Multi-energy supply collaborative optimization system and method of building energy storage system based on digital twinning

Abstract

The invention discloses a multi-energy supply collaborative optimization system and method for a building energy storage system based on digital twinning. The system comprises a wind and light power generation system, an energy storage system, a block chain space-time database, an intelligent driving system, a power system, a regional building power system and building unit power utilization which are connected with one another. The method comprises the following steps: establishing a target function, finely calculating energy loss in a charge-discharge process, and calculating carbon emission for measuring unit energy consumption; the method comprises the following steps: collecting and mathematical modeling, constructing a prediction model attention mechanism, optimizing an algorithm by sequencing temperature, irradiance and wind speed, and finally obtaining the minimum energy storage system loss rate and the minimum carbon emission intensity. By scheduling the energy storage system, the dependence of the area on the traditional power grid can be reduced, the resource configuration is optimized, the consumption of renewable energy sources is promoted, the operation cost of the building area is reduced, and the carbon footprint of the intelligent area is reduced.

Description

Technical Field

[0001] The present invention relates to the fields of energy management, digital twins, and intelligent control technologies, and in particular to a multi-energy supply collaborative optimization system and method for a building energy storage system based on digital twins. Background Art

[0002] As the global energy transition progresses, building energy consumption continues to grow as a proportion of total energy consumption. Traditional energy systems are no longer able to meet the demand for efficient, clean, and flexible energy supply. Against the backdrop of the "dual carbon" goals and the rapid development of new energy, a multi-energy supply collaborative optimization system for building energy storage systems based on digital twins can integrate multiple energy forms, achieve efficient energy utilization and flexible allocation, provide stable and reliable energy supply for buildings, and contribute to the realization of energy transition goals.

[0003] Due to the lag in monitoring and the lack of real-time monitoring capabilities in traditional systems, energy consumption and distribution cannot be monitored promptly, leading to energy waste. Data collection and processing are slow, making it difficult to quickly respond to changes in energy demand. Furthermore, due to the lack of collaborative optimization mechanisms, different energy sources often operate independently, this lack of collaborative optimization leads to inefficient energy utilization. Information sharing between systems prevents overall optimization, resulting in inefficient energy utilization. In terms of system reliability, traditional systems have weak fault prediction capabilities and insufficient backup power. Traditional systems lack the ability to predict and prevent equipment failures, often requiring repairs only after a failure occurs, impacting system operation. The lack of real-time monitoring and data analysis capabilities prevents the early identification of potential failures. Furthermore, in the event of grid failures or insufficient renewable energy generation, the lack of backup power can easily lead to energy supply interruptions. Energy storage systems have limited capacity and cannot meet energy demands in emergencies.

[0004] In terms of operational economics, high energy procurement and equipment maintenance costs are a problem. Traditional energy storage systems struggle to dynamically adjust charging and discharging strategies based on electricity price fluctuations, resulting in high energy procurement costs. Lack of intelligent control capabilities prevents the ability to capitalize on price fluctuations to generate economic benefits. Furthermore, frequent equipment starts and stops, coupled with excessive use, increases maintenance and replacement costs. A lack of precise maintenance strategies leads to accelerated equipment wear and tear.

[0005] In terms of intelligence, due to low levels of automation, traditional systems rely primarily on manual experience and fixed rules for energy management and control, making them difficult to adapt to complex and changing operating conditions. The lack of automated and intelligent management tools, advanced data analysis, and intelligent control capabilities makes it impossible to achieve automated and intelligent energy management. Decision-making relies on manual judgment and lacks data support, making it prone to errors and leading to inefficient management.

[0006] In terms of adaptability, the system's lack of flexibility makes it difficult to cope with complex operating conditions and flexibly adjust energy management and control strategies. This lack of flexibility and adaptability makes it impossible to meet diverse energy needs. Furthermore, when faced with external environmental changes (such as weather conditions and energy market price fluctuations), traditional systems lack the flexibility to make timely adjustments. The fixed system design makes it difficult to quickly respond to external changes.

[0007] Based on the above situation, the present invention proposes a multi-energy supply collaborative optimization system and method for a building energy storage system based on digital twins, aiming to achieve sustainable development in the construction field through digital twins and multi-energy supply technologies. Summary of the Invention

[0008] Purpose of the invention: The purpose of the present invention is to provide a multi-energy supply collaborative optimization system and method for a building energy storage system based on digital twins.

[0009] Technical solution: The multi-energy supply collaborative optimization system of the building energy storage system based on digital twin described in the present invention includes an interconnected wind and solar power generation system, an energy storage system, a blockchain spatiotemporal database, an intelligent drive system, a power system, a regional building power system, and electricity consumption of the building unit; the wind and solar system transmits electric energy to the energy storage system and transmits the data to the blockchain spatiotemporal database. The intelligent drive system collects multi-dimensional data such as battery status, photovoltaic output, load demand and electricity price in the building energy storage system in real time through the blockchain spatiotemporal database, and uses algorithms to accurately predict photovoltaic power generation fluctuations and dynamically adjust energy storage charging and discharging strategies. The energy storage system responds to demand and interacts with the power market and buildings.

[0010] Furthermore, the energy storage system uses digital twins to predict the system status for the next 24 hours, continuously optimizes control instructions, dynamically adjusts charging and discharging plans, simulates different energy storage configuration schemes through digital twins, optimizes initial investment and maps energy storage data to the building information model, and guides energy storage to shaving peaks and filling valleys through electricity price signals to reduce peak loads.

[0011] Furthermore, the intelligent drive system will collect information from the blockchain spatiotemporal database and the power database and execute corresponding programs, including an intelligent control module, a data analysis module, a monitoring module and a prediction module.

[0012] Furthermore, the blockchain spatiotemporal database builds a full life cycle management system for energy data. By deploying smart meters, temperature control sensors, and photovoltaic irradiance meters, it collects data such as energy storage, wind and solar output, and load curves in real time and stores them in a spatiotemporal cube model.

[0013] The multi-energy supply collaborative optimization method of a building energy storage system based on digital twins of the present invention comprises the following steps:

[0014] (1) Establish an objective function with energy storage operation efficiency and carbon emission intensity as optimization targets, and reduce energy loss during charging and discharging by improving the weight coefficient and maximizing energy storage operation efficiency and minimizing carbon emission intensity;

[0015] (2) Calculate the energy loss during the charging and discharging process and the carbon emissions per unit energy consumption;

[0016] (3) Dynamically adjust the weight coefficient according to the real-time grid carbon emission intensity and

[0017] (4) Collection and mathematical modeling, building a prediction model attention mechanism by sorting temperature, irradiance, and wind speed, importing data into the model structure and finally outputting the predicted data;

[0018] (5) Designing the state space to include real-time operating parameters and environmental conditions, ensuring that decisions take into account energy storage health, real-time supply and demand, and environmental impacts by fully covering system conditions;

[0019] (6) Designing the action space by adjusting the charging and discharging power and interacting with the grid;

[0020] (7) Optimize the algorithm, design the reward function to integrate efficiency and carbon emission targets, and add SOC stability constraints; smooth the instantaneous fluctuations of wind and solar power generation through second-level control, adjust the charge and discharge in real time through the PI controller to suppress wind and solar fluctuations; use EMA filtering to smooth the MODRL instructions and reduce battery loss;

[0021] (8) Dynamic weight adjustment and health feedback: When the energy storage health status SOH feedback is less than 80%, the upper limit of charge and discharge power is automatically reduced to 90%. When SOH is less than 70%, an early warning is triggered and maintenance is recommended.

[0022] (9) Finally, the minimized energy storage system loss rate and the minimized carbon emission intensity are obtained, that is, the minimized energy storage system loss rate and the minimized carbon emission intensity under the optimal fitness condition.

[0023] Furthermore, the objective function of step (1) is expressed as:

[0024]

[0025] in represents the weight coefficient 1, represents the weight coefficient 2, κ represents the energy storage loss rate, and λ represents the carbon emission intensity.

[0026] Furthermore, the energy loss in step (2) is calculated as follows:

[0027]

[0028] where η ch represents the charging efficiency, η dis Indicates discharge efficiency, P ch storage Indicates the charging power at time t, P dis storage (t) represents the discharge power, P load (t) represents the actual building load measured at time t;

[0029] Carbon emissions are calculated as:

[0030]

[0031] where ∈ grid represents the carbon emission factor of power grid supply, ∈ storage represents the carbon emission factor of energy storage charging and discharging, P grid (t) represents the net electric power exchange between the grid and the building at time t.

[0032] Furthermore, the step (4) includes constructing a prediction model attention mechanism by sorting the temperature, irradiance, and wind speed, importing the data into the model structure and finally outputting the prediction data.

[0033] α t =Softmax(W a tanh(W s h t LSTM ))

[0034] where α t Indicates the calculation of attention weight, W a Represents the linear transformation matrix for calculating attention weights, W s Indicates that the LSTM hidden state is mapped to the attention space, h t LSTM represents the LSTM hidden state at time t;

[0035]

[0036] where z t represents weighted fusion historical features;

[0037]

[0038] in It represents the predicted value including future load and wind and solar power output, W z Indicates mapping the fusion features to the output space, W o represents the weights of the fully connected layer that generates the final prediction, z tRepresents the weighted feature vector after attention mechanism fusion.

[0039] Furthermore, the action space design in step (6) is expressed as:

[0040] at=[ΔP ch , ΔP dis , Grid Interaction Flag]

[0041] ΔP ch / dis ∈[-10%, +10%] indicates the charging and discharging power adjustment step, Grid Interaction Flag∈{0, 1}, 0 indicates that only local renewable energy is used, and 1 indicates that grid interaction is allowed.

[0042] Furthermore, the energy storage health state SOH expression in step (8) is:

[0043]

[0044] where N cycle Indicates the cumulative number of cycles, D o D avg Indicates the average depth of discharge.

[0045] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0046] (1) In terms of energy distribution optimization, through real-time monitoring and optimization and multi-energy collaborative optimization, the operating status and energy consumption of various energy equipment in the building can be monitored in real time. Through data analysis and optimization algorithms, the supply of different energy forms can be reasonably allocated to ensure efficient use of energy in the building. Through real-time monitoring and multi-energy supply regulation, the digital twin system can flexibly adjust the charging and discharging strategy, optimize energy distribution, reduce energy waste, and improve energy utilization efficiency. In addition, through digital twin technology, the system integrates and collaboratively optimizes various energy equipment and systems in the building, realizes the complementary and collaborative operation of multiple energy forms, and further improves energy utilization efficiency.

[0047] (2) In terms of virtual commissioning and optimization, the digital twin system can monitor and optimize the operating status of the energy storage system in real time. By analyzing historical and real-time data, it can accurately predict energy demand, reasonably allocate energy, avoid waste, and improve efficiency. At the same time, it can simulate the operating conditions of different energy storage devices, including parameters such as battery capacity and charge and discharge rates, to help select the optimal energy storage device configuration in specific scenarios and improve the efficiency and stability of the energy storage system. Through multi-objective optimization algorithms, the system can achieve higher energy utilization efficiency, lower operating costs, stronger reliability, and better adaptability;

[0048] (3) In terms of multi-energy collaborative optimization, the simulation modeling and optimization control strategy of the multi-energy flow integrated energy system is realized through the construction of multi-energy flow virtual entities, a mirror collaborative interaction mechanism including physical digital space is established, and a multi-energy flow virtual entity including electricity, solar power and wind power is constructed. Through the distribution of equipment status information and deep neural network power prediction, the overall adjustment of the source end and the energy storage end is achieved; through the reasonable planning of batteries, the utilization rate of wind and solar energy is significantly improved. The system can accurately predict the fluctuation of grid load and electricity price trends, dynamically adjust the charging and discharging strategy, improve the utilization rate of new energy, and reduce the problem of wind and solar power abandonment. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Schematic diagram of the system structure of the present invention;

[0050] Figure 2 It is the algorithm flow chart of the present invention;

[0051] Figure 3 This is the distribution diagram of the digital twin system of the present invention. DETAILED DESCRIPTION

[0052] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0053] The multi-energy supply collaborative optimization system of the building energy storage system based on digital twin described in the present invention includes interconnected wind and solar power generation systems, energy storage systems, blockchain spatiotemporal databases, intelligent drive systems, power systems, regional building power systems, and building unit electricity consumption.

[0054] Wind and solar power systems integrate wind and photovoltaic technologies to efficiently convert wind and solar energy. Their complementarity manifests itself in multiple dimensions: Temporally, daily and seasonal complementarity ensures stable power generation. For example, solar energy is stronger during the day, while wind energy is stronger at night. In summer, solar radiation is strong but wind is weak, while in winter, the opposite is true. Energy storage can mitigate fluctuations. Spatially, microgrid-level complementarity, such as combining photovoltaic and wind turbines on urban rooftops, leverages height differences to achieve three-dimensional power generation, improving energy supply stability and efficiency.

[0055] As the core of multi-energy collaborative optimization, the energy storage system can smooth out fluctuations in wind and solar power and ensure a balance between supply and demand. It utilizes a lithium-ion battery storage BMS to monitor and regulate charging and discharging power in real time, reducing the standard deviation of power fluctuations. It uses digital twins to predict system status for the next 24 hours, optimize control instructions on a rolling basis, and dynamically adjust charging and discharging plans, thereby reducing the curtailment rate and achieving precise regulation and efficient operation of the energy system. Digital twins simulate the economic and reliability of different energy storage configurations, optimize initial investments, and map energy storage data to a building information model (BIM) to achieve three-dimensional visualization of "energy flow, space, and equipment." Furthermore, the energy storage system can be used for grid interaction and demand response, using electricity price signals to guide energy storage to shaving peak loads and fill valleys, reducing peak loads.

[0056] The blockchain spatiotemporal database builds a full lifecycle management system for energy data through three core functions: data ownership confirmation, trusted evidence storage, and spatiotemporal traceability. By deploying smart meters, temperature control sensors, photovoltaic irradiance meters, and other equipment, it collects real-time data on energy storage, wind and solar output, and load curves. This data is stored in a spatiotemporal cube model (a multidimensional data structure that integrates spatial and temporal dimensions), storing 3D spatial coordinate data with timestamps. This data is uploaded to each sensor node, generating a hash value for the data. This hash value is then distributed through IPFS (a decentralized distributed file system protocol), with only the hash value written to the blockchain. The spatiotemporal data is annotated with geotags and timestamps to ensure that the data cannot be tampered with. The data is then connected to the Oracle Bridge (a system that achieves comprehensive connectivity from the physical network layer to the application data layer through a layered architecture design), enabling data interoperability with external systems. The blockchain spatiotemporal database is used to construct a spatiotemporal index, the R-tree index is used to accelerate spatial queries, and the B+ tree index is established in the time dimension to optimize historical backtracking; then multi-source data fusion is performed to integrate photovoltaic power generation data (kW), energy storage charging and discharging power (kW), building load (kW) and other time series data to generate a building energy fingerprint.

[0057] The intelligent drive system consists of an intelligent control module, a data analysis module, a monitoring module, and a prediction module. This system will collect information from the blockchain spatiotemporal database and the power database and execute corresponding programs;

[0058] The intelligent control module and data analysis module utilize three core functions: real-time perception, dynamic optimization, and autonomous decision-making. These modules enable efficient collaboration between the energy storage system, wind and solar power generation, building loads, and the power grid. Using the blockchain's spatiotemporal database, they collect real-time environmental parameters from different areas within the building and analyze operational data from energy devices. Using optimization algorithms, they develop optimal energy management and control strategies. These strategies include energy storage device charging and discharging schedules, the allocation ratios of different energy sources, and adjustments to device operating modes to maximize energy efficiency, minimize energy costs, and maintain a stable energy supply. The intelligent control system communicates and interacts with the control systems of various energy devices within the building, converting the developed control strategies into specific control instructions and sending them to the corresponding devices for execution. By monitoring and analyzing energy market prices in real time, combined with the building's energy needs and the charging and discharging costs of energy storage devices, the intelligent control system can develop optimal energy procurement and utilization strategies, reducing energy costs and improving energy efficiency. For example, when electricity prices are low, the grid is prioritized for powering the energy storage devices. When electricity prices are high, the energy storage devices are prioritized for power, reducing reliance on the grid and thus reducing electricity bills.

[0059] The monitoring and prediction modules, through three core functions: spatiotemporal data modeling, dynamic simulation, and decision boundary prediction, form a predictive engine for energy flow. Based on data from the monitoring module and the blockchain's spatiotemporal database, the prediction module employs data mining, statistical analysis, and machine learning techniques to clean, preprocess, and extract features. It then constructs models for energy demand, energy supply, equipment performance, and the energy market. By analyzing multi-dimensional information such as historical building energy consumption data, occupant activity patterns, and meteorological data, the prediction module accurately estimates a building's electricity and other energy needs over different time periods. This facilitates proactive planning of energy production, distribution, and storage strategies, ensuring a dynamic balance between energy supply and demand and avoiding energy shortages or surpluses. For energy storage systems, the prediction system can incorporate meteorological forecast data to predict the power and timing of renewable energy generation. Furthermore, it can predict price fluctuations and supply stability of traditional energy sources (such as grid power), providing a basis for developing charging and discharging strategies for building energy storage systems. This helps optimize energy procurement costs and supply structures, and by rationally scheduling the charging and discharging of energy storage equipment, capitalizes on price differences to generate economic benefits, thereby improving the economic efficiency of building energy storage systems.

[0060] The multi-energy supply collaborative optimization system of the building energy storage system based on digital twin according to claim 1 is characterized in that the data analysis module is provided with the following objective function and the specific steps are as follows

[0061] S1: Establish an objective function with energy storage operation efficiency and carbon emission intensity as optimization targets. By improving the weight coefficients and maximizing energy storage operation efficiency and minimizing carbon emission intensity, we can reduce energy loss during charging and discharging, improve overall energy conversion efficiency, reduce dependence on high-carbon power grids, increase the local absorption rate of renewable energy, and reduce carbon emissions throughout the entire life cycle. The objective function is as follows

[0062]

[0063] in Weight coefficient 1, Weight coefficient 2, κ: energy storage loss rate, λ: carbon emission intensity

[0064] 1. Calculate the energy loss during the charging and discharging process

[0065]

[0066] where η ch : Charging efficiency, η dis : discharge efficiency, P ch storage t: charging power at all times, P dis storage (t): discharge power, Pload (t): Actual building load measured at time t (kw)

[0067] 2. Calculate carbon emissions per unit of energy consumption

[0068]

[0069] where ∈ grid : Carbon emission factor of grid power supply (kgCO2 / kWh), ∈ storage : Energy storage charging and discharging carbon emission factor (related to battery type), P grid (t) Net power exchange between the grid and the building at time t

[0070] 3. Weight coefficient Dynamic adjustment based on real-time grid carbon emission intensity

[0071]

[0072] 4. Collection and mathematical modeling, building a prediction model attention mechanism by sorting temperature, irradiance, and wind speed, importing data into the model structure and finally outputting the predicted data

[0073] α t =Softmax(W a tanh(W s h t LSTM ))

[0074] where α t : Calculate attention weight, W a : Calculate the linear transformation matrix of attention weight, W s : LSTM hidden state is mapped to the attention space, h t LSTM : LSTM hidden state at time t

[0075]

[0076] where z t : Weighted fusion of historical features

[0077]

[0078] in Contains the predicted value of future load and wind and solar power output, W z : Map the fusion features to the output space, W o : The fully connected layer weights that generate the final prediction, z t : Weighted feature vector after attention mechanism fusion

[0079] 5. Design the state space to include real-time operating parameters and environmental conditions, ensuring that decisions take into account energy storage health, real-time supply and demand, and environmental impacts by fully covering system conditions.

[0080] s t =[SOC(t),P PV (t), P Wind (t), P load (t), η ch / dis (t), SOH(t)]

[0081] Among them, SOH: online monitoring by electrochemical impedance spectroscopy (EIS);

[0082] 6. Design the action space by adjusting the charging and discharging power and interacting with the grid. Discrete the action space to reduce strategy complexity while retaining flexibility.

[0083] at==[ΔP ch , ΔP dis , Grid Interaction Flag]

[0084] where ΔP ch / dis ∈[-10%, +10%]: charge and discharge power adjustment step, Grid Interaction Flag∈{0, 1}: 0 means only using local renewable energy, 1 means allowing grid interaction

[0085] 7. Optimize the algorithm, design the reward function based on comprehensive efficiency and carbon emission targets, and add SOC stability constraints

[0086]

[0087] Where μ: SOC stability weight, which prevents SOC from being at extreme values ​​(such as <20% or >80%) for a long time

[0088] 8. Optimize the algorithm to perform second-level control (supercapacitor) to smooth out instantaneous fluctuations in wind and solar power generation (such as power drops caused by cloud cover). Use the PI controller to adjust charging and discharging in real time to suppress wind and solar fluctuations.

[0089]

[0090] 9. Optimize the algorithm and use EMA filtering to smooth the MODRL instruction to reduce battery loss P opt (t) = α·P RL (t)+(1-α)·P opt (t-1)

[0091] Where α = 0.2, which reduces the impact of power mutation on battery life.

[0092] 10. Dynamic weight adjustment and health feedback. When the energy storage health status (SOH) is less than 80%, the charge and discharge power limit is automatically reduced to 90%. When the SOH is less than 70%, an early warning is triggered and maintenance is recommended. By adaptively adjusting and optimizing target priorities, the energy storage life is extended and the system robustness is improved.

[0093]

[0094] Where N cycle : Cumulative number of cycles, D o D avg : Average depth of discharge.

[0095] 11. Finally, we get the minimized energy storage system loss rate and the minimized carbon emission intensity, that is, the minimized energy storage system loss rate and the minimized carbon emission intensity of F(x) under the optimal fitness condition, and the F(x) we want is the minimized energy storage system loss rate and the minimized carbon emission intensity, output.

Claims

1. A multi-energy supply collaborative optimization system for building energy storage systems based on digital twins, characterized by: It includes interconnected wind and solar power generation systems, energy storage systems, blockchain space-time databases, intelligent drive systems, power systems, regional building power systems, and electricity consumption of building units; the wind and solar systems transmit electrical energy to the energy storage system and transmit data to the blockchain space-time database. The intelligent drive system collects multi-dimensional data such as battery status, photovoltaic output, load demand and electricity price in the building energy storage system in real time through the blockchain space-time database, and uses algorithms to accurately predict photovoltaic power generation fluctuations and dynamically adjust energy storage charging and discharging strategies. The energy storage system responds to power market and buildings on demand to interact with electricity.

2. The multi-energy supply collaborative optimization system of the building energy storage system based on digital twin according to claim 1 is characterized in that: The energy storage system uses digital twins to predict the system status for the next 24 hours, continuously optimize control instructions, and dynamically adjust charging and discharging plans. It also simulates different energy storage configuration schemes through digital twins, optimizes initial investment, and maps energy storage data to the building information model. It uses electricity price signals to guide energy storage to shaving peaks and filling valleys, thereby reducing peak loads.

3. The multi-energy supply collaborative optimization system of the building energy storage system based on digital twin according to claim 1 is characterized in that: The intelligent drive system will collect information from the blockchain spatiotemporal database and the power database and execute corresponding programs, including an intelligent control module, a data analysis module, a monitoring module and a prediction module.

4. The multi-energy supply collaborative optimization system of the building energy storage system based on digital twin according to claim 1 is characterized in that: The blockchain spatiotemporal database builds a full life cycle management system for energy data. By deploying smart meters, temperature control sensors, and photovoltaic irradiance meters, it collects data such as energy storage, wind and solar output, and load curves in real time and stores them in a spatiotemporal cube model.

5. A multi-energy supply collaborative optimization method for a building energy storage system based on digital twins, characterized in that: The steps include: (1) Establish an objective function with energy storage operation efficiency and carbon emission intensity as optimization targets, and reduce energy loss during charging and discharging by improving the weight coefficient and maximizing energy storage operation efficiency and minimizing carbon emission intensity; (2) Calculate the energy loss during the charging and discharging process and the carbon emissions per unit energy consumption; (3) Dynamically adjust the weight coefficient according to the real-time grid carbon emission intensity and (4) Collection and mathematical modeling, building a prediction model attention mechanism by sorting temperature, irradiance, and wind speed, importing data into the model structure and finally outputting the predicted data; (5) Designing the state space to include real-time operating parameters and environmental conditions, ensuring that decisions take into account energy storage health, real-time supply and demand, and environmental impacts by fully covering system conditions; (6) Designing the action space by adjusting the charging and discharging power and interacting with the grid; (7) Optimize the algorithm, design the reward function to integrate efficiency and carbon emission targets, and add SOC stability constraints; smooth the instantaneous fluctuations of wind and solar power generation through second-level control, adjust the charge and discharge in real time through the PI controller to suppress wind and solar fluctuations; use EMA filtering to smooth the MODRL instructions and reduce battery loss; (8) Dynamic weight adjustment and health feedback: When the energy storage health status SOH feedback is less than 80%, the upper limit of charge and discharge power is automatically reduced to 90%. When SOH is less than 70%, an early warning is triggered and maintenance is recommended. (9) Finally, the minimized energy storage system loss rate and the minimized carbon emission intensity are obtained, that is, the minimized energy storage system loss rate and the minimized carbon emission intensity under the optimal fitness condition.

6. The multi-energy supply collaborative optimization method of a building energy storage system based on digital twin according to claim 5 is characterized in that: The objective function of step (1) is expressed as: in represents the weight coefficient 1, represents the weight coefficient 2, κ represents the energy storage loss rate, and λ represents the carbon emission intensity.

7. The multi-energy supply collaborative optimization method of a building energy storage system based on digital twin according to claim 5 is characterized in that: The energy loss in step (2) is calculated as follows: where η ch represents the charging efficiency, η dis Indicates discharge efficiency, P ch storage Indicates the charging power at time t, P dis storage (t) represents the discharge power, P load (t) represents the actual building load measured at time t; Carbon emissions are calculated as: where ∈ grid represents the carbon emission factor of power grid supply, ∈ storage represents the carbon emission factor of energy storage charging and discharging, P grid (t) represents the net electric power exchange between the grid and the building at time t.

8. The multi-energy supply collaborative optimization method of a building energy storage system based on digital twin according to claim 5 is characterized in that: The step (4) includes constructing a prediction model attention mechanism by sorting the temperature, irradiance, and wind speed, importing the data into the model structure and finally outputting the prediction data. α t =Softmax(W a ·tanh(W s h t LSTM )) where α t Indicates the calculation of attention weight, W a Represents the linear transformation matrix for calculating attention weights, W s Indicates that the LSTM hidden state is mapped to the attention space, h t LSTM represents the LSTM hidden state at time t; where z t represents weighted fusion historical features; in It represents the predicted value including future load and wind and solar power output, W z Indicates mapping the fusion features to the output space, W o represents the weights of the fully connected layer that generates the final prediction, z t Represents the weighted feature vector after attention mechanism fusion.

9. The multi-energy supply collaborative optimization method of a building energy storage system based on digital twin according to claim 5 is characterized in that: The step (6) represents the action space design as follows: at=[ΔP ch ,ΔP dis ,Grid Interaction Flag] ΔP ch / dis ∈[-10%, +10%] indicates the charging and discharging power adjustment step, Grid Interaction Flag∈{0, 1}, 0 indicates that only local renewable energy is used, and 1 indicates that grid interaction is allowed.

10. The multi-energy supply collaborative optimization method of a building energy storage system based on digital twin according to claim 5, characterized in that: The energy storage health state SOH expression in step (8) is: where N cycle Indicates the cumulative number of cycles, D o D avg Indicates the average depth of discharge.

Images