Energy consumption prediction and energy-saving control system

Abstract

The invention relates to the technical field of energy consumption control, in particular to an energy consumption prediction and energy-saving control system. According to the technical scheme, the system comprises a data acquisition and communication module, an energy consumption prediction and optimization module, an intelligent control execution module, an energy management and visualization module and a system cooperation mechanism, and data aggregation and transmission are realized through a multi-protocol bus. According to the method, building energy consumption is optimized through dynamic prediction and intelligent control in a cooperative mode, environment changes are responded in real time, operation of multiple devices is accurately coordinated, and the hysteresis and coupling interference limitation of a traditional model are broken through; the hidden danger is actively identified in combination with multi-source data fusion diagnosis, the operation and maintenance cost is reduced, meanwhile, construction management is simplified through standardized integrated design, visual regulation and control are provided, and energy efficiency improvement and a full-period closed loop of intelligent operation and maintenance are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy consumption control, and in particular to an energy consumption prediction and energy-saving control system. Background Art

[0002] Energy efficiency optimization of building automation systems has become a critical component in achieving the "dual carbon" goals. However, traditional building energy management systems face the challenge of achieving coordinated prediction, control, and diagnosis in multi-dimensional, dynamically coupled scenarios. The core technical bottleneck is the inability of existing systems to simultaneously capture dynamic load disturbances such as occupancy density and outdoor environment in real time, address the impact of multi-subsystem coupling interference on control accuracy, and implement intelligent multi-source data assessment of equipment energy efficiency.

[0003] Specifically, traditional energy consumption prediction models rely on historical data statistics and static load assumptions (such as a fixed heat coefficient K value and ignoring the dynamic impact of water quality TDS). These models are slow to respond to dynamic scenarios such as the sudden gathering of people (causing CO2 concentrations to surge by over 300 ppm) and extreme weather (where outdoor temperature and humidity fluctuate by over 15% hourly). This results in short-term energy consumption forecast errors generally exceeding 10%, directly causing the adjustment of operating parameters for air conditioning cooling and heating systems, water pumps, and fans to lag behind actual load demand, resulting in 15%-20% inefficient energy consumption. More critically, subsystems such as air conditioning temperature and humidity control, fresh air volume adjustment, and device group control are strongly coupled (for example, changes in fresh air valve opening simultaneously affect return air temperature and chilled water load). Traditional control strategies lack multivariable decoupling capabilities (by failing to establish a state-space model for return air temperature, CO2 concentration, and PM2.5 concentration). This leads to a "temperature overshoot-energy rebound" cycle during the control process, resulting in a 12%-18% decrease in the system's energy efficiency ratio (EER) compared to the theoretical optimal value. In addition, energy efficiency evaluation and fault diagnosis rely only on single-dimensional parameters such as current and voltage (e.g., only monitoring EER drop of 10% triggers an early warning), ignoring the current harmonic distortion rate (when THD>8%, the motor copper loss increases by 15%), vibration signals (RMS value of 2-3m / s in the initial stage of bearing wear), and the like. 2 ) and environmental parameters, resulting in a 3-5 day delay in discovering hidden energy efficiency losses of equipment (such as continuous energy waste caused by insulation aging), and an increase in operation and maintenance costs of more than 30%.

[0004] These issues essentially reflect the technical gaps in traditional systems within the "dynamic load sensing - coupled interference decoupling - multi-source data decision-making" chain: They are unable to dynamically modify load forecasting models using real-time environmental parameters (temperature, humidity, occupancy density, water quality indicators), struggle to achieve multivariable collaborative optimization during control execution, and lack the ability to conduct in-depth diagnostics of equipment health based on harmonic analysis and vibration monitoring. This results in building mechanical and electrical equipment operating in a state of "high energy consumption and low reliability" for extended periods, making it difficult to meet the requirements for reducing energy intensity.

[0005] In summary, this application proposes an energy consumption prediction and energy-saving control system. Summary of the Invention

[0006] The purpose of the present invention is to address the problem in the background technology that there is no refined management of building energy consumption and efficient energy saving, and to propose an energy consumption prediction and energy-saving control system.

[0007] The technical solution of the present invention: an energy consumption prediction and energy-saving control system, comprising:

[0008] Data acquisition and communication module, used to collect operating parameters, environmental parameters and equipment status of building mechanical and electrical equipment in real time, and realize data aggregation and transmission through multi-protocol bus;

[0009] Energy consumption prediction and optimization module, which is used to predict building energy consumption trends and generate optimization control instructions based on dynamic load calculation, machine learning algorithms and energy efficiency analysis models;

[0010] Intelligent control execution module, used to execute adaptive control strategies, equipment group control logic and preset energy-saving modes, and dynamically adjust the operating parameters of electromechanical equipment;

[0011] Energy management and visualization module, used for itemized energy consumption measurement, equipment failure diagnosis, and a visual monitoring interface;

[0012] The system coordination mechanism realizes two-way data interaction between modules through a real-time database, ensuring that the prediction results drive the dynamic adjustment of the control strategy.

[0013] Optionally, the data acquisition and communication module includes:

[0014] Distributed sensor unit, integrating temperature sensor, humidity sensor, CO2 concentration sensor, PM2.5 concentration sensor, current transformer, voltage transformer, liquid level sensor and flow sensor, including:

[0015] The temperature sensor uses PT1000 platinum resistance with an accuracy of ±0.1°C and is installed in the air conditioning return duct, chilled water supply and return pipes and outdoor environment monitoring points;

[0016] The current transformer is an open-type Rogowski coil with a range of 0-500A and an accuracy of 0.5. It is embedded in the power distribution cabinet of the fan, water pump and cold and heat source system;

[0017] The PM2.5 concentration sensor uses the principle of laser scattering and has a detection range of 0-1000μg / m 3 , resolution 1μg / m 3 , deployed at the air inlet of the fresh air unit and key indoor areas;

[0018] The equipment status monitoring unit is connected to the fan power distribution cabinet, water pump power distribution cabinet, lighting distribution box, cold and heat source system and elevator control cabinet through the LONWORKS fieldbus to collect the equipment's operating status, fault signals and energy efficiency parameters in real time, including:

[0019] The fan power distribution cabinet has a built-in motor protector to monitor the three-phase current imbalance rate, overload alarm and insulation resistance value;

[0020] The controller of the cold and heat source system uploads the chiller evaporator / condenser pressure, compressor operating frequency and energy efficiency ratio (COP) through the Modbus protocol;

[0021] The bus communication unit adopts a dual-redundant network architecture, including the LONWORKS fieldbus layer and the TCP / IP Ethernet layer, including:

[0022] The LONWORKS bus layer supports free topology, a transmission rate of 78kbps, a maximum number of 64 nodes, and is used for real-time control signal transmission at the device level;

[0023] The TCP / IP Ethernet layer connects the data collectors in each area through a fiber optic ring network. The transmission protocol is BACnet / IP and supports OPCUA data subscription and publishing.

[0024] The protocol converter is embedded in the PDC controller, converting LONWORKS data packets into JSON format and uploading them to the cloud database via the MQTT protocol.

[0025] Optionally, the energy consumption prediction and optimization module includes:

[0026] The dynamic load calculation unit calculates the cooling load of the air conditioning system in real time based on the chilled water supply and return temperature difference ΔT, the flow rate F, and the thermodynamic formula Q = K × F × ΔT, where K is the chilled water thermal coefficient, which is dynamically corrected based on the water quality test report using the formula K = 4186 × (1 - 0.0025 × TDS), where TDS is the total dissolved solids content (ppm). Historical load data is stored in a time series database, and load curves are generated at 15-minute intervals. Periodic features are extracted using a sliding window algorithm.

[0027] The machine learning prediction unit uses the ARIMA model to predict short-term energy consumption trends and combines it with an LSTM neural network to process nonlinear features. Specifically, input features include outdoor temperature, indoor occupancy density, cumulative equipment operating time, electricity price time intervals, and historical energy consumption data. The model is trained using the Adam optimizer with a learning rate of 0.001, a batch size of 64, and 500 training cycles. The error rate for predicting energy consumption over the next 24 hours is ≤3%.

[0028] The energy efficiency analysis unit evaluates the energy efficiency status of individual devices in real time by building an energy efficiency ratio (EER) model. The EER is calculated as EER = cooling capacity (kW) / input power (kW), where cooling capacity is calculated from the chilled water flow rate and the supply and return water temperature difference. The energy efficiency degradation warning trigger condition is: an EER drop of 10% for three consecutive hours or a current harmonic distortion rate exceeding THD>8%.

[0029] Optionally, the intelligent control execution module includes the following units:

[0030] The adaptive control unit uses a fuzzy PID algorithm to adjust the operating parameters of the air conditioning system, including:

[0031] The fuzzy rule base is set as "if the return air temperature deviation is large and the rate of change is fast, then increase the proportional coefficient KP", and the KP dynamic range is 0.5-2.0;

[0032] The opening of the fresh air valve is controlled in sections according to the CO2 concentration: 30% when CO2 < 800ppm, 50% when 800-1200ppm, and 100% when CO2 > 1200ppm;

[0033] The fan frequency conversion frequency is PID-adjusted according to the deviation between the return air temperature set value and the actual value, and the output frequency range is 10-50Hz;

[0034] The equipment group control unit performs group control of the chiller, circulating water pump and cooling tower, specifically including:

[0035] The round-robin strategy is based on the cumulative running time of the equipment. Each time the unit with the shortest running time is started, the unit with the longest running time is shut down first when shutting down.

[0036] The opening of the pressure differential bypass valve is controlled by PID, with the target value being a pressure differential of 0.2 MPa for the supply and return pipes, and an adjustment accuracy of ±0.01 MPa;

[0037] Small temperature difference compensation technology adjusts the cooling tower fan speed to make the cooling water return temperature close to the optimal set value of the chiller (32℃±0.5℃);

[0038] Energy-saving strategy library, which stores the following preset control modes:

[0039] Schedule control mode: Divide by weekdays / holidays, set the lighting system to fully open from 07:00 to 09:00, and control the lights in intervals from 22:00 to 06:00;

[0040] Infrared sensing control mode: garage lighting detects traffic flow through microwave radar sensors, with a trigger delay of 30 seconds and turns off 60 seconds after the car leaves;

[0041] Fire linkage mode: After receiving the fire alarm signal, the fresh air valve is forced to close, the smoke exhaust valve is opened, and the non-fire water pump is stopped.

[0042] Optionally, the energy management and visualization module includes the following units:

[0043] The energy consumption metering unit uses a 0.2S-level smart meter to measure the energy consumption of the equipment, including:

[0044] The energy consumption of the cooling and heating source systems is measured by a CT power meter, with a data update interval of 1 minute;

[0045] The lighting circuit energy consumption collects the current, voltage and power factor of each circuit through the RS485 bus;

[0046] The fault diagnosis unit uses an expert system rule engine with the following rules:

[0047] If the water pump current suddenly increases by 20% and lasts for 5 seconds, it is determined to be a blade jam fault;

[0048] If the three-phase current imbalance rate of the fan is greater than 15%, the bearing wear warning will be triggered;

[0049] The human-computer interaction unit builds a 3D visualization interface based on WebGL technology. Its specific functions include:

[0050] Dynamically display the water pipe topology of the cold and hot source system, and mark the high temperature alarm point in red;

[0051] Energy consumption ratio charts are displayed by air conditioning, lighting, water pumps, and elevators, and support drilling down to the sub-item equipment level;

[0052] The remote manual intervention interface provides forced start and stop, parameter modification and policy import functions. The operation permissions are divided into three levels: administrator, engineer and guest.

[0053] Optionally, the system coordination mechanism specifically includes:

[0054] The real-time database uses a time series database (InfluxDB), with storage granularity set to 1-second raw data. The data partitioning strategy is divided by device type. The data retention period for the cooling and heating source systems is 30 days, and the data retention period for the lighting system is 7 days.

[0055] The output instructions of the energy consumption prediction module are written into the PLC register of the control execution module through the OPCUA protocol. The specific parameter configuration is:

[0056] The OPCUA server address is opc.tcp: / / 192.168.1.100:4840, and the session timeout is set to 300 seconds;

[0057] The write register address range is 40001-40050, the data type is Float32, and the encoding rule is Little-Endian;

[0058] After receiving the feedback data from the control execution module, the energy management module uses a sliding window weighted average algorithm to update the energy efficiency analysis model weights. The formula is:

[0059] W new =α·W old +(1-α)·ΔW

[0060] Among them, W new is the updated model weight matrix, W old is the model weight matrix before updating, α is the forgetting factor (with a value of 0.7-0.9), ΔW is the model error gradient of the current period, and the weight matrix is ​​updated every hour.

[0061] Optionally, the edge computing function of the bus communication unit further includes:

[0062] The lightweight TensorFlow Lite model deployed on the power distribution cabinet controller takes as input features the device current I, voltage V, ambient temperature envTenv, and operating time t. The output is the load forecast value predPpred for the next hour. The model structure is as follows:

[0063] Input layer: 4 neurons;

[0064] Hidden layer: 2 layers, 8 neurons in each layer, activation function is ReLU;

[0065] Output layer: 1 neuron, activation function is linear;

[0066] The edge node calculation cycle is 5 minutes, and the prediction results are uploaded to the cloud through the MQTT protocol. The consistency check algorithm uses the root mean square error (RMSE) threshold judgment, and the formula is:

[0067]

[0068] Among them, P i is the load value predicted by the i-th edge node; P cloud is the load value predicted by the cloud model for the i-th time, n = 12, is the total number of data points in the verification period,

[0069] If RMSE > 5%, the cloud model resynchronization is triggered, otherwise the edge prediction result is accepted.

[0070] Optionally, the adaptive control unit further includes a multivariable decoupling control unit for resolving the coupling interference of return air temperature, CO2 concentration, and PM2.5 concentration in the air-conditioning system, specifically including:

[0071] Establish state space equations to describe multivariable relationships:

[0072]

[0073] Return air temperature change rate, ΔT: Return air temperature deviation, u valve : Water valve opening, C CO2 : Indoor CO2 concentration; a1, a2, a3: Temperature dynamic model coefficients; CO2 concentration change rate; u damper : Fresh air valve opening; Q vent : ventilation rate; b1, b2: CO2 dynamic model coefficients; PM2.5 concentration change rate; u filter : filter efficiency; c1, c2: PM2.5 dynamic model coefficients;

[0074] Feedforward-feedback composite control is adopted. The feedforward controller pre-adjusts the fresh air valve opening according to the change of CO2 concentration, and the feedback controller adjusts the water valve opening according to the return air temperature deviation. The control rate is:

[0075]

[0076] Among them, e CO2 =C CO2,set -C CO2 :CO2 concentration deviation, e T =T set -T return : Return air temperature

[0077] Degree deviation; K p , K i , K d : PID parameter for CO2 concentration control, K p ′,K i ′,K d ′: PID parameters for temperature control, dynamically adjusted by fuzzy rules.

[0078] Optionally, the fault diagnosis unit further includes a harmonic distortion rate analysis unit for detecting motor faults, specifically including:

[0079] The current signal is collected and fast Fourier transform (FFT) is performed to calculate the total harmonic distortion (THD). The formula is:

[0080]

[0081] Among them, I1 is the effective value of the fundamental current, I h is the effective value of the hth harmonic current, THD is the total harmonic distortion;

[0082] Optionally, set the fault determination rules:

[0083] If THD>8% and lasts for 10 minutes, the "motor winding insulation aging" alarm is triggered;

[0084] If the third harmonic component I3 / I1>5% and the fifth harmonic component I5 / I1>3%, it is judged as "power supply voltage unbalance";

[0085] Combined with the vibration sensor data, when THD exceeds the limit and the vibration acceleration RMS value is greater than 4m / s 2 When the fault occurs, it is determined to be a "combined fault of bearing wear and electrical failure".

[0086] Compared with the prior art, this application has at least one of the following beneficial technical effects:

[0087] By collecting multi-dimensional equipment and environmental data in real time, combined with dynamic load forecasting and machine learning algorithms, the real-time and accuracy of energy consumption forecasting can be significantly improved, overcoming the limitations of traditional models that rely on static assumptions.

[0088] Adopting multivariable decoupling control and adaptive strategies, it effectively coordinates the coupling relationship between subsystems such as air conditioning, lighting, and ventilation, and achieves global energy efficiency optimization of equipment group control;

[0089] Integrated harmonic analysis, vibration monitoring, and environmental parameter fusion diagnosis proactively identify hidden equipment failures and energy efficiency degradation, reducing passive operation and maintenance.

[0090] Standardized integrated power and weak current design simplifies the construction process and reduces the need for professional coordination. At the same time, the visualization platform provides intuitive energy consumption management and remote control capabilities.

[0091] The present invention collaboratively optimizes building energy consumption through dynamic prediction and intelligent control, responds to environmental changes in real time, and accurately coordinates the operation of multiple devices, breaking through the limitations of traditional models such as lag and coupling interference. It combines multi-source data fusion diagnosis to actively identify hidden dangers and reduce operation and maintenance costs. At the same time, standardized integrated design simplifies construction management, provides intuitive and visual control, and realizes a full-cycle closed loop of energy efficiency improvement and intelligent operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 This is the principle block diagram of the energy consumption prediction and energy-saving control system. DETAILED DESCRIPTION

[0093] The technical solution of the present invention is further described below with reference to the accompanying drawings and specific embodiments.

[0094] Example

[0095] 1. As Figure 1As shown, the energy consumption prediction and energy-saving control system proposed in the present invention includes a data acquisition and communication module, an energy consumption prediction and optimization module, an intelligent control execution module, an energy management and visualization module, and a system coordination mechanism. Each part is described in detail below.

[0096] 2. The data acquisition and communication module is used to collect the operating parameters, environmental parameters and equipment status of building mechanical and electrical equipment in real time, and realize data aggregation and transmission through a multi-protocol bus; specifically, it includes:

[0097] The distributed sensor unit integrates temperature sensor, humidity sensor, CO2 concentration sensor, PM2.5 concentration sensor, current transformer, voltage transformer, liquid level sensor and flow sensor, including:

[0098] The temperature sensor uses PT1000 platinum resistance with an accuracy of ±0.1°C and is installed in the air conditioning return duct, chilled water supply and return pipes and outdoor environment monitoring points;

[0099] The current transformer is an open-type Rogowski coil with a range of 0-500A and an accuracy of 0.5. It is embedded in the power distribution cabinet of the fan, water pump and cold and heat source system;

[0100] The PM2.5 concentration sensor uses the principle of laser scattering and has a detection range of 0-1000μg / m 3 , resolution 1μg / m 3 , deployed at the air inlet of the fresh air unit and key indoor areas;

[0101] The equipment status monitoring unit is connected to the fan power distribution cabinet, water pump power distribution cabinet, lighting distribution box, cold and heat source system and elevator control cabinet through the LONWORKS fieldbus to collect the equipment's operating status, fault signals and energy efficiency parameters in real time, including:

[0102] The fan power distribution cabinet has a built-in motor protector to monitor the three-phase current imbalance rate, overload alarm and insulation resistance value;

[0103] The controller of the cold and heat source system uploads the chiller evaporator / condenser pressure, compressor operating frequency and energy efficiency ratio (COP) through the Modbus protocol;

[0104] The bus communication unit adopts a dual-redundant network architecture, including the LONWORKS fieldbus layer and the TCP / IP Ethernet layer, including:

[0105] The LONWORKS bus layer supports free topology, a transmission rate of 78kbps, a maximum number of 64 nodes, and is used for real-time control signal transmission at the device level;

[0106] The TCP / IP Ethernet layer connects the data collectors in each area through a fiber optic ring network. The transmission protocol is BACnet / IP and supports OPCUA data subscription and publishing.

[0107] The protocol converter is embedded in the PDC controller, converting LONWORKS data packets into JSON format and uploading them to the cloud database via the MQTT protocol.

[0108] It should be noted that the edge computing functions of the bus communication unit include:

[0109] The lightweight TensorFlow Lite model deployed on the power distribution cabinet controller takes as input features the device current I, voltage V, ambient temperature envTenv, and operating time t. The output is the load forecast value predPpred for the next hour. The model structure is as follows:

[0110] Input layer: 4 neurons;

[0111] Hidden layer: 2 layers, 8 neurons in each layer, activation function is ReLU;

[0112] Output layer: 1 neuron, activation function is linear;

[0113] The edge node calculation cycle is 5 minutes, and the prediction results are uploaded to the cloud through the MQTT protocol. The consistency check algorithm uses the root mean square error (RMSE) threshold judgment, and the formula is:

[0114]

[0115] Among them, P i is the load value predicted by the i-th edge node; P cloud is the load value predicted by the cloud model for the i-th time, n = 12, is the total number of data points in the verification period,

[0116] If RMSE > 5%, the cloud model resynchronization is triggered, otherwise the edge prediction result is accepted.

[0117] The data acquisition and communication module integrates multiple sensor types (temperature, current, environmental quality, etc.) with distributed monitoring units to achieve comprehensive, real-time collection of building mechanical and electrical equipment operating parameters, environmental conditions, and equipment fault signals, resolving the data silos and protocol compatibility issues of traditional systems. A dual-redundant network architecture ensures stable and real-time data transmission, providing multi-dimensional, high-precision basic data support for subsequent modules, ensuring accurate perception of equipment status and environmental changes.

[0118] 3. In this embodiment, the energy consumption prediction and optimization module is used to predict building energy consumption trends and generate optimization control instructions based on dynamic load calculation, machine learning algorithms, and energy efficiency analysis models. Specifically, it includes:

[0119] The dynamic load calculation unit calculates the cooling load of the air conditioning system in real time based on the chilled water supply and return temperature difference ΔT, the flow rate F, and the thermodynamic formula Q = K × F × ΔT, where K is the chilled water thermal coefficient, which is dynamically corrected based on the water quality test report using the formula K = 4186 × (1 - 0.0025 × TDS), where TDS is the total dissolved solids content (ppm). Historical load data is stored in a time series database, and load curves are generated at 15-minute intervals. Periodic features are extracted using a sliding window algorithm.

[0120] The machine learning prediction unit uses the ARIMA model to predict short-term energy consumption trends and combines it with an LSTM neural network to process nonlinear features. Specifically, input features include outdoor temperature, indoor occupancy density, cumulative equipment operating time, electricity price time intervals, and historical energy consumption data. The model is trained using the Adam optimizer with a learning rate of 0.001, a batch size of 64, and 500 training cycles. The error rate for predicting energy consumption over the next 24 hours is ≤3%.

[0121] The energy efficiency analysis unit evaluates the energy efficiency status of individual devices in real time by building an energy efficiency ratio (EER) model. The EER is calculated as EER = cooling capacity (kW) / input power (kW), where cooling capacity is calculated from the chilled water flow rate and the supply and return water temperature difference. The energy efficiency degradation warning trigger condition is: an EER drop of 10% for three consecutive hours or a current harmonic distortion rate exceeding THD>8%.

[0122] This embodiment, based on dynamic load calculation and machine learning algorithms, breaks through the limitations of traditional static models, capturing in real time the impact of dynamic factors such as occupancy density and outdoor environment on energy consumption, and accurately predicting building energy consumption trends. Combined with energy efficiency analysis models, it assesses equipment energy efficiency in real time and generates optimized control instructions, providing a scientific basis for intelligent regulation. This transition from "historical data statistics" to "dynamic trend prediction" improves the foresight and targeted nature of energy management.

[0123] 4. Intelligent control execution module, which is used to execute adaptive control strategies, equipment group control logic and preset energy-saving modes, and dynamically adjust the operating parameters of electromechanical equipment; the intelligent control execution module includes the following units:

[0124] The adaptive control unit uses a fuzzy PID algorithm to adjust the operating parameters of the air conditioning system, including:

[0125] The fuzzy rule base is set as "if the return air temperature deviation is large and the rate of change is fast, then increase the proportional coefficient KP", and the KP dynamic range is 0.5-2.0;

[0126] The opening of the fresh air valve is controlled in sections according to the CO2 concentration: 30% when CO2 < 800ppm, 50% when 800-1200ppm, and 100% when CO2 > 1200ppm;

[0127] The fan frequency conversion frequency is PID-adjusted according to the deviation between the return air temperature set value and the actual value, and the output frequency range is 10-50Hz;

[0128] The adaptive control unit further includes a multivariable decoupling control unit to resolve the coupled interference of return air temperature, CO2 concentration and PM2.5 concentration in the air conditioning system, specifically including:

[0129] Establish state space equations to describe multivariable relationships:

[0130]

[0131] Return air temperature change rate, ΔT: Return air temperature deviation, u valve : Water valve opening, C CO2 : Indoor CO2 concentration; a1, a2, a3: Temperature dynamic model coefficients; CO2 concentration change rate; u damper : Fresh air valve opening; Q vent : ventilation rate; b1, b2: CO2 dynamic model coefficients; PM2.5 concentration change rate; u filter : filter efficiency; c1, c2: PM2.5 dynamic model coefficients;

[0132] Feedforward-feedback composite control is adopted. The feedforward controller pre-adjusts the fresh air valve opening according to the change of CO2 concentration, and the feedback controller adjusts the water valve opening according to the return air temperature deviation. The control rate is:

[0133]

[0134] Among them, e CO2 =C CO2,set -C CO2 :CO2 concentration deviation, e T =T set -T return : Return air temperature deviation; K p , K i , K d : PID parameter for CO2 concentration control, K p ′,K i ′,K d ′: PID parameters for temperature control, dynamically adjusted by fuzzy rules.

[0135] The equipment group control unit performs group control of the chiller, circulating water pump and cooling tower, specifically including:

[0136] The round-robin strategy is based on the cumulative running time of the equipment. Each time the unit with the shortest running time is started, the unit with the longest running time is shut down first when shutting down.

[0137] The opening of the pressure differential bypass valve is controlled by PID, with the target value being a pressure differential of 0.2 MPa for the supply and return pipes, and an adjustment accuracy of ±0.01 MPa;

[0138] Small temperature difference compensation technology adjusts the cooling tower fan speed to make the cooling water return temperature close to the optimal set value of the chiller (32℃±0.5℃);

[0139] Energy-saving strategy library, which stores the following preset control modes:

[0140] Schedule control mode: Divide by weekdays / holidays, set the lighting system to fully open from 07:00 to 09:00, and control the lights in intervals from 22:00 to 06:00;

[0141] Infrared sensing control mode: garage lighting detects traffic flow through microwave radar sensors, with a trigger delay of 30 seconds and turns off 60 seconds after the car leaves;

[0142] Fire linkage mode: After receiving the fire alarm signal, the fresh air valve is forced to close, the smoke exhaust valve is opened, and the non-fire water pump is stopped.

[0143] In this embodiment, adaptive control algorithms (such as fuzzy PID and multivariable decoupling) and device group control strategies effectively address the interference problem of multi-subsystem coupling, enabling dynamic adjustment and coordinated optimization of operating parameters for equipment such as air conditioning, ventilation, and lighting. Preset energy-saving modes (such as schedule control, sensor control, and fire linkage) automatically adapt to the needs of different scenarios, ensuring environmental comfort while reducing ineffective energy consumption, and promoting the transformation of equipment from "independent control" to "global intelligent collaboration."

[0144] 5. In this embodiment, the energy management and visualization module is used to measure energy consumption by item, diagnose equipment failures, and provide a visual monitoring interface; with the help of high-precision smart meters, energy consumption is measured by item, and combined with the expert system rule engine and multi-source data fusion analysis, equipment failures (such as bearing wear, winding aging) are accurately diagnosed and energy efficiency degradation is warned. The three-dimensional visualization interface presents energy consumption distribution, equipment operating status and fault location in real time, supports drill-down analysis and remote intervention from the system level to the equipment level, improves the level of refinement and decision-making efficiency of operation and maintenance management, and realizes the integrated management of "data visualization - fault diagnosis - strategy intervention". The energy management and visualization module includes the following units:

[0145] The energy consumption metering unit uses a 0.2S-class smart meter to measure the energy consumption of the equipment in sub-items, including: the energy consumption of the cold and heat source systems is measured by a CT power meter, with a data update interval of 1 minute;

[0146] The lighting circuit energy consumption collects the current, voltage and power factor of each circuit through the RS485 bus;

[0147] The fault diagnosis unit uses an expert system rule engine with the following rules: If the water pump current suddenly increases by 20% and lasts for 5 seconds, it is determined to be a blade jam fault; if the fan three-phase current imbalance rate is greater than 15%, a bearing wear warning is triggered. The fault diagnosis unit further includes a harmonic distortion rate analysis unit to detect motor faults, specifically including:

[0148] The current signal is collected and fast Fourier transform (FFT) is performed to calculate the total harmonic distortion (THD). The formula is:

[0149]

[0150] Among them, I1 is the effective value of the fundamental current, I h is the effective value of the hth harmonic current, THD is the total harmonic distortion;

[0151] Among them, set the fault judgment rules:

[0152] If THD>8% and lasts for 10 minutes, the "motor winding insulation aging" alarm is triggered;

[0153] If the third harmonic component I3 / I1>5% and the fifth harmonic component I5 / I1>3%, it is judged as "power supply voltage unbalance";

[0154] Combined with the vibration sensor data, when THD exceeds the limit and the vibration acceleration RMS value is greater than 4m / s 2 When the fault occurs, it is determined to be a "combined fault of bearing wear and electrical failure".

[0155] The human-computer interaction unit builds a 3D visualization interface based on WebGL technology. Its specific functions include:

[0156] Dynamically display the water pipe topology of the cold and hot source system, and mark the high temperature alarm point in red;

[0157] Energy consumption ratio charts are displayed by air conditioning, lighting, water pumps, and elevators, and support drilling down to the sub-item equipment level;

[0158] The remote manual intervention interface provides forced start and stop, parameter modification and policy import functions. The operation permissions are divided into three levels: administrator, engineer and guest.

[0159] 6. The system coordination mechanism realizes two-way data interaction between modules through a real-time database, ensuring that the prediction results drive the dynamic adjustment of the control strategy. Specifically, it includes:

[0160] The real-time database uses a time series database (InfluxDB), with storage granularity set to 1-second raw data. The data partitioning strategy is divided by device type. The data retention period for the cooling and heating source systems is 30 days, and the data retention period for the lighting system is 7 days.

[0161] The output instructions of the energy consumption prediction module are written into the PLC register of the control execution module through the OPCUA protocol. The specific parameter configuration is:

[0162] The OPCUA server address is opc.tcp: / / 192.168.1.100:4840, and the session timeout is set to 300 seconds;

[0163] The write register address range is 40001-40050, the data type is Float32, and the encoding rule is Little-Endian;

[0164] After receiving the feedback data from the control execution module, the energy management module uses a sliding window weighted average algorithm to update the energy efficiency analysis model weights. The formula is:

[0165] W new =α·W old +(1-α)·ΔW

[0166] Among them, W new is the updated model weight matrix, W old is the model weight matrix before updating, α is the forgetting factor (with a value of 0.7-0.9), ΔW is the model error gradient of the current period, and the weight matrix is ​​updated every hour.

[0167] In the implementation, a fully closed-loop collaborative system of "data acquisition - prediction and analysis - control execution - feedback optimization" is constructed through a real-time database and a two-way data exchange protocol, ensuring real-time data sharing and dynamic linkage between modules. A collaborative verification mechanism between edge computing and cloud-based models improves local data processing efficiency while ensuring the consistency of prediction results. This enhances the system's rapid response capabilities and overall robustness to complex scenarios, forming a highly efficient operating model of "prediction-driven control and control feedback optimization."

[0168] The present invention provides a complete chain from data perception, predictive analysis to intelligent control and visual management, breaking through the bottlenecks of traditional systems in data integration, prediction accuracy, control coordination and operation and maintenance efficiency, realizing refined management of building energy consumption, systematic improvement of equipment energy efficiency and intelligent transformation of operation and maintenance decision-making, providing solid technical support for green buildings and efficient energy conservation.

[0169] The above specific embodiments are merely several optional embodiments of the present invention. Based on the technical solutions of the present invention and the relevant inspirations of the above embodiments, those skilled in the art may make various alternative improvements and combinations to the above specific embodiments.

Claims

1. Energy consumption prediction and energy-saving control system, characterized in that: include: Data acquisition and communication module, used to collect operating parameters, environmental parameters and equipment status of building mechanical and electrical equipment in real time, and realize data aggregation and transmission through multi-protocol bus; Energy consumption prediction and optimization module, which is used to predict building energy consumption trends and generate optimization control instructions based on dynamic load calculation, machine learning algorithms and energy efficiency analysis models; Intelligent control execution module, used to execute adaptive control strategies, equipment group control logic and preset energy-saving modes, and dynamically adjust the operating parameters of electromechanical equipment; Energy management and visualization module, used for itemized energy consumption measurement, equipment failure diagnosis, and a visual monitoring interface; The system coordination mechanism realizes two-way data interaction between modules through a real-time database, ensuring that the prediction results drive the dynamic adjustment of the control strategy.

2. The energy consumption prediction and energy-saving control system according to claim 1, characterized in that: The data acquisition and communication module includes: Distributed sensor unit, integrating temperature sensor, humidity sensor, CO2 concentration sensor, PM2.5 concentration sensor, current transformer, voltage transformer, liquid level sensor and flow sensor; The equipment status monitoring unit is connected to the fan power distribution cabinet, water pump power distribution cabinet, lighting distribution box, cold and heat source system and elevator control cabinet through the LONWORKS fieldbus to collect the equipment's operating status, fault signals and energy efficiency parameters in real time. The bus communication unit adopts a dual redundant network architecture, including the LONWORKS fieldbus layer and the TCP / IP Ethernet layer.

3. The energy consumption prediction and energy-saving control system according to claim 1, characterized in that: The energy consumption prediction and optimization module includes: The dynamic load calculation unit calculates the cooling load of the air conditioning system in real time based on the chilled water supply and return temperature difference ΔT, the flow rate F, and the thermodynamic formula Q = K × F × ΔT, where K is the chilled water thermal coefficient, which is dynamically corrected based on the water quality test report using the formula K = 4186 × (1 - 0.0025 × TDS), where TDS is the total dissolved solids content. Historical load data is stored in a time series database, and load curves are generated at regular time intervals. Periodic features are extracted using a sliding window algorithm. The machine learning prediction unit uses the ARIMA model to predict short-term energy consumption trends and combines it with the LSTM neural network to process nonlinear features; The energy efficiency analysis unit evaluates the energy efficiency status of a single device in real time by building a device energy efficiency ratio model.

4. The energy consumption prediction and energy-saving control system according to claim 1, characterized in that: The intelligent control execution module includes the following units: Adaptive control unit, using fuzzy PID algorithm to adjust the operating parameters of the air conditioning system; Equipment group control unit, which performs group control on chillers, circulating water pumps and cooling towers; Energy-saving strategy library, which stores the following preset control modes: Schedule control mode: divided by working days / holidays, set the lighting system to be fully open during the day and the lights to be on at night; Infrared sensing control mode: The garage lighting detects the traffic flow through microwave radar sensors, triggers a delay time, and turns off at a fixed time after the car leaves; Fire linkage mode: After receiving the fire alarm signal, the fresh air valve is forcibly closed, the smoke exhaust valve is opened, and the non-fire water pump is stopped.

5. The energy consumption prediction and energy-saving control system according to claim 1, characterized in that: The energy management and visualization module includes the following units: Energy consumption metering unit uses 0.2S-level smart meter to measure equipment energy consumption in sub-items; The fault diagnosis unit uses an expert system rule engine with the following rules: If the water pump current suddenly increases by 20% and lasts for more than 3 seconds, it is determined to be a blade jam fault; If the three-phase current imbalance rate of the fan is greater than 15%, the bearing wear warning will be triggered; The human-computer interaction unit builds a three-dimensional visualization interface based on WebGL technology.

6. The energy consumption prediction and energy-saving control system according to claim 1, characterized in that: The system coordination mechanism specifically includes: The real-time database adopts a time series database, and the storage granularity is set to 1-second raw data. The data partitioning strategy is divided by device type. The data retention period of the cold and heat source system is 30 days, and the data retention period of the lighting system is 7 days. The output instruction of the energy consumption prediction module is written into the PLC register of the control execution module through the OPCUA protocol; After receiving the feedback data from the control execution module, the energy management module uses a sliding window weighted average algorithm to update the energy efficiency analysis model weights. The formula is: W new =α·W old +(1-a)·ΔW Among them, W new is the updated model weight matrix, W old is the model weight matrix before updating, α is the forgetting factor, ΔW is the model error gradient of the current period, and the weight matrix is ​​updated every hour.

7. The energy consumption prediction and energy-saving control system according to claim 2, characterized in that: The edge computing functions of the bus communication unit include: The lightweight TensorFlow Lite model deployed on the power distribution cabinet controller takes as input features the device current I, voltage V, ambient temperature envTenv, and operating time t. The output is the load forecast value predPpred for the next hour. The model structure is as follows: Input layer: 4 neurons; Hidden layer: 2 layers, 8 neurons in each layer, activation function is ReLU; Output layer: 1 neuron, activation function is linear; The edge node calculation cycle is 5 minutes, and the prediction results are uploaded to the cloud through the MQTT protocol. The consistency check algorithm uses the root mean square error threshold for judgment. The formula is: Among them, P i is the load value predicted by the i-th edge node; P cloud is the load value predicted by the cloud model for the i-th time, n = 12, is the total number of data points in the verification period, If RMSE > 5%, the cloud model resynchronization is triggered, otherwise the edge prediction result is accepted.

8. The energy consumption prediction and energy-saving control system according to claim 1, characterized in that: The adaptive control unit further includes a multivariable decoupling control unit for resolving the coupling interference of return air temperature, CO2 concentration and PM2.5 concentration in the air conditioning system, specifically including: Establish state space equations to describe multivariable relationships: in, Return air temperature change rate, ΔT: Return air temperature deviation, u valve : Water valve opening, C CO2 : Indoor CO2 concentration; a1, a2, a3: Temperature dynamic model coefficients; CO2 concentration change rate; u damper : Fresh air valve opening; Q vent : ventilation rate; b1, b2: CO2 dynamic model coefficients; PM2.5 concentration change rate; u filter : filter efficiency; c1, c2: PM2.5 dynamic model coefficients; Feedforward-feedback composite control is adopted. The feedforward controller pre-adjusts the fresh air valve opening according to the change of CO2 concentration, and the feedback controller adjusts the water valve opening according to the return air temperature deviation. The control rate is: Among them, e CO2 =C CO2,set -C CO2 :CO2 concentration deviation, e T =T set -T return : Return air temperature deviation; K p , K i , K d : PID parameter for CO2 concentration control, K p ′,K i ′,K d ′: PID parameters for temperature control, dynamically adjusted by fuzzy rules.

9. The energy consumption prediction and energy-saving control system according to claim 5, characterized in that: The fault diagnosis unit further includes a harmonic distortion rate analysis unit for detecting motor faults, specifically including: Collect the current signal and perform fast Fourier transform to calculate the total harmonic distortion rate. The formula is: Among them, I1 is the effective value of the fundamental current, I h is the effective value of the hth harmonic current, and THD is the total harmonic distortion.

10. The energy consumption prediction and energy-saving control system according to claim 9, characterized in that: Set the fault judgment rules: If THD>8% and lasts for 10 minutes, the "motor winding insulation aging" alarm is triggered; If the third harmonic component I3 / I1>5% and the fifth harmonic component I5 / I1>3%, it is judged as "power supply voltage unbalance"; Combined with the vibration sensor data, when THD exceeds the limit and the vibration acceleration RMS value is greater than 4m / s 2 When the fault occurs, it is determined to be a "combined fault of bearing wear and electrical failure".

Images