Multi-parameter cold source intelligent control system and method based on big data

Abstract

The application provides a multi-parameter cold source intelligent control system and method based on big data, wherein the system comprises a data acquisition and control module for acquiring cold source system related parameters; a comprehensive energy consumption standardization module for converting energy consumption parameters into unified standardized energy consumption indicators; a comfort degree algorithm construction module for generating a dynamic comfort degree index; an association rule mining module for mining multi-dimensional association rules from historical data; an intelligent decision module for generating an optimal control strategy based on the indicators, index and association rules; and a control instruction generation module for converting the optimal control strategy into equipment control parameters. The embodiments of the present application are based on a multi-parameter fusion cold source intelligent control technology based on deep reinforcement learning and an improved FP-Growth algorithm, and through deep mining and intelligent decision of multi-source data, efficient energy saving of the cold source system and precise regulation and control of indoor environmental comfort are realized.

Description

Technical Field

[0001] This application relates to the field of intelligent control technology, specifically to a multi-parameter intelligent control system and method for cold sources based on big data. Background Technology

[0002] As the proportion of global building energy consumption continues to rise, the dynamic optimization control of central air conditioning systems, as the core hub of building energy systems, has become a key breakthrough for achieving energy conservation and consumption reduction in buildings. The PID control technology used in traditional boilers and chiller units has significant limitations: on the one hand, its response speed lags behind actual load changes, leading to energy waste; on the other hand, relying on manual experience to adjust parameters is not only inefficient but also difficult to cope with the complex operating conditions of multi-variable coupling and nonlinear changes in modern buildings. It is worth noting that approximately 30% of the energy consumption during air conditioning system operation occurs under unnecessary conditions, such as continuous operation during unattended periods and excessive cooling / heating. However, traditional energy management methods can only monitor surface data and lack the ability to intelligently mine the deep-seated correlation rules between equipment operating parameters, environmental factors, and energy consumption, which severely restricts the full release of energy-saving potential.

[0003] CN118132968A, published under the patent number CN118132968A, proposes a data acquisition system and analysis method for energy consumption in tin smelting based on the Apriori algorithm. Although this application establishes the correlation rules between energy consumption and environmental factors through multi-data acquisition devices such as electricity, temperature, pressure, natural gas, and water consumption, and extracts regression coefficients to construct a multi-variable regression model, it can only achieve quantitative analysis of energy consumption patterns and lacks the ability to control equipment in real time. It cannot achieve energy-saving goals by dynamically optimizing equipment operating parameters.

[0004] Publication No. CN120295180A proposes an energy-saving intelligent control system for central air conditioning cold and heat sources. By incorporating the intelligent control system proposed in this solution, the system can effectively combine the user's control needs at specific times, the control temperature, and the ambient temperature to intelligently control and start the central air conditioning system. However, the system cannot automatically learn based on the user's energy consumption habits, and it cannot take into account other relevant factors affecting the efficiency of the central air conditioning system, such as working hours and personnel activities. Therefore, there is an urgent need for a multi-parameter integrated intelligent control technology for cold sources to achieve efficient energy saving of the cold source system and precise control of indoor environmental comfort. Summary of the Invention

[0005] This invention provides a multi-parameter intelligent control system and method for cold sources based on big data, which realizes efficient energy saving of cold source systems and precise control of indoor environmental comfort, and solves the problem mentioned in the background art of lacking the ability to intelligently mine the deep-level correlation rules between equipment operating parameters, environmental factors and energy consumption.

[0006] Firstly, this specification describes a multi-parameter intelligent cold source control system based on big data through several embodiments, including:

[0007] The data acquisition and control module is used to collect energy consumption parameters, environmental parameters, personnel activity parameters, and equipment operating status parameters of the cold source system, and execute control commands.

[0008] The integrated energy consumption standardization module, connected to the data acquisition and control module, is used to receive energy consumption parameters, convert them into unified standardized energy consumption indicators, and output energy efficiency assessment indicators.

[0009] The comfort algorithm construction module is connected to the data acquisition and control module to receive the environmental parameters and personnel activity parameters and generate a dynamic comfort index.

[0010] The association rule mining module is connected to the data acquisition and control module, the comprehensive energy consumption standardization module, and the comfort algorithm construction module. It is used to mine multidimensional association rules between the environmental parameters, standardized energy consumption indicators, and comfort index from historical data.

[0011] The intelligent decision-making module is connected to the comprehensive energy consumption standardization module, the comfort algorithm construction module, and the association rule mining module, and is used to generate the optimal control strategy based on the standardized energy consumption index, the comfort index, and the association rules.

[0012] The control command generation module, connected to the intelligent decision-making module and the data acquisition and control module, converts the optimal control strategy into equipment regulation parameters and sends them to the data acquisition and control module to drive the cooling source system to execute the control commands. By adopting the above technical solution, a multimodal data acquisition network is constructed to comprehensively and accurately monitor the cooling source's operating status, environmental parameters, and load demands. The comfort algorithm construction module accurately quantifies human thermal sensation according to international standards, collaboratively optimizing energy efficiency and comfort goals, significantly improving system decision-making accuracy and response speed. The intelligent decision-making module mines deep energy consumption patterns, outputs correlation rules, supports adaptive dynamic optimization of the control strategy, and achieves rapid and accurate control of equipment parameters. The control command generation module ensures efficient strategy execution, achieving rapid cooling and steady-state maintenance through intelligent adjustment, thus achieving the goal of energy saving and efficiency improvement.

[0013] Furthermore, the data acquisition and control module includes:

[0014] Energy metering units are used to monitor the consumption of electricity, natural gas, and water resources;

[0015] Thermodynamic state monitoring unit is used to monitor the three-dimensional temperature field of the evaporator and condenser, as well as the phase change state of the refrigerant;

[0016] The environmental sensing unit is used to monitor indoor light intensity, indoor carbon dioxide concentration, and personnel density and distribution. By adopting the above technical solution and integrating multi-dimensional units, comprehensive and refined monitoring of energy consumption, thermal state, and environmental parameters is achieved, providing a rich and accurate data foundation for subsequent intelligent decision-making and improving the system's sensing capabilities and control precision.

[0017] Furthermore, the comprehensive energy consumption standardization module constructs a unified multi-energy evaluation system to convert heterogeneous energy sources such as electricity, natural gas, and water into standard coal equivalents according to their calorific value. This unified multi-energy evaluation system is based on a dynamic conversion model, which is as follows: Where E represents the total standard coal equivalent, α_i represents the conversion factor for the i-th energy source, and Q_i represents the energy consumption of the i-th energy source. By constructing a unified multi-energy evaluation system and adopting a dynamic conversion model, standardized quantitative assessment of heterogeneous energy consumption such as electricity, natural gas, and water is achieved, providing a scientific indicator basis for accurate analysis and optimization of system energy efficiency.

[0018] Furthermore, the comprehensive energy consumption standardization module outputs the following energy efficiency evaluation indicators in real time: Where E_t represents the current energy consumption in standard coal equivalent, and E_{min} represents the historical minimum energy consumption in standard coal equivalent. By calculating energy efficiency evaluation indicators in real time, the system can dynamically reflect the gap between the current energy consumption level and the historical best level, providing an immediate and quantitative reference for energy efficiency monitoring, fault diagnosis, and optimization adjustment.

[0019] Furthermore, the comfort algorithm construction module calculates the dynamic comfort index based on the PMV-PPD model and constructs a comfort reward item based on this dynamic comfort index. A positive reward is given when the predicted proportion of dissatisfied personnel (PPD) is below a threshold, and a negative reward is given when the predicted proportion of dissatisfied personnel (PPD) is above the threshold. By adopting the PMV-PPD model and constructing a comfort reward mechanism, the thermal comfort of personnel is quantified and incorporated into the optimization objective, achieving accurate perception and positive guidance of comfort.

[0020] Furthermore, the association rule mining module employs an improved FP-Growth algorithm to mine high-frequency association rule sets and extract non-explicit performance-intensive association rules. By adopting the above technical solutions, the system can identify deep-seated operational patterns, providing interpretable and reliable data support for intelligent decision-making.

[0021] Furthermore, the intelligent decision-making module employs a deep reinforcement learning algorithm, using energy consumption and PMV comfort index as reward functions, to output compressor frequency and electronic expansion valve opening commands. By adopting a deep reinforcement learning algorithm, the system can adaptively generate key control commands such as compressor frequency and electronic expansion valve opening, with energy consumption and comfort as common optimization objectives, achieving precise control under complex operating conditions.

[0022] Furthermore, the reward function R is composed of a comfort reward term R_comfort and an energy reward term R_energy, weighted by the following formula: ω_1 and ω_2 are weighting coefficients. By designing a weighted reward function, the system can flexibly balance the two major objectives of energy saving and comfort, and adjust the weighting coefficients according to different scenarios to achieve configurable control strategies and multi-objective collaborative optimization.

[0023] Secondly, embodiments of this specification provide a multi-parameter intelligent control method for cold sources based on big data, including the following steps:

[0024] Collect energy consumption parameters, environmental parameters, personnel activity parameters, and equipment operating status parameters of the cold source system, and execute control commands;

[0025] Different types of energy consumption parameters will be uniformly replaced with standardized energy consumption indicators.

[0026] A dynamic comfort index is generated based on the environmental parameters and personnel activity parameters.

[0027] Based on historical operating data, we can explore multi-dimensional correlation rules between operating environment parameters, standardized energy consumption indicators, and comfort index.

[0028] Using the standardized energy consumption index and the comfort index as optimization targets, and combining the association rules, an optimal control strategy is generated through an intelligent learning algorithm.

[0029] The optimal control strategy is then translated into specific equipment control parameters and executed.

[0030] Furthermore, the intelligent learning algorithm is a deep reinforcement learning algorithm; when the mined association rules trigger a specific emergency situation, an emergency control strategy is activated; the deep reinforcement learning algorithm adopts a priority experience replay mechanism to assign higher weights to the training samples corresponding to the emergency situation.

[0031] The beneficial effects of the technical solutions provided in some embodiments of this specification include at least the following:

[0032] In several embodiments of this specification, a multi-parameter intelligent cold source control system and method based on big data is provided. By introducing a multimodal data acquisition network, it achieves comprehensive and high-precision perception of the cold source system's operating status, environmental conditions, and load changes. The comfort algorithm construction module quantifies human thermal comfort based on international standards, achieving a dual-objective balance optimization of energy efficiency and comfort, thus improving the accuracy and responsiveness of decision-making. The deep reinforcement learning decision module, combined with an improved FP-Growth algorithm, mines and outputs non-obvious performance consumption association rules, realizing adaptive dynamic optimization of the cold source system control strategy. It exhibits excellent performance in handling dynamic load changes and complex coupling relationships, enabling rapid response and precise adjustment of equipment operating parameters, achieving precise control of the cold source system, effectively reducing energy consumption, and improving energy utilization efficiency. Simultaneously, the control command generation module ensures the precise execution of the deep reinforcement learning strategy, achieving a smooth switch between rapid cooling and constant temperature maintenance through dynamic adjustment of the evaporation temperature setpoint.

[0033] Other features and advantages of various embodiments of this specification will be further revealed in the following detailed description and accompanying drawings. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this specification, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram of a multi-parameter intelligent cold source control system based on big data, provided as an embodiment of this specification.

[0036] Figure 2 This is a schematic diagram of the deep learning process in the intelligent decision-making module provided in the embodiments of this specification.

[0037] Figure 3 This is a schematic diagram of the multi-parameter intelligent control method for cold sources based on big data, provided in the embodiments of this specification.

[0038] Figure 4 This is a schematic diagram of the process for generating a dynamic comfort index provided in the embodiments of this specification. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. The terms "first," "second," "third," etc., in the specification, claims, and accompanying drawings are used to distinguish different objects, not to describe a specific order. In addition, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0040] In the following description, terms such as “inner,” “outer,” “upper,” “lower,” “left,” and “right” are used only to facilitate the description of the embodiments and to simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this specification.

[0041] Please see Figure 1 This manual provides a multi-parameter intelligent control system for cold sources based on big data. It comprises six core functional modules, which are tightly coupled and work together to form a closed-loop intelligent control system. The six core functional modules are: data acquisition and control module 100, comprehensive energy consumption standardization module 200, comfort algorithm construction module 300, association rule mining module 400, intelligent decision-making module 500, and control command generation module 600. The entire system responds in real-time to environmental and load changes and dynamically adjusts the operating status of the cold source equipment through data flow and iterative optimization between modules.

[0042] The data acquisition and control module 100 is equipped with a multimodal sensor array and actuators to collect energy consumption parameters, environmental parameters, personnel activity parameters, and equipment operating status parameters of the cold source system, and to receive and execute control commands from the control command generation module. The data acquisition and control module constructs a multimodal sensing network optimized for the cold source system. It collects the characteristic parameters required by the algorithm through an energy metering unit, a thermal state monitoring unit, and an environmental sensing unit. Specifically, it includes, but is not limited to, collecting temperature (indoor / outdoor temperature gradient), humidity (relative humidity and dew point temperature), personnel density (fusion of infrared sensing and access control data), solar radiation intensity (ultraviolet index and azimuth), and weather type (precipitation probability and wind speed). Furthermore, the energy metering unit is equipped with a smart meter with an accuracy of 0.5, a gas flow meter of G1.6, and an ultrasonic water meter with an accuracy of ±1%, for monitoring the consumption of electricity, natural gas, and water resources at the equipment level. The thermal state monitoring unit is equipped with a PT100 platinum resistance array with an accuracy of Class A and a digital pressure sensor with a range of 0-25MPa, for capturing the three-dimensional temperature field of the evaporator and condenser and tracking the phase change process of the refrigerant in real time. The environmental sensing unit integrates an illuminance sensor with a range of 0-200klux, a CO2 sensor based on the NDIR principle, and a binocular vision system with an accuracy of ±3%, for measuring indoor light intensity, monitoring indoor carbon dioxide concentration, and establishing an indoor personnel density and spatial distribution model. Specifically, the data acquisition is conducted through an industrial IoT architecture, using the MODBUS-TCP protocol for millisecond-level acquisition. The preprocessing includes accessing real-time weather data from the meteorological bureau via an HTTP RESTful API, and performing timestamp alignment and missing value filling on all data. The structured time-series dataset contains 15-dimensional state-space features and is pushed to a cloud-based time-series database via the MQTT protocol at a sampling frequency of 200Hz, meeting the requirements of deep reinforcement learning for high-dimensional and timely training data.

[0043] The integrated energy consumption standardization module 200, connected to the data acquisition and control module, receives energy consumption parameters, converts them into unified standardized energy consumption indicators, and outputs energy efficiency assessment indicators. Specifically, based on the GB / T2589-2020 standard, it uniformly converts the energy consumption of multiple sources such as electricity, natural gas, and water into kilograms of standard coal equivalent, thereby establishing a comparable multi-energy unified evaluation system. This multi-energy unified evaluation system is based on a dynamic conversion model, which is as follows: Where E represents the total standard coal equivalent, α_i represents the conversion factor for the i-th energy source, and Q_i represents the energy consumption of the i-th energy source. Through energy type normalization, the original 7-dimensional energy consumption parameters are compressed into a 1-dimensional standard coal equivalent index, reducing the state space dimension of deep reinforcement learning by 85.7%. As the core computational unit of the reward function, the above technology can output energy efficiency evaluation indicators in real time. This provides the gradient update direction for the policy network. Here, E_t represents the current energy consumption in standard coal equivalent, and E_{min} represents the historical minimum energy consumption in standard coal equivalent. This energy efficiency evaluation index provides the gradient update direction for the policy network. Furthermore, the comprehensive energy consumption standardization module, combined with the association rules mined by the FP-Growth algorithm, automatically corrects the conversion coefficient (e.g., seasonal correction factor β = 0.92-1.15), keeping the standard coal calculation error within ±2.3%. Actual testing shows that this module improves the convergence speed of reinforcement learning by 40%, and the total standard coal calculation can serve as an energy efficiency benchmark, supporting ISO50001 energy management system certification. This step not only achieves dimensionality reduction but also provides quantified energy consumption reward input for deep reinforcement learning.

[0044] The comfort algorithm construction module 300, connected to the data acquisition and control module, receives the environmental parameters and personnel activity parameters, and constructs a PMV-PPD model to quantify human thermal comfort and generate a dynamic comfort index by combining indoor and outdoor environmental parameters. The comfort algorithm construction module incorporates a multi-objective weighted aggregation reward function within a deep reinforcement learning framework. The reward function R is designed using a multi-objective weighted aggregation method, and its formula is as follows: The formula is as follows: ω_1 and ω_2 are weighting coefficients. The comfort reward R_comfort is based on the ISO7730:2023 international standard, which constructs a dynamic PMV-PPD model to calculate the predicted average vote value (PMV) and the predicted percentage of dissatisfaction (PPD). A positive reward is given when the PPD value is below 10%, and a negative penalty is applied at ΔPPD / 15 when it exceeds the threshold. The energy consumption reward R_energy achieves unified quantification of multiple energy sources through a standard coal equivalent conversion module. This module has a built-in conversion coefficient library of the "General Rules for Comprehensive Energy Consumption Calculation" (GB / T 2589-2020), which converts heterogeneous energy data such as electricity (0.1229 kgce / kWh) and natural gas (1.2143 kgce / m³) into comparable standard coal values ​​in real time, and uses a sliding window algorithm to calculate the energy consumption change rate over the past 24 hours as the basis for rewards and penalties. The FP-Growth algorithm mines association rules for equipment parameters from historical operating data (e.g., the frequent itemset support of cooling water flow rate and chilled water outlet temperature reaches 0.32). The generated feature vectors serve as dimensionality reduction inputs to the deep reinforcement learning state space, ultimately forming an adaptive control strategy that meets LEED v4.1 certification requirements. This step not only serves the purpose of dimensionality reduction but also provides quantified comfort reward inputs for deep reinforcement learning.

[0045] The association rule mining module 400, connected to the data acquisition and control module, the comprehensive energy consumption standardization module, and the comfort algorithm construction module, analyzes historical data using an improved frequent pattern mining algorithm (single scan + vertical data format) to mine multidimensional association rules between environmental parameters, standardized energy consumption indicators, and comfort indices. This module introduces a dynamic minimum support threshold function, adaptively adjusting the threshold based on the parameter dimensions (e.g., setting the personnel density weight coefficient to 0.3 and the temperature weight to 0.25). Matrix compression storage technology optimizes the FP-tree construction process, reducing memory usage by approximately 40%. The algorithm outputs high-frequency association rule sets, such as a 92% confidence level for a surge in cooling load under environmental conditions of >800W / m² sunshine and >70% humidity, and a 90% confidence level for a 35% increase in cooling demand under conditions of ≥28℃ temperature and >65% humidity.

[0046] The intelligent decision-making module 500 is connected to the comprehensive energy consumption standardization module, the comfort algorithm construction module, and the association rule mining module. This module employs a deep reinforcement learning algorithm, using energy consumption (standard coal equivalent) and the PMV comfort index (ISO7730:2023) as reward functions, and outputs action commands such as compressor frequency and electronic expansion valve opening. The intelligent decision-making module uses the output of the heating and cooling demand prediction model as input, combined with real-time indoor temperature, humidity, cleanliness, and airflow velocity parameters, and uses the comfort level assessed by the ISO7730:2023-04 international thermal comfort environment standard as the reward function. The reinforcement learning agent continuously learns and optimizes its decision-making strategy through interaction with the environment. When control is required, the intelligent agent outputs corresponding control actions, including adjusting the compressor's operating frequency and the electronic expansion valve opening, controlling the operating status of the HVAC system, and adjusting parameters such as temperature, humidity, cleanliness, and airflow velocity to achieve rapid cooling or constant temperature maintenance.

[0047] Please see Figure 2 The steps of deep reinforcement learning are as follows:

[0048] The reward function is designed based on energy consumption reward and comfort reward. It should be noted that the negative logarithm of standard coal consumption (kgce / h) is calculated in real time as the energy consumption reward item. The comfort reward item is a dynamic reward function based on the PMV index constructed according to the ISO7730:2023 standard. When PMV∈[-0.5,0.5], a positive reward is given.

[0049] When frequent pattern recognition triggers the rule [outdoor temperature > 35℃ ∩ sudden increase in pedestrian flow of 40%], the agent immediately activates the emergency control strategy, which is as follows:

[0050] The compressor frequency was gradually increased from 35Hz to 42Hz within 30 seconds (slope 0.23Hz / s).

[0051] Adjust the opening of the electronic expansion valve to 65%±2% synchronously;

[0052] COP = 4.8 ± 0.3 and PMV = 0.2 ± 0.1 were maintained through PID compensation;

[0053] For state-action space modeling, specifically,

[0054] The load data from the integrated heating and cooling demand forecasting model and real-time sensor data (temperature / humidity / cleanliness / airflow velocity) are used as the input layer;

[0055] The continuous control commands that generate the compressor frequency (20-50Hz) and the electronic expansion valve opening (10%-90%) are used as the output layer;

[0056] Online learning optimization is implemented, and a priority experience replay mechanism is adopted to assign 3 times the sampling weight to emergency working condition samples triggered by FP-Growth, thereby accelerating strategy convergence.

[0057] The intelligent decision-making module uses the high-frequency association rule set output by the association mining module as prior knowledge or state space dimensionality reduction input, and assigns higher learning weights to the working condition samples that trigger specific association rules. With the energy efficiency evaluation index and the dynamic comfort index as the core components of the reward function, it constructs a dual-objective optimization model based on deep reinforcement learning to generate an optimal control strategy that balances energy efficiency (lowest energy consumption) and comfort (PMV≈0).

[0058] The control command generation module 600, connected to the intelligent decision-making module and the data acquisition and control module, is used to convert the optimal control strategy output by the deep reinforcement learning decision-making module into specific equipment control parameters (such as compressor frequency and water pump flow rate), and send them to the data acquisition and control module to drive the cold source system to execute the control command. The system dynamically iterates in 15-minute cycles, and achieves adaptive control of the cold source system through real-time rule updates of FP-Growth and online strategy optimization of deep reinforcement learning.

[0059] Please see Figure 3 This specification also provides an embodiment of a multi-parameter intelligent control method for cold sources based on big data, including the following steps:

[0060] Step S201: Data acquisition and preprocessing. Real-time acquisition of operating environment parameters, equipment operating status parameters, energy consumption parameters, and personnel activity parameters of the cold source system through a multi-modal sensor array. Data cleaning, calibration, and synchronization are performed to form a structured time-series dataset.

[0061] Step S202: Convert different types of energy consumption parameters into standardized energy consumption indicators. Based on the national standard GB / T2589-2020, dynamically convert the energy consumption parameters collected in step S201, convert heterogeneous energy sources such as electricity, natural gas, and water into standard coal equivalent, and calculate energy efficiency assessment indicators.

[0062] Step S203: Comfort Index Construction. Based on the ISO 7730:2023 international standard, and using the environmental and human activity parameters collected in step S201, a dynamic comfort index is calculated and generated. Please refer to [link to relevant documentation]. Figure 4 The dynamic comfort index is generated by including:

[0063] The dynamic comfort index is calculated using the PMV-PPD model, and a comfort reward item is constructed based on this dynamic comfort index. A positive reward is given when the predicted proportion of dissatisfied personnel (PPD) is lower than a threshold, and a negative reward is given when the predicted proportion of dissatisfied personnel (PPD) is higher than the threshold. The PMV-PPD model calculates the predicted average vote PMV value using the following formula.

[0064] ,

[0065] Where M is the human body's energy metabolism rate; W is the amount of work done by the human body; H is the human body's convective heat loss; E is the human body's evaporative heat loss; C is the convective heat exchange between the human body and the surrounding environment; R is the radiative heat exchange between the human body and the surrounding environment; the PMV value is further converted into the predicted PPD value of the proportion of dissatisfied personnel using the following formula:

[0066] ,

[0067] The calculation of PMV and PPD values ​​is based on real-time collected environmental parameter data and user-personalized data. The dynamic comfort index is used to construct the comfort reward item R_comfort for deep reinforcement learning. Its reward mechanism is as follows: when the PPD value is lower than 10%, a positive reward is given; when the PPD value exceeds the threshold, a negative penalty is imposed by ΔPPD / 15.

[0068] Step S204: Association rule mining. The improved FP-Growth algorithm is used to analyze the historical operation data collected in step S201 to mine multidimensional association rules between environmental parameters, energy consumption indicators and comfort index, and output a set of high-frequency association rules.

[0069] Step S205: Intelligent Decision-Making. The energy efficiency assessment index calculated in Step S202 and the dynamic comfort index generated in Step S203 are used as the core components of the reward function. The high-frequency association rule set output in Step S204 is used as prior knowledge or state space dimensionality reduction input to construct an intelligent learning algorithm. This algorithm is a deep reinforcement learning model that learns online and generates an optimal control strategy that balances energy efficiency and comfort through interaction with the cold source system environment. The reward function of the deep reinforcement learning algorithm assigns the energy consumption target reward term to the negative logarithm of the standard coal consumption (kgce / h). Simultaneously, when the FP-Growth association rule mining result in Step S204 triggers a specific emergency operating condition rule, the deep reinforcement learning algorithm intelligently initiates the emergency control strategy immediately. Furthermore, the deep reinforcement learning model employs a priority experience replay mechanism, assigning a 3x sampling weight to the emergency operating condition samples triggered by FP-Growth to accelerate strategy convergence.

[0070] Step S206: Control command generation and execution. The optimal control strategy generated in step S205 is converted into specific equipment control parameters and sent to the actuator of the data acquisition and control module to adjust the cold source equipment parameters such as compressor frequency, electronic expansion valve opening, and water pump flow.

[0071] The above description is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in the claims, they should all fall within the protection scope of the present invention.

Claims

1. A multi-parameter intelligent control system for cold sources based on big data, characterized in that, include: The data acquisition and control module is used to collect energy consumption parameters, environmental parameters, personnel activity parameters, and equipment operating status parameters of the cold source system, and execute control commands. The integrated energy consumption standardization module, connected to the data acquisition and control module, is used to receive energy consumption parameters, convert them into unified standardized energy consumption indicators, and output energy efficiency assessment indicators. The comfort algorithm construction module is connected to the data acquisition and control module. It is used to receive the environmental parameters and personnel activity parameters, generate a dynamic comfort index, and construct a dynamic comfort reward item for deep reinforcement learning. The association rule mining module is connected to the data acquisition and control module, the comprehensive energy consumption standardization module, and the comfort algorithm construction module. It uses an improved FP-Growth algorithm to mine multidimensional association rules between the environmental parameters, standardized energy consumption indicators, and comfort index from historical data. The association rules introduce a dynamic minimum support threshold function and use matrix compression storage, and the mined rules are used as prior knowledge for the intelligent decision-making module. The intelligent decision-making module is connected to the comprehensive energy consumption standardization module, the comfort algorithm construction module, and the association rule mining module. It uses a deep reinforcement learning algorithm to construct a reward function based on the standardized energy consumption index and the dynamic comfort index, and combines the association rules to generate the optimal control strategy. The control command generation module, connected to the intelligent decision-making module and the data acquisition and control module, is used to convert the optimal control strategy into equipment regulation parameters and send them to the data acquisition and control module to drive the cold source system to execute the control commands.

2. The multi-parameter intelligent cold source control system based on big data according to claim 1, characterized in that, The data acquisition and control module includes: Energy metering units are used to monitor the consumption of electricity, natural gas, and water resources; Thermodynamic state monitoring unit is used to monitor the three-dimensional temperature field of the evaporator and condenser, as well as the phase change state of the refrigerant; The environmental sensing unit is used to monitor indoor light intensity, indoor carbon dioxide concentration, and personnel density and distribution.

3. The multi-parameter intelligent cold source control system based on big data according to claim 1, characterized in that, The comprehensive energy consumption standardization module constructs a unified multi-energy evaluation system to convert heterogeneous energy sources such as electricity, natural gas, and water into standard coal equivalents according to their calorific value. The unified multi-energy evaluation system is based on a dynamic conversion model, which is: E=Σ(α_i×Q_i), where E represents the total standard coal equivalent, α_i represents the conversion factor of the i-th energy source, and Q_i represents the energy consumption of the i-th energy source.

4. The multi-parameter intelligent cold source control system based on big data according to claim 1, characterized in that, The comprehensive energy consumption standardization module outputs the energy efficiency evaluation index in real time as: R_t=-(E_t-E_{min}) / E_{min}, where E_t represents the energy consumption standard coal equivalent at the current moment, and E_{min} represents the historical minimum energy consumption standard coal equivalent.

5. The multi-parameter intelligent cold source control system based on big data according to claim 1, characterized in that, The comfort algorithm construction module calculates the dynamic comfort index based on the PMV-PPD model, and constructs a comfort reward item based on the dynamic comfort index. When the predicted proportion of dissatisfied personnel PPD value is lower than the preset threshold, a positive reward is given; when the predicted proportion of dissatisfied personnel PPD value is higher than the preset threshold, a negative reward is given.

6. The multi-parameter intelligent cold source control system based on big data according to claim 1, characterized in that, The association rule mining module uses an improved FP-Growth algorithm to mine high-frequency association rule sets and extract non-explicit performance-intensive association rules.

7. The multi-parameter intelligent cold source control system based on big data according to claim 1, characterized in that, The intelligent decision-making module uses a deep reinforcement learning algorithm, with energy consumption and PMV comfort index as reward functions, to output compressor frequency and electronic expansion valve opening commands.

8. The multi-parameter intelligent cold source control system based on big data according to claim 7, characterized in that, The reward function R is composed of a comfort reward term R_comfort and an energy reward term R_energy, weighted by the formula: R = ω_1 × R_comfort + ω_2 × R_energy, where ω_1 and ω_2 are weighting coefficients.

9. A multi-parameter intelligent control method for cold sources based on big data, characterized in that, Including the following steps: Collect energy consumption parameters, environmental parameters, personnel activity parameters, and equipment operating status parameters of the cold source system, and execute control commands; Different types of energy consumption parameters are uniformly replaced with standardized energy consumption indicators; Based on the environmental parameters and personnel activity parameters, a dynamic comfort index is generated, and a dynamic comfort reward item for deep reinforcement learning is constructed. Based on historical operating data, an improved FP-Growth algorithm is used to mine multidimensional association rules between operating environment parameters, standardized energy consumption indicators and comfort index. These association rules introduce a dynamic minimum support threshold function and use matrix compression storage, and the mined rules are used as prior knowledge for the intelligent decision-making module. A deep reinforcement learning algorithm is used to construct a reward function based on the standardized energy consumption index and the dynamic comfort index, and combined with the association rules, an optimal control strategy is generated through an intelligent learning algorithm. The optimal control strategy is then translated into specific equipment control parameters and executed.

10. The multi-parameter intelligent control method for cold sources based on big data according to claim 9, characterized in that, The intelligent learning algorithm is a deep reinforcement learning algorithm; when the mined association rules trigger a specific emergency situation, an emergency control strategy is activated; the deep reinforcement learning algorithm adopts a priority experience replay mechanism to assign higher weights to the training samples corresponding to the emergency situation.

Images