Optical storage direct flexible system scheduling method based on multi-objective optimization and adaptive scheduling strategy

Abstract

According to the optical storage direct flexible system scheduling method based on multi-objective optimization and an adaptive scheduling strategy, monitoring devices are installed on a photovoltaic array, an energy storage unit and a load side, a data sensing network covering the whole link of'source-storage-load-network 'is constructed, and key data are collected in real time. A multi-objective optimization model with maximization of economical efficiency, reliability and clean energy consumption rate as objectives is established, and a hybrid optimization mechanism of a genetic algorithm and particle swarm optimization is adopted to generate a day-ahead scheduling reference scheme. And designing an adaptive scheduling algorithm and a dynamic parameter adjustment mechanism based on fuzzy logic, and combining a rolling optimization window to realize real-time optical storage coordination control and power real-time balance. According to the invention, the operation efficiency, stability and flexibility of the optical storage direct-flexible system are effectively improved, the capability of coping with emergencies is enhanced, and the optimization of the overall performance of the system is realized.

Description

Technical Field

[0001] The present invention relates to the field of energy management and optimization, and in particular to a scheduling method for a photovoltaic storage direct-flexible system based on multi-objective optimization and adaptive scheduling strategy. Background Art

[0002] In the energy field, with the rapid development of renewable energy and the continuous growth of energy demand, the PV-storage direct-flexible system has attracted widespread attention as an efficient and flexible energy supply solution. However, the existing PV-storage direct-flexible system still has many shortcomings in operation and management.

[0003] Traditional energy management systems often lack real-time and comprehensive monitoring of photovoltaic arrays, energy storage units and loads, resulting in untimely and inaccurate data collection, making it difficult to accurately grasp the operating status of the system. This makes the system lack a reliable basis for optimizing scheduling and unable to give full play to the advantages of the photovoltaic storage direct and flexible system.

[0004] In terms of optimization models, existing technologies mostly focus on optimizing a single objective, such as only considering economy or reliability, but fail to comprehensively consider multiple key objectives, such as clean energy consumption rate, etc. This results in the system being unable to achieve optimal overall performance in actual operation, which may cause resource waste or unstable operation.

[0005] In terms of algorithm application, traditional optimization algorithms are either computationally inefficient and unable to meet the needs of real-time scheduling; or they are easily trapped in local optimal solutions and cannot find the global optimal solution, affecting the overall benefits of the system.

[0006] In addition, the existing system's dispatching strategy is usually rigid and cannot be adjusted adaptively according to the real-time changing environment and load conditions. When there are emergencies such as a sudden drop in photovoltaic output and unexpected load demand, the system is often unable to respond quickly and effectively, making it difficult to ensure that the system always operates in a safe, stable and efficient state. Summary of the invention

[0007] To solve the above problems, the present invention proposes a scheduling method for a PV-storage direct-flexible system based on multi-objective optimization and adaptive scheduling strategy. The specific steps are as follows, and the characteristics are as follows:

[0008] Step 1: Deploy the real-time monitoring and data acquisition module of the PV-storage-direct-flexible system, and install monitoring devices on the PV array, energy storage unit and load side, including PV power sensors, energy storage SOC monitoring modules, load demand forecasting units and grid interaction interfaces; use edge computing nodes to collect PV power generation, energy storage charging and discharging status, load curves and grid electricity price signals in real time, build a data perception network covering the entire "source-storage-load-grid" link, and provide dynamic input parameters for multi-objective optimization;

[0009] Step 2: Construct a multi-objective optimization model and hybrid algorithm. Based on the system operation constraints, establish a mathematical model with the goal of maximizing economy, reliability and clean energy consumption rate. Use a hybrid optimization mechanism of genetic algorithm and particle swarm optimization to obtain a solution set through global search by genetic algorithm, and use particle swarm optimization to perform local refined solution to generate a benchmark day-ahead scheduling plan that meets multiple constraints.

[0010] Step 3: Design an adaptive dispatching strategy and dynamic parameter adjustment mechanism, develop an adaptive dispatching algorithm based on fuzzy logic, dynamically correct the weight coefficient of the dispatching model through real-time PV output forecast error analysis, load fluctuation detection and energy storage health status assessment; build a rolling optimization window, and update the dispatching instructions online in combination with ultra-short-term forecast data;

[0011] Step 4: Implement PV-storage coordinated control and real-time power balance, deploy distributed coordinated controllers, and prioritize the use of energy storage backup capacity when a sudden drop in PV output is detected; when load demand exceeds expectations, initiate demand-side flexible load regulation and simultaneously optimize the grid’s power purchase strategy to ensure that the system always operates within a safe domain.

[0012] The present invention is based on a multi-objective optimization and adaptive scheduling strategy for a PV-storage direct-flexible system scheduling method, and has beneficial effects. The technical effects of the present invention are:

[0013] 1. The present invention realizes accurate and real-time control of the operating status of the photovoltaic storage direct-flexible system. Through a full range of data acquisition and monitoring devices, key information of the photovoltaic array, energy storage unit and load side can be accurately obtained, providing a solid data foundation for the system's optimal scheduling, greatly enhancing the system's controllability and management convenience.

[0014] 2. The present invention significantly improves the overall performance of the system. The multi-objective optimization model takes into account economy, reliability and clean energy consumption rate, and the hybrid optimization algorithm ensures the efficiency and accuracy of the optimization scheme, so that the system can achieve a better balance in terms of operating costs, power supply stability and clean energy utilization.

[0015] 3. The present invention enhances the adaptive capability of the system. The adaptive scheduling algorithm based on fuzzy logic and the dynamic parameter adjustment mechanism can flexibly adjust the scheduling strategy according to the real-time changing system status and external conditions, effectively respond to situations such as photovoltaic output fluctuations and sudden changes in load demand, and ensure the stable operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the present invention;

[0017] Figure 2 This is a schematic diagram of the real-time monitoring and data acquisition module of the solar-storage direct-flexible system deployed in the present invention;

[0018] Figure 3 Schematic diagram of constructing a multi-objective optimization model and hybrid algorithm of the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0020] The present invention proposes a scheduling method for a PV-storage direct-flexible system based on multi-objective optimization and adaptive scheduling strategy. The flowchart of the invention is as follows: Figure 1 As shown, the steps of the present invention are described in detail below.

[0021] Step 1: Deploy real-time monitoring and data collection modules for the PV-storage direct-flexible system

[0022] Monitoring devices are installed on the photovoltaic array, energy storage unit and load side, including photovoltaic power generation sensors, energy storage SOC monitoring modules, load demand forecasting units and grid interaction interfaces; photovoltaic power generation, energy storage charging and discharging status, load curves and grid electricity price signals are collected in real time through edge computing nodes to build a data perception network covering the entire "source-storage-load-grid" link, providing dynamic input parameters for multi-objective optimization, and deploying the real-time monitoring and data acquisition module of the photovoltaic storage direct-flexible system. Figure 2 As shown;

[0023] Step 1.1 System architecture and deployment

[0024] Sensors, monitoring modules, and edge computing nodes are deployed in the key components of the system to form an integrated perception network. The specific deployment is as follows:

[0025] Photovoltaic array deployment: Photovoltaic voltage and current sensors and irradiance and temperature sensors are installed on the DC side of the photovoltaic array output and the AC side of the grid connection. The sensors collect real-time data on photovoltaic power generation power output and environmental changes, and send the information to the edge computing node for processing.

[0026] Energy storage unit deployment: In the energy storage system, the BMS battery management system integrates multi-level SOC sensors to monitor the battery's state of charge, temperature, internal resistance and other data in real time, especially the charging and discharging status, and transmit the data to the edge computing node for processing.

[0027] Load side deployment: Smart meters and edge AI prediction nodes are installed on the user load side. Smart meters are used to collect real-time power consumption data from the user side. Edge computing nodes generate short-term load demand forecasts based on load curves, historical data, weather information, and user behavior. For adjustable loads such as air conditioners and charging piles, communication modules are installed to collect information such as adjustable power, priority, and interruption duration.

[0028] Grid interactive deployment: At the grid connection point, a bidirectional energy metering device is configured to obtain the grid’s electricity price signal, grid dispatch signal, frequency deviation, etc. in real time. The edge computing node collects this data in real time and feeds it back to the system dispatch.

[0029] Step 1.2 Edge computing nodes

[0030] The edge computing node is the core of the data collection and processing system, completing real-time data collection, preprocessing and analysis at each key link. The specific implementation process is as follows:

[0031] Data collection: The edge computing node exchanges data with the communication interfaces of various monitoring devices and sensors to collect various monitoring data in real time. The data includes photovoltaic power, environmental parameters, energy storage SOC, temperature, charge and discharge status, load power demand and its adjustment information, power grid electricity price and dispatch signals, etc.

[0032] Data preprocessing and filtering: The edge computing node performs preliminary processing on the collected data. Since the data collected by the sensor may be affected by noise, the edge computing node will clean and correct the data to ensure the accuracy and reliability of the data. The Kalman filter algorithm is used to estimate the true SOC of the energy storage system to avoid inaccurate data caused by noise or sensor errors.

[0033] Data fusion and preliminary analysis: Edge computing nodes fuse data from different sources. Edge computing nodes combine photovoltaic power generation data with environmental information to evaluate the real-time fluctuation trend of photovoltaic power generation, combine grid electricity price signals with load forecast data, and dynamically adjust the scheduling strategy. Edge computing nodes will perform preliminary analysis and calculations to provide useful information for subsequent optimization decisions.

[0034] Step 1.3 Real-time communication and data transmission

[0035] All collected data is transmitted through an efficient communication network. Data is transmitted between edge computing nodes and each monitoring device through local area networks and wireless networks.

[0036] Step 1.4 Data processing and feedback mechanism

[0037] Edge computing nodes also perform real-time data analysis to provide support for subsequent optimization decisions. Specifically:

[0038] Short-term prediction and real-time response: Based on the real-time collected photovoltaic output data, load data and grid electricity prices, the edge computing node performs short-term power prediction and demand prediction. Based on historical photovoltaic output data and environmental change trends, the fluctuation of photovoltaic power generation in the future is predicted.

[0039] Dynamic adjustment and optimization feedback: The edge computing node inputs all collected data into the optimization algorithm to generate real-time dispatch instructions. Through linkage with the dispatch system and energy storage control system, the system operation strategy is adjusted in real time to optimize the power grid purchase strategy or dispatch energy storage to participate in frequency regulation and other services.

[0040] Step 1.5 Data fusion and system collaboration

[0041] Finally, all data and information collected by edge computing nodes will enter the cloud platform scheduling system for global data analysis and multi-objective optimization. Data fusion is not limited to local nodes, but also combines centralized processing with global data analysis to achieve more efficient resource allocation and scheduling decisions.

[0042] Step 2: Build a multi-objective optimization model and hybrid algorithm

[0043] Based on the system operation constraints, a mathematical model is established with the goal of maximizing economy, reliability and clean energy consumption rate. A hybrid optimization mechanism of genetic algorithm and particle swarm optimization is adopted. The genetic algorithm is used to perform a global search to obtain a solution set, and particle swarm optimization is used to perform local refinement to generate a day-ahead dispatch benchmark plan that meets multiple constraints. A multi-objective optimization model and a hybrid algorithm schematic diagram are constructed as shown in the figure. Figure 3 As shown;

[0044] In step 1, the real-time monitoring and data acquisition module of the PV-storage-direct-flexible system is deployed, and a data perception network covering the entire process of "source-storage-load-network" is established, ensuring that the system can obtain the status information of each link in real time and provide dynamic input parameters for subsequent multi-objective optimization. In step 2, a multi-objective optimization model is established based on the system operation constraints, and a hybrid optimization algorithm is used to solve it, and a multi-constrained day-ahead scheduling benchmark solution is obtained and embedded in the scheduling system.

[0045] Step 2: Construct a multi-objective optimization model and hybrid algorithm solution framework

[0046] Step 2.1 Multi-objective optimization model

[0047] The goal of multi-objective optimization is to optimize the economy, reliability and clean energy consumption rate of the system at the same time. During the dispatch cycle, the system's operation decisions include energy storage charging and discharging strategy, photovoltaic power generation utilization strategy, power grid purchase strategy and load regulation strategy. Under the constraints of decision variables, the following three goals are optimized:

[0048] Step 2.1.1. Economic efficiency: This goal aims to reduce the total operating cost of the system by minimizing the cost of electricity procurement, energy storage charging and discharging, and system operating costs. Its mathematical expression is:

[0049] C total =C purchase+C charge +C discharge

[0050] Among them, C purchase It represents the cost of purchasing electricity from the power grid, and the formula is:

[0051]

[0052] Where P grid (t) is the power purchased from the grid at time t, P price (t) is the electricity price of the power grid at that moment, and T is the time. charge Represents the energy storage charging cost, the formula is:

[0053]

[0054] Where P charge (t) is the energy storage charging power, C charge,unit is the cost of charging each unit of energy storage.

[0055] C discharge It represents the energy storage discharge cost, and the formula is:

[0056]

[0057] Where P discharge (t) is the energy storage discharge power, C discharge,unit is the cost per unit of energy storage discharged.

[0058] Step 2.1.2. Reliability

[0059] This goal ensures that the system can stably supply load demand and there will be no power gap at any time. Its mathematical expression is:

[0060]

[0061] Among them, P load (t) represents the load demand at time t, P pv (t) represents the photovoltaic power generation at time t, P discharge (t) represents the energy storage discharge power at time t, P grid (t) represents the power purchased from the grid at time t. This goal ensures that the system always has sufficient power supply, and the load demand is met while avoiding power shortage.

[0062] Step 2.1.3. Maximize clean energy consumption rate

[0063] This goal aims to maximize the use of photovoltaic power generation and minimize the waste of photovoltaic power generation, which is mathematically expressed as:

[0064]

[0065] Among them, P pv (t) represents the photovoltaic power generation at time t, P charge (t) represents the energy storage charging power at time t, P discharge (t) represents the energy storage discharge power at time t, P grid (t) represents the power purchased from the grid at time t. This goal optimizes the efficiency of photovoltaic power generation, reduces unnecessary grid power purchases, and increases the proportion of clean energy used in the system.

[0066] Step 2.2 Constraints

[0067] During the optimization process, it is necessary to meet the constraints while ensuring the physical feasibility and operational safety of the system. The main constraints include:

[0068] Power balance constraint: At any time, the system output power needs to balance the load demand and the charging and discharging capacity of the energy storage system:

[0069] P load (t) = P pv (t)+P discharge (t)+P grid (t)

[0070] Energy storage charge and discharge rate constraints: The charge and discharge power of the energy storage unit has certain limitations and cannot exceed the maximum charge and discharge rates of the device:

[0071] 0≤P charge (t)≤P charge,max ,0≤P discharge (t)≤P discharge,max

[0072] Among them, P charge,max and P discharge,max are the maximum charging and discharging powers of the energy storage, respectively.

[0073] Energy storage capacity constraints: the state of charge of energy storage must be kept within a certain range to avoid overcharging or over-discharging:

[0074] SOC min ≤SOC(t)≤SOC max

[0075] Among them, SOC(t) is the energy storage charge state at time t, SOC min and SOC max are the minimum and maximum states of charge for energy storage, respectively.

[0076] Step 2.3 Hybrid Optimization Algorithm

[0077] In order to solve this multi-objective optimization problem, a hybrid optimization mechanism of genetic algorithm and particle swarm optimization is adopted. Genetic algorithm can provide global search capability to ensure the breadth of solution space, while particle swarm optimization can find local optimal solution in solution space through refined search, thereby improving optimization efficiency. Both genetic algorithm and particle swarm optimization algorithm need to strictly meet power balance constraints, energy storage charge and discharge rate constraints and energy storage capacity constraints in solution space, and ensure the feasibility of the solution through constraint processing mechanism.

[0078] Step 2.3.1 Initialize decision variables

[0079] Decision variables include: Energy storage charging power P charge (t), energy storage discharge power P discharge (t), power P purchased from the power grid grid (t), photovoltaic power generation power P pv (t), the decision variables are randomly generated in the feasible domain.

[0080] Step 2.3.2 Genetic algorithm fitness evaluation

[0081] Each group of decision variables is evaluated by the fitness function, which is an objective function that comprehensively considers economy, reliability, clean energy consumption rate and constraint rules. For each individual, its four objective values ​​are calculated, and the fitness is obtained based on the weighted sum of the objective function:

[0082] F total =α·C total +β·Reliability+γ·CleanEnergyUtilization+λ∑ constraint

[0083] Among them, C total is the operating cost, Reliability is the power supply reliability, CleanEnergyUtilization is the clean energy consumption rate, α, β, γ, λ are the weight coefficients of each target, respectively, which control the contribution of different targets to fitness.

[0084] Step 2.3.3 Constraint handling for particle swarm optimization

[0085] The genetic algorithm provides a global valid solution set of decision variables, and the particle swarm optimization performs a refined search for each valid solution, and finally outputs the decision variables of the day-ahead scheduling solution that meets multiple constraints.

[0086] Step 3: Design adaptive scheduling strategy and dynamic parameter adjustment mechanism

[0087] Develop an adaptive dispatching algorithm based on fuzzy logic, dynamically correct the weight coefficient of the dispatching model through real-time photovoltaic output forecast error analysis, load fluctuation detection and energy storage health status assessment; build a rolling optimization window, and update the dispatching instructions online in combination with ultra-short-term forecast data;

[0088] In step 2, a multi-objective optimization model was constructed, and a hybrid genetic algorithm and particle swarm optimization algorithm were used to solve the decision variables. In step 3, an adaptive scheduling algorithm based on fuzzy logic was designed to dynamically adjust the scheduling strategy to further improve the operating efficiency and stability of the system.

[0089] Step 3.1 Design of fuzzy logic controller

[0090] Step 3.1.1 Define input variables. Use the following input variables to evaluate the system status:

[0091] Photovoltaic power generation error e pv (t): The difference between the actual photovoltaic power generation and the target photovoltaic power generation decided in step 2.

[0092] Load demand error load (t): The difference between the actual load demand and the target load decided in step 2.

[0093] Energy storage SOC error e SOC (t): The difference between the current SOC of the energy storage system and the target SOC decided in step 2.

[0094] Step 3.1.2 Definition of output variables

[0095] Based on the input variables, the fuzzy logic controller outputs the scheduling adjustment:

[0096] Energy storage charging and discharging power adjustment ΔP charge (t): Adjust the energy storage charging and discharging power to balance the supply and demand deviation.

[0097] Power grid purchase power adjustment ΔP grid (t): Dynamically correct the power purchased by the power grid to reduce the abandoned power rate and optimize economic efficiency.

[0098] Photovoltaic power adjustment ΔP pv (t): Adjust the output of the photovoltaic inverter to maximize the absorption rate.

[0099] Step 3.1.3 Input variable membership function

[0100] Using the triangular membership function, five fuzzy sets are defined: {negative big (NB), negative small (NS), zero (ZE), positive small (PS), positive big (PB)}.

[0101] The output variable membership function is consistent with the input variable membership function.

[0102] Step 3.1.4 Construction of fuzzy rule base

[0103] Fuzzy rules are formulated based on expert experience and system requirements.

[0104] Step 3.1.5 Fuzzy reasoning mechanism

[0105] The Mamdani fuzzy reasoning method is used to match the membership of the input variables with the rule base to generate fuzzy output.

[0106] Step 3.1.6 Defuzzification

[0107] Use the center of gravity method to convert the fuzzy output into a specific control quantity ΔP control :

[0108]

[0109] Among them, μ is the membership degree of the fuzzy output, and x is the corresponding control quantity.

[0110] Step 3.2 Dynamic parameter adjustment mechanism

[0111] In order to enable the fuzzy logic controller to dynamically adjust the scheduling strategy according to the system operation status, an adaptive mechanism is introduced:

[0112] Step 3.2.1 Real-time error analysis and weight correction

[0113] Calculate the input variable error e every 5 minutes pv (t), e load (t), e SOC (t) and its rate of change. The error change rate is calculated as follows:

[0114] Photovoltaic power generation error change rate:

[0115] Load demand error change rate:

[0116] Energy storage SOC error change rate:

[0117] Let the initial weight coefficients of economy, reliability and clean energy consumption rate be α 0 ,β 0 ,γ 0 , according to the error analysis results, dynamically adjust the weight coefficients α(t), β(t), and γ(t).

[0118]

[0119] in, is the adjustment rate factor.

[0120] Step 3.2.2 Weight adjustment strategy: According to the size and change trend of the error, the weight coefficient is adjusted using the following strategy:

[0121] When the error increases, the weight coefficient of the corresponding target is increased to emphasize the importance of the target.

[0122] When the error decreases, the weight coefficient of the corresponding target is reduced to reduce the attention paid to the target.

[0123] If the photovoltaic error continues to increase, the weight of clean energy consumption rate will be increased, giving priority to photovoltaic consumption.

[0124] If the load error exceeds the threshold, the reliability weight is increased to ensure power supply reliability.

[0125] Step 3.2.3 Rolling Optimization Window Design

[0126] Adapt to system changes by optimizing within a fixed time window and rolling updates over time. Set the window length to 30 minutes to cover the ultra-short-term forecast period. Set the rolling step length to 5 minutes to ensure real-time performance. In each rolling cycle, recalculate the weight coefficient of each target and perform scheduling optimization. The rolling optimization steps are as follows:

[0127] Step 3.2.3.1 Data update: Get the latest PV output forecast, load forecast and grid electricity price signal every 5 minutes.

[0128] Step 3.2.3.2 Weight update: Based on the error analysis results of step 3.2.1, dynamically correct the multi-objective weights.

[0129] Step 3.2.3.3 Hybrid algorithm solution: Call the genetic-particle swarm hybrid algorithm in step 2 to generate a scheduling plan for the next 30 minutes.

[0130] Step 3.2.3.4 Instruction issuance: Send the optimization results to the local controller for execution.

[0131] Step 4: Implement PV-storage coordinated control and real-time power balance

[0132] Deploy distributed coordination controllers to prioritize the use of energy storage backup capacity when a sudden drop in PV output is detected; when load demand exceeds expectations, initiate demand-side flexible load regulation and simultaneously optimize the grid's power purchase strategy to ensure that the system always operates within a safe domain;

[0133] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A scheduling method for a PV-storage direct-flexible system based on multi-objective optimization and adaptive scheduling strategy, the specific steps are as follows, and the characteristics are as follows: Step 1: Deploy the real-time monitoring and data acquisition module of the PV-storage-direct-flexible system, and install monitoring devices on the PV array, energy storage unit and load side, including PV power sensors, energy storage SOC monitoring modules, load demand forecasting units and grid interaction interfaces; use edge computing nodes to collect PV power generation, energy storage charging and discharging status, load curves and grid electricity price signals in real time, build a data perception network covering the entire "source-storage-load-grid" link, and provide dynamic input parameters for multi-objective optimization; Step 2: Construct a multi-objective optimization model and hybrid algorithm. Based on the system operation constraints, establish a mathematical model with the goal of maximizing economy, reliability and clean energy consumption rate. Use a hybrid optimization mechanism of genetic algorithm and particle swarm optimization to obtain a solution set through global search by genetic algorithm, and use particle swarm optimization to perform local refined solution to generate a benchmark day-ahead scheduling plan that meets multiple constraints. Step 3: Design an adaptive dispatching strategy and dynamic parameter adjustment mechanism, develop an adaptive dispatching algorithm based on fuzzy logic, dynamically correct the weight coefficient of the dispatching model through real-time PV output forecast error analysis, load fluctuation detection and energy storage health status assessment; build a rolling optimization window, and update the dispatching instructions online in combination with ultra-short-term forecast data; Step 4: Implement PV-storage coordinated control and real-time power balance, deploy distributed coordinated controllers, and prioritize the use of energy storage backup capacity when a sudden drop in PV output is detected; when load demand exceeds expectations, initiate demand-side flexible load regulation and simultaneously optimize the grid’s power purchase strategy to ensure that the system always operates within a safe domain.

2. The method for scheduling a PV-storage direct-flexible system based on multi-objective optimization and adaptive scheduling strategy according to claim 1 is characterized by: The real-time monitoring and data acquisition module of the PV-storage direct-flexible system deployed in step 1 can be expressed as: Step 1.1 System architecture and deployment: Deploy sensors, monitoring modules, and edge computing nodes in the key components of the system to form an integrated perception network. The deployment is as follows: Photovoltaic array deployment: photovoltaic voltage and current sensors and irradiance and temperature sensors are installed on the DC side of the photovoltaic array output and the AC side of the grid connection. The sensors collect real-time data on photovoltaic power output and environmental changes, and send the information to the edge computing node for processing. Energy storage unit deployment: In the energy storage system, the BMS battery management system integrates multi-level SOC sensors to monitor the battery's state of charge, temperature, internal resistance and other data in real time, especially the charging and discharging status, and transmit the data to the edge computing node for processing; Load-side deployment: Smart meters and edge AI prediction nodes are installed on the user load side. Smart meters are used to collect real-time electricity consumption data on the user side. Edge computing nodes generate short-term load demand forecasts based on load curves, historical data, weather information, and user behavior. For adjustable loads such as air conditioners and charging piles, communication modules are installed to collect information such as adjustable power, priority, and interruption duration. Grid interaction deployment: At the grid connection point, a bidirectional energy metering device is configured to obtain the grid’s electricity price signal, grid dispatch signal, frequency deviation, etc. in real time; the edge computing node collects this data in real time and feeds it back to the system dispatch; Step 1.2 Edge computing node. The edge computing node is the core of the data collection and processing system. It completes real-time data collection, preprocessing and analysis at each key link. The specific implementation process is as follows: Data collection: The edge computing node interacts with the communication interfaces of various monitoring devices and sensors to collect various monitoring data in real time; the data includes photovoltaic power, environmental parameters, energy storage SOC, temperature, charging and discharging status, load power demand and its adjustment information, power price and dispatching signals of the power grid, etc. Data preprocessing and filtering: The edge computing node performs preliminary processing on the collected data. Since the data collected by the sensor may be affected by noise, the edge computing node will clean and correct the data to ensure the accuracy and reliability of the data. The Kalman filter algorithm is used to estimate the actual SOC of the energy storage system to avoid inaccurate data caused by noise or sensor errors. Data fusion and preliminary analysis: Edge computing nodes fuse data from different sources; edge computing nodes combine photovoltaic power generation data with environmental information to evaluate the real-time fluctuation trend of photovoltaic power generation, combine grid electricity price signals with load forecast data, and dynamically adjust the scheduling strategy; edge computing nodes will perform preliminary analysis and calculations to provide useful information for subsequent optimization decisions; Step 1.3 Real-time communication and data transmission: all collected data are transmitted through an efficient communication network; edge computing nodes and each monitoring device transmit data through local area networks and wireless networks; Step 1.4 Data processing and feedback mechanism: The edge computing node also performs real-time data analysis to provide support for subsequent optimization decisions: Short-term prediction and real-time response: Based on the real-time collected photovoltaic output data, load data and grid electricity prices, the edge computing node performs short-term power prediction and demand prediction; based on historical photovoltaic output data and environmental change trends, it predicts future fluctuations in photovoltaic power generation; Dynamic adjustment and optimization feedback: The edge computing node inputs all collected data into the optimization algorithm to generate real-time dispatch instructions. By linking with the dispatch system and energy storage control system, the system's operating strategy is adjusted in real time to optimize the grid's power purchase strategy or dispatch energy storage to participate in frequency regulation and other services. Step 1.5: Data fusion and system collaboration: All data and information collected by edge computing nodes enter the cloud platform scheduling system for global data analysis and multi-objective optimization. Data fusion is not limited to local nodes, but is also combined with centralized processing and global data analysis to achieve more efficient resource allocation and scheduling decisions.

3. The method for scheduling a PV-storage direct-flexible system based on multi-objective optimization and adaptive scheduling strategy according to claim 1 is characterized by: The multi-objective optimization model and hybrid algorithm constructed in step 2 can be expressed as follows: Step 2.1 Multi-objective optimization model The goal of multi-objective optimization is to optimize the economy, reliability and clean energy consumption rate of the system at the same time. During the dispatching period, the operation decisions of the system include energy storage charging and discharging strategy, photovoltaic power generation utilization strategy, power grid purchase strategy and load regulation strategy. Under the constraints of decision variables, the following three goals are optimized: Step 2.1.

1. Economics: This goal aims to reduce the total operating cost of the system by minimizing the cost of electricity procurement, energy storage charging and discharging, and system operating costs; its mathematical expression is: C total =C purchase +C charge +C discharge Among them, C purchase It represents the cost of purchasing electricity from the power grid, and the formula is: Where P grid (t) is the power purchased from the grid at time t, P price (t) is the electricity price of the power grid at that moment, T is the time; C charge It represents the energy storage charging cost, and the formula is: Where P charge (t) is the energy storage charging power, C charge,unit is the cost of charging per unit of energy storage; C discharge It represents the energy storage discharge cost, and the formula is: Where P discharge (t) is the energy storage discharge power, C discharge,unit is the cost per unit of energy storage discharge; Step 2.1.

2. Reliability This goal ensures that the system can stably supply load demand and that there is no power gap at any time; its mathematical expression is: Among them, P load (t) represents the load demand at time t, P pv (t) represents the photovoltaic power generation at time t, P discharge (t) represents the energy storage discharge power at time t, P grid (t) represents the power purchased from the grid at time t. This goal ensures that the system always has sufficient power supply, and the load demand is met while avoiding power shortage; Step 2.1.

3. Maximize clean energy consumption rate This goal aims to maximize the use of photovoltaic power generation and minimize the waste of photovoltaic power generation, which is mathematically expressed as: Among them, P pv (t) represents the photovoltaic power generation at time t, P charge (t) represents the energy storage charging power at time t, P discharge (t) represents the energy storage discharge power at time t, P grid (t) represents the power purchased from the grid at time t; this goal optimizes the efficiency of photovoltaic power generation, reduces unnecessary grid power purchases, and increases the proportion of clean energy used in the system; Step 2.2 Constraints During the optimization process, it is necessary to meet the constraints while ensuring the physical feasibility and operational safety of the system. The main constraints include: Power balance constraint: At any time, the system output power needs to balance the load demand and the charging and discharging capacity of the energy storage system: P load (t)=P pv (t)+P discharge (t)+P grid (t) Energy storage charge and discharge rate constraints: The charge and discharge power of the energy storage unit has certain limitations and cannot exceed the maximum charge and discharge rates of the device: 0≤P charge (t)≤P charge,max ,0≤P discharge (t)≤P discharge,max Among them, P charge,max and P discharge,max are the maximum charging and discharging powers of the energy storage, respectively; Energy storage capacity constraints: the state of charge of energy storage must be kept within a certain range to avoid overcharging or over-discharging: SOC min ≤SOC(t)≤SOC max Among them, SOC(t) is the energy storage charge state at time t, SOC min and SOC max are the minimum and maximum states of charge of energy storage, respectively; Step 2.3 Hybrid Optimization Algorithm In order to solve this multi-objective optimization problem, a hybrid optimization mechanism of genetic algorithm and particle swarm optimization is adopted; the genetic algorithm can provide global search capabilities to ensure the breadth of the solution space, while the particle swarm optimization can find the local optimal solution in the solution space through refined search, thereby improving the optimization efficiency; both the genetic algorithm and the particle swarm optimization algorithm must strictly meet the power balance constraints, energy storage charge and discharge rate constraints, and energy storage capacity constraints in the solution space, and ensure the feasibility of the solution through the constraint processing mechanism; Step 2.3.1 Initialize decision variables Decision variables include: Energy storage charging power P charge (t), energy storage discharge power P discharge (t), power P purchased from the power grid grid (t), photovoltaic power generation power P pv (t), the decision variables are randomly generated in the feasible domain; Step 2.3.2 Genetic algorithm fitness evaluation Each group of decision variables is evaluated by the fitness function, which is an objective function that comprehensively considers economy, reliability, clean energy consumption rate and constraint rules. For each individual, its four target values ​​are calculated, and the fitness is obtained according to the weighted sum of the objective function: F total = α·C total + β·Reliability + γ·CleanEnergyUtilization + λ∑ Constraints Among them, C total is the operating cost, Reliability is the power supply reliability, CleanEnergyUtilization is the clean energy consumption rate, α, β, γ, λ are the weight coefficients of each target, respectively, controlling the contribution of different targets to fitness; Step 2.3.3 Constraint handling for particle swarm optimization The genetic algorithm provides a global valid solution set of decision variables, and the particle swarm optimization performs a refined search for each valid solution, and finally outputs the decision variables of the day-ahead scheduling solution that meets multiple constraints.

4. The method for scheduling a PV-storage direct-flexible system based on multi-objective optimization and adaptive scheduling strategy according to claim 1 is characterized by: The adaptive scheduling strategy and dynamic parameter adjustment mechanism designed in step 3 are expressed as follows: Step 3.1 Design of fuzzy logic controller Step 3.1.1 Define input variables. Use the following input variables to evaluate the system status: Photovoltaic power generation error e pv (t): the difference between the actual photovoltaic power generation and the target photovoltaic power generation decided in step 2; Load demand error load (t): the difference between the actual load demand and the target load decided in step 2; Energy storage SOC error e SOC (t): the difference between the current SOC of the energy storage system and the target SOC decided in step 2; Step 3.1.2 Definition of output variables Based on the input variables, the fuzzy logic controller outputs the scheduling adjustment: Energy storage charging and discharging power adjustment ΔP charge (t): Adjust the energy storage charging and discharging power to balance the supply and demand deviation; Power grid purchase power adjustment ΔP grid (t): Dynamically correct the power purchased by the power grid to reduce the abandoned solar power rate and optimize economic efficiency; Photovoltaic power adjustment ΔP pv (t): Adjust the output of the photovoltaic inverter to maximize the consumption rate; Step 3.1.3 Input variable membership function Using the triangular membership function, define five fuzzy sets: {negative large, negative small, zero, positive small, positive large}; The membership function of the output variable is consistent with the membership function of the input variable; Step 3.1.4 Construction of fuzzy rule base Formulate fuzzy rules based on expert experience and system requirements; Step 3.1.5 Fuzzy reasoning mechanism The Mamdani fuzzy reasoning method is used to match the membership of the input variables with the rule base to generate fuzzy output; Step 3.1.6 Defuzzification Use the center of gravity method to convert the fuzzy output into a specific control quantity ΔP control : Among them, μ is the membership degree of fuzzy output, and x is the corresponding control quantity; Step 3.2 Dynamic parameter adjustment mechanism In order to enable the fuzzy logic controller to dynamically adjust the scheduling strategy according to the system operation status, an adaptive mechanism is introduced: Step 3.2.1 Real-time error analysis and weight correction Calculate the input variable error e every 5 minutes pv (t), e load (t), e SOC (t) and its rate of change. The error change rate is calculated as follows: Photovoltaic power generation error change rate: Load demand error change rate: Energy storage SOC error change rate: Let the initial weight coefficients of economy, reliability and clean energy consumption rate be α0, β0, γ0, and dynamically adjust the weight coefficients α(t), β(t), γ(t) according to the error analysis results; in, is the adjustment rate factor; Step 3.2.2 Weight adjustment strategy: According to the size and change trend of the error, the weight coefficient is adjusted using the following strategy: When the error increases, the weight coefficient of the corresponding target is increased to emphasize the importance of the target; When the error decreases, the weight coefficient of the corresponding target is reduced to reduce the attention paid to the target; If the photovoltaic error continues to increase, the weight of clean energy consumption rate will be increased, giving priority to photovoltaic consumption; If the load error exceeds the threshold, the reliability weight is increased to ensure power supply reliability; Step 3.2.3 Rolling Optimization Window Design By optimizing within a fixed time window and rolling updates over time, the system can adapt to changes. The window length is set to 30 minutes to cover the ultra-short-term forecast period. The rolling step is set to 5 minutes to ensure real-time performance. In each rolling cycle, the weight coefficient of each target is recalculated and scheduling optimization is performed. The rolling optimization steps are as follows: Step 3.2.3.1 Data update: Get the latest PV output forecast, load forecast and grid price signal every 5 minutes; Step 3.2.3.2 Weight update: Based on the error analysis results of step 3.2.1, dynamically correct the multi-objective weights; Step 3.2.3.3 Hybrid algorithm solution: Call the genetic-particle swarm hybrid algorithm in step 2 to generate a scheduling plan for the next 30 minutes; Step 3.2.3.4 Instruction issuance: Send the optimization results to the local controller for execution.

Images