Load calculation method based on electricity consumption prediction of subway air conditioner and participating in virtual power plant

Abstract

The load calculation method based on subway air conditioner power consumption prediction and participating in the virtual power plant comprises the steps that influence factor data information of subway air conditioner loads is collected; according to a thermodynamic formula and an empirical formula, establishing a thermoelectricity calculation model of the subway air conditioner by combining air conditioner electricity utilization characteristics; the precision and the accuracy of a thermoelectric calculation model of the subway air conditioner are improved by utilizing a database on which subway load actual measurement is based and utilizing a machine learning technology; a thermoelectric calculation model of the metro air conditioner is utilized to obtain metro air conditioner electricity consumption prediction curves of a plurality of points in the future based on weather forecast published by meteorological data and the simulated passenger flow volume of an operation company; constructing an underground station air conditioner cluster scheduling strategy of a main substation area according to the scheduling load curve issued by the urban power grid virtual power plant platform and in combination with the subway air conditioner electricity consumption prediction curve and aiming at the air conditioner load levels of different stations; and further simulating and building a subway air conditioner virtual power plant load prediction and management cloud platform model, and carrying out real-time load prediction and scheduling.

Description

A load calculation method based on subway air conditioning power consumption forecasting and participation in virtual power plants. Technical Field

[0001] This invention relates to the field of virtual power plant technology, and in particular to a method for load calculation based on subway air conditioning power consumption forecasting and its participation in virtual power plant calculation. Background Technology

[0002] With the continuous advancement of the national dual-carbon strategy and the construction of new power systems, my country's virtual power plants have entered a period of comprehensive construction. Subways, with their large loads and high electricity consumption, play a crucial role in urban power grids, making the participation of subways in virtual power plant regulation a consideration. The prerequisite for subway participation in virtual power plant regulation is accurate load forecasting. Knowing the exact load at the next point will determine the subway's load and electricity consumption planning.

[0003] The load of a subway power supply system can be divided into two main categories: traction load and power and lighting load. From the perspective of load forecasting, traction load is determined by the train timetable and is relatively easy to forecast. However, the power and lighting load is largely comprised of air conditioning in underground stations. Subway air conditioning operation is influenced by many factors, including weather, passenger flow, and the parameters of the air conditioning's operating cycle. Underground stations also exhibit heat dissipation lag and thermal inertia. Therefore, accurately forecasting subway air conditioning load becomes both a key focus and a challenge in subway virtual power plant load forecasting.

[0004] The information disclosed in the background section is only for enhancing the understanding of the background of this invention, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] To address the aforementioned technical issues, this invention considers the randomness and volatility of photovoltaic power generation. When photovoltaic power is used in the regulation of the subway virtual power plant, it is significantly affected by weather. Unexpected factors can lead to substandard regulation and deviations from the predicted curve, necessitating timely adjustments to meet the standards of the sampling points. Furthermore, underground station air conditioning accounts for a large proportion of the overall subway system's power and lighting load, and its operation is relatively stable and regular. Therefore, this invention considers coupling the subway's photovoltaic and dynamic lighting systems for adjustment. When the subway's photovoltaic power supply fails to meet the load aggregation margin of the virtual power plant, emergency compensation is provided using the subway's air conditioning system, achieving joint scheduling of subway photovoltaic power supply and air conditioning.

[0006] Therefore, this invention discloses a load calculation method based on subway air conditioning power consumption prediction and participation in a virtual power plant, characterized by the following steps:

[0007] S100. Collect data on factors affecting subway air conditioning load, and establish a thermoelectric calculation model for subway air conditioning based on thermodynamic formulas and empirical formulas, combined with the characteristics of air conditioning power consumption.

[0008] S200: Utilizing a database based on actual metro load measurements and employing machine learning techniques, the accuracy and precision of the thermoelectric calculation model for metro air conditioning are improved.

[0009] S300: Using the thermoelectric calculation model of subway air conditioning, based on weather forecasts published by meteorological data and simulated passenger flow by the operating company, the predicted power consumption curves of subway air conditioning at 96 points in the future are obtained.

[0010] S400. Based on the dispatch load curve issued by the urban power grid virtual power plant platform and combined with the predicted electricity consumption curve of subway air conditioning, construct a cluster dispatch strategy for air conditioning in underground stations in the main substation area for different station air conditioning load levels.

[0011] S500. Based on the aforementioned air conditioning cluster scheduling strategy, machine learning technology is used to simulate and build a cloud platform model for load prediction and management of the subway air conditioning virtual power plant. Real-time load prediction and scheduling calculations are performed to achieve clustered joint management of the load status of the subway air conditioning virtual power plant.

[0012] Compared with the prior art, the present invention has the following advantages:

[0013] This invention addresses the issues of predictive calculation models for subway air conditioning and the participation of subway air conditioning cluster scheduling in the construction of virtual power plants in the power grid. It analyzes the influencing factors and electricity consumption characteristics of subway air conditioning load, establishes an accurate predictive calculation model for subway air conditioning load through simulation analysis, focuses on the participation of subway air conditioning in virtual power plant scheduling, and proposes a load calculation method based on subway air conditioning electricity consumption prediction and participation in virtual power plants. This enables real-time load prediction and scheduling calculation to achieve clustered joint management of the load status of subway air conditioning virtual power plants. Attached Figure Description

[0014] Various other advantages and benefits of the present invention will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The drawings are for illustrative purposes only and are not intended to limit the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can derive other drawings from these drawings without creative effort. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0015] Figure 1 is a flowchart illustrating a method for calculating load based on subway air conditioning power consumption prediction and participation in a virtual power plant, according to one embodiment of the present invention. Detailed Implementation

[0016] Specific embodiments of the present invention will now be described in detail with reference to the figures. While specific embodiments of the invention are shown in the figures, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.

[0017] It should be noted that certain terms are used in the specification and claims to refer to specific components. Those skilled in the art will understand that different terms may be used to refer to the same component. This specification and claims do not distinguish components based on differences in terminology, but rather on differences in function. The terms "comprising" or "including" used throughout the specification and claims are open-ended and should be interpreted as "comprising but not limited to." The following descriptions are preferred embodiments for carrying out the invention; however, these descriptions are for the purpose of understanding the general principles of the specification and are not intended to limit the scope of the invention. The scope of protection of this invention is determined by the appended claims.

[0018] To facilitate understanding of the embodiments of the present invention, the following will provide further explanation and description with reference to the figures and specific embodiments, and the figures do not constitute a limitation on the embodiments of the present invention.

[0019] In one embodiment, the present invention discloses a load calculation method based on subway air conditioning power consumption forecasting and participation in a virtual power plant, characterized by comprising the following steps:

[0020] S100. Collect data on factors affecting subway air conditioning load, and establish a thermoelectric calculation model for subway air conditioning based on thermodynamic formulas and empirical formulas, combined with the characteristics of air conditioning power consumption.

[0021] S200: Utilizing a database based on actual metro load measurements and employing machine learning techniques, the accuracy and precision of the thermoelectric calculation model for metro air conditioning are improved.

[0022] S300: Using the thermoelectric calculation model of subway air conditioning, based on weather forecasts published by meteorological data and simulated passenger flow by the operating company, the predicted power consumption curves of subway air conditioning at 96 points in the future are obtained.

[0023] S400. Based on the dispatch load curve issued by the urban power grid virtual power plant platform and combined with the predicted electricity consumption curve of subway air conditioning, construct a cluster dispatch strategy for air conditioning in underground stations in the main substation area for different station air conditioning load levels.

[0024] S500. Based on the aforementioned air conditioning cluster scheduling strategy, machine learning technology is used to simulate and build a cloud platform model for load prediction and management of the subway air conditioning virtual power plant. Real-time load prediction and scheduling calculations are performed to achieve clustered joint management of the load status of the subway air conditioning virtual power plant.

[0025] In another embodiment, the method for load calculation based on subway air conditioning power consumption forecasting and participation in a virtual power plant, wherein,

[0026] The subway air conditioning cluster dispatch strategy is based on the urban power grid virtual power plant dispatch curve and the daily subway air conditioning power consumption forecast curve for dispatch calculation. This allows the subway to respond to demand-side issues by controlling the on / off status of underground station air conditioning in the dispatch area through cluster control on the subway virtual power plant dispatch platform. It also corrects deviations by comparing the forecast curve with the actual consumption direction, and monitors and manages the virtual power plant load of large subway users in real time.

[0027] In another embodiment, the load calculation method based on subway air conditioning power consumption prediction and participation in a virtual power plant includes dynamic influencing factors such as outdoor temperature and real-time passenger flow, as well as static factors such as station volume, geographical depth, and the number of station entrances and exits.

[0028] In another embodiment, the method for load calculation based on subway air conditioning power consumption forecasting and participation in a virtual power plant, wherein in step S100:

[0029] Since there are many factors affecting the operation of subway air conditioning, including weather, passenger flow, and the parameters of the air conditioning cycle, and underground stations also have heat dissipation lag and thermal inertia, black box modeling is adopted to extract empirical formulas for subway load heat and power calculation.

[0030] In another embodiment, the method for calculating the load of a virtual power plant based on the prediction of subway air conditioning power consumption involves converting the thermal energy changes of subway stations into the subway air conditioning power load through the model, and studying the subway air conditioning load modeling problem by combining the power consumption characteristics of air conditioning.

[0031] In another embodiment, the method for calculating the load of a virtual power plant based on the prediction of electricity consumption for subway air conditioning includes the following step S200: collecting data on factors affecting subway air conditioning load using a temperature and humidity collector and a subway load remote recording device to form a database based on actual subway load measurements; and using machine learning technology to improve the accuracy and precision of the thermoelectric calculation model for subway air conditioning.

[0032] In another embodiment, the method for calculating the load of a virtual power plant based on the prediction of subway air conditioning power consumption includes step S300: based on weather forecast data provided by meteorological monitoring and passenger flow forecast data of the subway operating company, combined with the air conditioning load thermoelectric calculation model, the future subway air conditioning power consumption at multiple points is predicted.

[0033] In another embodiment, step S100 includes:

[0034] S101: Static and Dynamic Data Acquisition

[0035] Static data collection (one-time acquisition, valid for a long time):

[0036] Station volume V (unit: m) 3 ): The internal space volume of the station is calculated from the architectural design drawings and is used to calculate the heat capacity.

[0037] Internal surface area A (unit: m) 2 ): The total area of ​​the station's internal enclosure structure, including walls, ceilings, and floors, is used to calculate the heat transfer area.

[0038] Overall thermal conductivity of the wall (k) (unit: W / (m) 2 ·K): Reflects the overall thermal conductivity of wall materials, obtained through material thermal engineering handbooks or actual measurements.

[0039] Average burial depth of the station ℎh (unit: m): The average vertical distance from the top of the station to the ground, which affects the soil temperature lag.

[0040] Rated power of air conditioning system (unit: kW): The maximum input electrical power of the air conditioning unit under standard operating conditions.

[0041] Air conditioner rated energy efficiency ratio (COPrated) (dimensionless): The ratio of cooling capacity to input electrical power under rated operating conditions.

[0042] Dynamic data acquisition (continuous real-time acquisition):

[0043] Outdoor dry-bulb temperature Tout(t) (unit: °C): The temperature of the air outside the station, collected every 5 minutes by a meteorological sensor.

[0044] Real-time passenger flow within the station, Npassenger(t) (unit: people / minute): Statistically recorded in real time through the gate counting and video recognition system.

[0045] Station interior temperature Tin(t) (unit: °C): average value of multiple sensors in the station hall and platform.

[0046] Real-time electrical power of the air conditioning system, Pelec(t) (unit: kW): recorded every 15 minutes by a smart meter.

[0047] S102: The heat balance equation is established based on energy conservation. The instantaneous heat load Qload(t) (unit: kW) of the station is:

[0048]

[0049] Heat transfer load of building envelope Qenv(t):

[0050]

[0051] Where Tsoil(t) is the soil temperature (in °C) at depth h at time t, estimated by an empirical model:

[0052]

[0053] Tavg: Local annual average temperature (unit: °C); τ: Soil temperature lag time (unit: hours); λ: Soil thermal diffusivity (unit: m), obtained from geological data.

[0054] Heat load of people, Qpeople(t):

[0055]

[0056] q p Sensible heat dissipation power per person (unit: kW / person), taken as 0.08 kW / person (light activity).

[0057] Equipment heat load Qequip(t):

[0058]

[0059] P equip Total rated power of lighting, escalators, and other equipment (unit: kW); f equip (t): Equipment simultaneous utilization rate (dimensionless), determined according to the operating schedule.

[0060] S103: The air conditioning power Pelec(t) (unit: kW) in the thermoelectric conversion model is:

[0061]

[0062] COP(t): Real-time energy efficiency ratio of the air conditioner, obtained by fitting the performance curve.

[0063]

[0064] Qrated: Rated cooling capacity of the air conditioner (unit: kW); a, b, c, d: Fitting coefficients, obtained from regression of equipment performance data.

[0065] Pstandby: Standby power of the air conditioning system (unit: kW), including control cabinet, circulating fan, etc.

[0066] S104: The model parameter calibration uses the least squares method to calibrate parameters k, qp, a, b, c, d, etc., based on historical data, so as to minimize the mean square error between the model calculated value Pelec_calc(t) and the actual measured value, for example, Pelec_meas(t).

[0067] S105: Model validation uses historical data from another period to validate the model, calculating the mean absolute percentage error (MAPE). If MAPE < 15%, the model passes validation; otherwise, return to S104 to recalibrate or adjust the model structure.

[0068] The S100 step described above establishes a physical model based on thermodynamic principles, quantifying the complex heat-to-electricity conversion process of subway air conditioning and providing a calculable basis for load forecasting. Its technical contribution lies in the fact that, within the scenario addressed and the technical problem to be solved, the above steps comprehensively consider static building characteristics, dynamic environment, and operational factors. In particular, the introduction of a soil temperature lag model and the dynamic characteristics of air conditioning energy efficiency ratio makes the model more realistic and physically interpretable and adaptable compared to traditional empirical formula methods.

[0069] In another embodiment, step S200 includes:

[0070] S201: Dataset Construction Collect at least one year of historical data to construct a sample set {(Xi,yi)}, where:

[0071] The eigenvector Xi includes:

[0072] Time characteristics: current hour (1-24), current day of the week (1-7), holiday marker (0 / 1).

[0073] Environmental characteristics: current outdoor temperature, forecast temperature for the next hour, and outdoor humidity.

[0074] Operational characteristics: current passenger flow and predicted passenger flow for the next hour.

[0075] Historical load characteristics: Air conditioning power at the previous moment Pelec(t−1), average power at the previous time period.

[0076] Station status characteristics: Current indoor temperature Tin(t).

[0077] It should be noted that the above feature vectors are clearly multi-dimensional feature vectors, which serve as the source of the multi-dimensional feature vectors received by the input layer in the subsequent S203 step of this invention. Among them,

[0078] The current hour (24-hour clock) reflects intraday cyclical patterns, such as morning and evening rush hours;

[0079] The current day of the week reflects the cyclical pattern within the week, such as the difference between weekdays and weekends;

[0080] Does it reflect load pattern changes on special dates such as holidays?

[0081] The current outdoor temperature is the most critical environmental factor that directly affects the air conditioning load;

[0082] The temperature forecast for the next hour takes into account the lag effect and trend of temperature changes;

[0083] Current passenger flow is a human factor that directly affects the heat load inside the station;

[0084] The predicted passenger flow for the next hour takes into account the trend of passenger flow changes and the delay effect;

[0085] The air conditioner's power consumption at the previous moment is used to capture the time autocorrelation of the load;

[0086] The average power over the past hour is used to reflect the short-term statistical characteristics of the load;

[0087] The current indoor temperature is used to reflect the current operating status and adjustment needs of the air conditioning system;

[0088] Tag yi: Actual air conditioning power at the current moment, Pelec(t).

[0089] S202: Feature engineering preprocesses the original features of the above feature vectors:

[0090] Apart from holiday markers or binary logic of 0 and 1, all other features in the feature vector are scaled to the [0,1] interval.

[0091] In addition to the current time t, lag features are constructed, such as Pelec(t−2) and Pelec(t−3), to use historical values ​​in the time series as features for prediction at the current time.

[0092] In addition, construct mobile statistical features such as the average temperature over the past 3 hours and the rate of change in passenger flow over the past 1 hour.

[0093] It should be noted that, for the problem of predicting subway air conditioning load, constructing lag features can capture the time autocorrelation of the load, reflect the short-term patterns of load changes, and improve the model's ability to model time dependence.

[0094] S203: Model selection and training employ a Long Short-Term Memory (LSTM) network, with the following structure:

[0095] Input layer: Receives multidimensional feature vectors.

[0096] First LSTM layer: 64 neurons, return sequence.

[0097] The second LSTM layer has 32 neurons and returns the output at the last time step.

[0098] Dropout layer: Dropout rate 0.2 to prevent overfitting.

[0099] Fully connected output layer: 1 neuron, outputting the predicted load value. Training uses the mean squared error (MSE) loss function, Adam optimizer (learning rate 0.001), batch size 32, training epochs 150, and early stopping.

[0100] S204: Model fusion involves weighted fusion of the physical model output Pelec_physics(t) and the LSTM model output Pelec_LSTM(t).

[0101]

[0102] The weights w1 and w2 satisfying w1+w2=1 are determined on the validation set through grid search. For example, the optimal values ​​are w1=0.3 and w2=0.7.

[0103] S205: The model update mechanism uses the latest data to incrementally train the LSTM model every quarter, updating the network weights to adapt to seasonal changes and device performance degradation.

[0104] The S200 step described above utilizes machine learning to mine complex nonlinear patterns of load changes from historical data, compensating for the shortcomings of pure physical models in addressing unmodeled factors. Its technical contribution lies in proposing a fusion framework of physical and data-driven models, which preserves the interpretability of physical principles while improving prediction accuracy, and enables the model to adapt through a periodic update mechanism.

[0105] In another embodiment, step S300 includes:

[0106] S301: Input Data Preparation

[0107] Obtain the 24-hour weather forecast released by the meteorological department, with a time resolution of 15 minutes, and a temperature sequence of 24×4=96 points {Toutfc(1),…,Toutfc(96)}.

[0108] Obtain the passenger flow forecast sequence {Nfc(1),…,Nfc(96)} for the next 24 hours provided by the operating company, also with a 15-minute granularity.

[0109] Perform anomaly detection and interpolation on the input data.

[0110] S302: Multi-step rolling forecasting uses a rolling forecasting strategy to generate 96 points sequentially.

[0111] Starting from the true feature vector X(t0) at the current time t0;

[0112] Inputting X(t0) into the fusion model yields... ;use And construct X(t0+1) from the forecast data at time t0+1;

[0113] Repeat the previous step until you get... .

[0114] S303: Uncertainty quantification is based on historical prediction error statistics. The standard deviation σ(t) of the error at each prediction point is calculated, and the 95% confidence interval is:

[0115]

[0116] S304: Post-processing and smoothing: The predicted sequence is filtered by moving average (window width 3 points) to eliminate random fluctuations and make the curve smoother and more reasonable.

[0117] S305: Prediction result storage stores the 96-point prediction curve and its confidence interval into the time series database for subsequent scheduling.

[0118] The specific implementation of step S300 involves applying the trained model to future scenarios to generate high-temporal-resolution load forecast curves, providing direct input for scheduling decisions. Its technical contribution lies in achieving rolling forecasting and uncertainty quantification, providing not only point estimates but also forecast intervals, thus enhancing the robustness of the scheduling system.

[0119] In another embodiment, step S400 includes:

[0120] S401: Load Baseline Delineation and Station Classification

[0121] Calculate the hourly average load curve of each station for a typical historical day (such as weekdays and weekends) as the baseline Pbase(t);

[0122] Classification based on the degree of deviation between predicted load and baseline:

[0123] High-load stations: Forecasted load continues > 1.2 × Pbase(t);

[0124] Medium-load stations: The predicted load is between 0.8×Pbase(t) and 1.2×Pbase(t);

[0125] Low-load stations: Forecasted load is consistently < 0.8 × Pbase(t).

[0126] S402: The scheduling rule base design defines rules for the types of scheduling instructions issued by virtual power plants.

[0127] Peak shaving command (requiring a reduction in load ΔPtarget):

[0128] Level 1 Response: The air conditioning temperature setting of all stations will be increased by 0.5℃ (estimated reduction of 8-10%).

[0129] Level 2 response: Air conditioning in non-core areas of low- and medium-load stations will operate at reduced frequencies (further reduced by 5-8%).

[0130] Level 3 Response: High-load stations implement a rotational stop strategy (run for 15 minutes, stop for 5 minutes, and then reduce by 10-15%).

[0131] Valley filling command (requiring an increase in load): Reverse operation, such as adjusting the set temperature or pre-cooling.

[0132] S403: The optimization model aims to minimize discomfort and scheduling cost. A mixed-integer linear programming model is established, where:

[0133] Decision variables xi,j,t: Whether station i implements level j measures during time period t (0 or 1);

[0134] Objective function:

[0135]

[0136] α,β: weighting coefficients; Tset: set temperature (e.g., 26℃); cj: scheduling cost coefficient of the j-th level measure.

[0137] Constraints:

[0138] The total reduction meets the requirement: ∑i,jeffi,j⋅xi,j,t≥ΔPtarget(t);

[0139] Temperature comfort constraint: Tmin ≤ Tin, i(t) ≤ Tmax;

[0140] Equipment operation constraints: such as minimum start-up and shutdown time, maximum adjustment frequency; where effi,j is the baseline power reduction of station i when implementing measure j, calibrated through simulation.

[0141] S404: Strategy Simulation and Evaluation. Run the scheduling strategy in a simulated environment. Evaluation metrics include: instruction tracking error, duration of temperature exceedance, and number of equipment actions. If metrics are not met, adjust the rules or optimize the parameters.

[0142] S405: Generate executable control instructions by translating the optimization results into a specific set of control instructions, for example:

[0143] Station A: 10:00-10:15, set temperature increased to 27℃.

[0144] Station B: 10:00-10:05, air conditioning unit No. 1 will be turned off.

[0145] The specific implementation of the S400 steps outlines a differentiated cluster scheduling strategy and employs optimization algorithms to achieve a balance between meeting grid demand and passenger comfort. Its technical contribution lies in proposing a hierarchical response mechanism and a multi-objective optimization model, enabling flexible control of subway air conditioning load and transforming it into a dispatchable resource similar to a virtual power plant.

[0146] In another embodiment, step S500 includes:

[0147] S501: The system architecture adopts a microservice architecture, with core services including: data acquisition service, load forecasting service, scheduling decision service, equipment control service, and visualization service.

[0148] Data acquisition service: Access sensor and meter data via MQTT / OPC UA protocol, and obtain meteorological and passenger flow data via API.

[0149] Load forecasting service: Loads the fusion model, performs rolling forecasts periodically, and generates and updates the 96-point curve.

[0150] Dispatch decision service: Receives instructions from virtual power plants, runs optimization algorithms, and generates control strategies.

[0151] Equipment control service: Interacts with the station's BAS system via the BACnet / IP protocol to issue temperature settings and start / stop commands.

[0152] Visualization service: Provides a web monitoring interface based on Vue.js + ECharts.

[0153] Message bus: Use RabbitMQ to implement asynchronous communication between services.

[0154] S502: Real-time data stream processing uses Apache Flink to build the stream processing pipeline.

[0155] Access raw data streams in real time from the data source.

[0156] Perform data cleaning, format conversion, and outlier removal.

[0157] Calculate real-time features (such as moving averages and rates of change).

[0158] The processed data stream is written to a time series database (InfluxDB) and pushed to the prediction service.

[0159] S503: Implementation of Core Platform Functions

[0160] A prediction task is triggered every 15 minutes to update the curve and evaluate the accuracy in order to achieve automatic prediction.

[0161] After the virtual power plant platform issues the dispatch curve, the dispatch decision service completes the optimization calculation and generates the control instruction set within 1 minute;

[0162] The equipment control service executes commands and monitors actual load changes in real time. If the deviation exceeds a threshold (e.g., 10%), an alarm is triggered and a backup strategy is activated.

[0163] It stores all operation logs, prediction results, and actual data, and supports historical queries and analysis.

[0164] S504: Platform Deployment and Maintenance

[0165] Each service is packaged as a Docker image and managed through Kubernetes orchestration to achieve automatic scaling. It also integrates Prometheus + Grafana to monitor the status of each service, resource utilization, prediction error and other metrics, and sets thresholds for alerts. At the same time, it uses the ELK stack (Elasticsearch, Logstash, Kibana) to centrally manage logs.

[0166] S505: Security and Access Management

[0167] The transport layer uses TLS / SSL to encrypt stored data; user authentication is implemented based on OAuth 2.0, and operation permissions are assigned according to roles (dispatcher, administrator, read-only user); in addition, all critical operations are recorded to meet security audit requirements.

[0168] The specific implementation of the S500 steps mentioned above is used to integrate all the aforementioned algorithms, steps and strategies into a runnable and maintainable cloud platform, realizing an automated closed loop from data to control. Its technical contribution lies in the design of a highly available and scalable microservice architecture to achieve standard docking with the virtual power plant platform, enabling the subway air conditioning cluster to participate in grid regulation as a plug-and-play flexible load resource.

[0169] Although the embodiments of the present invention have been described above in conjunction with the figures, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A method for load calculation based on subway air conditioning power consumption forecasting and participation in a virtual power plant, characterized in that, The process includes the following steps: S100, collecting data on factors influencing subway air conditioning load, and establishing a thermoelectric calculation model for subway air conditioning based on thermodynamic and empirical formulas and the characteristics of air conditioning power consumption; S200, using a database based on actual subway load measurements and machine learning technology to improve the accuracy and precision of the thermoelectric calculation model for subway air conditioning; S300, using the thermoelectric calculation model for subway air conditioning, based on weather forecasts published by meteorological data and simulated passenger flow by the operating company, obtaining predicted power consumption curves for subway air conditioning at 96 points in the future; S400, based on the dispatch load curve issued by the urban power grid virtual power plant platform and combined with the predicted power consumption curve for subway air conditioning, constructing a cluster dispatch strategy for underground station air conditioning in the main substation area for different station air conditioning load levels; S500, based on the aforementioned cluster dispatch strategy for air conditioning, using machine learning technology to simulate and build a cloud platform model for subway air conditioning virtual power plant load prediction and management, performing real-time load prediction and dispatch calculations to achieve clustered joint management of the load status of the subway air conditioning virtual power plant.

2. The load calculation method based on subway air conditioning power consumption forecasting and participation in a virtual power plant according to claim 1, characterized in that, The subway air conditioning cluster dispatch strategy is based on the urban power grid virtual power plant dispatch curve and the daily subway air conditioning power consumption forecast curve for dispatch calculation. This allows the subway to respond to demand-side issues by controlling the on / off status of underground station air conditioning in the dispatch area through cluster control on the subway virtual power plant dispatch platform. It also corrects deviations by comparing the forecast curve with the actual consumption direction, and monitors and manages the virtual power plant load of large subway users in real time.

3. The load calculation method based on subway air conditioning power consumption forecasting and participation in a virtual power plant according to claim 1, characterized in that, The data on factors affecting air conditioning load includes dynamic factors such as outdoor temperature and real-time passenger flow, as well as static factors such as station volume, geographical location depth, and the number of station entrances and exits.

4. The load calculation method based on subway air conditioning power consumption forecasting and participation in a virtual power plant according to claim 1, characterized in that, In step S100: Since there are many factors affecting the operation of subway air conditioning, including weather, passenger flow, and parameters of the air conditioning working cycle, and underground stations also have heat dissipation lag and thermal inertia, black box modeling is adopted to extract empirical formulas for subway load heat and power calculation.

5. The load calculation method based on subway air conditioning power consumption forecasting and participation in a virtual power plant according to claim 4, characterized in that, The aforementioned subway load thermoelectric calculation involves converting the thermal energy changes of subway stations into the electrical load of subway air conditioning through this model, and combining the electrical characteristics of air conditioning to study the modeling problem of subway air conditioning load.

6. The load calculation method based on subway air conditioning power consumption forecasting and participation in a virtual power plant according to claim 1, characterized in that, In step S200: Data on factors affecting subway air conditioning load are collected using a temperature and humidity collector and a subway load remote recording device to form a database based on actual subway load measurements. Machine learning technology is then used to improve the accuracy and precision of the thermoelectric calculation model for subway air conditioning.

7. The load calculation method based on subway air conditioning power consumption forecasting and participation in a virtual power plant according to claim 1, characterized in that, In step S300: Based on weather forecast data provided by meteorological monitoring and passenger flow forecast data from the subway operating company, combined with the air conditioning load heat and power calculation model, the future subway air conditioning power consumption at multiple points is predicted.

Images