with the smoothed cooling load benchmark value to output a dynamic cooling load demand that reflects the spatial thermal distribution characteristics of the building and the changing trend in the time dimension.

[0009] Furthermore, based on the dynamic cooling load demand, a dynamic adjustment model for the number of chillers in operation is constructed. The fuzzy PID control algorithm is used to calculate the matching degree between the actual cooling load value and the cooling capacity of a single chiller, and the final unit operation combination plan is generated, including: The dynamic cooling load demand is divided into a cooling load fluctuation sequence according to the time dimension. Based on the distribution characteristics of the cooling load peak and valley values, the cooling load fluctuation level is divided and the cooling load deviation threshold is defined. Based on the cooling load fluctuation sequence, a dynamic adjustment model for the number of operating chillers is constructed. The real-time cooling load value is dynamically matched with the cooling capacity of each chiller unit through a fuzzy PID control algorithm. The cooling load fluctuation range is quantified based on the membership function, and a cooling load deviation index is generated. When the cooling load deviation index exceeds the preset threshold, the total cooling capacity redundancy or shortfall of the currently operating units is calculated based on the time distribution characteristics of the cooling load demand. Combined with the historical start-up and shutdown energy consumption data of the chillers, the energy consumption evaluation index of the candidate unit combination is constructed. Based on the energy consumption evaluation index, the initial chiller unit operation number combination scheme is determined from the candidate unit combination, and the initial chiller unit operation number combination scheme is coupled with the chilled water flow demand forecast value for verification, and the final unit operation number combination scheme that matches the chilled water pump energy consumption constraint is output.

[0010] Furthermore, based on the final combination plan for the number of operating units, the real-time value of the chilled water supply and return pressure difference is used as feedback, combined with the flow demand forecast value, and the chilled water pump variable frequency parameters are optimized through genetic algorithms to output dynamic control instructions for the chilled water cycle, including: Based on the final combination plan of the number of operating units, a chilled water flow demand benchmark curve is mapped and generated, in which the flow demand benchmark value is associated with the number of operating units and the rated flow of each unit; The core parameter is the deviation between the real-time value of the chilled water supply and return pressure difference and the benchmark curve. Combined with the predicted value of chilled water flow demand, a fitness function is constructed, and the pressure difference stability threshold range and the water pump energy consumption weight ratio are set. The pump speed parameter is encoded into a binary chromosome population, the population size is initialized, and the dynamic crossover probability and mutation probability are set. The crossover probability and mutation probability are dynamically adjusted according to the pressure difference deviation value. In each round of iteration, the individuals in the population are evaluated based on the fitness function, and the roulette wheel selection strategy is used to screen the parent individuals to generate the offspring population. The offspring are subjected to crossover and mutation operations until the maximum number of iterations is reached, and the optimized speed parameters are obtained. According to the optimized speed parameters, the pressure difference fluctuation range and the water pump energy consumption reduction rate are calculated, the final speed parameter combination is determined, and the dynamic control instructions of the chilled water cycle including the target speed value and the adjustment time window are generated.

[0011] Furthermore, in each iteration, the individuals in the population are evaluated based on the fitness function, and the roulette wheel selection strategy is used to screen the parent individuals to generate the offspring population. The offspring are subjected to crossover and mutation operations until the maximum number of iterations is reached, and the optimized speed parameters are obtained, including: Based on the fitness function, the fitness of the binary-coded pump speed parameter population is evaluated, and the absolute value of the pressure difference deviation and the weighted sum of energy consumption of each individual are calculated to generate a fitness value sequence. According to the fitness value sequence, the roulette wheel selection strategy is used to screen the parent individuals and generate the offspring population, where the crossover probability and mutation probability are dynamically adjusted according to the current pressure difference deviation value; Perform crossover and mutation operations on the offspring population, determine the individuals that enter the next generation population according to the fitness value sequence, and iterate the optimization until the maximum number of iterations is reached to obtain the optimized speed parameters.

[0012] Furthermore, based on the dynamic control instructions of the chilled water cycle, the temperature and flow data of the condenser outlet and the cooling tower inlet are collected in real time. The feedback correction coefficient is extracted by fitting the dynamic temperature-flow relationship curve of the cooling water. The cooling water demand temperature is predicted by combining the neural network algorithm, and the cooling tower fan speed and bypass valve opening are dynamically adjusted to generate the cooling water cycle control results, including: Based on the dynamic control instructions of the chilled water cycle, the temperature and flow data of the condenser outlet and the cooling tower inlet are collected in real time. The dynamic temperature-flow relationship curve of the cooling water is fitted according to the time series, and the slope of the curve is extracted as the feedback correction coefficient; Based on the feedback correction coefficient, a neural network prediction model is constructed, and the feedback correction coefficient is used as the input layer. Combined with the real-time data collected by the ambient temperature and humidity sensor, a fully connected neural network consisting of an input layer, a hidden layer, and an output layer is constructed. The weight parameters of the hidden layer are trained using historical cooling water operation data, and the output layer predicts the cooling water demand temperature; The difference between the cooling water demand temperature predicted by the neural network and the cooling tower inlet temperature collected in real time by the temperature sensor is calculated to generate a temperature deviation value; Using the proportional-integral control algorithm, the temperature deviation value is linearly superimposed with the accumulated deviation value within the preset integral time window to generate the cooling tower fan speed adjustment value. The bypass valve opening adjustment value is dynamically calculated based on the slope change direction and amplitude of the feedback correction coefficient. The cooling tower fan speed adjustment and bypass valve opening adjustment are integrated into the cooling water circulation control result.

[0013] Furthermore, based on the results of equipment operating parameter optimization, a model for balancing equipment operating time was established. Based on the historical operating time data of each chiller, a rotation priority algorithm was used to dynamically allocate unit operating tasks to achieve balanced operating time within the equipment group, including: Based on the cooling water circulation control results, establish the equipment operation time balance distribution model; Input the historical operating time data of each chiller into the equipment operating time balance distribution model, calculate the operating time standard deviation of each unit, and generate the priority weight according to the size of the standard deviation; When the cooling load demand triggers the start and stop of the unit, the candidate units are sorted according to the priority weight, and the target units are selected for start-up in descending order of weight value, while the redundant units are shut down in descending order of weight value; Through the sliding time window algorithm, the latest running time data is intercepted with the preset time window length, and the standard deviation and priority weight of the group running time are recalculated; If the recalculated running time standard deviation exceeds the preset threshold, the equipment running time balance allocation model is triggered to recalculate the weight and adjust the group task allocation until the standard deviation is less than the preset threshold, thus achieving dynamic balance of running time within the equipment group.

[0014] In a second aspect, a high-efficiency control system for a central air-conditioning refrigeration room includes: Dynamic cooling load module, used to obtain chilled water supply and return temperature difference and flow data, and combined with thermal imaging data of occupant distribution on each floor, calculate the dynamic cooling load demand of the building; The operation adjustment module is used to build a dynamic adjustment model for the number of chillers in operation based on the dynamic cooling load demand, and calculate the matching degree between the actual cooling load value and the cooling capacity of a single chiller through the fuzzy PID control algorithm to generate the final unit operation combination plan; The frequency conversion optimization module is used to optimize the frequency conversion parameters of the chilled water pumps using a genetic algorithm based on the final unit operation number combination plan generated by the chiller operation adjustment module, using the real-time value of the chilled water supply and return water pressure difference as feedback and combining it with the flow demand forecast value, and output the dynamic control instructions for the chilled water cycle; The circulation control module is used to dynamically control the chilled water circulation based on the output of the chilled water pump frequency conversion optimization module. It collects the temperature and flow data of the condenser outlet and the cooling tower inlet in real time, extracts the feedback correction coefficient by fitting the dynamic temperature-flow relationship curve of the cooling water, and uses the neural network algorithm to predict the cooling water demand temperature. It dynamically adjusts the cooling tower fan speed and bypass valve opening to generate the cooling water circulation control results. The balanced allocation module is used to establish a balanced allocation model for equipment operating time through the equipment operating parameter optimization results generated by each module. Based on the historical operating time data of each chiller, it adopts a rotation priority algorithm to dynamically allocate unit operating tasks to achieve balanced operating time within the equipment group.

[0015] According to a third aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0016] In a fourth aspect, a computer-readable storage medium stores a program, which implements the method when executed by a processor.

[0017] The above solution of the present invention includes at least the following beneficial effects: By integrating chilled water system data with thermal imaging data on occupant distribution, the system accurately captures dynamic heat load changes within the building, avoiding the lag inherent in traditional fixed-parameter calculations. This allows cooling load calculations to better align with actual needs, reducing energy waste or supply shortages caused by load misjudgment. A regulation model based on dynamic cooling loads is constructed, utilizing a fuzzy PID algorithm to dynamically match cooling load and cooling capacity. This allows for flexible adjustment of the number of operating chillers based on real-time load, preventing inefficient operation or frequent starts and stops. This improves system response speed while reducing overall energy consumption and ensuring stable cooling efficiency.

[0018] Combining the unit's operating plan with the chilled water system's parameters, a genetic algorithm is used to optimize the chilled water pump's variable frequency parameters, precisely matching the pump's speed to actual flow requirements. This reduces energy consumption from idling or overload while ensuring efficient chilled water delivery. Pressure differential stabilization control improves system operational stability and reduces equipment wear. Cooling water data is dynamically collected based on chilled water circulation instructions, and cooling tower operating parameters are adjusted using curve fitting and a neural network algorithm. This allows the cooling water system to quickly adapt to load changes and environmental fluctuations. While ensuring efficient condenser heat exchange, over-adjustment of the fan and bypass valve is avoided, achieving energy-saving and stable operation of the cooling water circulation system. By establishing an equipment operating time balancing model and a rotation algorithm, unit operating tasks are dynamically allocated to avoid uneven wear caused by excessive operation of some units. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The present invention provides a flow chart of a high-efficiency control method for a central air-conditioning refrigeration room.

[0020] Figure 2 The present invention provides a schematic diagram of a high-efficiency control system for a central air-conditioning refrigeration room. DETAILED DESCRIPTION

[0021] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0022] like Figure 1 As shown, an embodiment of the present invention provides a high-efficiency control method for a central air-conditioning refrigeration room, the method comprising the following steps: Step 1: Obtain the chilled water supply and return temperature difference and flow data, and combine it with the thermal imaging data of the occupant distribution on each floor to calculate the dynamic cooling load demand of the building; Step 2: Based on the dynamic cooling load demand, a dynamic adjustment model for the number of chillers in operation is constructed. The matching degree between the actual cooling load value and the cooling capacity of a single chiller is calculated using a fuzzy PID control algorithm to generate the final unit operation combination plan. Step 3: Based on the final unit operation number combination plan, the real-time value of the chilled water supply and return water pressure difference is used as feedback, combined with the flow demand forecast value, the chilled water pump frequency conversion parameters are optimized through genetic algorithm, and the dynamic control instructions of the chilled water cycle are output; Step 4: Based on the dynamic control instructions for the chilled water cycle, the temperature and flow data of the condenser outlet and the cooling tower inlet are collected in real time. The feedback correction coefficient is extracted by fitting the dynamic temperature-flow relationship curve of the cooling water. The cooling water demand temperature is predicted by combining the neural network algorithm. The cooling tower fan speed and bypass valve opening are dynamically adjusted to generate the cooling water cycle control results. Step 5: Based on the equipment operating parameter optimization results, a model for balancing the equipment operating time is established. Based on the historical operating time data of each chiller, a rotation priority algorithm is used to dynamically allocate unit operating tasks to achieve balanced operating time within the equipment group.

[0023] In an embodiment of the present invention, the dynamic cooling load demand is calculated by integrating the chilled water supply and return water temperature difference, flow data, and personnel distribution thermal imaging data. This can capture the heat changes generated by personnel activities in the building while taking into account the actual operating status of the chilled water system. Compared with the traditional fixed load calculation method, it can more realistically reflect the thermal distribution characteristics of the building space and the trend of changes in the time dimension. The number of operating chillers is flexibly adjusted according to the real-time cooling load to avoid frequent start-up and shutdown or inefficient operation of the units. By quantifying the cooling load fluctuation range and generating a deviation index, it is possible to respond to load changes in a timely manner. Combined with the energy consumption evaluation index, the final unit combination is determined to reduce the overall energy consumption of the chiller and improve energy utilization efficiency.

[0024] Precisely matching the chilled water pump speed to actual demand effectively reduces pump energy consumption while ensuring efficient chilled water delivery. By setting a fitness function and dynamically adjusting parameters, it is possible to balance pressure differential stability and energy consumption, avoiding energy consumption and equipment wear caused by excessive pump operation, and achieving energy-efficient operation of the chilled water circulation system. Dynamically adjusting the cooling tower fan speed and bypass valve opening allows the cooling water system to quickly adapt to changes in the chilled water system and fluctuating environmental conditions. Compared to traditional fixed adjustment methods, this dynamic regulation can effectively reduce cooling tower fan energy consumption and ensure efficient condenser operation. It prevents some chillers from over-operating while others are under-utilized, effectively balancing the operating time of each unit within the equipment group.

[0025] In a preferred embodiment of the present invention, step 1, obtaining chilled water supply and return water temperature difference and flow data, and calculating the building's dynamic cooling load demand in combination with thermal imaging data of occupant distribution on each floor, may include: Step 100: Identify the boundary of the gathering area based on the infrared radiation intensity distribution of each area in the thermal imaging data, and calculate the dynamic heat source weight coefficient of each floor based on the heat source density per unit area. Step 101: The dynamic heat source weight coefficient is integrated with the chilled water supply and return water temperature difference and flow data to calculate the instantaneous cooling load value of each zone layer by layer; Step 102: Dynamically smooth the instantaneous cooling load value, set a fixed time window length, extract the cooling load value sequence within the window, and generate a smoothed cooling load reference value; Step 103: Based on the smoothed cooling load reference value and the current temperature data collected in real time by the ambient temperature sensor, the historical average temperature of the same period in the past preset time is obtained, the deviation between the current temperature and the historical average temperature is calculated, and a temperature compensation factor is generated proportionally according to the deviation. Step 104 : The temperature compensation factor is superimposed and corrected with the smoothed cooling load reference value to output a dynamic cooling load demand that reflects the spatial heat distribution characteristics of the building and the changing trend in the time dimension.

[0026] In this embodiment of the present invention, after obtaining thermal imaging data of each floor, it is necessary to set the infrared radiation intensity threshold based on the statistical characteristics of the data. The specific operation is as follows: Calculate the mean (reflecting the overall radiation level) and standard deviation (reflecting the degree of data dispersion) of the thermal imaging data for each floor of the entire building. For example, if the mean is 50 units and the standard deviation is 30 units, then divide the intervals based on the mean and the standard deviation: Low radiation zone: intensity ≤ mean - standard deviation (e.g., 20 units), corresponding to areas with few or no people; High radiation area: intensity ≥ mean + standard deviation (e.g. 80 units), corresponding to densely populated areas; Middle radiation zone: between the two, corresponding to the personnel flow or semi-dense area.

[0027] Using the thresholds set above, the thermal imaging data is segmented and judged point by point. For each data point, if its infrared radiation intensity falls within the high-radiation zone, it is marked as a potential point in a gathering area; if it falls within the low-radiation zone, it is marked as a point in a non-gathering area; points in the medium-radiation zone are temporarily retained. Next, using the connected domain analysis algorithm in image recognition, adjacent potential points in the high-radiation zone and connected retained points in the medium-radiation zone are connected and merged, eliminating isolated points and small areas of noise. This allows the boundaries of gathering areas to be identified and accurately determined.

[0028] Next, determine the heat source density per unit area, which is determined based on factors such as the average heat dissipation of personnel and the heat dissipation of equipment. For example, in an ordinary office scenario, the heat source density per unit area can be set to 50-80 watts per square meter. Combined with the identified area of the gathering area, calculate the dynamic heat source weight coefficient of each floor. In the specific calculation, taking a certain floor as an example, if the area of the gathering area is S square meters, the heat source density per unit area is D watts / square meter, and the total area of the floor is T square meters, then the dynamic heat source weight coefficient of the floor is W= The baseline density value is the reference value of the average heat source density per unit area of the entire building, and its value range is [0.1, 1]. The more people and activities in an area, the larger the corresponding weight coefficient, and the weight coefficient range is [0, 1], where 0 indicates no heat source impact in the area, and 1 indicates the area has reached the maximum heat source impact.

[0029] Step 101 combines the dynamic heat source weight coefficient calculated in step 100 with the chilled water supply and return temperature difference and flow rate data. The building is first divided into zones based on floor layout and functional areas, such as office areas, conference rooms, and rest areas. For each zone, the chilled water supply and return temperature difference reflects the amount of heat transferred within that zone; a larger temperature difference indicates greater heat transfer. The flow rate represents the chilled water delivery volume; a larger flow rate indicates greater heat transport. Combined with the zone's dynamic heat source weight coefficient, the instantaneous cooling load is calculated using specific calculation logic. For example, if the supply and return temperature difference of a zone's chilled water is ΔT (unit: °C), the flow rate is q (unit: cubic meters / hour), and the dynamic heat source weight coefficient for that zone is W, then according to the heat calculation formula q = cmΔT (c is the specific heat capacity of water, and m is the mass of water, which can be calculated by converting the flow rate and the water density), and considering the impact of the weight coefficient on heat transfer and delivery, the instantaneous cooling load value for that zone is L = W × k × cmΔT, where k is a correction factor used to adjust the calculated result to actual demand and is in the range of [0.8, 1.2]. This layer-by-layer calculation yields the cooling load demand for each zone at the current moment.

[0030] Step 102: Dynamically smooth the instantaneous cooling load value to eliminate fluctuations caused by accidental factors. A fixed time window length, such as 5 minutes, is set. From the collected cooling load data sequence, the cooling load value sequence within the 5-minute window is extracted, taking the current time as the benchmark. Smoothing is performed using the moving average method, which adds the cooling load values within the window and divides them by the number of data points within the window (e.g., if data is collected once per minute within 5 minutes, the number of data points is 5) to obtain an average value. This average value is used as the cooling load value at the corresponding moment after smoothing. The data at each moment in the window is processed sequentially, abnormal fluctuations are removed, and a smoothed cooling load baseline value is calculated. In this way, the cooling load data can better reflect the actual stable demand trend and reduce data fluctuations caused by accidental factors such as the temporary entry and exit of personnel and the instantaneous start and stop of equipment.

[0031] In step 103, based on the smoothed cooling load baseline value obtained in step 102 and the current temperature data collected in real time by the ambient temperature sensor, the historical average temperature for the same period within a preset period of time (e.g., the past week) is obtained. The deviation between the current temperature and the historical average temperature is calculated: Δt = current temperature - historical average temperature. This determines whether the current temperature is above or below the historical average. Based on pre-set rules, a temperature compensation factor is generated proportionally to the deviation. For example, when |Δt| ≤ 1°C, the temperature compensation factor α is set to 0; when 1°C < |Δt| ≤ 3°C, α is set to 0.1 × sign(Δt) (1 when Δt is greater than 0, -1 when Δt is less than 0); when 3°C < |Δt| ≤ 5°C, α is set to 0.2 × sign(Δt), and so on. If the current temperature is higher than the historical average temperature, the compensation factor is positive and is used to increase the cooling load demand; if the current temperature is lower than the historical average temperature, the compensation factor is negative and is used to reduce the cooling load demand. The temperature compensation factor range is between [-0.5, 0.5].

[0032] In step 104, the temperature compensation factor generated in step 103 is superimposed and corrected with the smoothed cooling load baseline value. Specifically, the calculation is: final cooling load demand value L' = smoothed cooling load baseline value + temperature compensation factor × correction coefficient β, where β is a coefficient used to adjust the degree of temperature compensation's impact on the cooling load, and its value range is between [0.8, 1.2]. In this way, the dynamic heat source generated by human activities, the operation of the chilled water system, and the impact of ambient temperature changes are comprehensively considered. The final output is a dynamic cooling load demand that fully reflects the spatial thermal distribution characteristics of the building and its temporal trend. This provides accurate basic data for subsequent central air conditioning system control, enabling the system to more accurately meet the building's actual cooling load demand and achieve energy-saving and efficient operation.

[0033] By calculating the building's dynamic cooling load demand, the accuracy of cooling load calculations has been improved. Firstly, by combining thermal imaging data of occupant distribution, this method fully accounts for human activity, a significant dynamic heat source. Compared to traditional calculations that rely solely on fixed parameters, this method more accurately reflects the actual heat load within the building. Secondly, through dynamic smoothing and temperature compensation, it effectively eliminates interference caused by data fluctuations and ambient temperature changes. This precise cooling load calculation helps achieve energy-efficient system operation, avoiding energy waste and irrational equipment operation.

[0034] In a preferred embodiment of the present invention, the above step 2, based on the dynamic cooling load demand, constructs a dynamic adjustment model for the number of chillers in operation, calculates the matching degree between the actual cooling load value and the cooling capacity of a single chiller unit through a fuzzy PID control algorithm, and generates a final unit operation combination plan, which may include: Step 200: Divide the dynamic cooling load demand into a cooling load fluctuation sequence according to the time dimension, classify the cooling load fluctuation level based on the distribution characteristics of the cooling load peak and valley values, and define the cooling load deviation threshold; Step 201: Based on the cooling load fluctuation sequence, a dynamic adjustment model for the number of operating chillers is constructed. The real-time cooling load value is dynamically matched with the cooling capacity of each chiller unit through a fuzzy PID control algorithm. The cooling load fluctuation range is quantified based on a membership function to generate a cooling load deviation index. Step 202: When the cooling load deviation index exceeds a preset threshold, the total cooling capacity redundancy or shortfall of the currently operating units is calculated based on the time distribution characteristics of the cooling load demand, and the energy consumption evaluation index of the candidate unit combination is constructed in combination with the historical start-up and shutdown energy consumption data of the chillers. Step 203: Based on the energy consumption evaluation index, determine the initial chiller unit number combination plan from the candidate unit combination, couple and verify the initial chiller unit number combination plan with the chilled water flow demand forecast value, and output the final unit number combination plan that matches the chilled water pump energy consumption constraint.

[0035] In an embodiment of the present invention, the dynamic cooling load demand is divided into a continuous sequence of cooling load values at fixed time intervals (such as 10 minutes), forming a fluctuation curve with the time point as the horizontal axis and the cooling load value as the vertical axis. For example, if the current time is 14:00, the cooling load value every 10 minutes from 13:00 to 14:00 can be extracted to obtain a fluctuation sequence consisting of 12 data points. The peak (maximum value) and valley (minimum value) in the cooling load fluctuation are identified by a sliding window algorithm (such as taking the data of the previous hour). For example, if the maximum cooling load in the window is 800kW and the minimum is 300kW, then this period is defined as a "high fluctuation period". The levels are divided according to the difference between the peak and valley values (fluctuation amplitude): Low fluctuation: amplitude ≤ 20% of average cooling load (e.g. average load 500kW, amplitude ≤ 100kW); Medium fluctuation: 20%<amplitude≤50% average cooling load; High fluctuation: amplitude >50% average cooling load.

[0036] Cooling load deviation thresholds are set for different levels. For example, at low fluctuation levels, the cooling load is allowed to deviate from the rated cooling capacity of a single unit by ±10% (e.g., for a single unit with a cooling capacity of 200kW, the threshold is ±20kW). At high fluctuation levels, the threshold is tightened to ±5% to quickly respond to load changes.

[0037] Step 201: Taking the cooling load fluctuation sequence as input, the model reads the current cooling load value (e.g., the current value is 750kW) and the cooling capacity of a single chiller (e.g., a single chiller is 200kW) in real time, and calculates the theoretical number of chillers currently required ( ≈3.75 units). The deviation between the actual cooling load and the theoretical number of units (e.g. 3.75 The current number of operating units (3 = +0.75 units) and the rate of change of the deviation (e.g., an increase of 0.1 units per minute) are converted into fuzzy linguistic variables, such as "small deviation," "medium deviation," and "large deviation." A control signal is generated based on a preset rule (e.g., "If the deviation is large and the rate of change is positive, add one unit"), converting the fuzzy signal into a specific unit adjustment (e.g., output +1 unit).

[0038] The triangular membership function is used to divide the cooling load fluctuation range into three parts: "below demand", "close to demand" and "above demand". The specific process is as follows: First, the key nodes of the triangular membership function are determined based on the cooling capacity of a single chiller. , defining the boundaries of three intervals: "Below demand" range: When the cooling load Lower than When , it completely belongs to this interval, and the membership degree is When the cooling load reaches When , the membership degree decreases linearly to That is, when the cooling load changes from Increase to In the process of Gradually decrease to For example, if the cooling capacity of a single unit is , when the cooling load is When the membership degree is When the cooling load is When the membership degree is ,exist The membership degree is based on the linear equation (in is the cooling load value, is the membership degree) calculation.

[0039] “Near demand” range: cooling load is arrive When the cooling load is When the cooling load is When , the membership reaches the maximum value of 1. In this interval, the membership increases or decreases from the two ends to the middle in a straight line relationship. For example, for , cooling load is When the membership is 0, the cooling load is When the membership is 1, The membership between the two groups is based on the linear equation calculate, The membership between the two groups is based on the linear equation calculate.

[0040] "Above demand" range: When the cooling load exceeds When the cooling load drops to When , the membership degree decreases linearly to 0. That is, the cooling load changes from Reduce to In the process, the membership degree gradually changes from 1 to 0 according to the linear relationship. For example, if , cooling load is When the membership is 1, the cooling load is When the membership is 0, The membership degree is based on the linear equation calculate.

[0041] By setting such a triangular membership function, different states of cooling load fluctuation can be attributed to corresponding intervals in a clear and quantitative manner.

[0042] Deviation Index Calculation: The index (ranging from 0 to 1) is generated by weighted summation of the membership degrees in each interval. The closer the index is to 1, the more significant the deviation of the cooling load from the cooling capacity of the individual unit. For example, if the current cooling load is 130% of the cooling capacity of the individual unit, the membership degree for the "above demand" interval is 0.8, and the "close to demand" interval is 0.2, then the index = 0.8 × 1 + 0.2 × 0.5 = 0.9.

[0043] Step 202, Redundancy / Deficit Calculation: The total cooling capacity of the currently operating chillers = number of chillers in operation × cooling capacity per chiller (e.g., 3 chillers × 200kW = 600kW). Comparing this with the current cooling load (e.g., 750kW) yields a 150kW deficit (needing to add one chiller) or a redundancy (e.g., 100kW redundancy at a cooling load of 500kW, requiring the reduction of one chiller). The target chiller's start-up and shutdown energy consumption data for the past 30 days is retrieved. For example, a chiller consumes 5kWh per startup and 150kWh per hour. If the candidate solution is "add one chiller," the chiller's startup energy consumption plus its operating energy consumption is evaluated (e.g., for a projected 4-hour operation, total energy consumption = 5 + 150 × 4 = 605kWh). With the goal of minimizing total energy consumption, the projected energy consumption of each candidate combination (e.g., "maintain 3 chillers," "add one chiller," "reduce one chiller") is calculated, prioritizing the combination with the lowest incremental energy consumption. For example, adding one unit will increase energy consumption by 605kWh, while removing one unit may result in a decrease in efficiency due to insufficient load, increasing energy consumption by 800kWh. In this case, adding one unit is preferred.

[0044] In step 203, based on the initial unit combination plan (e.g., four units), the chilled water flow requirement is calculated as: number of units × rated flow rate per unit (e.g., 50 m³ / h per unit, total flow rate 200 m³ / h). The predicted flow rate is compared with the current maximum adjustable flow rate of the chilled water pump (e.g., the pump's rated flow rate is 250 m³ / h, and the current variable frequency drive adjustment range allows 200 m³ / h). If the predicted flow rate is within the allowable range (200 ≤ 250), the plan is feasible. If it is (e.g., predicted flow rate is 260 m³ / h), the unit combination is adjusted (e.g., reducing one unit to reduce the flow rate to 150 m³ / h). Ensure that the adjusted unit combination is linked to pump energy consumption optimization. For example, as the number of units is reduced, the pump speed is also reduced. The final output is a comprehensive plan that includes the number of units (e.g., four), operating time (e.g., four hours), and pump speed (e.g., 80% of rated speed).

[0045] By segmenting cooling loads over time and classifying fluctuations into different levels, we can analyze the timing and regularity of cooling load changes, avoiding delayed unit regulation due to sudden load changes. By dynamically matching cooling load with cooling capacity using a fuzzy PID control algorithm and combining it with membership functions to quantify deviations, we can reduce frequent unit starts and stops (e.g., reducing regulation frequency by 30%) while also avoiding energy waste caused by "one-size-fits-all" regulation. We evaluate candidate solutions based on historical energy consumption data to avoid energy consumption spikes caused by blindly adding or removing units. By coupling chilled water flow with unit combinations, we ensure coordinated operation of the cooling and water systems, reducing the risk of pump idling or overload.

[0046] In a preferred embodiment of the present invention, step 3, based on the final unit operation number combination plan, uses the real-time value of the chilled water supply and return water pressure difference as feedback, combines it with the flow demand forecast value, optimizes the chilled water pump variable frequency parameters through a genetic algorithm, and outputs the chilled water cycle dynamic control instructions, which may include: Step 300: Based on the final combination plan of the number of operating units, a chilled water flow demand reference curve is mapped and generated, wherein the flow demand reference value is associated with the number of operating units and the rated flow of each unit; Step 301: Using the deviation between the real-time chilled water supply and return pressure difference and the reference curve as the core parameter, combined with the chilled water flow demand forecast value, construct a fitness function, and set the pressure difference stability threshold range and the water pump energy consumption weight ratio; Step 302: Encode the pump speed parameter into a binary chromosome population, initialize the population size, and set dynamic crossover probability and mutation probability, where the crossover probability and mutation probability are dynamically adjusted according to the pressure difference deviation value; Step 303: In each iteration, the individuals in the population are evaluated based on the fitness function, and the roulette wheel selection strategy is used to select the parent individuals to generate the offspring population. The offspring are subjected to crossover and mutation operations until the maximum number of iterations is reached to obtain the optimized speed parameter. Specifically, the following steps are performed: Step 3030: Based on the fitness function, perform fitness evaluation on the binary-coded water pump speed parameter population, calculate the absolute value of the pressure difference deviation and the weighted sum of the energy consumption of each individual, and generate a fitness value sequence; Step 3031: Based on the fitness value sequence, a roulette wheel selection strategy is used to select parent individuals to generate a child population, wherein the crossover probability and mutation probability are dynamically adjusted according to the current pressure difference deviation value; Step 3032: Perform crossover and mutation operations on the offspring population, determine the individuals that will enter the next generation population based on the fitness value sequence, and iterate the optimization until the maximum number of iterations is reached to obtain the optimized speed parameter; Step 304 : Calculate the pressure difference fluctuation range and the pump energy consumption reduction rate based on the optimized speed parameters, determine the final speed parameter combination, and generate a chilled water cycle dynamic control instruction including a target speed value and an adjustment time window.

[0047] In this embodiment of the present invention, based on the finalized chiller operation combination plan, the rated flow rate of each chiller is first determined. For example, if the rated flow rate of a single chiller is 50 cubic meters per hour, when the final operating combination is 4 units, the total flow rate requirement baseline value is 4 × 50 = 200 cubic meters per hour.

[0048] Using time as the horizontal axis, the operating process is divided into multiple time periods (e.g., 15-minute periods). Within each time period, the corresponding flow demand baseline value is calculated based on the number of units currently operating. These baseline values are then connected sequentially to form a chilled water flow demand baseline curve. This curve reflects the ideal chilled water flow rate at different points in time to meet the cooling needs of the units.

[0049] Step 301: monitor the chilled water supply and return pipes in real time and obtain the supply and return pressure differential value through the pressure sensor. At the same time, compare the flow demand reference curve generated in step 300 to find the theoretical pressure differential at the corresponding time point. For example, assuming that the current time is 10:15 am, the reference curve shows that the theoretical pressure differential at this moment should be 0.2MPa, and the real-time pressure differential measured by the sensor is 0.25MPa. The pressure differential deviation value is the real-time pressure differential minus the theoretical pressure differential, that is, 0.25MPa 0.2MPa=0.05MPa.

[0050] Chilled water flow demand forecast: Retrieve chilled water flow data for the hour around 10:00 AM on all weekdays over the past quarter and organize the data into 10-minute intervals. For example, statistics show that during the same period over the past 12 weeks, the average flow rate in the first 10 minutes was 120 cubic meters / hour, and the average flow rate in the second 10 minutes was 125 cubic meters / hour, showing a stable upward trend. Calculate the standard deviation of the data to assess the flow fluctuation. A standard deviation of 3 cubic meters / hour indicates relatively regular flow fluctuations during that period. Use thermal imaging data from each floor to count the number of people in each area at the current moment. For example, if 50 people are currently gathered in a large conference room, based on the heat dissipation per person and the conference room area, estimate that the additional cooling load in this area is equivalent to an increase in chilled water flow demand of 8 cubic meters / hour. Check the operating status of heat-generating equipment such as the server room and large printers. If three new servers are added to the server room, it is calculated that the cooling load in the corresponding area will increase, resulting in an increase in chilled water flow demand of 5 cubic meters / hour.

[0051] Comprehensive forecast calculation: Superimpose the changing trend of historical data with the current load factor. If historical data shows a flow rate increase of 5 cubic meters / hour every 10 minutes during this period, and the flow rate demand caused by current personnel and equipment increases by 13 cubic meters / hour, the estimated chilled water flow rate demand for the first 10 minutes over the next 30 minutes will be 120 + 5 + 13 = 138 cubic meters / hour, and will then increase by 5 cubic meters / hour every 10 minutes based on the historical trend.

[0052] Pressure difference stability threshold range setting: Review the chilled water system design documents in detail to identify the design parameters for the supply and return water pressure differential. For example, the design requires that when the system is operating at full load, the supply and return water pressure differential must be maintained between 0.18 and 0.22 MPa to ensure the normal heat exchange efficiency of the terminal air conditioning equipment. Use this range as a basic reference. Extract the pressure differential operating data for the past six months and draw a line graph showing the pressure differential change over time. Use data statistics software to calculate the 90% data distribution range. Assume that, after calculation, 90% of the pressure differential data is concentrated between 0.19 and 0.21 MPa, which is a baseline value of 0.2 MPa ± 0.01 MPa. However, considering the stability of the system under extreme operating conditions, the range is appropriately relaxed to ± 0.03 MPa. Establish a detailed correspondence table between pressure differential deviation and penalty value. When the deviation is within ±0.03 MPa, the penalty is 0; when the deviation reaches 0.04 MPa, the penalty is set to 20; when it reaches 0.05 MPa, the penalty increases by 20 for every 0.01 MPa increase, that is, the penalty is 40; if the deviation exceeds 0.08 MPa (seriously out of range), the penalty is directly set to 100 to strongly constrain pressure fluctuations.

[0053] Pump energy consumption weight ratio setting: The goal of reducing chilled water pump energy consumption by 10% this quarter was further broken down by month. Assuming this quarter lasted three months, the monthly energy consumption reduction required was approximately 3.3%. Analysis of the current pump energy consumption revealed that the frequency conversion regulation phase accounted for 60% of energy consumption. Therefore, the focus was on optimizing the frequency conversion parameters and setting weights. Data on pump operating energy consumption and costs over the past year were compiled to create an energy-cost relationship curve. For example, if energy consumption decreased from 10,000 kWh to 9,900 kWh, system operating costs would decrease from 5,000 yuan to 4,950 yuan, meaning that every 1% reduction in energy consumption resulted in a 50 yuan cost savings. Furthermore, the potential increase in equipment maintenance costs associated with unstable pressure differentials was estimated. For example, if excessive pressure differential fluctuations damaged pipe joints, each repair would cost approximately 2,000 yuan. Multiple simulations were conducted to evaluate the system's performance under different weighting combinations. When the energy consumption weight is 0.6 and the pressure difference stability weight is 0.4, the simulation results show that it can not only meet the monthly energy saving target of 3.3%, but also control the number of times the pressure difference exceeds the threshold to no more than 2 times per month, effectively balancing energy saving and stability needs.

[0054] Fitness function calculation: Simulated operation tests were conducted for different combinations of pump speed parameters. For example, speed parameter combinations A (70% of rated speed) and B (75% of rated speed) were set, and the actual pressure differential and energy consumption data for each combination were recorded. If the pressure differential during operation of combination A was 0.24 MPa, a deviation of 0.04 MPa from the baseline value of 0.2 MPa, the corresponding penalty score was 20 according to the penalty value rules; if the pressure differential during operation of combination B was 0.22 MPa, a deviation of 0.02 MPa, the penalty score was 0. The energy consumption values were converted into scores proportionally. Assuming that the energy consumption of the system at full load is 100 points, combination A's energy consumption of 70 kWh corresponds to a score of 70 points; combination B's energy consumption of 75 kWh corresponds to a score of 75 points.

[0055] Based on the set weights, the fitness value of each combination is calculated. Combination A's fitness value = 20 × 0.4 + 70 × 0.6 = 50; Combination B's fitness value = 0 × 0.4 + 75 × 0.6 = 45. This shows that Combination B has a lower fitness value and is more in line with the requirements for differential pressure stability and energy conservation.

[0056] In step 302, the pump speed parameter is binary-encoded, mapping the speed range (e.g., minimum speed is 30% of rated speed, maximum is 100%) to a binary string of a certain length. For example, using 8-bit binary encoding can represent 256 different speed states, corresponding to the fine-grained adjustment from lowest to highest speed. Initialize the population size, selecting 30-50 individuals to form the initial population to ensure sufficient parameter combinations for optimization search. Set the dynamic crossover probability and mutation probability: When the deviation value of the chilled water supply and return water pressure difference is large (for example, exceeding 0.05 MPa), the crossover probability is increased (from the initial value of 0.6 to 0.8), which increases gene exchange between individuals in the population and speeds up the search for a better solution; at the same time, the mutation probability is reduced (from 0.01 to 0.005) to avoid excessive mutation that destroys high-quality gene combinations.

[0057] When the pressure difference deviation is small (such as within ±0.02MPa), the crossover probability is reduced (to 0.4), gene exchange is reduced, and the current optimal solution is retained; the mutation probability is increased (to 0.02), population diversity is increased, and the algorithm is prevented from falling into a local optimum.

[0058] Step 3030: For each binary-coded individual in the population, first determine its code length (e.g., 8-bit binary code) and the mapping range of the speed parameter (e.g., the minimum speed is 30% of the rated speed, and the maximum is 100%). Convert the binary number to decimal, and then calculate the actual speed parameter based on the mapping range. For example, the binary code of an individual is 10101010, which is converted to decimal as 170. Assuming that the 8-bit code corresponds to the speed range of 30%-100%, then the actual speed = 30% + ( )×(100%-30%)≈76.7% rated speed.

[0059] Pressure differential deviation calculation: Based on the pump speed parameters calculated above, combined with the fluid mechanics characteristics of the chilled water system (such as the relationship between speed and flow, flow and pressure differential), estimate the corresponding chilled water supply and return pressure differential at that speed. For example, according to historical data, for every 10% decrease in pump speed, the pressure differential decreases by approximately 0.02 MPa. If the pressure differential is 0.2 MPa at the base speed, the pressure differential corresponding to the current speed of 76.7% is approximately 0.2 × 0.02 ≈ 0.153 MPa. Comparing the estimated pressure difference with the theoretical pressure difference (assuming it is 0.18 MPa) at the corresponding time point of the flow demand reference curve generated in step 300, the absolute value of the pressure difference deviation is obtained as |0.153-0.18| = 0.027 MPa. Energy consumption calculation: Based on the empirical relationship between pump speed and energy consumption (energy consumption is proportional to the cube of the speed), calculate the energy consumption at the current speed. For example, if the pump's energy consumption at rated speed is 100kW and the current speed is 76.7% of the rated speed, the estimated energy consumption is 100 × (76.7%)³ = 45.1.

[0060] Weighted summation of fitness values: According to the energy consumption weight of 0.6 and the pressure difference stability weight of 0.4 set in step 301, the absolute value of the pressure difference deviation and the energy consumption are weighted. Assuming that every 0.01MPa pressure difference deviation corresponds to a 10-point penalty, and every 1kW of energy consumption corresponds to a 1-point penalty, then: Pressure difference deviation score = ×10=27 points, energy consumption score = 44.7×1=44.7 points, fitness value = 27×0.4+45.1×0.6=37.86.

[0061] Step 3031, calculate the sum of the fitness values of all individuals in the population (assuming that the population has 5 individuals with fitness values of 37.86, 40.2, 35.1, 38.8, and 42.5, respectively, then the sum is 37.86+40.2+35.1+38.8+42.5=194.46). For each individual, the probability of being selected = 1-( ).

[0062] Roulette wheel selection process: The selection probability of each individual is mapped to a range of 0-1, forming a roulette wheel-like partitioning. For example, the probability ranges for five individuals are [0, 0.819], [0.819, 0.945], [0.945, 0.987], [0.987, 0.999], and [0.999, 1]. A number between 0 and 1 (such as 0.85) is randomly generated, and the range in which it falls is determined. The individual in the corresponding range is selected as the parent. This process is repeated until the required number of parents is selected to generate offspring (for example, if the population size is 30, 30 parents are selected). During the selection process, the deviation of the current chilled water supply and return pressure difference from the baseline value is obtained in real time (for example, the current deviation is 0.04 MPa). If the pressure difference deviation exceeds a threshold (e.g., 0.03 MPa), indicating insufficient system stability, the crossover probability is increased (from an initial 0.6 to 0.8) to promote gene exchange between individuals and accelerate the search for a better solution. At the same time, the mutation probability is reduced (from 0.01 to 0.005) to prevent excessive mutation from destroying stable gene combinations. If the pressure difference deviation is within the threshold, the crossover probability is reduced to 0.4 and the mutation probability is increased to 0.02 to increase population diversity and prevent the algorithm from falling into a local optimum.

[0063] Step 3032: Crossover and mutation: A crossover operation is performed on the selected parent individuals. A crossover point is randomly selected, and partial gene segments of the two parent individuals are exchanged to generate offspring individuals. For example, if the binary code of parent individual 1 is 10101010 and the code of parent individual 2 is 01010101, a crossover is performed after the fourth bit, resulting in offspring individual 1 being 10100101 and offspring individual 2 being 01011010. A mutation operation is performed on some of the offspring individuals, randomly changing certain bits in the binary code (e.g., changing 0 to 1), introducing new gene combinations and increasing population diversity. Based on the fitness value sequence, offspring individuals with higher fitness are selected to enter the next generation population. The fitness evaluation, selection, crossover, and mutation process is repeated until the preset maximum number of iterations (e.g., 100) is reached, resulting in the optimized pump speed parameter.

[0064] In step 304, based on the optimized speed parameters, the fluctuation range of the chilled water supply and return pressure differential at that speed is calculated. For example, the optimized speed results in a pressure differential fluctuation between 0.18 and 0.22 MPa. The pump energy consumption reduction rate is also calculated. The final speed parameter combination is determined by comprehensively considering the pressure differential stability and energy consumption reduction. A dynamic control instruction for the chilled water cycle is generated, specifying the target speed value (e.g., 70% of the rated speed) and the adjustment time window (e.g., gradually adjusting the speed to the target value within the next 30 minutes) to guide the actual operation and adjustment of the chilled water pump.

[0065] A baseline chilled water flow demand curve is generated by mapping the number of operating units, ensuring precise matching of chilled water flow with unit cooling requirements. This prevents insufficient flow from impacting cooling efficiency or excessive flow from wasting energy. A fitness function that comprehensively considers pressure differential stability and energy consumption provides an evaluation criterion for chilled water pump parameter optimization, effectively reducing pump energy consumption while ensuring stable system operation. Using binary encoding and dynamically adjusted crossover and mutation probabilities, the genetic algorithm rapidly adapts to changes in system operating conditions, improving the efficiency and accuracy of parameter optimization and avoiding local optimal solutions. Through iterative optimization using the genetic algorithm, optimal pump speed parameter combinations are continuously identified, further improving chilled water system performance and achieving the dual goals of pressure differential stability and energy conservation. Control instructions, including target speed values and adjustment time windows, are generated, providing specific and actionable guidance for actual chilled water pump operation, ensuring stable system operation.

[0066] In a preferred embodiment of the present invention, the above step 4, based on the dynamic control instruction of the chilled water cycle, collects the temperature and flow data of the condenser outlet and the cooling tower inlet in real time, extracts the feedback correction coefficient by fitting the dynamic temperature-flow relationship curve of the cooling water, predicts the cooling water demand temperature by combining the neural network algorithm, dynamically adjusts the cooling tower fan speed and bypass valve opening, and generates the cooling water cycle control result, which may include: Step 400: Based on the dynamic control instructions for the chilled water cycle, the temperature and flow data of the condenser outlet and the cooling tower inlet are collected in real time, a cooling water dynamic temperature-flow relationship curve is fitted in time series, and the slope of the curve is extracted as a feedback correction coefficient; Step 401: Based on the feedback correction coefficient, a neural network prediction model is constructed. The feedback correction coefficient is used as the input layer. In combination with the real-time data collected by the ambient temperature and humidity sensor, a fully connected neural network is constructed, which includes an input layer, a hidden layer, and an output layer. The hidden layer is trained with weight parameters using historical cooling water operation data, and the output layer predicts the cooling water demand temperature. Step 402 , performing a difference calculation between the cooling water demand temperature predicted by the neural network and the cooling tower inlet temperature collected in real time by the temperature sensor to generate a temperature deviation value; Step 403: Using a proportional-integral control algorithm, the temperature deviation value is linearly superimposed with the accumulated deviation value within a preset integral time window to generate a cooling tower fan speed adjustment value. The bypass valve opening adjustment value is dynamically calculated based on the slope change direction and amplitude of the feedback correction coefficient. Step 404 : Integrate the cooling tower fan speed adjustment amount and the bypass valve opening adjustment amount into a cooling water circulation control result.

[0067] In the embodiment of the present invention, the condenser outlet temperature (T1) and the cooling tower inlet temperature (T2) are collected in real time by a temperature sensor (accuracy ±0.5°C), and the cooling water flow rate (F) is collected by an electromagnetic flowmeter (accuracy ±1%). The collection frequency is once per minute to form time series data (for example, at 14:00:00, T1=32°C, T2=28°C, F=150m³ / h).

[0068] Time Series Fitting: Using a one-hour window, the collected temperature-flow rate data (T2-F) is plotted as a scatter plot, and a dynamic relationship curve is fitted using the least squares method. For example, if the flow rate increases from 120 m³ / h to 180 m³ / h over a certain period, the inlet temperature drops from 30°C to 26°C. The fitted curve is T2 = -0.067F + 38, with a slope of -0.067 (the feedback correction factor J), indicating that for every 1 m³ / h increase in flow rate, the temperature drops by 0.067°C. After each window is fitted, the feedback correction factor is immediately updated for subsequent control calculations.

[0069] Step 401 processes three months of historical cooling water operation data. This data includes the feedback correction coefficient J, ambient temperature Ts, ambient humidity Hs, and the actual cooling water demand temperature Td. Because different data have different dimensions and ranges (e.g., J ranges from -0.1 to 0.1, Ts ranges from 0 to 50°C, and Hs ranges from 0 to 100%), the data needs to be normalized to facilitate neural network learning. For example, to normalize J to a range of 0 to 1, the actual J value is divided by 0.1 and added to 0.5; for Ts, it is divided by 50 (e.g., if Ts = 35°C, the normalized value is 35 ÷ 50 = 0.7); and Hs is simply divided by 100. The actual cooling water demand temperature Td is also normalized to a range of 0 to 1 using a similar method.

[0070] Initialization of neural network structure: The neural network consists of an input layer, two hidden layers (10 nodes each), and an output layer. The three nodes in the input layer correspond to the normalized feedback correction coefficient J, the ambient temperature Ts, and the ambient humidity Hs, respectively. The nodes in the hidden layers are responsible for feature extraction and processing of the input data. The single node in the output layer outputs the predicted cooling water demand temperature. Initialize the weights of the connections between nodes in each layer of the neural network. The weight values are typically randomly selected between -1 and 1. Each node is also assigned an initial bias value, also randomly set between -1 and 1.

[0071] Forward propagation calculation: Real-time input data (for example, normalized J=0.165, Ts=0.7, and Hs=0.6) is fed into the input layer nodes. These input layer nodes pass the data to the first hidden layer. During this process, each hidden layer node receives data from the input layer nodes and performs calculations based on the connection weights and bias values. For example, a node in the first hidden layer receives data from three input layer nodes, multiplies each by its corresponding weights, adds the sum, and then adds the node's bias value to produce an intermediate result. This intermediate result is then processed using an activation function (such as a sigmoid function) to convert it into a value between 0 and 1, which serves as the node's output. After all nodes in the first hidden layer complete this calculation, the output result is passed to the second hidden layer, which repeats the above calculation process. Finally, the output of the second hidden layer is passed to the output layer, which, after similar calculations, produces a predicted value between 0 and 1, which is the normalized cooling water demand temperature forecast.

[0072] Error calculation: The normalized predicted value obtained by the output layer is converted to an actual temperature value through denormalization (for example, if the predicted value is 0.55, the denormalized value is 0.55 × 50 = 27.5°C). This predicted temperature value is compared with the actual cooling water demand temperature to calculate the error. The error is calculated using the mean square error (MSE), which squares the difference between the predicted and actual values and then takes the average. For example, if the actual temperature is 27°C and the predicted temperature is 27.5°C, the error is (27.5 - 27)² = 0.25.

[0073] Back propagation and weight adjustment: Based on the calculated error, backpropagation begins at the output layer, passing the error back to the hidden and input layers layer by layer. During backpropagation, the weights and bias values between nodes in each layer are adjusted based on the error. This adjustment is based on the principle of gradually reducing the error, which is achieved through the gradient descent algorithm. Specifically, the weights and bias values are modified according to a certain learning rate (such as 0.1) in the direction of fastest error decrease. For example, if a weight causes a large error, the value of that weight is reduced according to the learning rate, reducing the error in the next calculation. By repeating the forward propagation, error calculation, and backpropagation process, the weights and bias values are continuously adjusted until the mean squared error between the predicted temperature output by the model and the actual temperature is less than 0.5°C. At this point, the model training is considered complete.

[0074] The trained model can be used for real-time prediction. The feedback correction coefficient J, ambient temperature Ts, and ambient humidity Hs, which are collected and normalized in real time, are input into the model. After forward propagation through the trained neural network, the output layer directly obtains the normalized cooling water demand temperature forecast. After denormalization, the actual cooling water demand temperature forecast value Tdpred is obtained, such as a forecast value of 27.5°C.

[0075] Step 402 : Read the cooling tower water inlet temperature T2real (eg, 28° C.) collected in real time by the temperature sensor.

[0076] Deviation calculation: Temperature deviation = Tdpred - T2real = 27.5°C - 28°C = -0.5°C, indicating that the current inlet water temperature is 0.5°C higher than the predicted demand and needs to be cooled.

[0077] In step 403, based on the temperature deviation, a proportional adjustment is generated: Kp × temperature deviation. Kp is the proportional coefficient (for example, Kp = -10, where the negative sign indicates a speed reduction when the temperature is too high). Therefore, the proportional adjustment is -10 × (-0.5) = 5 rpm. The cumulative temperature deviation over the past 10 minutes is calculated (for example, -2°C / min). The integral adjustment is calculated as Ki × the cumulative temperature deviation. Ki is the integral coefficient (for example, Ki = 0.5). Therefore, the integral adjustment is 0.5 × (-2) = -1 rpm. The total adjustment is the proportional adjustment + the integral adjustment, which equals 5 - 1 = 4 rpm. This means that the fan speed needs to be increased by 4 rpm to improve heat dissipation. Analyzing the change in the slope of the feedback correction coefficient (for example, from -0.05 to -0.067, the absolute value of the slope increases by 0.017), indicates that the cooling water temperature is more sensitive to flow rate changes, and the bypass valve opening needs to be increased to increase flow rate. According to the slope change range, it is set that every 0.01 slope change corresponds to a 2% increase in the bypass valve opening. The current opening needs to be increased by 3.4% (rounded to 3%).

[0078] In step 404, the fan speed adjustment (+4 rpm) and the bypass valve opening adjustment (+3%) are combined into a control instruction and sent to the cooling tower control system. The adjustment time (e.g., 14:05:00) and the current operating data are also recorded for reference in the next round of control.

[0079] By fitting the temperature-flow curve in real time, the changes in the heat exchange characteristics of the cooling water system are dynamically captured (for example, when scaling causes a decrease in heat exchange efficiency, the slope of the curve will slow down), making the feedback correction coefficient more in line with the actual working conditions and improving the control accuracy. The neural network model can learn the temperature requirements in complex environments (such as the ambient temperature and humidity, and equipment load jointly affect the cooling demand). The PI control algorithm combined with dynamic slope adjustment can quickly respond to the current temperature deviation (proportional part) and eliminate long-term accumulated errors (integral part), while flexibly adjusting the bypass valve according to changes in system characteristics. The integrated control instructions realize the coordinated optimization of the fan and the bypass valve, avoiding system imbalance caused by single parameter adjustment (for example, adjusting only the fan may cause water pressure fluctuations), and improving the stability and energy efficiency of the cooling water circulation system.

[0080] In a preferred embodiment of the present invention, step 5, establishing a balanced distribution model for equipment operating time based on the equipment operating parameter optimization results, and dynamically allocating unit operating tasks using a rotation priority algorithm based on the historical operating time data of each chiller to achieve balanced operating time within the equipment group, may include: Step 500: Establishing a model for balancing equipment operating time based on the cooling water circulation control result; Step 501: Input the historical operating time data of each chiller into the equipment operating time balance distribution model, calculate the operating time standard deviation of each chiller, and generate a priority weight based on the standard deviation; Step 502: When the cooling load demand triggers the start and stop of the unit, the candidate units are sorted according to the priority weights, and the target units are selected for start-up in descending order of weight value, while the redundant units are shut down in descending order of weight value; Step 503: Using a sliding time window algorithm, the latest running time data is intercepted with a preset time window length, and the standard deviation of the group running time and the priority weight are recalculated; In step 504, if the recalculated operating time standard deviation exceeds the preset threshold, the equipment operating time balance allocation model is triggered to recalculate the weights and adjust the group task allocation until the standard deviation is less than the preset threshold, thereby achieving dynamic balance of operating time within the equipment group.

[0081] In this embodiment of the present invention, the current operating load status of the chiller is determined based on cooling water circulation control results (e.g., parameters such as cooling tower fan speed and bypass valve opening). For example, when cooling water control instructions require increased heat dissipation efficiency, this indicates that the number of chillers in operation may need to be increased to cope with the increased load. In this case, basic unit information must be imported during model initialization, including unit number (e.g., units A, B, C, and D, totaling four units), rated cooling capacity of each unit, and historical fault records.

[0082] Step 501 , extracting the daily operating time data of each unit for the past 30 days (e.g., unit A has accumulated 450 hours of operation, unit B has accumulated 400 hours, unit C has accumulated 500 hours, and unit D has accumulated 430 hours).

[0083] Mean and standard deviation calculation: Calculate the average operating time of the four units, that is, =445 hours.

[0084] Calculate the deviation of each unit's operating time from the mean: Unit A: 450−445=+5 hours; Unit B: 400 − 445 = − 45 hours; Unit C: 500−445=+55 hours; Unit D: 430−445=−15 hours.

[0085] The standard deviation is calculated by summing the squared deviations. Assume the calculated standard deviation is 35 hours. Based on the degree of imbalance reflected by the standard deviation, units with operating hours below the mean are given higher priority. For example, units B (-45 hours) and D (-15 hours) have shorter operating hours, so their weight coefficients are set to 0.9 and 0.7, respectively. Units A (+5 hours) and C (+55 hours) have weights of 0.5 and 0.3 (the weights are inversely proportional to the operating time deviation).

[0086] Step 502: When the cooling load increases and a unit needs to be started: Sorting of candidate units: Sorting by weight from high to low is unit B (0.9) > D (0.7) > A (0.5) > C (0.3), and unit B is selected for startup first.

[0087] Redundant unit shutdown: If the cooling load decreases and one unit needs to be shut down, the order of weight from low to high is C (0.3) > A (0.5) > D (0.7) > B (0.9), and unit C is shut down first.

[0088] Through this strategy, it is ensured that the units with short operating time are given priority in obtaining operating tasks, and the accumulated time of each unit is gradually balanced.

[0089] In step 503, a sliding window is preset to 7 days. At 1:00 AM each day, the last 7 days of unit operating time data (e.g., operating time from day 1 to day 7) is automatically captured. For example, after 7 days, unit A has accumulated 70 hours of operation, B has accumulated 50 hours, C has accumulated 80 hours, and D has accumulated 60 hours. The new average operating time is calculated to be 65 hours, with a standard deviation of 12 hours (significantly lower than the previous average of 35 hours). Weights are regenerated based on the new data: B (50 < 65, weight 0.8), D (60 < 65, weight 0.7), A (70 > 65, weight 0.6), and C (80 > 65, weight 0.5).

[0090] Step 504: The default threshold for the standard deviation of the running time is 15 hours (which can be adjusted according to equipment maintenance requirements). If the standard deviation obtained in a calculation is 18 hours (exceeding the threshold), the model is triggered to recalculate the weights and adjust the task allocation: Prioritize starting unit B (weight 0.8), which has the shortest operating time, and shutting down unit C (weight 0.5), which has the longest operating time. Continue monitoring subsequent data until the standard deviation drops below 15 hours. For example, after two rounds of adjustments, the standard deviation drops to 10 hours, achieving balanced operating time.

[0091] Quantifying unit load differences through historical operating data avoids the randomness of manual experience-based allocation and makes priority weight generation more scientific. A weighted start-stop strategy can reduce the run time deviation of newly started units by 40% compared to previous ones. For example, the deviation of Unit C, which originally had the longest run time (+55 hours), was reduced to +10 hours after three adjustments. A sliding window algorithm combined with threshold judgment achieves dynamic balancing adjustment, stabilizing the standard deviation of run time within a device group within a preset threshold (e.g., 15 hours), thereby preventing increased wear caused by excessive operation of a single unit. By balancing run time, the frequency of failures caused by uneven equipment aging is reduced, while also optimizing unit energy efficiency.

[0092] like Figure 2 As shown, an embodiment of the present invention further provides a high-efficiency control system for a central air-conditioning refrigeration room, comprising: Dynamic cooling load module, used to obtain chilled water supply and return temperature difference and flow data, and combined with thermal imaging data of occupant distribution on each floor, calculate the dynamic cooling load demand of the building; The operation adjustment module is used to build a dynamic adjustment model for the number of chillers in operation based on the dynamic cooling load demand, and calculate the matching degree between the actual cooling load value and the cooling capacity of a single chiller through the fuzzy PID control algorithm to generate the final unit operation combination plan; The frequency conversion optimization module is used to optimize the frequency conversion parameters of the chilled water pumps using a genetic algorithm based on the final unit operation number combination plan generated by the chiller operation adjustment module, using the real-time value of the chilled water supply and return water pressure difference as feedback and combining it with the flow demand forecast value, and output the dynamic control instructions for the chilled water cycle; The circulation control module is used to dynamically control the chilled water circulation based on the output of the chilled water pump frequency conversion optimization module. It collects the temperature and flow data of the condenser outlet and the cooling tower inlet in real time, extracts the feedback correction coefficient by fitting the dynamic temperature-flow relationship curve of the cooling water, and uses the neural network algorithm to predict the cooling water demand temperature. It dynamically adjusts the cooling tower fan speed and bypass valve opening to generate the cooling water circulation control results. The balanced allocation module is used to establish a balanced allocation model for equipment operating time through the equipment operating parameter optimization results generated by each module. Based on the historical operating time data of each chiller, it adopts a rotation priority algorithm to dynamically allocate unit operating tasks to achieve balanced operating time within the equipment group.

[0093] It should be noted that this system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0094] An embodiment of the present invention further provides a computing device comprising: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the above-described method. All implementations in the above-described method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0095] The embodiment of the present invention further provides a computer-readable storage medium storing instructions, which, when executed on a computer, causes the computer to execute the above-described method. All implementations in the above-described method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0096] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A high-efficiency control method for a central air-conditioning refrigeration room, characterized in that: The method comprises: Step 1: Obtain the chilled water supply and return temperature difference and flow data, and combine it with the thermal imaging data of the occupant distribution on each floor to calculate the dynamic cooling load demand of the building; Step 2: Based on the dynamic cooling load demand, a dynamic adjustment model for the number of chillers in operation is constructed. The matching degree between the actual cooling load value and the cooling capacity of a single chiller is calculated using a fuzzy PID control algorithm to generate the final unit operation combination plan. Step 3: Based on the final unit operation number combination plan, the real-time value of the chilled water supply and return water pressure difference is used as feedback, combined with the flow demand forecast value, the chilled water pump frequency conversion parameters are optimized through genetic algorithm, and the dynamic control instructions of the chilled water cycle are output; Step 4: Based on the dynamic control instructions for the chilled water cycle, the temperature and flow data of the condenser outlet and the cooling tower inlet are collected in real time. The feedback correction coefficient is extracted by fitting the dynamic temperature-flow relationship curve of the cooling water. The cooling water demand temperature is predicted by combining the neural network algorithm. The cooling tower fan speed and bypass valve opening are dynamically adjusted to generate the cooling water cycle control results. Step 5: Based on the equipment operating parameter optimization results, a model for balancing the equipment operating time is established. Based on the historical operating time data of each chiller, a rotation priority algorithm is used to dynamically allocate unit operating tasks to achieve balanced operating time within the equipment group.

2. The high-efficiency control method for a central air-conditioning refrigeration room according to claim 1, characterized in that: Obtain chilled water supply and return temperature difference and flow data, combined with thermal imaging data of occupancy distribution on each floor, to calculate the building's dynamic cooling load demand, including: Based on the infrared radiation intensity distribution of each area in the thermal imaging data, the boundary range of the personnel gathering area is identified, and the dynamic heat source weight coefficient of each floor is calculated based on the heat source density per unit area; The dynamic heat source weight coefficient is integrated with the chilled water supply and return water temperature difference and flow data to calculate the instantaneous cooling load value of each zone layer by layer; Dynamically smooth the instantaneous cooling load value, set a fixed time window length, extract the cooling load value sequence within the window, and generate a smoothed cooling load benchmark value; Based on the smoothed cooling load reference value and the current temperature data collected in real time by the ambient temperature sensor, the historical average temperature of the same period in the past preset time is obtained, the deviation between the current temperature and the historical average temperature is calculated, and the temperature compensation factor is generated proportionally according to the size of the deviation; The temperature compensation factor is superimposed and corrected with the smoothed cooling load benchmark value to output a dynamic cooling load demand that reflects the spatial thermal distribution characteristics of the building and the changing trend in the time dimension.

3. The high-efficiency control method for a central air-conditioning refrigeration room according to claim 2, characterized in that: Based on the dynamic cooling load demand, a dynamic adjustment model for the number of chillers in operation is constructed. The fuzzy PID control algorithm is used to calculate the matching degree between the actual cooling load value and the cooling capacity of a single chiller, and the final unit operation combination plan is generated, including: The dynamic cooling load demand is divided into a cooling load fluctuation sequence according to the time dimension. Based on the distribution characteristics of the cooling load peak and valley values, the cooling load fluctuation level is divided and the cooling load deviation threshold is defined. Based on the cooling load fluctuation sequence, a dynamic adjustment model for the number of operating chillers is constructed. The real-time cooling load value is dynamically matched with the cooling capacity of each chiller unit through a fuzzy PID control algorithm. The cooling load fluctuation range is quantified based on the membership function, and a cooling load deviation index is generated. When the cooling load deviation index exceeds the preset threshold, the total cooling capacity redundancy or shortfall of the currently operating units is calculated based on the time distribution characteristics of the cooling load demand. Combined with the historical start-up and shutdown energy consumption data of the chillers, the energy consumption evaluation index of the candidate unit combination is constructed. Based on the energy consumption evaluation index, the initial chiller unit operation number combination scheme is determined from the candidate unit combination, and the initial chiller unit operation number combination scheme is coupled with the chilled water flow demand forecast value for verification, and the final unit operation number combination scheme that matches the chilled water pump energy consumption constraint is output.

4. The high-efficiency control method for a central air-conditioning refrigeration room according to claim 3, characterized in that: Based on the final unit operation combination plan, the real-time value of the chilled water supply and return pressure difference is used as feedback, combined with the flow demand forecast value, the chilled water pump frequency conversion parameters are optimized through genetic algorithms, and the dynamic control instructions of the chilled water cycle are output, including: Based on the final combination plan of the number of operating units, a chilled water flow demand benchmark curve is mapped and generated, in which the flow demand benchmark value is associated with the number of operating units and the rated flow of each unit; The core parameter is the deviation between the real-time value of the chilled water supply and return pressure difference and the benchmark curve. Combined with the predicted value of chilled water flow demand, a fitness function is constructed, and the pressure difference stability threshold range and the water pump energy consumption weight ratio are set. The pump speed parameter is encoded into a binary chromosome population, the population size is initialized, and the dynamic crossover probability and mutation probability are set. The crossover probability and mutation probability are dynamically adjusted according to the pressure difference deviation value. In each round of iteration, the individuals in the population are evaluated based on the fitness function, and the roulette wheel selection strategy is used to screen the parent individuals to generate the offspring population. The offspring are subjected to crossover and mutation operations until the maximum number of iterations is reached, and the optimized speed parameters are obtained. According to the optimized speed parameters, the pressure difference fluctuation range and the water pump energy consumption reduction rate are calculated, the final speed parameter combination is determined, and the dynamic control instructions of the chilled water cycle including the target speed value and the adjustment time window are generated.

5. The high-efficiency control method for a central air-conditioning refrigeration room according to claim 4, characterized in that: In each iteration, the individuals in the population are evaluated based on the fitness function, and the roulette wheel selection strategy is used to select the parent individuals to generate the offspring population. The offspring are subjected to crossover and mutation operations until the maximum number of iterations is reached, and the optimized speed parameters are obtained, including: Based on the fitness function, the fitness of the binary-coded pump speed parameter population is evaluated, and the absolute value of the pressure difference deviation and the weighted sum of energy consumption of each individual are calculated to generate a fitness value sequence. According to the fitness value sequence, the roulette wheel selection strategy is used to screen the parent individuals and generate the offspring population, where the crossover probability and mutation probability are dynamically adjusted according to the current pressure difference deviation value; Perform crossover and mutation operations on the offspring population, determine the individuals that enter the next generation population according to the fitness value sequence, and iterate the optimization until the maximum number of iterations is reached to obtain the optimized speed parameters.

6. The high-efficiency control method for a central air-conditioning refrigeration room according to claim 5, characterized in that: Based on the dynamic control instructions of the chilled water cycle, the temperature and flow data of the condenser outlet and the cooling tower inlet are collected in real time. The feedback correction coefficient is extracted by fitting the dynamic temperature-flow relationship curve of the cooling water. The cooling water demand temperature is predicted by combining the neural network algorithm. The cooling tower fan speed and bypass valve opening are dynamically adjusted to generate the cooling water cycle control results, including: Based on the dynamic control instructions of the chilled water cycle, the temperature and flow data of the condenser outlet and the cooling tower inlet are collected in real time. The dynamic temperature-flow relationship curve of the cooling water is fitted according to the time series, and the slope of the curve is extracted as the feedback correction coefficient; Based on the feedback correction coefficient, a neural network prediction model is constructed, and the feedback correction coefficient is used as the input layer. Combined with the real-time data collected by the ambient temperature and humidity sensor, a fully connected neural network consisting of an input layer, a hidden layer, and an output layer is constructed. The weight parameters of the hidden layer are trained using historical cooling water operation data, and the output layer predicts the cooling water demand temperature; The difference between the cooling water demand temperature predicted by the neural network and the cooling tower inlet temperature collected in real time by the temperature sensor is calculated to generate a temperature deviation value; Using the proportional-integral control algorithm, the temperature deviation value is linearly superimposed with the accumulated deviation value within the preset integral time window to generate the cooling tower fan speed adjustment value. The bypass valve opening adjustment value is dynamically calculated based on the slope change direction and amplitude of the feedback correction coefficient. The cooling tower fan speed adjustment and bypass valve opening adjustment are integrated into the cooling water circulation control result.

7. The high-efficiency control method for a central air-conditioning refrigeration room according to claim 6, characterized in that: Based on the results of equipment operating parameter optimization, a model for balancing equipment operating time is established. Based on the historical operating time data of each chiller, a rotation priority algorithm is used to dynamically allocate unit operating tasks to achieve balanced operating time within the equipment group, including: Based on the cooling water circulation control results, establish the equipment operation time balance distribution model; Input the historical operating time data of each chiller into the equipment operating time balance distribution model, calculate the operating time standard deviation of each unit, and generate the priority weight according to the size of the standard deviation; When the cooling load demand triggers the start and stop of the unit, the candidate units are sorted according to the priority weight, and the target units are selected for start-up in descending order of weight value, while the redundant units are shut down in descending order of weight value; Through the sliding time window algorithm, the latest running time data is intercepted with the preset time window length, and the standard deviation and priority weight of the group running time are recalculated; If the recalculated running time standard deviation exceeds the preset threshold, the equipment running time balance allocation model is triggered to recalculate the weight and adjust the group task allocation until the standard deviation is less than the preset threshold, thus achieving dynamic balance of running time within the equipment group.

8. A high-efficiency control system for a central air-conditioning refrigeration room, the system implementing the method according to any one of claims 1 to 7, characterized in that: include: Dynamic cooling load module, used to obtain chilled water supply and return temperature difference and flow data, and combined with thermal imaging data of occupant distribution on each floor, calculate the dynamic cooling load demand of the building; The operation adjustment module is used to build a dynamic adjustment model for the number of chillers in operation based on the dynamic cooling load demand, and calculate the matching degree between the actual cooling load value and the cooling capacity of a single chiller through the fuzzy PID control algorithm to generate the final unit operation combination plan; The frequency conversion optimization module is used to optimize the frequency conversion parameters of the chilled water pumps using a genetic algorithm based on the final unit operation number combination plan generated by the chiller operation adjustment module, using the real-time value of the chilled water supply and return water pressure difference as feedback and combining it with the flow demand forecast value, and output the dynamic control instructions for the chilled water cycle; The circulation control module is used to dynamically control the chilled water circulation based on the output of the chilled water pump frequency conversion optimization module. It collects the temperature and flow data of the condenser outlet and the cooling tower inlet in real time, extracts the feedback correction coefficient by fitting the dynamic temperature-flow relationship curve of the cooling water, and uses the neural network algorithm to predict the cooling water demand temperature. It dynamically adjusts the cooling tower fan speed and bypass valve opening to generate the cooling water circulation control results. The balanced allocation module is used to establish a balanced allocation model for equipment operating time through the equipment operating parameter optimization results generated by each module. Based on the historical operating time data of each chiller, it adopts a rotation priority algorithm to dynamically allocate unit operating tasks to achieve balanced operating time within the equipment group.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images