A User-Side Power Load Response Control Method Considering Building Energy Storage

Abstract

This invention belongs to the field of building energy management technology, specifically relating to a user-side power load response control method that considers building energy storage. By real-time monitoring and analysis of user demand and power system load, it dynamically adjusts temperature-controlled power load to optimize energy consumption and balance power load. Through the integration of advanced sensors and IoT devices, it can accurately capture user comfort needs and environmental changes, and make flexible adjustments based on real-time fluctuations in power load. During periods of high power demand, the intelligent agent can intelligently optimize temperature load settings to reduce energy consumption; while during periods of low electricity prices, it automatically restores a comfortable temperature, thereby optimizing load output.

Description

Technical Field

[0001] This invention belongs to the field of building energy management technology, specifically relating to a user-side power load response control method that takes into account building energy storage. Background Technology

[0002] In existing technologies, temperature-controlled power load management for electricity demand response mainly includes the following representative solutions:

[0003] Traditional direct load control is a common approach. This method typically relies on power companies or system operators to remotely adjust temperature-controlled electricity load during peak grid load periods, or to forcibly set fixed temperature thresholds at the user side (e.g., locking air conditioning temperatures to 26°C or higher in summer) to quickly reduce load. The system architecture usually includes load management terminals (such as smart energy systems) deployed at the user side to collect electricity consumption data and transmit it to a master station via 4G / 5G wireless communication networks. The master station then issues one-way control commands to the terminals. This method is simple to implement and reacts quickly, but its crude control approach has significant drawbacks. Another mainstream approach is closed-loop control technology based on PID (proportional-integral-derivative) regulation. This technology uses real-time collected ambient temperature and humidity parameters as feedback, compares them with a preset target temperature, and uses a PID algorithm to calculate the temperature-controlled electricity load (e.g., adjusting chiller outlet water temperature, chilled water pump frequency, or fan speed), achieving dynamic and continuous adjustment of the air conditioning system load. Its typical architecture relies on a field-mounted programmable logic controller (PLC) to build a local control loop, connecting sensors and actuators via interfaces such as RS485. PID control can maintain a relatively stable environment in the target monitoring area, but its control parameters need to be pre-tuned and fixed. Its main focus is on stable control itself rather than load response optimization, and it is usually only suitable for adjusting a single device or a local system.

[0004] To coordinate large-scale, distributed temperature-controlled electricity loads, hierarchical aggregation-based centralized demand response systems have been developed. These systems typically follow a logical architecture of "regional control - hierarchical control - local control," aggregating distributed temperature-controlled electricity loads across different regions on a central platform such as a virtual power plant. Based on load forecasting models and system simulations (e.g., applying chiller plant efficiency optimization algorithms), the platform generates a global load regulation strategy and decomposes it into control commands for each region and level. These commands are ultimately distributed to the user-side temperature-controlled electricity load system via edge controllers for specific load adjustment operations (e.g., adjusting water valve openings, fan frequencies, or initiating chilled tank storage / melting processes). This approach achieves large-scale load management and collaborative optimization, but it is highly dependent on centralized decision-making and complex communication networks.

[0005] In addition, ice storage and load shifting solutions combining energy storage technology are also important technological directions. This solution utilizes the off-peak electricity period (low electricity price period) to operate the refrigeration unit for ice production (energy storage), and stops or reduces the operation of the refrigeration unit during the peak electricity period (high electricity price period), instead relying on ice melting to release cooling capacity to meet the temperature-controlled electricity load demand, thus achieving peak shaving and valley filling. Modern systems often use AI algorithms to optimize the ice storage rate, ice melting volume, and the operating combination strategy of the refrigeration unit and ice tank. Its core system architecture is to add intelligent monitoring and control equipment to the central cooling station, and dynamically adjust the operating strategy in conjunction with time-of-use electricity pricing. This solution is economical and suitable for large-scale centralized cooling systems. Although the above-mentioned existing technologies provide multiple ways to flexibly regulate temperature-controlled electricity load, they generally have some key defects or limitations: First, the user experience is significantly sacrificed: the direct load control method is too simple and crude, and the forced uniform setpoints and power outage operations seriously sacrifice the comfort requirements of individual users, lacking differentiated treatment for different user preferences and tolerances, which can easily cause user resistance and damage the user base for demand response implementation. Second, there is a lack of personalization and adaptability: Whether it's PID control or centralized demand response systems, their control strategies mainly rely on predefined algorithm models or centralized optimization results, typically employing a "one-size-fits-all" approach or strategies based on regional average levels. They struggle to dynamically learn and adapt to the unique usage habits, temperature sensitivity, and dynamic changes in comfort levels of specific users (such as differences in demand at different times and under different activity states), making it difficult to provide truly personalized services. Third, the response mechanism is rigid and lacks flexibility: Existing solutions have limited responsiveness to real-time fluctuations in power system signals. Direct load control has a fast response speed but its actions are singular and cannot be fine-tuned; PID closed-loop control focuses on local stability rather than global optimal response; hierarchical centralized systems have long decision chains, with time delays in strategy generation and command transmission; ice storage solutions are limited by the capacity and dynamic characteristics of ice storage equipment, potentially limiting response speed and failing to cover all load control scenarios. Overall, there is a lack of highly flexible and rapid fine-grained adjustment capabilities. Fourth, there is a lack of comprehensive economic considerations: many solutions primarily focus on load reduction itself or the stability of system operation, failing to fully integrate real-time electricity price information or effectively consider the need to minimize user-side operating costs (such as not fully utilizing off-peak electricity for equipment condition recovery or pre-cooling). While ice storage considers economics, its application scenarios are limited. Fifth, the level of system intelligence needs improvement: PID control relies on fixed parameters; the updating and prediction of centralized optimization models may have biases; and the scope of AI strategy optimization for ice storage is relatively limited. Systems generally lack the ability to continuously learn and iteratively optimize control strategies based on actual operating data, making it difficult to continuously improve long-term energy efficiency and response performance. Summary of the Invention

[0006] The technical problem to be solved by the present invention is that the existing load adjustment methods are too simplistic, cannot be adjusted for different conditions, the control strategies do not take into account all factors, and the decision-making ability of the intelligent agent is poor. The purpose is to provide a user-side power load response control method that takes into account building energy storage, thereby solving the above-mentioned technical problems.

[0007] This invention is achieved through the following technical solution:

[0008] This invention provides a user-side power load response control method considering building energy storage, specifically including the following steps:

[0009] Based on building energy storage and electricity load data, a database was constructed. Using this database, a building physics simulation model, a temperature-controlled electricity load simulation model, a building energy storage capacity model, and a building load prediction model were built. The building energy storage capacity model was used to simulate the building's heat storage and release process, heat transfer process, and temperature-controlled electricity load regulation process during different electricity price peak periods, outputting model analysis results. These results include: building energy storage capacity values, building energy storage capacity, and environmental parameter trends in the temperature-controlled electricity load energy consumption zone. An optimal energy storage strategy was generated based on building energy storage capacity and electricity price data during off-peak hours, and based on PMV (Power, Volume, and Value). - The PPD model and environmental parameter change trends generate a temperature-controlled power load reduction operation strategy; using the database, model analysis results, optimal energy storage strategy, temperature-controlled power load reduction operation strategy, output of building physics simulation model, output of temperature-controlled power load simulation model, and output of building load prediction model, the state space of the agent is generated; the action space of the agent is constructed, and the agent is trained using the state space, action space, and reward function; the trained agent is deployed to the BAS control system, and the agent is debugged based on the comparison results of baseline data and agent feedback data; the optimal control strategy is output based on the debugged agent.

[0010] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0011] By monitoring and analyzing user demand and power system load in real time, the system dynamically adjusts its operation to optimize energy consumption and balance power load. Integrating advanced sensors and IoT devices, it accurately captures user comfort needs and environmental changes (such as temperature and humidity), making flexible adjustments based on real-time fluctuations in power load. During periods of high power demand, the intelligent agent optimizes the load and reduces energy consumption; during periods of low electricity prices, it automatically restores comfortable temperatures, maximizing energy savings and reducing electricity costs. Furthermore, employing a self-learning algorithm, it progressively optimizes control strategies based on user habits and preferences, providing personalized and efficient load management. Through coordination with the power dispatching system, this user-side power load response control method, which considers building energy storage, not only improves the energy efficiency and ease of use of load equipment but also provides stable load response support for power companies while achieving energy optimization. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0013] Figure 1 This is a system block diagram of a user-side power load response control method considering building energy storage in an embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. The illustrative embodiments and descriptions of this invention are for illustrative purposes only and are not intended to limit the invention. The embodiments described below are some, but not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0015] This embodiment provides a user-side power load response control method considering building energy storage, consisting of data acquisition, an intelligent decision-making layer, and an execution feedback layer. The following is combined with... Figure 1 The data acquisition, intelligent decision-making, and execution feedback processes, as well as their respective internal functions, are explained and described separately.

[0016] I. Data Collection

[0017] Data acquisition is used to collect building thermal model parameters, operating parameters of temperature-controlled electrical load equipment, environmental parameters of temperature-controlled electrical load energy consumption areas, system model parameters of temperature-controlled electrical load, meteorological data, population density change data, and electricity price data for different time periods to establish a database.

[0018] Data acquisition is a high-precision data collection system for sensing building load equipment and the environment. By analyzing the composition of the building's energy system and combining it with the distribution of different functional rooms within the building, based on the data collected from the building's intelligent system, it further clarifies the key data collection points for the building's main temperature-controlled electrical load energy consumption areas, achieving high-precision load data collection for the building from both temporal and spatial dimensions.

[0019] The required high-precision load data includes: building thermal model parameters—building geometric parameters, functional zoning information, thermal parameters of each building envelope, and multiple thermal bridge areas. Operating parameters of temperature-controlled electrical load equipment—supply / return air temperature and humidity, air volume / fresh air volume / exhaust air volume, chilled / hot water supply and return water temperature and flow rate, valve opening, and real-time power and cumulative power consumption of equipment such as fans / pumps / chillers / heat pumps / boilers. Environmental parameters of the temperature-controlled electrical load energy consumption area—time-varying sequences of air temperature, relative humidity, CO2 concentration, wind speed, and personnel density in the monitoring area. Model parameters of the temperature-controlled electrical load system—model parameters, performance curves, and control logic of each temperature-controlled electrical load device. Meteorological data—atmospheric temperature, solar radiation intensity, wind speed, and relative humidity. Personnel density change data—access control system data or meeting reservation system data for each building. Power grid data—real-time electricity spot price (including time-of-use price curves) and building-side load forecast data. The load forecast data is the output of the building load forecast model (see "Intelligent Decision Layer—Fourth Model Building Module"). The building load forecast data was used in subsequent agent training and policy solving to jointly optimize the constraints of comfort and energy consumption costs.

[0020] Building a data collection system involves the following five deployment steps:

[0021] Firstly, high-precision sensors (including temperature sensors, humidity sensors, CO2 concentration sensors, and wind speed sensors) are deployed in various temperature-controlled electrical load energy consumption areas of the building (such as large open-plan target areas and multi-functional large conference rooms) to expand the dimensions of data collection and achieve high-precision load data collection based on building time and space. Secondly, independent metering instruments (including electricity acquisition instruments, power acquisition instruments, and flow acquisition instruments) are installed in each temperature-controlled electrical load energy consumption area to achieve zonal statistics and analysis of energy consumption data. Thirdly, personnel density change data is obtained through the building's access control system or conference reservation system to provide richer data support for modeling temperature-controlled electrical load energy consumption areas. Fourthly, building thermal model parameters (including building geometric parameters, functional zoning information, thermal parameters of each building envelope, and multiple thermal bridge areas) are obtained by analyzing building drawings to provide data support for building a building physical simulation model. Fifthly, model parameters, performance curves, and control logic of each temperature-controlled electrical load device in the temperature-controlled electrical load system are collected to provide data support for building a temperature-controlled electrical load simulation model.

[0022] Based on the various sensors and measuring instruments mentioned above, the data acquisition dimensions include spatial and temporal dimensions.

[0023] In terms of spatial dimensions: the deployment of various sensors and instruments follows the building's functional zoning and floor structure, ensuring that each temperature-controlled power load energy consumption zone is covered. The collected data includes operating parameters of temperature-controlled power load equipment, power grid data, and environmental parameters of the temperature-controlled power load energy consumption zone. Simultaneously, it integrates meteorological data (including atmospheric temperature, solar radiation intensity, wind speed, and relative humidity), personnel density changes during the operation of temperature-controlled power load equipment, and meeting reservation system data to construct a multi-dimensional load data acquisition system. This provides a comprehensive and accurate data foundation for load forecasting, enabling the incorporation of high-precision data collection needs into planning. In actual operation, the system dynamically adjusts the acquisition strategy through an intelligent sensor network, breaking down the data bottleneck from energy dispatch to load forecasting.

[0024] In terms of time: For the first temperature-controlled power load energy consumption zone (meeting rooms, densely populated target areas), the data collection frequency is on the order of seconds; for the second temperature-controlled power load energy consumption zone (all other temperature-controlled power load energy consumption zones besides the first one), the data collection frequency is on the order of minutes, achieving high-precision monitoring of temperature-controlled power load energy consumption zones. Simultaneously, a real-time electricity price data acquisition module is connected to collect electricity spot prices in real time, providing data support for subsequent electricity price-based strategy formulation.

[0025] It should be noted that the division between the first and second temperature-controlled power load energy consumption zones is based on personnel density. ρ or space occupancy rate kpopulation density ρ Threshold and space occupancy rate k The threshold can be determined based on the spatial size of different temperature-controlled power load energy consumption zones. For example, in this embodiment, the densely populated target area can be defined as... ρ ≥0.15 or k ≥0.7. The remaining temperature control power load energy consumption areas, excluding the first temperature control power load energy consumption area, are defined as the second temperature control power load energy consumption area.

[0026] II. Intelligent Decision Making

[0027] Intelligent decision-making includes four aspects: First, building a building physics simulation model, a temperature-controlled electricity load simulation model, a building energy storage capacity model, and a building load prediction model; second, analyzing the models to obtain the building energy storage capacity value, building energy storage capacity, and the environmental parameter change trends of the temperature-controlled electricity load energy consumption area; third, generating energy storage strategies and operation strategies based on the model analysis results; and fourth, developing intelligent agents.

[0028] The following sections will explain each functional module of the intelligent decision-making layer in light of the four aspects of the work described above.

[0029] (a) Establish building physics simulation model, temperature control power load simulation model, building energy storage capacity model and building load prediction model.

[0030] On the one hand, high-precision building physics simulation models and temperature-controlled electricity load simulation models are built to accurately reflect the annual dynamic load variation characteristics of buildings and the operating status of temperature-controlled electricity load systems under different control strategies. The models analyze the energy consumption characteristics and influencing factors of different temperature-controlled electricity load energy consumption zones, providing a basis for formulating targeted energy management strategies and achieving precise control over the energy consumption zones of load equipment in the target monitoring area of ​​the building. On the other hand, the building energy storage capacity model, based on real-time meteorological data (hourly temperature, solar radiation intensity, and relative humidity), dynamically simulates the heat storage and release processes of building structures such as walls, roofs, and floors, as well as the heat transfer between the air and the building envelope in the target monitoring area. This allows for accurate analysis of the building's energy storage capacity based on the simulation results. Furthermore, by building a building load prediction model, a correlation model between electricity prices and load impact is formed.

[0031] This embodiment uses a model building module to build a building physics simulation model, a temperature-controlled electricity load simulation model, a building energy storage capacity model, and a building load prediction model. Specifically, the model building module includes the following functions:

[0032] 1. Building Physics Simulation Model Construction: The building physics simulation model is constructed using building thermal model parameters, operating parameters of temperature-controlled electrical load equipment, and environmental parameters of the temperature-controlled electrical load energy consumption zone. Specific methods for constructing the building physics simulation model can be found in paragraphs

[0013] to

[0038] of the specification of the invention patent "Model-Data Dual-Driven Method for Creating a Training Environment for Building Energy Management Intelligent Agents" (publication number CN120180575B).

[0033] 2. Construction of a Temperature-Controlled Electrical Load Simulation Model: This model is used to construct a temperature-controlled electrical load simulation model using the parameters of the temperature-controlled electrical load system model. Specific methods for constructing the temperature-controlled electrical load simulation model can be found in paragraphs

[0039] to

[0061] of the invention patent "Model-Data Dual-Driven Method for Creating a Training Environment for Building Energy Management Intelligent Agents" (publication number CN120180575B). The temperature-controlled electrical load simulation model can be a heating, ventilation, and air conditioning (HVAC) system simulation model or other models capable of performing the same function.

[0034] 3. Building Physics Simulation Model Calibration: Building physics simulation model calibration is used to calibrate the building physics simulation model using real-time building operation data.

[0035] The real-time building operation data mentioned in this embodiment refers to the physical quantities and operating parameters collected in real time by on-site sensors, monitoring systems, or operation management systems during actual operation. These data reflect the true operating status of the building and can be categorized as follows:

[0036] (1) Measured environmental data of the target monitoring area, including: measured temperature data, measured humidity data, measured CO2 concentration data, and measured wind speed / direction data. Among them, measured temperature data includes temperature data of temperature measuring points at different locations in the target monitoring area (such as return air vents, work areas, and near walls); measured humidity data includes humidity data of relative humidity measuring points (such as air conditioning supply vents and the center point of typical rooms); measured CO2 concentration data is used to reflect personnel activity and ventilation levels; measured wind speed / direction data includes wind speed and direction at the main airflow channels or supply vent locations in the target monitoring area. The target monitoring area can be indoors or an area designated for monitoring.

[0037] (2) Real-time operating parameters of HVAC system, including: supply and return air temperature and humidity of temperature control power load system, fresh air volume, supply air volume, exhaust air volume, terminal equipment switch status and load level, heat pump / boiler / chiller operating power or cooling / heating capacity.

[0038] (3) Building energy consumption and equipment data, including: metering data of sub-items such as electricity, cooling, heating and lighting (hourly or higher frequency), and equipment operating status (start-up, shutdown, load rate).

[0039] (4) Real-time monitoring data of the external environment, including: temperature and humidity, wind speed and direction, and solar radiation intensity of the target monitoring area.

[0040] Based on the aforementioned real-time building operation data, the calibration of the building physical simulation model includes the following steps:

[0041] Step 1: Compare the real-time building operation data monitored above with the simulation output data of the building physics simulation model at the same time, calculate the deviation of various data, and form an error distribution or residual sequence as the basis for model adjustment.

[0042] Step 2: Identify the source of the deviation and generate deviation correction items. For example, if the deviation is concentrated in the thermal response of the building envelope, it indicates that the thermal conductivity, transmittance, etc. of the material need to be corrected; if the deviation is concentrated in the temperature control electrical load, it indicates that the internal heat, flow rate, or equipment power needs to be corrected; if the deviation is concentrated in the wind speed distribution, it indicates that the air supply volume or CFD boundary conditions need to be corrected; if the deviation changes in a consistent trend over time but in different magnitudes, it indicates that the control strategy parameters or system efficiency need to be corrected.

[0043] Step 3: Based on the deviation correction terms, perform online or periodic corrections on key parameters in the building physics simulation model one by one. For example: adjust the equivalent thermal resistance of the walls or the transmittance of the windows; update the internal load curves; correct operating boundaries such as supply air temperature and air volume; and correct the control setpoint and system response delay. It should be noted that deviation correction can employ the direct substitution method (replacing model parameters with measured values) or the parameter inversion / optimization method. That is, the optimal parameters are identified by minimizing the error between the simulation output and the measured data (e.g., least squares, Kalman filtering, Bayesian inversion).

[0044] Step 4: Re-input the corrected parameters into the building physics simulation model, perform simulation iterations, and return to Step 1 until the error meets the set convergence criteria. A mechanism for periodic (e.g., daily / hourly) automatic updates can also be set to achieve dynamic online calibration.

[0045] 4. Calibration of the temperature control electrical load simulation model

[0046] The calibration of the temperature-controlled electrical load simulation model is used to calibrate the temperature-controlled electrical load simulation model using real-time operating data of the temperature-controlled electrical load system. Real-time operating data of the temperature-controlled electrical load system refers to the operating parameters and environmental response data of the target monitoring area collected in real time by field sensors, monitoring systems, or operation management systems during the actual operation of the temperature-controlled electrical load system. This data reflects the true operating status of the temperature-controlled electrical load system and can be categorized as follows:

[0047] (1) Environmental measured data

[0048] The measured environmental data reflects the "results" of the temperature-controlled electrical load system's operation, including: air temperature and relative humidity in the target monitoring area, temperature, humidity, and solar radiation in the target monitoring area, room CO2 concentration (reflecting personnel activity and ventilation needs), and wind speed and airflow field in the target monitoring area (key supply air vents, return air vents, and work areas). The measured environmental data is compared with the model output of the temperature-controlled electrical load simulation model. The comparison results are used to assess whether the environmental state after the temperature-controlled electrical load system's operation has met expectations.

[0049] (2) Operating parameters of temperature-controlled electrical load equipment

[0050] The operating parameters of temperature-controlled electrical load equipment are the key boundary conditions and control inputs of the temperature-controlled electrical load simulation model. These include: air supply system data (supply air temperature, supply air humidity, supply air volume, air outlet velocity, return air temperature, return air humidity); cold and heat source system data (chilled water supply and return water temperature and flow rate, hot water / steam supply and return temperature and flow rate, actual cooling / heating capacity of equipment such as chillers, heat pumps, and boilers); and operating data of terminal equipment (such as fan coil units and air conditioning units), including: fan frequency, valve opening, control signals of electric regulating valves or frequency converters, and inlet and outlet water temperatures and flow rates of the coils. These data directly determine the "boundary conditions" and "control strategies" of the temperature-controlled electrical load simulation model.

[0051] (3) Energy consumption and system efficiency data

[0052] This includes: system component power consumption (such as the energy consumption of the temperature control equipment main unit, cooling water pump, supply and exhaust fan, and terminal equipment), calorimeter or chiller readings, COP (coefficient of performance), and equipment load rate. These data are used to verify whether the energy consumption prediction of the temperature control power load simulation model is consistent with the actual energy consumption, and are an important basis for calibration.

[0053] (4) Control and scheduling information

[0054] This includes: control strategies for setting temperature, humidity, and fresh air volume (set value, variable value, energy-saving mode), equipment start-up and shutdown plans, load distribution strategies, ventilation modes for target monitoring areas, and nighttime ventilation strategies.

[0055] Based on the real-time operation of the temperature-controlled electrical load system described above, the calibration of the temperature-controlled electrical load simulation model includes the following steps:

[0056] Step 1: Compare the real-time collected temperature, humidity, energy consumption, and supply air temperature of the target monitoring area with the corresponding indicators output by the temperature control power load simulation model hourly, calculate the temperature deviation, humidity deviation, energy consumption deviation, and supply air temperature deviation, and obtain the deviation distribution of the time series to provide a basis for subsequent parameter correction.

[0057] Step 2: Based on the type of deviation, identify the model parameters that cause the model error and generate deviation correction items. For example, if there is a deviation in the equipment performance parameters, it means that the chiller COP, heat exchange efficiency, and supply air temperature control deviations need to be corrected; if there is a deviation in the air volume and water volume, it means that the supply air volume, return air ratio, and coil water flow need to be corrected; if there is a deviation in the control strategy, it means that the control logic, regulating valve hysteresis, and set temperature curve need to be corrected; if there is a deviation in the load estimation, it means that the personnel density and internal heat source estimation need to be corrected.

[0058] Step 3: Correct the identified key parameters based on the deviation correction terms. For example: replace the simulation setpoint with the actual supply air temperature, adjust the chiller part-load performance curve, and correct the duct resistance coefficient to match the actual air volume.

[0059] Similarly, direct replacement (simple real-time replacement of key boundary parameters) or optimization inversion (such as automatic adjustment of model parameters based on least squares, Kalman filtering or Bayesian inversion) can be used.

[0060] Step 4: Input the corrected parameters into the temperature control electrical load simulation model, perform simulation iterations, and return to Step 1 until the error meets the set convergence criteria. A mechanism for periodic (e.g., daily / hourly) automatic updates can also be set to achieve dynamic online calibration.

[0061] 5. Building Energy Storage Capacity Model Construction

[0062] Building energy storage capacity model construction is used to build a building energy storage capacity model using meteorological data.

[0063] The energy storage capacity model described in this embodiment refers to a model that calculates the temporal temperature change of a building under disturbances such as external climate and internal loads through heat transfer equations and boundary conditions. A typical form of building energy storage capacity model is the RC network model (Resistance-Capacitance Network). The RC network model is a commonly used building thermal dynamic response model in the engineering field. It is a simplified physical model used to simulate the "heat transfer" and "heat storage" processes of the building envelope (such as walls, roof, and ground) and the air in the target monitoring area. Its core is to use the dynamic relationship of "current-voltage-capacitance" in a circuit to analogize the thermal characteristics of "heat flow-temperature difference-heat capacity" in a building, thereby quantifying the building's energy storage capacity (i.e., the building's ability to absorb, store, and release heat) and its buffering effect on the temperature of the target monitoring area.

[0064] Building a building energy storage capacity model based on meteorological data includes the following steps:

[0065] Step 1: Preprocess meteorological data. The original meteorological data may contain "missing values ​​and outliers" (such as sudden temperature rises / falls caused by sensor failures). Missing values ​​need to be filled, outliers need to be removed, and data format needs to be converted. (1) Missing values ​​are filled. For short-term missing values ​​(such as missing values ​​for 1-2 hours), the "linear interpolation method" is used; for long-term missing values ​​(such as missing values ​​for more than 1 day), the "data from the same period in the adjacent year" is used to replace them (such as missing data on July 10, 2023, which is replaced by data from the same period on July 10, 2022, and the annual temperature deviation needs to be corrected). (2) Outlier removal. Set reasonable thresholds (such as the temperature in the target monitoring area cannot be lower than -30℃ or higher than 45℃, and the solar radiation intensity cannot exceed 1200W / m²). Data exceeding the threshold is marked as outliers; the "3σ principle" (statistical method, removing data that deviates from the mean by 3 times the standard deviation) is used to further screen outliers to ensure that the data conforms to the local meteorological laws. (3) Data format conversion. Meteorological data is converted into a format recognizable by simulation software (such as EnergyPlus's "EPW format" and TRNSYS's "TMY format"), including year / month / day / hour timestamps and corresponding parameter values, making it easier for models to call.

[0066] Step 2: Integrate the building thermal model parameters and preprocessed meteorological data into the RC network model, and realize dynamic simulation through software tools (EnergyPlus). The implementation process is as follows: Step 2.1: Select a single-node RC model, a two-node RC model, or a multi-node RC model according to the modeling objectives and data accuracy. In order to obtain high-precision energy storage analysis results, this embodiment selects a multi-node RC model. Step 2.2: Model building and parameter input. (1) Import the building thermal model parameters and preprocessed meteorological data into the EnergyPlus software. The EnergyPlus software automatically calls the meteorological parameters according to "hourly". (2) Set the simulation parameters, including: simulation time, time step and output parameters. The simulation time is a typical meteorological year (8760 hours) or an extreme period (such as the cold wave period from January 1 to 7); the time step is 1 hour / step (balancing accuracy and computational efficiency, high-precision simulation can be set to 15 minutes / step); the output parameters include: hourly temperature of each node of the building (outer layer / inner layer), hourly heat storage and hourly heat flow transfer, which are the core indicators for evaluating energy storage capacity.

[0067] Step 2.3: Coupling Meteorological Parameters with the RC Model. Meteorological parameters are dynamically linked to the RC network through a "heat balance equation." Taking the nodal model as an example, the core coupling logic is as follows: Heat balance of the outer nodes (outside the building envelope): Heat gain of the outer nodes = Solar radiation heat gain (from meteorological data) + Thermal resistance of the inner nodes. R1; Heat loss at outer nodes = convective heat dissipation from the target monitoring area; Temperature change at outer nodes = (heat gain / loss) / outer layer heat capacity C1 (core formula of the RC model). Thermal balance of intermediate nodes (internal intermediate layer of the building envelope): Heat dissipates through the internal thermal resistance of the building envelope. R 2. Heat transferred from outer nodes to middle nodes; Heat loss at middle nodes = Thermal resistance of middle nodes R 2. Heat transferred to the inner node; Temperature change of the middle node = (Heat gain - Heat loss) / Heat capacity of the middle layer C2. Thermal balance of the inner node (air in the target monitoring area + inside the enclosure structure): Heat gain of the inner node = Heat gained through the thermal resistance of the middle node. R 3. Heat transferred + heat dissipation by personnel / equipment; heat loss of inner node = air conditioning cooling capacity + heat dissipation from ventilation in the target monitoring area; temperature change of inner node = (heat gain - heat loss) / inner heat capacity C3.

[0068] Through the above dynamic equations, the model can calculate the "outer node temperature, inner node temperature, and total building heat storage" hourly, realizing a complete simulation of "meteorological conditions - heat transfer - energy storage changes".

[0069] 6. Building load prediction model construction

[0070] A building load forecasting model is built to utilize historical load data, meteorological data, and population density change data to establish a predictive model for building load. Deep learning algorithms from artificial intelligence are employed to construct the model. The collected historical load data, meteorological data, and population density change data are analyzed to establish an accurate load forecasting model covering short-term (1-24 hours), medium-term (1-7 days), and long-term (monthly, quarterly) forecasts. Simultaneously, real-time collected building operation load data is continuously input into the building load forecasting model for verification and error feedback, improving forecast accuracy. This breaks down the bottleneck in load forecasting from design to actual operation, providing reliable load forecasting data support for energy dispatch.

[0071] The system employs a "data-driven + deep learning" approach, consisting of four steps, to build a framework covering short-, medium-, and long-term predictions, as detailed below:

[0072] (1) Data acquisition and preprocessing, historical load data: collect hourly load data of temperature-controlled electrical equipment over the past year ( kW Outliers were removed (e.g., a sudden 50kW increase due to equipment failure, following the 3σ principle), and missing values ​​were filled in (linear interpolation). Meteorological data: hourly temperature (°C) and solar radiation (°C) of the target monitoring area during the same period. W / m ²), relative humidity (%), format converted to EPW; personnel density data: hourly personnel density (persons / m²) during the same period, from access control / conference system, classified by "weekdays 9:00-18:00 / other time periods".

[0073] (2) Feature engineering, extract 4 types of features (32 dimensions in total): Time features: hours (0-23), weekdays (1-7), months (1-12), holidays (0 / 1); Meteorological features: current temperature, temperature of the previous 1 / 3 / 6 hours, solar radiation, humidity; Personnel features: current density, density change rate of the previous 1 hour; Load features: load of the previous 1 / 6 / 12 / 24 hours, load of the previous 7 days.

[0074] (3) Model construction: LSTM (Long Short-Term Memory Network) is adopted, and the structure is as follows: Input layer: 32-dimensional features; Hidden layer: 2 layers, each with 128 neurons, activation function ReLU; Output layer: short-term (1-24 hours) output 24 load values, medium-term (1-7 days) output 7 daily average load values, long-term (monthly) output 1 monthly average load value; Loss function: MAE (mean absolute error) is selected, and the optimizer is Adam.

[0075] (4) Training and optimizing the model, dataset division: 70% training set, 15% validation set, 15% test set; training parameters: epochs=100, batch_size=32, early stopping method (stop if val_loss does not decrease for 5 consecutive rounds); hyperparameter tuning: grid search to optimize learning rate (0.001-0.01) and number of hidden layer neurons (64-256).

[0076] (II) The model analysis yields the building's energy storage capacity, energy storage capacity, and environmental parameter trends for the temperature-controlled electricity load energy consumption zone. This is achieved through the model analysis module. The model analysis module simulates the building's heat storage and release process, heat transfer process, and temperature-controlled electricity load regulation process during different electricity price peak periods using the building's energy storage capacity model, outputting the model analysis results. The model analysis results include: the building's energy storage capacity, energy storage capacity, and environmental parameter trends for the temperature-controlled electricity load energy consumption zone. Specifically, the model analysis module includes the following functions:

[0077] 1. Simulation of heat storage and release and heat transfer

[0078] The heat storage and heat transfer simulation is used to simulate the heat storage and heat transfer processes of a building using a building energy storage capacity model, and outputs the building's energy storage capacity value. The building energy storage capacity model (based on an RC network model) achieves the simulation of heat storage and heat transfer processes and the quantification of energy storage capacity through three steps: "parameter input - sub-stage simulation - energy calculation," as detailed below:

[0079] (1) Model input parameter preparation. The input parameters include two types of core data. One type is building thermal parameters: the thermal resistance of the building envelope (walls, roof, ground) is obtained from the analysis of building drawings. R (such as exterior walls) R=1.5 m 2 · K / W ), heat capacity C (such as concrete roof) C =180 kJ / m 2 · K ), and the equivalent heat capacity of air and furniture in the target monitoring area. C air (such as target area) C air =50 kJ / m 2 · K Another type is preprocessed meteorological data: hourly target monitoring area temperature (e.g., 25-35℃ in summer) and solar radiation intensity (e.g., 0-1000W / m²) after missing value completion (linear interpolation), outlier removal (3σ principle), and format conversion (EPW format). m ²), wind speed (e.g., 0.5-2) m / s).

[0080] (2) Simulation of heat storage and heat transfer process based on RC network model. A multi-node RC model (outer node: outside of the building envelope; middle node: inside of the building envelope; inner node: air in the target monitoring area) is adopted. Dynamic simulation is performed through heat balance equation: In the building thermal RC network model (multi-node), the heat transfer of the outer node (outer of the building envelope), middle node (inside of the building envelope), and inner node (air in the target monitoring area and near the wall area) follows the dynamic cycle logic of "temperature difference drives heat flow - heat flow changes node temperature - new temperature difference generates new heat flow". Combining heat balance equation, Fourier's law of thermal conduction and engineering simulation method, the complete heat transfer calculation process can be divided into seven steps: "parameter initialization - outer heat balance calculation - outer-middle heat transfer - middle heat balance calculation - middle-inner heat transfer - inner heat balance calculation - iterative update", as follows: Step 1: Initialization of core parameters (taking the exterior wall of an office building in summer as an example) (as shown in Table 1):

[0081]

[0082] Step 2: Thermal balance calculation of outer node (finding) T_Outer (Real-time value)

[0083] The temperature change of the outer node is determined by the net difference between "total heat" and "total heat loss". First, calculate the net heat of the outer layer, and then solve for the real-time temperature using the heat balance equation. T_Outer .

[0084] 1) Calculate the total heat of the outer nodes (Q_Outer_In ),

[0085] The heat gain in the outer layer comes from three types of inputs from the environmental conditions of the target monitoring area, ranked by "heat flux density". q_sol "× heat transfer area" is converted into total heat (heat transfer area) A =100m², unified unit is kJ / h: 1W=3.6kJ / h): solar radiation heat gain Q_sol : q_sol × A ×3.6=700W / m²×100m²×3.6=252000kJ / h; Convective heat gain in the target monitoring area Q_conv_out : Convection heat transfer coefficient h_out =15W / (m²•K), heat flux density q_conv_out = h_out ×( T_out - T_Outer 0); Total calories: Q_ conv_out = q_conv_out × A ×3.6=15×(35-33)×100×3.6=10800kJ / h; Long-wave radiation heat gain Q_ lwr_out The simplified calculation is the difference between the long-wave radiation of the target monitoring area and the outer radiation. In engineering, a fixed value is often used (e.g., Q_lwr_out =5000 kJ / h); In summary: Q_Outer_In = Q_sol + Q_conv_out + Q_lwr_out =252000+10800+5000=267800kJ / h.

[0086] 2) Calculate the total heat loss at the outer nodes ( Q_Outer_Out )

[0087] The heat loss in the outer layer is only the conductive heat flow towards the middle node (because the outer layer temperature is higher than the middle layer in summer, there is no reverse heat flow). Calculated according to Fourier's law: Conductive heat flux density: q_cond_Outer - Middle =( T_Outer 0- T_Middle 0) / R 1 = (33-29) / 2 = 2W / (m²•K); Total heat loss through conduction: Q_cond_Outer - Middle = q_cond_Outer - Middle × A ×3.6 = 2 × 100 × 3.6 = 720 kJ / h; In summary: Q_Outer_Out = Q_cond_Outer - Middle =720kJ / h (no other heat loss pathways).

[0088] 3) Solve for the real-time temperature of the outer layer T_Outer

[0089] According to the heat balance equation: C_Outer ×(T_Outer - T_Outer 0)=( Q_Outer_In - Q_Outer_Out )×Δ t Substitute the data: 2112kJ / K×( T_Outer -33℃)=(267800-720)kJ / h×0.25h=66770; T_Outer -33 = 66770 / 2112 ≈ 31.6℃; T_Outer ≈33+31.6=64.6℃ (In summer, solar radiation is strong at noon, and the outer temperature is higher than the temperature of the target monitoring area).

[0090] Step 3: Calculate the heat transfer between the outer and middle layers (determine the heat transfer in the middle layer).

[0091] Heat transfer between the outer and middle layers is determined by the "real-time temperature of the outer layer". T_Outer With the initial temperature of the middle layer T_Middle Driven by the "difference of 0", following Fourier's law of thermal conduction, the calculation result is the core heat of the middle node.

[0092] Outer and middle layers conduct heat flow q_cond_Outer - Middle =( T_Outer - T_Middle 0) / R 1 = (64.6 - 29) / 2 = 17.8 W / (m²•K); Total conductive heat flux (heat in the middle layer) Q_Outer_Middle = q_cond_Outer - Middle × A ×3.6=17.8×100×3.6=6408kJ / h; Note: Because T_Outer (64.6℃) is much higher than T_Middle 0 (29℃), heat flow direction from the outer layer to the middle layer, Q_Outer — Middle This represents the total heat in the middle layer (without other heat sources).

[0093] Step 4: Thermal balance calculation of mid-level nodes (to find T_Middle (Real-time value)

[0094] The temperature change at the middle node is determined by the net difference between the heat flow input to the outer layer and the heat flow output to the inner layer. First, calculate the net heat in the middle layer, and then solve for the real-time temperature. T_Middle This provides the driving force for the transfer of heat flow from the middle layer to the inner layer.

[0095] 1) Calculate the total heat loss of the middle layer nodes ( Q_Middle_Out )

[0096] The heat loss in the middle layer refers only to the conductive heat flow to the inner nodes, calculated according to Fourier's law (based on the initial temperature of the middle layer). T_ Middle 0 and the initial temperature of the inner layer T_Inner 0): Conductive heat flux density q_cond_Middle - Inner =( T_Middle0 - T_ Inner0) / R 2 = (29-26) / 0.025 = 120 W / (m²•K); Total heat loss through conduction: Q_cond_Middle - Inner = q_cond_ Middle - Inner × A ×3.6 = 120 × 100 × 3.6 = 43200 kJ / h; In summary: Q_Middle_Out = Q_cond_Middle - Inner =43200kJ / h.

[0097] 2) Solve for the real-time temperature of the middle layer T_Middle

[0098] According to the heat balance equation: C_Middle ×( T_Middle - T_Middle 0)=( Q_Outer — Middle - Q_ Middle_Out )×Δ t Substitute the data: 240kJ / K×( T_Middle -29℃)=(6408-43200)kJ / h×0.25h; 240×( T_Middle -29) = (-36792) × 0.25 = -9198; T_Middle -29 = -9198 / 240 ≈ -38.3℃; T_Middle ≈29-38.3=-9.3℃ (This is an extreme example. In reality, the thermal resistance of the middle layer R1 is much greater than R2, and the heat flow input from the outer layer is insufficient to offset the heat loss to the inner layer. Dynamic iterative correction is required. The actual temperature of the middle layer in summer is usually 25-30℃).

[0099] Step 5: Calculate the heat transfer between the middle and inner layers (determine the heat transfer in the inner layer)

[0100] Heat transfer between the middle and inner layers is determined by the "real-time temperature of the middle layer". T_Middle With the initial temperature of the inner layer T_Inner Driven by the "difference of 0", the calculation result is the main heat of the inner node (the heat dissipation of personnel / equipment in the target monitoring area needs to be superimposed).

[0101] 1) Calculate the heat flow conducted between the middle and inner layers ( Q_Middle — Inner )

[0102] Conductive heat flux density q_cond_Middle - Inner =( T_Middle - T_Inner 0) / R2=(-9.3-26) / 0.025=-1412W / (m²•K); Total conductive heat flux Q_Middle — Inner = q_cond_Middle - Inner × A×3.6 = (-1412) × 100 × 3.6 = -508320 kJ / h; Note: The negative sign indicates that the heat flow direction is from the inner layer to the middle layer (because the real-time temperature of the middle layer is lower than that of the inner layer). In reality, the temperature of the middle layer is higher than that of the inner layer in summer. Q_Middle—Inner It should be a positive value (heat in the inner layer); the extreme values ​​here are only for demonstration purposes of the calculation logic.

[0103] 2) The superimposed heat from other inner layers ( Q_Inner_Other )

[0104] The inner heat gain also includes the heat dissipation from personnel and equipment in the target monitoring area (e.g., for 10 people working in an office, the heat dissipation per person is 100W, and the heat dissipation from equipment is 500W, totaling...). Q_Inner_Other =(10×100+500)×3.6=5400kJ / h); In summary, the total heat of the inner layer is: Q_Inner_In = Q_Middle — Inner + Q_Inner_Other (summer Q_Middle — Inner (Positive value, heat gain of inner layer is superimposed).

[0105] Step 6: Calculation of thermal balance at inner nodes (to find) T_Inner (Real-time value)

[0106] The temperature change of the inner node is determined by the net difference between "total heat and total heat loss" (heat loss is mainly due to air conditioning cooling capacity). The final result directly reflects the thermal comfort state of the target monitoring area and is also the core basis of the air conditioning control strategy in the patent.

[0107] 1) Calculate the total heat loss of the inner nodes ( Q_Inner_Out )

[0108] The heat loss in the inner layer is mainly due to the cooling capacity of the air conditioner. Q_ac According to the "load reduction operation strategy" in the patent (e.g., 30% load reduction in summer, rated cooling capacity...), Q_ac_rated =100000kJ / h, then Q_ac =70000 kJ / h); In summary: Q_Inner_Out = Q_ac =70000kJ / h.

[0109] 2) Solve for the real-time temperature of the inner layer T_Inner

[0110] According to the heat balance equation: C_Inner ×( T_Inner - T_Inner 0)=( Q_Inner_In -Q_ Inner_Out )×Δt (assuming summer) Q_Inner_In=15000kJ / h, substituting the data): 863.6kJ / K×(T_Inner-26℃)=(15000-70000)kJ / h×0.25h; 863.6×(T_Inner-26)=(-55000)×0.25=-13750; T_Inner-26=-13750 / 863.6≈-15.9℃; T_Inner≈26-15.9=10.1℃ (This is an extreme example. In real scenarios, the air conditioner's cooling capacity will dynamically match the heat, and the inner temperature will be maintained at 24-28℃).

[0111] Step 7: Dynamic iterative updates to achieve continuous heat transfer simulation

[0112] The above calculations represent a static process with a single time step (15 minutes). Actual heat transfer is continuous and dynamic, requiring iteration according to the following logic:

[0113] 1) The calculation for this time T_Outer (64.6℃) T_Middle (-9.3℃) T_Inner (10.1℃) is used as the "initial temperature" for the next time step; 2) Environmental parameters for the next time step are collected (e.g., after 15 minutes). T_out =34℃, G =950W / m²); 3) Repeat steps 2 to 6 to calculate the temperature of the three-layer nodes at the next moment; 4) Iterate hourly according to the time step (e.g., 96 steps a day, 8760 steps a year) to obtain the complete heat transfer time series of the outer layer-middle layer-inner layer (e.g., the inner layer temperature rises slowly during the day and falls slowly at night in summer).

[0114] (3) Calculate the building's energy storage capacity

[0115] Energy storage capacity is defined as the maximum amount of heat a building can store under a unit temperature change, and the calculation formula is: Q max =( C_ Outer + C_Middle + C_Inner )×Δ T allow , where Δ T allow The allowable temperature fluctuation range for the target monitoring area (e.g., Δ) during summer when the temperature is 24-28℃. T allow (4℃).

[0116] It should be noted that: Q max This refers to "available energy storage capacity," not "inherent energy storage capacity." Inherent energy storage capacity corresponds to the building's total heat capacity. Ctotal = C outer + C middle + C in Energy storage capacity (or energy storage capacity) is an inherent property of a building (related only to materials and volume), measured in kJ / K, and reflects the amount of heat that can be stored per unit temperature change. Usable energy storage capacity is the maximum amount of heat a building can actually store while still meeting human comfort requirements, and is subject to Δ... T allow (The allowable temperature fluctuation range of the target monitoring area, such as 24-28℃ in summer, Δ) T allow =4℃) Limit - If temperature fluctuation exceeds Δ T allow Even if the building has heat storage capacity, it cannot be used (it will cause discomfort). Therefore Q max It refers to available capacity, not inherent capacity, and Δ T allow This is because comfort constraints need to be taken into account.

[0117] For example: If C outer + C middle + C in =18000+150+5000=23150 kJ / K ,but Q max =23150×4=92600 kJ That is, the building's maximum energy storage capacity is 92,600. kJ .

[0118] Furthermore, lowering the temperature only increases the available energy storage capacity without changing the inherent energy storage capacity: on the one hand, lowering the temperature—Δ T allow Increase (e.g., change from 26-28℃ to 22-28℃, Δ) T allow (from 2℃ to 6℃), then Q max = C total ×Δ T allow Increase; on the other hand, the total heat capacity of the building C total The inherent energy storage capacity remains unchanged; only the available energy storage space increases, not the energy storage capacity itself is enhanced.

[0119] 2. Full-load operation scenario simulation

[0120] Full-load operation scenario simulation is used to simulate the full-load operation of temperature-controlled electrical load equipment during off-peak electricity price periods using a building energy storage capacity model, and output the building's energy storage capacity. Specifically, it includes the following functions:

[0121] (1) Multi-strategy operation simulation: Multi-strategy operation simulation is used to simulate the operation scenarios of temperature-controlled electrical load equipment under various energy storage strategies using the building energy storage capacity model. The energy storage strategy refers to: starting the temperature-controlled electrical load equipment to operate at full load for a preset period of time before the start of off-peak electricity prices at night. Taking the cooling mode as an example, the building energy storage capacity model is used to simulate that the temperature-controlled electrical load equipment is started 1-2 hours in advance before the start of off-peak electricity prices at night and operates at 100% cooling capacity.

[0122] (2) Temperature data acquisition: Temperature data acquisition is used to collect virtual temperature data output by the building energy storage capacity model in real time.

[0123] Specifically, the building's energy storage capacity model monitors temperature changes in various areas in real time by using virtual temperature sensors placed in rooms on different floors and facing different directions to obtain hourly temperature data.

[0124] (3) Energy storage acquisition: Energy storage acquisition is used to draw the temperature-time change curve under each energy storage strategy using virtual temperature data, and to obtain the building energy storage under each energy storage strategy based on the temperature-time change curve.

[0125] Taking the energy storage strategy of "air conditioning pre-cooling at full load during off-peak electricity price periods (22:00-24:00)" as an example, the specific steps are as follows:

[0126] 1) Obtain the temperature-time curve, simulate it using the building energy storage capacity model, output the temperature change data of the target monitoring area over time under this strategy (as shown in Table 2), and plot the temperature-time curve (the horizontal axis is time, and the vertical axis is the temperature of the target monitoring area).

[0127]

[0128] 2) Calculate the cumulative temperature change. Using the initial temperature (22:00, 26.0℃) as the baseline, calculate the temperature change Δ for each time interval. T i And sum them up to get the cumulative temperature change Δ T total 22:00-22:30: Δ T 1=26.0-24.5=1.5℃; 22:30-23:00: Δ T 2=24.5-23.2=1.3℃; 23:00-23:30: Δ T3=23.2-22.5=0.7℃; 22:00-22:30: Δ T 4 = 22.5 - 22.0 = 0.5℃; therefore, Δ T total =1.5+1.3+0.7+0.5=4℃.

[0129] 3) Calculate the building's energy storage capacity, taking into account the building's total heat capacity. C total (precedent C total =23,150 kJ / K And with heat loss correction, the energy storage formula is: Q store = C total ×Δ T total - Q loss ,in, Q loss To account for building heat loss during the energy storage period, Q loss It can be derived from model simulation, such as Q loss =2150 kJ Substitute the data: Q store =23150×4.0−2,150=92600−2150=90450 kJ, Approximately 25.1 kWh That is, the building's energy storage capacity under this energy storage strategy is 90450. kJ .

[0130] 3. Load Reduction Operation Scenario Simulation: This simulation uses a building energy storage capacity model to model the load reduction operation of temperature-controlled electrical load equipment during peak electricity price periods, outputting the environmental parameter change trends of the temperature-controlled electrical load energy consumption zone. Specifically, it includes the following functions:

[0131] (1) Area division: The area division is used to divide the target area into three types of target areas: Type I, Type II and Type III. Type I target areas are those with a population density greater than or equal to the second threshold, Type II target areas are those with a population density greater than or equal to the third threshold, and Type III target areas are those with a population density greater than or equal to the fourth threshold.

[0132] Based on the actual personnel density in office settings (typically 0.1-0.5 people / m²) and GB / T50785-2012 "Evaluation Standard for Thermal and Humid Environment of Target Monitoring Areas in Civil Buildings", thresholds and classifications can be set as shown in Table 3:

[0133]

[0134] Threshold setting basis: when ρ When the load is >0.4, heat dissipation from personnel accounts for >25% of the total load (requiring high cooling capacity); when the load is <0.25, the heat dissipation from personnel accounts for >25% of the total load. ρ When the concentration of heat generated by personnel is ≤0.4, the heat dissipation accounts for 15%-25% of the total heat loss. ρ ≤0.25% of personnel heat dissipation is less than 15% (reducing load has little impact on comfort).

[0135] (2) Multi-zone operation simulation: Multi-zone operation simulation is used to simulate the environmental parameter change trends of Class I, Class II, and Class III target areas during daytime office hours and after the temperature-controlled electrical load equipment is operated at a reduced load according to a proportional ratio, using the building energy storage capacity model. For example, the air conditioning cooling capacity is set to be reduced by 20%-30%. For the target area, the focus is on monitoring the environmental parameter change trends of the target monitoring area (Class I, Class II, and Class III target areas) after the equipment is operated at a reduced load during the daytime office hours on weekdays, simulating the change trends of environmental parameters such as temperature, humidity, and air velocity in the target monitoring area. The details are as follows:

[0136] 1) Based on the regional classification in Table 3 above, the core parameters for each region are set as shown in Table 4:

[0137]

[0138] 2) Hourly thermal balance calculation using the RC model: The RC model is run separately for each region, calculating temperature and humidity in 15-minute increments: Temperature change = Heat dissipation from personnel + Solar radiation heat transfer + Q_Middle — Inner - Cooling capacity after load reduction - ventilation and heat dissipation; humidity change rate = moisture loss from personnel (100g / person•h) - air conditioning dehumidification capacity (positively correlated with cooling capacity, dehumidification capacity after load reduction = original dehumidification capacity × load reduction ratio).

[0139] 3) Output of changing trends: Based on the hourly calculation results, the parameter trend curves of each region are plotted: Region 1: People have a large heat dissipation, and the temperature rises rapidly after load reduction (0.5℃ in 15 minutes, 3℃ in 2 hours), and the humidity rises to 65%; Region 3: People have a small heat dissipation, and the temperature rises slowly (0.2℃ in 15 minutes, 1℃ in 2 hours), and the humidity rises to 60%; Finally, the changing trends of environmental parameters in different regions are output to provide a basis for optimizing the load reduction strategy.

[0140] (iii) Generate energy storage strategies and operation strategies based on model analysis results.

[0141] This is achieved through a strategy generation module. The strategy generation module is used to generate optimal energy storage strategies based on building energy storage and electricity price data during off-peak hours, and to generate load shedding strategies for equipment based on the PMV-PPD model and environmental parameter change trends. Specifically, the strategy generation module includes the following functions:

[0142] The functions used to generate the optimal energy storage strategy include:

[0143] 1. Relationship Quantification

[0144] Relationship quantification is used to calculate the energy consumption of temperature-controlled electrical load equipment under each energy storage strategy based on the power curve of the temperature-controlled electrical load equipment, and to establish a quantitative relationship between the building's energy storage capacity and energy consumption under each energy storage strategy.

[0145] Relationship quantification is used to calculate the energy consumption of temperature-controlled electrical load equipment under each energy storage strategy based on the power curve of the temperature-controlled electrical load equipment, and to establish a quantitative relationship between the building's energy storage capacity and energy consumption under each energy storage strategy.

[0146] Specifically, by combining the power curves of the equipment, the energy consumption of the air conditioning system under different operating durations and cooling intensities is calculated, and the relationship between cold storage capacity and energy consumption is quantitatively analyzed. For example, the cold storage capacity of the temperature-controlled electrical load equipment is calculated when it starts up 1 hour in advance and runs at full load for 2 hours, and the relationship between this cold storage capacity and energy consumption is quantified; the cold storage capacity of the temperature-controlled electrical load equipment is calculated when it starts up 2 hours in advance and runs at full load for 1 hour, and the relationship between this cold storage capacity and energy consumption is quantified. The details are as follows:

[0147] Quantification is achieved through "energy consumption calculation - cold storage capacity calculation - efficiency ratio establishment", with the core indicator being "cold storage energy consumption efficiency ratio (CSP)". EER_cool ()”, specifically as follows:

[0148] (1) Calculate the energy consumption of temperature control electrical load equipment

[0149] Based on the device power curve (e.g., the full-load power of an air conditioner) P =2.5 kW (Partial load power is linearly related to load factor) Calculate the total energy consumption during the energy storage period. ,in, P i For the first i Equipment power during the time period t i For the first i Duration of the time period. For example: Energy storage period 22:00-24:00 (2 hours), air conditioning operating at full load ( P =2.5 kW ),but E total =2.5×2=5 kWh =18,000 kJ .

[0150] (2) Calculate the cold storage capacity

[0151] Similar to "Energy Storage Calculation Based on Temperature-Time Curves", the cold storage capacity under this strategy in the previous example... Q cool =90450 kJ .

[0152] (3) Establish quantitative relationships ( EER_cool )

[0153] EER_cool Defined as the ratio of cold storage capacity to total energy consumption, it reflects the cold storage efficiency per unit of energy consumption. The formula is: EER_cool = E total / Q cool Substitute the data: EER_cool =90450 / 18000≈5.03 kJ / kJ (or converted to coefficient of performance) COP = Q cool / ( E total (×3600)=90450 / (5×3600)≈5.03).

[0154] Further establish different strategies EER_cool Matrix (as shown in Table 5) is used to achieve quantitative comparison:

[0155]

[0156] 2. Efficiency Evaluation

[0157] Efficiency assessment is used to extract the building energy storage capacity of each energy storage strategy under the same energy consumption based on quantitative relationships, and to obtain the building energy storage efficiency of each energy storage strategy based on the extracted building energy storage capacity.

[0158] Building energy storage efficiency η Defined as "the ratio of available effective cold storage capacity during peak hours to total energy consumption during off-peak hours," it reflects the actual utilization efficiency of the energy storage strategy. The calculation steps are as follows:

[0159] (1) Define core parameters

[0160] Effective cold storage capacity Q eff : The amount of cooling capacity actually used to maintain the temperature of the target monitoring area during peak hours (e.g., 14:00-16:00) (excluding heat loss during peak hours); Total energy consumption of energy storage E total Total electrical energy consumed during off-peak hours.

[0161] (2) Calculate the effective cold storage capacity

[0162] Simulate the chilled water release process during peak hours through the building energy storage capacity model and calculate Q eff = Q store − Q loss，speak , where Q store is the chilled water storage volume during low valley hours (such as 90450 kJ in the previous example), Q loss，speak is the heat loss during peak hours, which can be obtained by model simulation. For example Q loss，speak = 6450 kJ, then Q eff = 90450 - 6450 = 84000 kJ .

[0163] Heat loss during peak hours Q loss，speak includes "heat transfer loss of the enclosure structure" and "ventilation loss", which are calculated jointly through the RC model and the ventilation formula:

[0164] 1) Heat transfer loss of the enclosure structure ( Q 1), based on the multi-layer heat transfer calculation of the RC model: Heat transfer between the outer layer and the target monitoring area: Q _outer - room = h × A ×( Touter−Tout_speak )( Tout_speak is the temperature of the peak target monitoring area); Heat transfer between the outer layer and the middle layer: Q _outer - middle =( Touter−Tmiddle ) / R 2; Heat transfer between the middle layer and the inner layer: Q _middle - inner =( Tmiddle − Tin_speak ) / R 3 ( Tin_speak is the temperature of the peak target monitoring area); Total heat transfer loss of the enclosure structure Q 1 = Q _middle - inner (i.e., the heat dissipated from the inner layer to the outside through the enclosure structure).

[0165] 2) Ventilation loss ( Q 2), based on the air heat balance formula: Q 2 = L_speak × ρ × C ×( Tin_speak − Tout_ speak ) Where: L_speak : Ventilation volume during peak hours (such as 100 m³ / h, determined by the building ventilation standard); ρ Air density (1.2 kg / m³); C Specific heat capacity of air (1.01 kJ / kg•K); when Tin_speak < Tout_speak hour, Q 2 is positive, meaning that the cold energy in the target monitoring area is dissipated to the target monitoring area through ventilation.

[0166] 3) Total heat loss Q loss，speak , Q loss，speak = Q 1+ Q 2. Furthermore, calibration using historical data is required: If the simulated value deviates from the actual monitored "air conditioning supplementary heat" (actual air conditioning cooling capacity during peak hours - load demand) by more than 5%, then the thermal resistance parameters of the RC model should be adjusted (e.g., ...). R 2. R 3) Ensure simulation accuracy.

[0167] (3) Calculate the energy storage efficiency

[0168] The formula is: η = Q eff / ( E total (×3600)×100%. Substituting the data, we get: E total =5 kWh =18000 kJ ,but η =84,000 / (18,000)×100%≈466.7%. Since the energy stored in the cold storage includes the energy stored in the building's heat capacity, the efficiency is greater than 100%, which is consistent with actual physical logic. It should be noted that the reason the efficiency is greater than 100% is because "the effective energy storage includes free cooling from the environment," not because the electrical energy conversion efficiency exceeds 100%. The specific physical logic is as follows:

[0169] 1) Differences in energy form and source E total This refers to the electrical energy consumed by the air conditioner during off-peak hours (used to drive the compressor to produce cooling capacity). Q eff This refers to the cooling capacity stored in a building (the energy unit is the same as for electrical energy, both in kJ), which comes from two sources: cooling capacity converted from electrical energy used in air conditioning. Q_ref = Etotal × CO P( COP The coefficient of performance (COP) of an air conditioner is typically 3-5, meaning 1 kWh of electricity can produce 3-5 kWh of cooling capacity. Free cooling capacity for the environment: Q_env (During off-peak hours, the temperature in the target monitoring area is low, and the building dissipates heat to the target monitoring area through its building envelope, i.e., it absorbs ambient cold.)

[0170] 2) The essence of efficiency calculation, the "efficiency" described in this embodiment η = Q eff / ( E total "×3600)×100%" is not "electrical energy conversion efficiency," but rather "cold storage cost-effectiveness efficiency." For example, E total =5kWh (18000kJ), COP =5— Q_ref =5×5=25kWh (90000kJ) Q_env =2000kJ, Q loss，speak =7600kJ— Q_ref =90000+2000-7600=84400kJ; therefore η =84400 / 18000≈4.69 (469%), this value is greater than 100% because Q eff Includes "free environmental cooling capacity" Q_env "and" COP The "excess cooling capacity" is not an increase in energy created out of thin air by electrical energy; it conforms to the law of conservation of energy.

[0171] 3. Strategy Selection

[0172] Strategy selection is used to calculate the input-output ratio (ROI) of each energy storage strategy based on power data, and then selects the energy storage strategy with the highest ROI. ROI (Input-Output Ratio) ROI The energy storage cost (ESC) is defined as the ratio of energy-saving benefits during peak hours to energy storage costs during off-peak hours, reflecting the economic efficiency of the energy storage strategy. The calculation steps are as follows:

[0173] (1) Obtain electricity spot price data and time-of-use pricing from the power grid dispatch platform, such as off-peak pricing. p low =0.35 yuan / kWh, peak electricity price p high =1.2 yuan / kWh (2) Calculate the energy storage input, which is the electricity cost consumed during off-peak hours. The formula is: C in = E total × p low Substituting the data, we get: E total =5 kWh ,butC in =5×0.35=1.75 yuan. (3) Calculate the energy-saving output, which is "the electricity cost saved during peak hours by using cold storage instead of air conditioning", the formula is: C out = E save ×( p high - p low ),in, E save The energy consumption for air conditioning during peak hours is calculated using a building load forecasting model. The calculation steps are as follows:

[0174] 1) Predict peak temperature control electricity load without cold storage: Use the LSTM building load prediction model in the document, input peak hour meteorological data (temperature, solar radiation) and population density, and predict the hourly air conditioning load without cold storage. P peak (e.g., 2.5kW), and peak duration t peak (e.g., 2 hours), then the peak total energy consumption without cold storage is: E peak = P peak × t peak =2.5×2=5kWh.

[0175] 2) Calculate the amount of cold storage release during peak periods that can be replaced by cold storage. Q eff (e.g., 84000kJ = 23.33kWh), the cooling capacity that an air conditioner can provide per hour is... Q_air = P peak × COP =2.5×5=12.5kWh / h ( COP =5). Therefore, the air conditioning operating time that can be replaced by cold storage is: t_replace = Q eff / Q_air =23.33 / 12.5≈1.87 hours.

[0176] 3) Calculation E save After cold storage replacement, the actual operating time of air conditioning during peak hours t_actual = t peak - t_replace =2 - 1.87 = 0.13 hours; Actual energy consumption of air conditioning during peak hours: E_actual = Ppeak × t_actual =2.5 × 0.13 ≈ 0.33 kWh; therefore, the energy saved is: E save = E peak - E_actual =5-0.33≈4.67kWh.

[0177] (4) Calculate the input-output ratio ROI, The formula is: ROI = C out / C in Substituting the data, we get: ROI =4.25 / 1.75≈2.43, meaning that for every 1 yuan invested in energy storage, 2.43 yuan of energy-saving output can be obtained.

[0178] The strategies used to generate load reduction operation strategies for equipment include:

[0179] 1. Thermal Comfort Acquisition: Thermal comfort acquisition is used to obtain the basic human comfort level of people in each target area based on the PMV-PPD model and the changing trends of environmental parameters. The PMV-PPD model was proposed by Danish scientist P.Fanger based on the "thermal balance theory" and "a large amount of human thermal comfort experimental data". It is a classic model used to quantitatively evaluate the thermal comfort state of the human body in a specific environment, and it is also one of the most widely used thermal comfort evaluation standards in the world (such as ISO7730, GB / T50785, etc., which are based on it).

[0180] (1) The PMV-PPD model consists of two interrelated indicators, which describe the thermal comfort level from the two dimensions of “average thermal sensation” and “population dissatisfaction rate” respectively.

[0181] 1) PMV (Predicted Mean Vote) is an index derived from the concept of "human thermal balance" to predict the average thermal sensation of a large population in a specific environment. It assumes that the human body is in a state of thermal equilibrium (heat production = heat dissipation) and quantifies thermal sensation into a standardized vote value by calculating the difference in heat exchange between the human body and the environment.

[0182] 2) The calculation logic of PMV: The thermal balance of the human body is determined by 6 key factors, which are also the core input parameters of PMV, as shown in Table 6 below:

[0183]

[0184] 3) The range and meaning of PMV values

[0185] The PMV value ranges from -3 to +3, and each value corresponds to a specific thermal sensation description, as shown in Table 7:

[0186]

[0187] (2) PPD is an auxiliary indicator derived from PMV, used to predict the proportion of people who are dissatisfied with their thermal sensation in the current environment. It solves the limitation of PMV in describing only the "average sensation" - even if PMV is close to 0 (neutral), some people will still feel uncomfortable due to individual differences (such as metabolism, clothing, preferences).

[0188] 1) The value logic and meaning of PPD

[0189] The PPD (Potential Physical Discomfort) ranges from 0% to 100%, with higher values ​​indicating a greater proportion of dissatisfied individuals. According to international standards (such as ISO 7730), the core requirement for a thermal comfort environment is: PPD ≤ 10%. At this point, the corresponding PMV (Potential Physical Value) range is approximately -0.5 to +0.5, falling within the "acceptable thermal comfort zone," meaning that over 90% of people are satisfied with their thermal sensations. When PMV = 0 (ideal neutral state), PPD ≈ 5% (meaning approximately 5% of people will still feel uncomfortable), which is the optimal target for human thermal comfort.

[0190] 2) The correspondence between PMV and PPD is shown in Table 8:

[0191]

[0192] Based on GB / T50785-2012 and considering the changing trends of environmental parameters, comfort is calculated in four steps:

[0193] (1) Determine the PMV calculation parameters. Six types of parameters are required, of which three types come from model simulation and two types are default values ​​for the office scenario: Environmental parameters (model output): Temperature of the target monitoring area t (e.g., 26℃), relative humidity φ (e.g., 60%), wind speed v (e.g., 0.2 m / s); Personnel parameters (default): Metabolic rate M =1.2 met (Seated office posture), clothing thermal resistance I cl =0.5 clo (Summer clothing); Other: Difference between air temperature and mean radiant temperature in the target monitoring area Δ T rad = 1℃ (default).

[0194] (2) Substitute into the PMV calculation formula, adopt the PMV calculation formula of Fanger thermal comfort model, and refer to the table in Appendix A of GB / T50785. When t=26℃, φ =60% v =0.2m / s M =1.2 met、I cl =0.5 clo hour, PMV=0.3 .

[0195] (3) Obtain the PPD value. PPD and PMV The correspondence is PPD =100−95 e −0.03353PMV4 -0.2179 PMV 2 Substitute PMV= 0.3, obtained PPD =7%.

[0196] (4) Determine comfort level by combining the trend of environmental parameter changes. If the trend of environmental parameter changes is " t "From 25℃ to 27℃", then PMV From 0.1 to 0.8, PPD From 5% to 22%: When PMV ∈[-0.5,0.5], and PPD ≤10%, judged as comfortable; when PMV ∈[-1.0,1.0], and PPD ≤30% is considered basically comfortable (i.e., meeting basic needs); when PMV ∈[-1.0,1.0] or PPD >30% is considered uncomfortable.

[0197] 2. Strategy Optimization

[0198] The strategy optimization aims to minimize the total energy consumption of temperature-controlled electrical load equipment in the entire building while ensuring basic human comfort. For target areas with population density below the fourth threshold, the current operating power of temperature-controlled electrical load equipment is reduced proportionally. For target areas with high population density, the operating time of temperature-controlled electrical load equipment is extended. The strategy outputs the load reduction operation strategy of temperature-controlled electrical load equipment in each type of target area.

[0199] In this embodiment, the fourth threshold is set to 0.2 people / m². The basis for this is: when... ρ When the number of people is less than 0.2 people / m², the heat dissipation from people accounts for less than 12% of the total load. After load reduction (such as a 30% reduction in cooling capacity), the PMV can still be maintained within [-1.0, 1.0].

[0200] Through the simulation and analysis of different electricity price periods described above, a series of detailed data reports are generated, including: operating parameters of temperature-controlled electrical load equipment for each period (such as cooling capacity, heating capacity, power, and operating time), energy consumption data, changes in building energy storage, environmental parameter change curves in the target monitoring area, and economic evaluation index values. These data comprehensively reflect the relationship between the operation of temperature-controlled electrical load equipment and building energy consumption and the environment of the target monitoring area under different electricity price periods. This provides intuitive and accurate data for the virtual power plant of the building complex to formulate electricity price-based energy dispatch strategies, enabling it to rationally arrange the operating status of temperature-controlled electrical load equipment according to real-time electricity price fluctuations and load demand, thereby maximizing the economic and flexible value of energy use.

[0201] (iv) Developing intelligent agents

[0202] The intelligent agent has a monitoring and dynamic adjustment protection mechanism for temperature-controlled power load systems. It is developed through reinforcement learning algorithms and outputs a series of optimal control strategies for cold and heat sources and transmission and distribution equipment.

[0203] Includes the following functional modules:

[0204] 1. State space generation module: Utilizes database, model analysis results, optimal energy storage strategy, load reduction operation strategy, output of building physics simulation model, output of temperature control power load simulation model, and output of building load prediction model to generate the state space of the intelligent agent.

[0205] 2. Action Space Generation Module: This module generates the action space for the intelligent agent. Furthermore, it needs to translate each action in the action space into a corresponding control command for the temperature-controlled electrical load equipment. The agent's action space is defined around the core strategy of load reduction and transfer. Actions include: adjusting the temperature setpoint of the temperature-controlled electrical load equipment (discrete adjustment within ±2℃ range), changing the water pump frequency, controlling the start / stop of the chiller / heat pump unit, adjusting the fresh air volume, adjusting the fan speed, adjusting the opening of hot and cold water valves, and adjusting the energy release rate of the energy storage equipment. Each action corresponds to a specific control command for the temperature-controlled electrical load equipment, ensuring that the agent's decisions can be directly translated into equipment operation adjustments.

[0206] 3. Agent Training Module: This module trains the agent using state space, action space, and reward function. It operates on a "state-action-reward" framework, training the agent in five steps:

[0207] (1) Define the core elements of reinforcement learning

[0208] State space S: Environmental parameters (temperature of the target monitoring area) t relative humidity φ CO2 concentration, personnel density), equipment status (load power, valve opening), grid parameters (electricity price, load forecast), model output (energy storage capacity, current energy storage); action space M: temperature setting (24℃ / 25℃ / 26℃ / 27℃ / 28℃), pump frequency (25Hz / 30Hz / 35Hz / 40Hz / 45Hz / 50Hz), chiller start / stop; reward function Y : Y= 10× C save -5× PPD err −2× E excess .in, C save To save on electricity bills, PPD err =max(0, PPD -30%) E excess This refers to energy consumption exceeding the predicted load.

[0209] (2) Training data preparation

[0210] The dataset includes: simulation data: SAY samples generated from the building physics / temperature control power load system model (e.g., 100,000 samples, covering scenarios such as extreme high temperatures and sudden increases in electricity prices); historical data: hourly data collected on-site for one year (e.g., 8,760 samples), used for adaptation to real-world scenarios.

[0211] (3) The DQN algorithm is used to train the model.

[0212] Network structure: Q-network (input state space S, output Q-values ​​for all actions), 2 hidden layers (128 neurons); Training steps:

[0213] 1) Initialize the Q-network and the target Q-network; 2) Sample 64 samples from the empirical replay pool; 3) Calculate the target Q-value. Q target = Y +0.9× max ( Q ( S , A )); 4) Minimize Q ( S , A )and Q target 5) Update the target Q network every 100 steps.

[0214] (4) Training assessment

[0215] Evaluation indicators: average Y Value > 2.0 PPD The compliance rate is ≥90%, and the energy consumption reduction rate is ≥15%; if the compliance rate is not met, the weight of the reward function will be adjusted (e.g., the reward function will be adjusted in the following ways). PPD err The weight increased from 5 to 8.

[0216] III. Execution Feedback Layer

[0217] The execution feedback layer is used to deploy the trained agent to the BAS control system and to debug the agent based on the comparison results between the baseline data and the agent's feedback data.

[0218] A Building Automation System (BAS) is an intelligent system based on computer technology, automatic control technology, and network communication technology. It centrally monitors, automatically adjusts, and optimizes the management of various equipment within a building, such as air conditioning, lighting, water supply and drainage, elevators, and security systems. Its core objective is to improve building operating efficiency, reduce energy consumption, ensure environmental comfort and safety, and simplify equipment management processes through automated control.

[0219] After the intelligent agent is trained, it is seamlessly integrated and deployed through existing intelligent systems, BAS control systems, and power distribution systems. Optimized control strategies are executed based on real-time data input, and the building's energy consumption data, building environmental parameters, and system response are continuously monitored to verify the agent's control effectiveness and reliability. Furthermore, through real-time data interaction, the intelligent agent can continuously and automatically adjust equipment control strategies based on the building's dynamic operating status, constantly fine-tuning and optimizing. This effectively addresses the complex tasks of load system management and equipment control in the target monitoring area of ​​the office complex, achieving intelligent management and control of the temperature-controlled power load system.

[0220] This embodiment also includes the following functional modules:

[0221] (I) Communication Interface Construction

[0222] The communication interface building module is used to build a data communication interface between the intelligent agent and the temperature-controlled electrical load system.

[0223] To achieve stable data transmission, a robust data communication interface needs to be built, enabling seamless connection between the agent's output strategy and the sensors and actuators of the temperature-controlled electrical load system. Interface development requires the agent to acquire environmental status data in real time and dynamically adjust its control strategy based on this data. Simultaneously, the interface must support bidirectional communication, allowing the agent to not only send control commands to the temperature-controlled electrical load system but also receive feedback data to adjust future decisions. The interface design must consider the real-time performance, reliability, and security of data transmission to avoid system performance degradation due to data delays or loss.

[0224] For example, using a "layered architecture + industrial protocol" approach, it can be built in five steps:

[0225] 1. The interface architecture is designed in three layers, with each layer having a clearly defined function:

[0226] Physical layer: RJ45 industrial Ethernet (100Mbps), connecting the intelligent agent server and PLC / sensor gateway; Data link layer: Modbus-TCP (control commands) + MQTT (sensor data), ensuring real-time performance; Application layer: JSON data format + CRC32 checksum, ensuring data integrity.

[0227] 2. Protocol and Data Format

[0228] Modbus-TCP: Used for control commands (such as temperature setpoints) and register address mapping (such as temperature-controlled electrical load—register 0x0001); MQTT: Used for sensor data (temperature per second) and topic definitions (such as "sensor / AC-001 / temp"). MQTT (Message Queuing Telemetry Transport) is a lightweight instant messaging protocol based on a publish / subscribe model, developed by IBM, primarily used in the Internet of Things (IoT) field to achieve efficient message passing between devices in low-bandwidth, unstable network environments.

[0229] 3. Security Design

[0230] Authentication: Username and password + digital certificate (two-way authentication between agent and PLC); Data encryption: TLS 1.3 encrypted transmission; Access control: RBAC permissions (agents only have control / read permissions). RBAC stands for Role-Based Access Control. It is a commonly used access control strategy. The core idea is to associate permissions with roles, and users obtain the permissions of those roles by becoming members of the appropriate roles.

[0231] 4. Exception Handling

[0232] Data loss: MQTTQoS=2 (ensure message delivery), resend if lost.

[0233] MQTT QoS, or MQTT Quality of Service, refers to the quality of service levels in the MQTT protocol, used to define the reliability and guarantee level of message delivery. It is divided into three levels:

[0234] QoS0 (At Most Once): The message sender makes every effort to send the message, but does not guarantee that the message will reach the receiver, and will not resend the message. Message loss may occur. The advantage is that the transmission overhead is minimal.

[0235] QoS1 (At Least Once): The message sender ensures that the message is received by the receiver at least once. If no acknowledgment is received from the receiver, the message sender will resend the message. In this mode, messages may be duplicated, but they will not be lost.

[0236] QoS2 (Exactly Once): This is the highest quality of service level, ensuring that a message is received by the receiver only once. It avoids message loss and duplication through complex message acknowledgment and retransmission mechanisms. In the construction scheme, QoS=2 is adopted to ensure that the message is delivered and retransmitted if lost.

[0237] Communication interruption: The agent automatically switches to the local backup strategy (maintaining the current state) and triggers an alarm.

[0238] 5. Interface Testing

[0239] Performance testing: 100 sensors transmitting simultaneously, latency <100ms, packet loss rate <0.1%; Compatibility testing: compatible with mainstream PLCs (Siemens S7-1200) and sensors (Schneider temperature and humidity sensors).

[0240] (ii) Deployment of intelligent agents

[0241] The agent deployment module is used to deploy the trained agents to the BAS control system and transmit multidimensional data to the agents through the data communication interface.

[0242] Deploying an agent to a real-world BAS control system involves transferring the agent from a simulation environment to a real operating environment, requiring fine-tuning to adapt it to the specific system. During deployment, compatibility with existing control systems must be considered to ensure the model can operate effectively in the real-world environment. Furthermore, deployment includes mechanisms to support real-time updates and optimization of the agent, enabling it to continuously adapt to environmental changes and system requirements. Fault detection and recovery capabilities should also be included to handle potential system failures or anomalies. Development should also prioritize system scalability, allowing for the addition of new functional modules or adaptation to new hardware devices in the future to support long-term system operation and upgrades.

[0243] For example, deployment is carried out in four phases to ensure compatibility with on-site systems:

[0244] 1. Pre-deployment preparation and environmental survey: Confirm the PLC model and BAS system version for the temperature control power load system; Data import: Import building thermal parameters and historical load data, and update RC model parameters; Hardware deployment: Connect the intelligent agent server to the industrial switch. 2. Simulation environment verification and semi-physical platform: Connect the on-site PLC and sensors, simulate the temperature of the target monitoring area (e.g., 35℃), and set the personnel density (e.g., 0.3 people / m²); Testing: Run for 72 hours to verify command execution (e.g., "temperature 26℃") and backup strategies (maintaining status during network outages). 3. On-site regional deployment: Phase 1: Deployment in Category III target areas (few personnel), run for 24 hours, monitoring temperature / energy consumption; Phase 2: Expand to Category II / Category I target areas, configuring PLC command priority (intelligent agent > manual, < emergency stop). 4. Joint commissioning and acceptance: Single device joint commissioning: Test the "pump frequency 40Hz" command, and the feedback frequency error is <0.5Hz; Full system joint commissioning: After 7 days of operation, energy consumption is reduced by 15% (from 1000kWh to 850kWh).

[0245] (III) Agent Debugging

[0246] The intelligent agent debugging module is used to monitor the laboratory temperature, humidity, time to reach temperature and humidity control indicators, temperature and humidity fluctuation rate, energy consumption and power consumption in real time during the operation of the intelligent agent. It compares and analyzes the data obtained from real-time monitoring with the baseline data before the deployment of the intelligent agent, and debugs the intelligent agent based on the analysis results.

[0247] In practical applications, the overall effectiveness of the intelligent agent's control will be verified through continuous monitoring and data analysis. After the system is operational, it is necessary to monitor in real time indicators such as laboratory temperature, humidity, time to reach temperature and humidity control targets, temperature and humidity fluctuation rates, energy consumption, and power consumption, and compare these with baseline data before deployment. The purpose of verification is to evaluate whether the intelligent agent's performance in the actual environment meets expectations and to identify potential optimization areas. The initial focus of verification is to verify whether the intelligent agent's performance in the actual environment is consistent with the simulation phase, and to continuously debug and optimize the intelligent agent's parameters to ensure stable operation under different working conditions. Simultaneously, a series of functional tests are required, including the system's response speed, control accuracy, and ability to handle abnormal situations. During this process, the real-time data collected will be used to further optimize the intelligent agent's control strategy, ensuring that it still achieves optimal results in complex environments.

[0248] For example, debug in the order of "single machine - region - whole system" to ensure that the indicators meet the standards:

[0249] 1. Adjustment parameters and settings for thermal comfort: PMV ∈[-1.0,1.0], PPD1. **Standalone Debugging:** ≤30%, temperature fluctuation rate ≤±0.5℃ / h; Energy consumption: daily average reduction rate ≥12%, energy consumption per unit area ≤50kWh / (m²·year); System: command delay <200ms, fault recovery <5min. 2. **Single-unit Debugging:** Air conditioning debugging: Send a "26℃" command, reach the desired temperature within 30 minutes (adjust PID proportional coefficient if timeout occurs); Water pump debugging: "40Hz" command, pressure deviation <0.1MPa (calibrate frequency-pressure curve if deviation is large). 3. **Regional Debugging:** Three target areas: 24 hours of operation, temperature fluctuation ±0.8℃; Comfort debugging: If feedback is "too cold," adjust the temperature baseline from 25℃ to 26℃. 4. **Full System Debugging:** Extreme scenario testing: When the target monitoring area reaches 40℃, the agent automatically increases cooling capacity, temperature ≤27℃; Stability testing: 30 consecutive days, energy consumption reduction rate drift <0.5% (incremental training if drift is large).

[0250] In summary, this embodiment provides a user-side power load response control method considering building energy storage. By monitoring and analyzing user demand and power system load in real time, it dynamically adjusts the load operation of the temperature-controlled power load system to optimize energy consumption and balance the power load. This system integrates advanced sensors and IoT devices to accurately capture user comfort needs and environmental changes, and makes flexible adjustments based on real-time fluctuations in power load. During periods of high power demand, the intelligent agent can intelligently increase the temperature setting to reduce energy consumption; while during periods of low electricity prices, it automatically restores a comfortable temperature, thereby maximizing energy savings.

[0251] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A user-side power load response control method considering building energy storage, characterized in that, Specifically, the following steps are included: A database was constructed based on building energy storage data and electricity load data. Utilize databases to build building physics simulation models, temperature-controlled power load simulation models, building energy storage capacity models, and building load prediction models; The building energy storage capacity model is used to simulate the building's heat storage and release process, heat transfer process, and temperature-controlled electricity load regulation process during different peak electricity price periods. The model analysis results are output, including the building's energy storage capacity value, the building's energy storage capacity, and the environmental parameter variation trends of the temperature-controlled electricity load energy consumption area. The optimal energy storage strategy is generated based on building energy storage and off-peak electricity price data, and the temperature-controlled load reduction operation strategy is generated based on the PMV-PPD model and environmental parameter change trends. Specifically, this includes: The energy consumption of the temperature-controlled electrical load under each energy storage strategy is calculated based on the power curve of the temperature-controlled electrical load, and a quantitative relationship between the building's energy storage capacity and energy consumption under each energy storage strategy is established. Based on the quantitative relationship, the building energy storage capacity of each energy storage strategy under the same energy consumption is extracted, and the building energy storage efficiency of each energy storage strategy is obtained based on the extracted building energy storage capacity. By combining power data, the input-output ratio of each energy storage strategy is calculated, and the energy storage strategy with the highest input-output ratio is selected as the optimal energy storage strategy. The basic human comfort level of people in each target area is obtained based on the PMV-PPD model and the changing trends of environmental parameters; The constraints are minimizing the total energy consumption of temperature control power load and meeting the basic human comfort of personnel. For target areas where the population density is below a preset threshold, the current operating power of the temperature control power load is reduced by a preset ratio; For target areas where the population density is higher than or equal to the preset threshold, extend the operating time of the temperature-controlled power load; Output the optimal temperature-controlled power load reduction operation strategy for each type of target area; The state space of the intelligent agent is generated by utilizing the database, model analysis results, optimal energy storage strategy, temperature-controlled power load reduction operation strategy, output of building physics simulation model, output of temperature-controlled power load simulation model and output of building load prediction model; Construct the action space of the agent, train the agent using the state space, action space and reward function, deploy the trained agent to the BAS control system, and debug the agent based on the comparison results of baseline data and agent feedback data. The optimal control strategy is output based on the debugged agent.

2. The user-side power load response control method considering building energy storage according to claim 1, characterized in that, The building energy storage-related data and electricity load-related data include: Thermal model parameters of the target temperature-controlled power load and its surrounding building envelope, operating status parameters of the temperature-controlled power load, environmental parameters of the energy consumption zone where the temperature-controlled power load is located, system model parameters of the power load, meteorological data, population density change data, and electricity price data for different time periods; The operating status parameters of the temperature-controlled electrical load include: temperature and humidity, flow rate, power consumption, and energy consumption of the temperature-controlled electrical load. The meteorological data includes: atmospheric temperature, solar radiation intensity, and relative humidity; The environmental parameters of the energy consumption zone where the temperature-controlled power load is located include: temperature, carbon dioxide concentration and personnel density change data of the target monitoring area; The temperature-controlled power load energy consumption zones include: a first temperature-controlled power load energy consumption zone and a second temperature-controlled power load energy consumption zone; the first temperature-controlled power load energy consumption zone includes: meeting rooms and target areas with personnel density greater than or equal to a first threshold; the second temperature-controlled power load energy consumption zone includes all other temperature-controlled power load energy consumption zones except the first temperature-controlled power load energy consumption zone; the data acquisition frequency for the first temperature-controlled power load energy consumption zone is on the order of seconds; the data acquisition frequency for the second temperature-controlled power load energy consumption zone is on the order of minutes; the electricity price data is collected in real time.

3. The user-side power load response control method considering building energy storage according to claim 2, characterized in that, The aforementioned use of databases to build building physics simulation models, temperature-controlled electricity load simulation models, building energy storage capacity models, and building load prediction models specifically includes: A building physical simulation model is built using building thermal model parameters, temperature-controlled power load operating parameters, and environmental parameters of the temperature-controlled power load energy consumption zone. A simulation model of a temperature-controlled electrical load is built using the parameters of a temperature-controlled electrical load system model. Utilize meteorological data to build a model of building energy storage capacity; A building load forecasting model was established using historical load data, meteorological data, and population density change data. in, Building physics simulation model and temperature-controlled power load simulation model are used to accurately reflect the dynamic load change characteristics of buildings throughout the year and the operating status of temperature-controlled power load system under different control strategies. The building energy storage capacity model is used to dynamically simulate the heat storage and release process of building structures based on real-time meteorological data, as well as the heat transfer between the air and the building envelope in the target monitoring area, and to analyze the building's energy storage capacity based on the simulation results. Building load forecasting models are used to reflect the correlation between electricity prices and load impact.

4. A user-side power load response control method considering building energy storage according to claim 1 or 2, characterized in that, After building the model, a model validation step is also included, which includes: The building physical simulation model is calibrated using real-time building operation data; and The temperature-controlled power load simulation model is calibrated using real-time operating data from the temperature-controlled power load system.

5. A user-side power load response control method considering building energy storage according to claim 2, characterized in that, The model analysis specifically includes: The building's heat storage and release process and heat transfer process are simulated using a building energy storage capacity model, and the building's energy storage capacity value is output. The building's energy storage capacity model is used to simulate the full-load operation scenario of temperature-controlled electricity load during off-peak electricity price periods, and the building's energy storage capacity is output. The building's energy storage capacity model is used to simulate the reduced load operation scenario of temperature-controlled electricity load during peak electricity price periods, and the environmental parameter change trend of the temperature-controlled electricity load energy consumption area is output.

6. A user-side power load response control method considering building energy storage according to claim 5, characterized in that, The method of using a building energy storage capacity model to simulate the building's heat storage and release process and heat transfer process, and outputting the building's energy storage capacity value, specifically includes: Use the temperature values ​​of the inner, middle and outer nodes calculated at the current time step as the initial temperature for the next time step; Calculate the conduction heat flux based on the initial temperatures of the inner, middle, and outer layers at the next time step; The heat flow is continuously iterated according to a set time step to generate complete time series data of the temperature changes of the inner, middle and outer layers over time, which is output as the building's energy storage capacity value.

7. A user-side power load response control method considering building energy storage according to claim 5, characterized in that, The method of using a building energy storage capacity model to simulate the full-load operation scenario of temperature-controlled electricity load during off-peak electricity price periods and outputting the building's energy storage capacity specifically includes: The full-load operation scenario under various candidate energy storage strategies was simulated; wherein, the candidate energy storage strategy is: before the start of the off-peak electricity price period, the temperature-controlled power load is started in advance and it is operated at full load for a preset time; During the simulation, virtual temperature data of each target monitoring area is collected from the output of the building's energy storage capacity model. Based on the virtual temperature data, calculate the building's energy storage capacity under each candidate energy storage strategy; The calculation steps for the building's energy storage are as follows: plot the temperature-time change curve based on the virtual temperature data; calculate the cumulative temperature change based on the initial temperature; and calculate the building's energy storage based on the cumulative temperature change, the building's total heat capacity, and the heat loss during the simulation process.

8. The user-side power load response control method considering building energy storage as described in claim 7, characterized in that, The method utilizes a building energy storage capacity model to simulate the reduced load operation scenario of temperature-controlled electricity loads during peak electricity price periods, and outputs the environmental parameter change trends of the temperature-controlled electricity load energy consumption area, specifically including: The target area is divided into three categories: Category I, Category II, and Category III. Category I target areas are those with a population density greater than or equal to the second threshold, Category II target areas are those with a population density greater than or equal to the third threshold, and Category III target areas are those with a population density greater than or equal to the fourth threshold. The environmental parameter change trends of Class I, Class II, and Class III target areas were simulated using a building energy storage capacity model after daytime office hours and proportionally reduced temperature-controlled power load. Calculate hourly heat balance; Based on the hourly calculation results, plot the parameter trend curves for each region; Trends in environmental parameters related to the energy consumption of temperature-controlled electrical loads in different regions.

9. A user-side power load response control method considering building energy storage according to claim 1, characterized in that, The debugging agent specifically includes: Obtain a simulation dataset generated by the building physics simulation model and the temperature control electrical load simulation model. The dataset contains samples of state S, action A, and reward Y. Historical operational datasets collected from the field; The model is trained using a deep Q-network algorithm: a Q-network is constructed, which takes state S as input and outputs the Q-value of each action in the action space; the parameters of the Q-network and the target Q-network are initialized; training samples are sampled from the experience replay pool; Calculate the target Q-value, and update the Q-network parameters by minimizing the mean squared error loss between the Q-network output and the target Q-value; update the parameters of the target Q-network periodically; The training effect is evaluated, and the evaluation indicators include the average reward value Y, the percentage of unsatisfactory predictions, the rate of achieving the target, and the energy consumption reduction rate; if the evaluation results are not up to standard, the weight coefficients of the corresponding terms in the reward function are adjusted.

Images