Efficient refrigeration machine room air conditioner energy-saving intelligent control system and method

Abstract

The application discloses an energy-saving intelligent control system and method for a high-efficiency refrigeration machine room air conditioner, belongs to the technical field of air conditioner energy saving, and comprises a prediction layer, an optimization layer and a control layer in a three-level hierarchical collaborative architecture, wherein the prediction layer is used for receiving environmental data, building load data and historical operation data uploaded by a data acquisition unit and outputting building heat load prediction results in a future preset time length; the optimization layer is used for receiving the heat load prediction results, taking the minimum system total energy consumption as an optimization target, combining equipment rated parameters, safety operation constraints and pipe network hydraulic balance constraints, solving optimal operation parameter combinations of each device of the refrigeration machine room, and delivering the optimal operation parameter combinations to the control layer; the control system effectively solves control problems caused by system large lag, strong nonlinearity and time-varying characteristics, the method is obviously superior to a traditional scheme in temperature control precision and energy-saving effect, and is easy to implement in engineering and high in long-term operation reliability.

Description

Technical Field

[0001] This invention belongs to the field of air conditioning energy-saving technology, specifically relating to a high-efficiency intelligent control system and method for energy-saving air conditioning in refrigeration rooms. Background Technology

[0002] In the energy consumption structure of public buildings, the central air conditioning system accounts for 40% to 60% of the total building energy consumption. The chiller room, as the core power unit of the central air conditioning system, includes key equipment such as chillers, chilled water pumps, cooling water pumps, and cooling towers. Its energy consumption accounts for more than 60% of the total energy consumption of the air conditioning system, making it a core target for building energy-saving renovations. With the continuous increase in the number of large public buildings, the operational energy efficiency of the chiller room directly affects the overall carbon emission level of the building. Developing efficient and intelligent energy-saving control technologies has significant economic value and environmental benefits.

[0003] Currently, mainstream control schemes for chiller rooms are mainly divided into two categories. The first category is traditional PID control. This method is widely used in the field of industrial control due to its advantages such as simple structure, clear physical meaning of parameters, and ease of engineering implementation. However, once the parameters of a traditional PID controller are tuned, they remain fixed and cannot adaptively adjust to the dynamic changes in outdoor temperature and humidity, building heat load, and equipment operating status. In scenarios with large load fluctuations in the morning and evening and drastic changes in personnel density, problems such as temperature overshoot and regulation lag are prone to occur, causing the chiller unit to deviate from its high-efficiency operating range and the system COP (coefficient of performance) to decrease by 5% to 15%. At the same time, fixed-parameter PID is difficult to adapt to the long-term characteristic drift of the system due to equipment aging, pipe scaling, valve wear, etc. The longer the operating time, the more significant the degradation in control accuracy and energy-saving effect.

[0004] The second category is model-based predictive control (MPC). This method establishes a mathematical model of the system and solves a constrained optimization problem in each control cycle to achieve rolling optimization control. Theoretically, MPC can achieve better control performance, but its engineering implementation faces three major bottlenecks: First, it highly depends on an accurate system mechanism model, while chiller rooms have characteristics such as strong nonlinearity, multivariable coupling, large inertia, and large lag, making accurate modeling extremely difficult. Model mismatch will directly lead to a deterioration in control performance. Second, each optimization step requires solving a complex constrained optimization problem, resulting in high computational complexity and high requirements for controller hardware performance, making it difficult to meet the needs of second-level real-time control. Third, the model parameters are fixed and cannot adapt to slow time-varying characteristics such as equipment aging and changes in pipeline resistance, resulting in insufficient long-term operational reliability. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art. Firstly, it provides a high-efficiency intelligent control system for air conditioning in refrigeration rooms, comprising a three-tiered collaborative architecture of a prediction layer, an optimization layer, and a control layer, wherein: The prediction layer is used to receive environmental data, building load data and historical operation data uploaded by the data acquisition unit, and output the building heat load prediction result for a preset time period in the future. The optimization layer is used to receive the heat load prediction results, take the minimum total energy consumption of the system as the optimization objective, and combine the rated parameters of the equipment, safe operation constraints and hydraulic balance constraints of the pipeline network to solve for the optimal combination of operating parameters of each equipment in the chiller room, and then send it down to the control layer. The control layer is used to receive the optimal combination of operating parameters as control setpoints, combine them with real-time temperature feedback data, adopt a fuzzy PID control strategy with online self-tuning parameters, and output control signals to the actuator unit to perform real-time closed-loop regulation of the chiller, water pump and cooling tower equipment.

[0006] Furthermore, the data collected by the data acquisition unit is divided into three categories: Environmental and load data, including outdoor dry-bulb temperature, outdoor wet-bulb temperature, indoor area temperature, occupancy density, and power consumption for lighting and equipment; System operating data includes chilled water supply and return water temperature and flow rate, cooling water supply and return water temperature and flow rate, chiller unit inlet and outlet pressure, unit operating power, water pump operating frequency, and cooling tower fan operating frequency; Equipment status data includes equipment runtime, cumulative number of starts and stops, fault alarm information, and valve opening feedback.

[0007] Furthermore, the prediction layer employs a long short-term memory network model with an attention mechanism to perform rolling prediction of heat load, with a prediction duration covering 15 to 60 minutes. The prediction results are sent to the optimization layer for global energy efficiency optimization and also serve as a feedforward signal input to the control layer to counteract the large hysteresis characteristics of the water system.

[0008] Furthermore, the optimization layer uses an improved whale optimization algorithm to solve for the optimal combination of operating parameters. The optimal combination of operating parameters includes the number of operating chiller units and load rate allocation, chilled water supply temperature setpoint, cooling water supply temperature setpoint, optimal operating frequency of water pumps, and optimal operating frequency of cooling tower fans. The optimization cycle of the optimization layer is 5 minutes.

[0009] Furthermore, the fuzzy PID control strategy of the control layer uses temperature deviation and deviation change rate as fuzzy inference inputs, and corrects the proportional, integral, and derivative control parameters online in real time through preset fuzzy rules; the control cycle of the control layer is 10 seconds.

[0010] Furthermore, the initial reference parameters of the fuzzy PID control strategy are obtained offline by an improved whale optimization algorithm; during system operation, parameter optimization is re-executed using the latest operating data on a 24-hour cycle to adaptively adapt to system characteristic drift caused by equipment aging and pipe scaling.

[0011] Furthermore, the system adopts a deployment architecture that combines edge controllers and a cloud platform. The edge controllers are deployed locally in the chiller room and are responsible for real-time data acquisition, low-level closed-loop control, and local fault protection. They can operate independently and safely when the network is down. The cloud platform is deployed in the energy management center and undertakes tasks such as load forecasting, global energy efficiency optimization, parameter self-learning, and remote monitoring.

[0012] Secondly, the present invention provides a high-efficiency intelligent control method for energy-saving air conditioning in a computer room, employing a high-efficiency intelligent control system for energy-saving air conditioning in a computer room as described in the first aspect, comprising the following steps: Step 1: Collect environmental parameters, system operating parameters, and equipment status parameters of the refrigeration room in real time, and upload the data after cleaning and preprocessing. Step 2: Based on historical operating data and real-time collected data, perform rolling predictions of building heat load for a preset period of time in the future to generate a heat load prediction sequence; Step 3: Using the heat load prediction sequence as input and the overall system energy efficiency as the objective, combined with the equipment safety operation constraints, the optimal operating parameter settings for each piece of equipment in the refrigeration room are obtained. Step four: Using the optimal operating parameter setpoint as the control target, and combining real-time temperature feedback data, a fuzzy PID control method with online parameter self-tuning is adopted to output control commands, drive the operation of each actuator, and realize closed-loop control of temperature and energy efficiency. Step 5: Optimize and update the control parameters using the latest operating data at fixed intervals to adapt to changes in the system's long-term operating characteristics.

[0013] The beneficial effects of this invention include: This invention addresses the shortcomings of traditional PID control in chiller rooms, namely poor adaptability and complex MPC modeling. It proposes an adaptive energy-saving control system that integrates load forecasting and intelligent optimization. A three-layer architecture decouples global optimization from real-time control. Attention LSTM is employed to improve load forecasting accuracy, and an improved whale algorithm optimizes fuzzy PID parameters and supports online self-learning. This effectively solves the control challenges posed by large system time lag, strong nonlinearity, and time-varying characteristics. Simulation results demonstrate that this method significantly outperforms traditional solutions in both temperature control accuracy and energy saving. Furthermore, it is easy to implement in engineering, exhibits high long-term operational reliability, and has broad prospects for widespread application. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the structure of an energy-saving intelligent control system for a high-efficiency refrigeration room air conditioner according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating an energy-saving intelligent control method for a high-efficiency refrigeration room air conditioner according to an embodiment of the present invention. Detailed Implementation

[0016] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0017] A typical refrigeration room system consists of three main parts: a chiller unit, a chilled water system, and a cooling water system. The chiller unit produces low-temperature chilled water through a compression refrigeration cycle. The chilled water pump delivers the chilled water to the terminal air conditioning equipment to absorb indoor heat, and the heated chilled water returns to the unit for further cooling. The cooling water system dissipates the heat generated by the chiller unit's condenser to the outdoor environment through cooling water pumps and a cooling tower.

[0018] Under normal operating conditions, chiller units account for 60% to 70% of the total energy consumption of the computer room, while water pumps and cooling towers account for 30% to 40%. Determined by the building's heat load, the core objective of energy-saving control is to maximize energy savings by optimizing the operating parameters of each piece of equipment while meeting cooling requirements.

[0019] The core challenges of controlling a refrigeration room lie in three aspects: Multivariable strong coupling nonlinearity: Parameters such as chilled water supply and return water temperature difference, cooling water inlet temperature, unit load rate, and water pump frequency are coupled with each other, and the equipment efficiency curve is strongly nonlinear, making it difficult for traditional single-loop control to achieve global optimization.

[0020] Large inertia and large lag characteristics: The water system has a large capacity and a long pipeline transmission distance. Temperature regulation has a significant lag, and the pure lag time can reach several minutes, which can easily lead to PID control oscillation.

[0021] Load dynamics and time-varying nature: The building heat load is affected by multiple factors such as outdoor meteorological parameters, human activities, and equipment heat dissipation. The daily fluctuation range can reach more than 40%, and fixed parameter control is difficult to adapt to all working conditions.

[0022] Based on the above characteristics, the intelligent control system needs to meet three core requirements: first, high-precision temperature control to ensure that the chilled water supply temperature is stable within the set value ±0.5℃ range; second, energy-saving optimization under all operating conditions to maintain efficient system operation under different load rates; and third, self-adaptation and self-learning capabilities to adapt to slow drift of system characteristics without manual readjustment.

[0023] Example 1 like Figure 1-2 As shown, this invention discloses an efficient intelligent control system for air conditioning in a refrigeration room. The proposed intelligent control system adopts a layered architecture of prediction layer, optimization layer, and control layer, which takes into account both global optimization accuracy and real-time response at the bottom layer.

[0024] The perception layer data acquisition involves the system collecting full-dimensional operational data through a sensor network, categorized into three types: Environmental and load data: outdoor dry-bulb temperature, outdoor wet-bulb temperature, indoor temperature in various areas, personnel density, and power consumption for lighting and equipment; System operating data: chilled water supply and return water temperature and flow rate, cooling water supply and return water temperature and flow rate, chiller unit inlet and outlet pressure, unit operating current and power, water pump operating frequency, cooling tower fan frequency; Equipment status data: equipment runtime, cumulative number of starts and stops, fault alarm information, valve opening feedback.

[0025] All data is preprocessed by the edge gateway and then uploaded to the control platform at a frequency of 1Hz, providing a data foundation for prediction and control algorithms.

[0026] Prediction Layer: Short-term heat load forecasting. Based on historical operating data and real-time meteorological data, the prediction layer uses an attention-enhanced LSTM network to perform rolling forecasts of building heat load for the next 15-60 minutes. The forecast results are fed to the optimization layer for global energy efficiency optimization, and also used as a feedforward signal to the lower-level controller to offset the system's large lag characteristics and achieve "on-demand cooling".

[0027] Optimization Layer: This layer performs global optimization of system energy efficiency. Using predicted heat load as input and minimizing total system energy consumption as the objective, it considers constraints such as equipment rated parameters, safe operation limitations, and network hydraulic balance. An improved whale optimization algorithm is used to solve for the optimal combination of operating parameters, including the optimal number and load rate allocation of chiller units, chilled water supply temperature setpoints, cooling water supply temperature setpoints, and the optimal operating frequencies of pumps and cooling towers. The optimization cycle is set to 5 minutes to ensure both operational condition tracking capability and controlled computational overhead.

[0028] Control Layer: Adaptive fuzzy PID real-time control. The control layer receives setpoints from the optimization layer and employs a fuzzy PID controller optimized with an improved whale algorithm to perform real-time closed-loop control of actuators such as chilled water pump frequency, cooling water pump frequency, cooling tower fan frequency, and chiller unit guide vane opening. The control cycle is set to 10 seconds to meet real-time adjustment requirements. The fuzzy inference module corrects the PID parameters online based on the current temperature deviation and the rate of change of the deviation, achieving full-condition adaptive control.

[0029] The core algorithm design and mathematical derivation involve an LSTM-based heat load prediction algorithm based on an attention mechanism. The basic principle of the LSTM unit is explained. The Long Short-Term Memory (LSTM) network solves the gradient vanishing problem of traditional RNNs through a gating mechanism, making it suitable for processing heat load data with time-series characteristics. An LSTM unit consists of three gating structures: an input gate, a forget gate, and an output gate. Their mathematical expressions are as follows: The forgetting gate controls the degree to which the cell's state from the previous moment is preserved:

[0030] The input gate controls the amount of information written to the cell state at the current moment:

[0031] Cell status update:

[0032] The output gate controls the output of the cell state to the hidden layer:

[0033] in, Let be the input vector at time t. This is the output of the hidden layer from the previous time step. In cellular state, This is the weight matrix. For bias terms, is the Sigmoid activation function, and ⊙ is the Hadamard product.

[0034] Traditional LSTMs treat the hidden layer outputs at all times equally, but in heat load forecasting, the impact of different historical times on future loads varies. This paper introduces an attention mechanism to assign dynamic weights to the hidden states at each time step, highlighting key influencing factors.

[0035] Attention weights are calculated as follows:

[0036] Weighted context vector:

[0037] in, For trainable parameters of the attention network, Let be the attention weights of the hidden state at time t. The input time step is denoted as c. Finally, the context vector c is input into the fully connected layer to obtain the predicted heat load values ​​for future time steps.

[0038] The algorithm's advantage lies in its ability to automatically capture the time-varying influence weights of factors such as outdoor temperature and historical load, improving prediction accuracy by 8% to 12% compared to ordinary LSTM, thus providing a reliable basis for feedforward control.

[0039] An improved fuzzy PID control algorithm based on the whale algorithm, using incremental PID as its foundation. Incremental PID is commonly used in refrigeration room control, where the output is the increment of the control quantity, avoiding integral accumulation problems. The mathematical expression is:

[0040] Where e(k) is the temperature deviation at the k-th sampling time. These are the proportional, integral, and derivative coefficients, respectively. Traditional PID parameters are fixed and difficult to adapt to changes in operating conditions. Therefore, fuzzy inference is introduced to achieve online parameter self-tuning.

[0041] Fuzzy PID controllers use temperature deviation (e) and the rate of change of deviation (ec) as inputs, and correct the increments of the three PID parameters online using fuzzy rules. The corrected parameters are: Fuzzy inference employs the Mamdani method, dividing both input and output variables into seven fuzzy subsets: {NB, NM, NS, ZO, PS, PM, PB}, corresponding to negative large, negative medium, negative small, zero, positive small, positive medium, and positive large. A total of 49 fuzzy rules are established, with the core rules following the physical laws: When the deviation is large, increase Accelerate response, reduce Suppress overshoot, limit Avoid integration saturation; When the deviation is moderate, reduce it appropriately. Keep it moderate and ; When the deviation is small, increase Improve steady-state accuracy, increase Suppress oscillations.

[0042] After obtaining the optimal initial parameters of the fuzzy PID through offline optimization using IWOA, the system is put into operation. During operation, the optimization layer re-executes parameter optimization every 24 hours using the latest operating data to achieve slow time-varying adaptive control of controller parameters, effectively coping with system characteristic drift such as equipment aging and pipe scaling.

[0043] The system adopts an edge controller and cloud platform deployment architecture. The edge controller is deployed locally in the chiller room, using an industrial-grade ARM processor and equipped with multiple AI / AO and DI / DO interfaces. It directly connects to sensors and actuators, and is responsible for real-time data acquisition, low-level PID control, and local logic protection, ensuring the system can still operate safely in the event of a network outage. The cloud platform is deployed in the building energy management center, undertaking computationally intensive tasks such as load forecasting, global optimization, parameter self-learning, data visualization, and remote monitoring.

[0044] At the actuator level, all water pumps and cooling tower fans are equipped with frequency converters, and the chiller units support Modbus or BACnet communication protocols, enabling remote adjustment of parameters such as guide vane opening and target water supply temperature. High-precision temperature sensors (accuracy ±0.1℃) and electromagnetic flow meters are installed in key components to ensure control feedback accuracy.

[0045] The system software comprises six functional modules: Data acquisition and preprocessing module: Implements multi-protocol data parsing, outlier removal, data completion, and normalization. Load prediction module: Loads the trained attention LSTM model, performs rolling prediction, and supports incremental model updates; Energy efficiency optimization module: calls the IWOA algorithm to solve for the optimal operating parameters and outputs the setpoints to the control layer; Adaptive control module: runs a fuzzy PID control algorithm and outputs a 4~20mA signal or communication command to the actuator; Fault diagnosis and protection module: Real-time monitoring of equipment operating status; automatic degradation protection and alarm when abnormalities occur; Visualization and Reporting Module: Displays real-time operating curves, energy consumption statistics, and energy-saving analysis reports, and supports historical data backtracking.

[0046] The complete system operation process is as follows: After the system starts up, the perception layer collects real-time environmental and device data and uploads it to the edge controller and the cloud; The forecast layer outputs minute-by-minute heat load forecasts for the next 60 minutes based on data from the past 24 hours and future weather forecasts. The optimization layer takes the predicted load as input and aims to maximize the system COP to solve for the optimal operating parameters of each device under the current operating conditions. The optimized setpoint is sent to the control layer, and the fuzzy PID controller uses the temperature deviation as feedback and combines it with the load feedforward signal to adjust the actuator output in real time. After each optimization cycle, the actual running effect is compared with the predicted value, and the algorithm model is updated; a full parameter self-learning is performed once every morning to correct the controller parameters. When a device malfunction or communication interruption is detected, the system automatically switches to the traditional PID backup control mode to ensure the safe and continuous operation of the system.

[0047] The three-level collaborative architecture of prediction, optimization and control decouples short-term load forecasting, global energy efficiency optimization and real-time adaptive control in layers. It solves the problem of large system lag through prediction feedforward and balances global optimization accuracy and real-time control response speed through layered design, thus overcoming the shortcomings of MPC in terms of complex calculation and poor real-time performance.

[0048] Attention Mechanism LSTM Load Forecasting Algorithm: Based on the traditional LSTM, an attention mechanism is introduced to automatically identify the time-varying weights of different influencing factors. The mean absolute percentage error (MAPE) of the forecast is controlled within 3%, providing reliable data support for feedforward control and optimized scheduling, and effectively reducing control overshoot caused by load fluctuations.

[0049] An improved fuzzy PID control strategy based on the whale algorithm: By improving the whale algorithm with nonlinear convergence factors and adaptive weights, the optimization efficiency is increased by more than 30%. Combined with fuzzy inference, online self-tuning of PID parameters is achieved, while supporting daily offline rolling optimization of parameters. This not only solves the problem of insufficient adaptive capability of traditional PID, but also adapts to the long-term characteristic drift of the system through a self-learning mechanism, eliminating the need for repeated manual tuning.

[0050] Improvements over existing technologies Compared to traditional PID, it solves the problem that fixed parameters cannot adapt to fluctuations in operating conditions, reduces system overshoot by more than 60%, shortens settling time by 40%, improves temperature control accuracy to ±0.3℃, and increases the system average COP by 10%~15%.

[0051] Compared to traditional MPC: there is no need to build a precise mechanism model, and optimized control can be achieved based on data-driven approaches, significantly reducing modeling costs; the computational complexity is reduced by an order of magnitude, making it feasible with ordinary industrial controllers, and significantly enhancing engineering applicability.

[0052] Compared to ordinary fuzzy PID, it automatically tunes fuzzy rules and quantization factors through intelligent optimization algorithms, avoiding the subjectivity and limitations of manual experience tuning, resulting in better control quality; it also has long-term self-learning capabilities, solving the problem of control effect degradation caused by equipment aging.

[0053] To verify the effectiveness of the proposed method, a simulation model was built based on the actual operating data of a refrigeration room in a large commercial building. The refrigeration room is equipped with three screw chillers (each with a rated cooling capacity of 1200RT), along with three chilled water pumps, three cooling water pumps, and three cooling towers, using a one-pump-to-one-unit correspondence.

[0054] Comparative tests were conducted using typical summer day operating data, employing three control schemes: traditional PID, ordinary fuzzy PID, and the IWOA-fuzzy PID proposed in this paper. The test results are shown in the table below:

[0055] Experimental results show that, compared with traditional PID control, the proposed method reduces temperature overshoot by 70.8%, shortens settling time by 59.4%, and significantly improves steady-state control accuracy; daily power consumption is reduced by 17.4%, and the system's average COP is increased by 20.9%, demonstrating significant energy-saving effects. Furthermore, in load change tests, the proposed method can quickly track load changes without significant oscillations, verifying the algorithm's adaptive capability and robustness.

[0056] This paper addresses the pain points of traditional PID control in chiller rooms, namely poor adaptability and complex MPC modeling, by proposing an adaptive energy-saving control system that integrates load forecasting and intelligent optimization. A three-layer architecture decouples global optimization from real-time control. Attention LSTM is employed to improve load forecasting accuracy, and an improved whale algorithm is used to optimize fuzzy PID parameters and support online self-learning. This effectively solves the control challenges posed by the system's large time lag, strong nonlinearity, and time-varying characteristics. Simulation results demonstrate that this method significantly outperforms traditional solutions in both temperature control accuracy and energy saving. Furthermore, it is easy to implement in engineering, exhibits high long-term operational reliability, and has broad prospects for widespread application. Future research will further integrate digital twin technology to enhance the system's optimization and maintenance capabilities throughout its entire lifecycle.

[0057] Example 2 This embodiment provides a high-efficiency intelligent control method for energy saving in a computer room air conditioning system, employing a high-efficiency intelligent control system for energy saving in a computer room air conditioning system as described in Embodiment 1, including the following steps: Step 1: Collect environmental parameters, system operating parameters, and equipment status parameters of the refrigeration room in real time, and upload the data after cleaning and preprocessing. Step 2: Based on historical operating data and real-time collected data, perform rolling predictions of building heat load for a preset period of time in the future to generate a heat load prediction sequence; Step 3: Using the heat load prediction sequence as input and the overall system energy efficiency as the objective, combined with the equipment safety operation constraints, the optimal operating parameter settings for each piece of equipment in the refrigeration room are obtained. Step four: Using the optimal operating parameter setpoint as the control target, and combining real-time temperature feedback data, a fuzzy PID control method with online parameter self-tuning is adopted to output control commands, drive the operation of each actuator, and realize closed-loop control of temperature and energy efficiency. Step 5: Optimize and update the control parameters using the latest operating data at fixed intervals to adapt to changes in the system's long-term operating characteristics.

[0058] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A high-efficiency intelligent control system for energy-saving refrigeration room air conditioning, characterized in that, It includes a three-tiered collaborative architecture consisting of a prediction layer, an optimization layer, and a control layer, wherein: The prediction layer is used to receive environmental data, building load data and historical operation data uploaded by the data acquisition unit, and output the building heat load prediction result for a preset time period in the future. The optimization layer is used to receive the heat load prediction results, take the minimum total energy consumption of the system as the optimization objective, and combine the rated parameters of the equipment, safe operation constraints and hydraulic balance constraints of the pipeline network to solve for the optimal combination of operating parameters of each equipment in the chiller room, and then send it down to the control layer. The control layer is used to receive the optimal combination of operating parameters as control setpoints, combine them with real-time temperature feedback data, adopt a fuzzy PID control strategy with online self-tuning parameters, and output control signals to the actuator unit to perform real-time closed-loop regulation of the chiller, water pump and cooling tower equipment.

2. The high-efficiency refrigeration room air conditioning energy-saving intelligent control system according to claim 1, characterized in that, The data collected by the data acquisition unit is divided into three categories: Environmental and load data, including outdoor dry-bulb temperature, outdoor wet-bulb temperature, indoor area temperature, occupancy density, and power consumption for lighting and equipment; System operating data includes chilled water supply and return water temperature and flow rate, cooling water supply and return water temperature and flow rate, chiller unit inlet and outlet pressure, unit operating power, water pump operating frequency, and cooling tower fan operating frequency; Equipment status data includes equipment runtime, cumulative number of starts and stops, fault alarm information, and valve opening feedback.

3. The high-efficiency energy-saving intelligent control system for refrigeration room air conditioning according to claim 1, characterized in that, The prediction layer uses a long short-term memory network model with an attention mechanism to perform rolling prediction of heat load, with a prediction duration of 15 to 60 minutes. The prediction results are sent to the optimization layer for global energy efficiency optimization and also used as a feedforward signal to the control layer to counteract the large hysteresis characteristics of the water system.

4. The high-efficiency energy-saving intelligent control system for refrigeration room air conditioning according to claim 1, characterized in that, The optimization layer uses an improved whale optimization algorithm to solve for the optimal combination of operating parameters. The optimal combination of operating parameters includes the number of operating chiller units and load rate allocation, chilled water supply temperature setpoint, cooling water supply temperature setpoint, optimal operating frequency of water pumps, and optimal operating frequency of cooling tower fans. The optimization cycle of the optimization layer is 5 minutes.

5. The high-efficiency energy-saving intelligent control system for refrigeration room air conditioning according to claim 1, characterized in that, The fuzzy PID control strategy of the control layer uses temperature deviation and deviation change rate as fuzzy inference inputs, and corrects the proportional, integral and derivative control parameters online in real time through preset fuzzy rules; the control cycle of the control layer is 10 seconds.

6. The high-efficiency energy-saving intelligent control system for refrigeration room air conditioning according to claim 1, characterized in that, The initial baseline parameters of the fuzzy PID control strategy are obtained offline by an improved whale optimization algorithm.

7. The high-efficiency energy-saving intelligent control system for refrigeration room air conditioning according to claim 6, characterized in that, During system operation, the latest operating data is used to re-execute parameter optimization every 24 hours to adaptively adapt to system characteristic drift caused by equipment aging and pipe scaling.

8. The high-efficiency energy-saving intelligent control system for refrigeration room air conditioning according to claim 1, characterized in that, The system adopts a deployment architecture that combines edge controllers and a cloud platform. The edge controllers are deployed locally in the chiller room and are responsible for real-time data acquisition, low-level closed-loop control and local fault protection. They can operate independently and safely when the network is disconnected.

9. The high-efficiency energy-saving intelligent control system for refrigeration room air conditioning according to claim 8, characterized in that, The cloud platform is deployed in the energy management center and is responsible for load forecasting, global energy efficiency optimization, parameter self-learning, and remote monitoring.

10. A highly efficient intelligent control method for energy saving in refrigeration room air conditioning, characterized in that, The high-efficiency energy-saving intelligent control system for refrigeration room air conditioning as described in claim 1 includes the following steps: Step 1: Collect environmental parameters, system operating parameters, and equipment status parameters of the refrigeration room in real time, and upload the data after cleaning and preprocessing. Step 2: Based on historical operating data and real-time collected data, perform rolling predictions of building heat load for a preset period of time in the future to generate a heat load prediction sequence; Step 3: Using the heat load prediction sequence as input and the overall system energy efficiency as the objective, combined with the equipment safety operation constraints, the optimal operating parameter settings for each piece of equipment in the refrigeration room are obtained. Step four: Using the optimal operating parameter setpoint as the control target, and combining real-time temperature feedback data, a fuzzy PID control method with online parameter self-tuning is adopted to output control commands, drive the operation of each actuator, and realize closed-loop control of temperature and energy efficiency. Step 5: Optimize and update the control parameters using the latest operating data at fixed intervals to adapt to changes in the system's long-term operating characteristics.

Images