Cooperative control method and system of light storage and charging integrated system

Abstract

The invention provides a cooperative control method and system for an optical storage and charging integrated system, relates to the field of optical storage and charging, and solves the technical problems of extensive energy management, low operation economy and large impact on a power distribution network caused by lack of high-precision cooperative prediction, single optimization target and rigidity of an instruction execution mechanism in the prior art. The method comprises the following steps: collecting state data of the optical storage system; based on the numerical weather forecast data and the historical operation data, predicting a photovoltaic output prediction sequence and a charging load prediction sequence in a future preset time period; constructing a multi-time-scale collaborative optimization model, and solving an optimal power distribution instruction sequence in a future control time domain; and an instruction corresponding to the current control period is extracted from the optimal power distribution instruction sequence, and energy flow among the photovoltaic power generation unit, the energy storage battery unit, the power grid and the charging pile unit is regulated and controlled through the instruction. The method is used in the light storage and charging cooperative control process.

Description

Technical Field

[0001] This application relates to the field of photovoltaic energy storage and charging, and in particular to a collaborative control method and system for an integrated photovoltaic energy storage and charging system. Background Technology

[0002] With the large-scale deployment of distributed photovoltaic power generation, electrochemical energy storage, and electric vehicle charging facilities, integrated photovoltaic-storage-charging systems are widely used in scenarios such as industrial park microgrids, public charging stations, and highway service areas. These systems aim to improve the local consumption capacity of renewable energy, reduce electricity costs, and enhance the regulation flexibility of local power grids. Currently, mainstream energy management strategies mostly employ rule-based control logic or single-time-domain economic dispatch methods, such as setting energy storage state-of-charge thresholds for charge / discharge switching, prioritizing photovoltaic power supply to local charging piles, or formulating 24-hour operation plans solely based on day-ahead time-of-use pricing.

[0003] However, the aforementioned existing technologies have a significant technical deficiency: the system lacks the ability to accurately predict the fluctuations in photovoltaic output and the randomness of charging load, and it does not embed a constraint mechanism for the rate of change of grid interaction power in the optimization decision-making process. This leads to frequent problems in actual operation, such as drastic jumps in power commands, repeated start-ups and shutdowns of energy storage devices, and excessive power fluctuations injected into the distribution network. Especially in scenarios where multiple electric vehicles are fast-charging simultaneously, the lack of a feedback compensation mechanism for execution-layer disturbances (such as communication delays and sudden changes in user behavior) can easily cause interruptions to high-priority charging tasks and may also lead to voltage over-limits or harmonic amplification, making it difficult to meet the power quality requirements of the grid for distributed resource access.

[0004] Therefore, there is an urgent need for an intelligent control method that can integrate high-precision prediction, multi-timescale collaborative optimization, and robust command execution to solve the core technical problems of existing photovoltaic-storage-charging systems, such as extensive energy management, limited economic benefits, and significant impact on the power grid. Summary of the Invention

[0005] This application provides a collaborative control method and system for an integrated photovoltaic, energy storage and charging system, which solves the technical problems in the prior art, such as extensive energy management, low operating economy and large impact on the distribution network, caused by the lack of high-precision collaborative prediction, single optimization target and rigid command execution mechanism.

[0006] To achieve the above objectives, this application adopts the following technical solution: Firstly, a collaborative control method for an integrated photovoltaic, energy storage, and charging system is provided, including: The system collects real-time status data of the photovoltaic-storage system, including photovoltaic power generation, energy storage status of charge, total power demand of charging piles, and time-of-use electricity price of the grid. Based on numerical weather forecast data and historical operating data, a preset deep time-series prediction model generates a photovoltaic output prediction sequence and a charging load prediction sequence for a preset future time period. The numerical weather forecast data is a future meteorological element prediction sequence obtained by numerically solving the atmospheric dynamic-thermodynamic equations. Based on the state data, the photovoltaic output prediction sequence, and the charging load prediction sequence, a multi-timescale collaborative optimization model is constructed, which includes economic objectives and grid smoothing constraints. The model is then solved to generate the optimal power allocation command sequence in the future control time domain. The command sequence includes energy storage charging and discharging power commands, grid interaction power commands, and charging pile power adjustment commands. Extract the instruction corresponding to the current control cycle from the optimal power allocation instruction sequence, and regulate the energy flow between the photovoltaic power generation unit, the energy storage battery unit, the power grid and the charging pile unit through the instruction.

[0007] Based on the above technical solutions, the collaborative control method for an integrated photovoltaic-storage-charging system provided in this application addresses the shortcomings of existing systems, such as rigid scheduling strategies, limited operational benefits, and significant power fluctuations in the distribution network. These shortcomings primarily stem from weak predictive capabilities, limited optimization dimensions, and a lack of adaptability in the execution mechanism. To address these issues, this application constructs a closed-loop collaborative architecture of "perception-prediction-optimization-execution": a deep time-series model driven by numerical weather prediction and historical operational data generates high-precision photovoltaic output and charging load prediction sequences; a joint optimization model is established at both day-ahead and intraday time scales, guiding economical operation not only through time-of-use pricing but also by introducing a dynamic penalty term for the grid interaction power change rate to proactively mitigate power surges; and a dynamic pointer mechanism intelligently selects executable instructions for the current period from the optimal instruction sequence, effectively addressing actual disturbances such as communication delays and sudden changes in user fast-charging demand. This solution significantly improves the local consumption level of renewable energy and the overall economic benefits of the system, while substantially reducing the impact on the grid, achieving refined, robust, and user-friendly energy management.

[0008] In conjunction with the first aspect above, in one possible implementation, obtaining the photovoltaic power output prediction sequence and the charging load prediction sequence for a future preset time period includes: Numerical weather forecast data is collected, including total surface radiation at time steps within a preset future period. Ambient temperature and cloud cover index ; Acquire historical operating data, which includes photovoltaic power generation, charging load power, date type identifier, and time-of-use electricity price sequence for the same time period in the past N days; The numerical weather forecast data and historical operational data are respectively input into the pre-trained deep time series prediction model; Among them, the photovoltaic output prediction branch adopts a long short-term memory network with an encoder-decoder structure, and its input is... The output is a photovoltaic power output prediction sequence. ; For future time variables, For photovoltaic installed capacity, The irradiance-power conversion function is based on component temperature correction. This is the nonlinear deviation compensated by the cloud cover dynamic disturbance term; To indicate the number within the history window The measured photovoltaic output value at each moment, , The length of the history window; The charging load forecasting branch employs a time-fusion transformer model. Its inputs include historical load sequences, current day type, future time-of-use electricity price sequences, and the number of online charging piles. The output is the charging load forecast sequence. .

[0009] In conjunction with the first aspect above, in one possible implementation, the multi-timescale collaborative optimization model includes a day-ahead planning layer and an intraday rolling optimization layer; The day-ahead planning layer uses a preset long-period time period as the optimized time domain and the first time resolution as the discrete step length, based on the photovoltaic power output prediction sequence. and charging load prediction sequence Solve the following economic scheduling problem: ; in, For a moment The grid purchase price of electricity, For a moment The power purchase capacity, This is the discount factor for the on-grid electricity price. For a moment The on-grid electricity price For a moment Internet power, For a moment Energy storage daytime charging and discharging power, For a moment Power grid interaction power; t represents the current time, and T represents the preset optimized time domain; The intraday rolling optimization layer performs periodic optimization with a rolling time domain shorter than the long period and a second time resolution; wherein the second time resolution is greater than the first time resolution; Based on current state data, updated photovoltaic output forecast sequences, and short-term charging load forecast sequences, an objective function with grid smoothing penalty is constructed: ; in, For the k-th discrete control time, , To control the time step, To control variables, For a moment Energy storage charging and discharging power, For a moment Grid interaction power, This represents the total number of steps in the rolling time domain. For a moment Electricity purchase price from the grid For a moment The power purchase capacity, For a moment The on-grid electricity price For a moment Internet power, For a moment Net interaction power with the power grid, For a moment Net interaction power with the power grid, For a moment Penalty coefficient for the rate of change of power grid.

[0010] In conjunction with the first aspect above, in one possible implementation, the method for obtaining the penalty coefficient of the power grid change rate includes: Real-time acquisition of grid voltage signals at grid connection points Calculate the current time voltage fluctuation level ; Obtain the rate of change of grid interaction power within the previous control step. ; Based on the voltage fluctuation level and power change rate, the current penalty coefficient is calculated using the following formula: ; in, As the baseline penalty coefficient, This is an adjustable sensitivity coefficient.

[0011] In conjunction with the first aspect above, in one possible implementation, the voltage fluctuation level Methods for obtaining [the information] include: Obtain the past grid connection point Calculate the normalized standard deviation of voltage amplitude fluctuation and the mean of normalized voltage change rate; sum the standard deviation of fluctuation and the mean of change rate with weighted coefficients to obtain the voltage fluctuation level. .

[0012] In conjunction with the first aspect above, in one possible implementation, the constraints of the multi-timescale collaborative optimization model include: Energy balance constraint: satisfies at any time t ; Dynamic constraints on the state of charge of energy storage: ;in, For a moment The state of charge of energy storage, For the rated capacity of energy storage, The charge / discharge efficiency function; State of charge safety boundary constraints: ;in, At minimum state of charge, It is at its maximum state of charge; Equipment power limit constraints: , ;in, The maximum permissible power for energy storage charging and discharging. This is the maximum allowed power for internet access; The intraday rolling optimization layer also includes a soft constraint for tracking day-ahead plans: ;in, This is the reference power output of the day-ahead planning layer. This is the allowable deviation threshold.

[0013] In conjunction with the first aspect above, in one possible implementation, the optimal power allocation command sequence in the future control time domain is obtained by solving the interior-point method or a quadratic programming solver.

[0014] In conjunction with the first aspect above, in one possible implementation, extracting the instruction corresponding to the current control cycle from the optimal power allocation instruction sequence includes: The future obtained by the solution The optimal instruction sequence is cached in the local instruction queue and marked as such. ;in, , ; Setting a dynamic pointer initial value ; At the start of the current control cycle, read the pointer. The instruction it points to As instructions to be issued; Dynamically update the pointer based on the following conditions : If the instruction from the previous control cycle was successfully executed and the system state deviation is less than the preset threshold, then let... The system state deviation includes the absolute deviation between the actual value and the reference value of the energy storage state of charge and the absolute deviation between the actual value and the command value of the grid interaction power, both of which are less than a preset threshold. and ; like And it was detected that the charging station was performing a high-priority fast charging task, while meeting the requirements. Then maintain the pointer. The instructions from the previous control cycle remain unchanged. If there is a delay in the communication link Then let The next instruction in the instruction queue is executed sequentially to compensate for timing misalignment; The finalized instructions The data is sent to the energy storage converter, grid-connected inverter, and charging pile controller to complete the energy regulation of this control cycle.

[0015] In conjunction with the first aspect above, in one possible implementation, the high-priority fast charging task is one where the charging power of the current charging pile is greater than or equal to a preset charging power, and the battery state of charge is less than a preset state of charge.

[0016] Secondly, this application provides a collaborative control system for an integrated photovoltaic, energy storage, and charging system, comprising: an acquisition module, an optimization module, and an adjustment module; wherein, the acquisition module is used to collect real-time status data of the photovoltaic and energy storage system, and, based on numerical weather forecast data and historical operating data, generate a photovoltaic output prediction sequence and a charging load prediction sequence for a preset future time period through a preset deep time series prediction model; the optimization module is used to construct a multi-time-scale collaborative optimization model containing economic objectives and grid smoothing constraints based on the status data, the photovoltaic output prediction sequence, and the charging load prediction sequence, and solve the model to generate an optimal power allocation command sequence in the future control time domain; the adjustment module is used to extract the command corresponding to the current control cycle from the optimal power allocation command sequence, and regulate the energy flow between the photovoltaic power generation unit, the energy storage battery unit, the grid, and the charging pile unit through the command.

[0017] This application provides a collaborative control method and system for an integrated photovoltaic, energy storage, and charging system, which can effectively solve the technical problems existing in the current system, such as extensive energy dispatch, insufficient economic efficiency, and significant impact on the distribution network. The method first collects real-time data on photovoltaic power generation, energy storage state of charge, total power demand of charging piles, and time-of-use electricity prices. It then integrates numerical weather prediction (obtained by numerically solving atmospheric dynamic-thermodynamic equations) with historical operating data to generate a high-precision future photovoltaic output and charging load prediction sequence through a pre-set deep time-series prediction model. Next, it constructs a multi-timescale collaborative optimization model including a day-ahead planning layer and an intraday rolling optimization layer. The day-ahead layer aims for long-cycle economic dispatch, while the intraday layer introduces a grid interaction power change rate penalty term to improve grid friendliness. Energy storage charging and discharging, grid interaction, and charging pile power adjustment are used as joint decision variables. Finally, a dynamic pointer mechanism intelligently extracts executable instructions for the current control cycle from the optimal instruction sequence, ensuring the continuity of high-priority fast charging while adapting to communication delays and state deviations. This solution significantly improves the local consumption rate of renewable energy and the economic benefits of the system, while greatly reducing the impact of power fluctuations on the power grid, and achieving safe, efficient and flexible energy collaborative management.

[0018] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0019] Figure 1 A system architecture diagram of a collaborative control system for an integrated photovoltaic, energy storage, and charging system provided in this application embodiment; Figure 2 A flowchart illustrating a collaborative control method for an integrated photovoltaic, energy storage, and charging system provided in this application embodiment; Figure 3 This is a flowchart illustrating a method for predicting photovoltaic power output and charging load, provided in an embodiment of this application. Detailed Implementation

[0020] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The collaborative control method for an integrated photovoltaic, energy storage, and charging system provided in this application embodiment can be applied to, for example... Figure 1 In the collaborative control system of the integrated photovoltaic, energy storage and charging system shown, the communication system includes: an acquisition module, an optimization module and an adjustment module; The acquisition module is used to collect real-time status data of the photovoltaic and energy storage system, and, based on numerical weather forecast data and historical operating data, to generate a photovoltaic power output prediction sequence and a charging load prediction sequence for a preset period of time through a preset deep time series prediction model. The optimization module is used to construct a multi-timescale collaborative optimization model that includes economic objectives and grid smoothing constraints based on state data, photovoltaic output prediction sequence and charging load prediction sequence, and solve the model to generate the optimal power allocation command sequence in the future control time domain. The regulation module is used to extract the instruction corresponding to the current control cycle from the optimal power allocation instruction sequence, and regulate the energy flow between the photovoltaic power generation unit, the energy storage battery unit, the power grid and the charging pile unit through the instruction.

[0022] To address the technical problems in existing technologies, such as inefficient energy management, low operational economy, and significant impact on the power distribution network due to the lack of high-precision collaborative forecasting, single optimization objectives, and rigid command execution mechanisms, this application provides a collaborative control method for an integrated photovoltaic-storage-charging system. This method includes: real-time acquisition of the photovoltaic-storage system's state data, including photovoltaic power generation, energy storage state of charge, total power demand of charging piles, and time-of-use electricity price; and, based on numerical weather forecast data and historical operational data, generating a photovoltaic output prediction sequence and a charging load prediction sequence for a preset future time period using a pre-defined deep time-series prediction model. The numerical weather forecast data is obtained by numerically solving the atmospheric dynamic-thermodynamic equations. This application uses a deep time-series model to generate high-precision photovoltaic and charging load forecasts by deeply integrating numerical weather prediction and historical operational data. Based on state data, photovoltaic output forecasts, and charging load forecasts, a multi-time-scale collaborative optimization model is constructed, incorporating economic objectives and grid smoothing constraints. This model is then solved to generate the optimal power allocation command sequence within the future control time domain. The command sequence includes energy storage charging and discharging power commands, grid interaction power commands, and charging pile power adjustment commands. Commands corresponding to the current control cycle are extracted from the optimal power allocation command sequence. These commands regulate the energy flow between photovoltaic power generation units, energy storage battery units, the grid, and charging pile units. Currently, existing photovoltaic-storage-charging systems often suffer from poor economic performance and low photovoltaic absorption rates due to inaccurate forecasts and inefficient scheduling, and are prone to causing severe fluctuations in grid power. This application utilizes a deep time-series model to generate high-precision photovoltaic and charging load forecasts by deeply integrating numerical weather prediction and historical operational data. Based on this, a day-ahead and intraday two-layer collaborative optimization model is constructed. While minimizing electricity purchase costs, a grid power change rate penalty term is introduced to actively suppress sudden changes in interaction power. Furthermore, a dynamic pointer mechanism is used to intelligently extract current control commands, taking into account practical constraints such as the continuity of fast charging services and communication delays. This method realizes the transformation from "passive response" to "active collaboration", which not only improves the self-consumption rate and operating benefits, but also significantly reduces the impact on the distribution network, effectively solving the three core problems of extensive energy management, low efficiency and poor grid friendliness.

[0023] like Figure 2 As shown in the embodiment of this application, a collaborative control method for an integrated photovoltaic, energy storage, and charging system includes: S201. Real-time acquisition of the status data of the photovoltaic and energy storage system, and based on numerical weather forecast data and historical operation data, generating a photovoltaic output prediction sequence and a charging load prediction sequence for a preset future time period through a preset deep time series prediction model.

[0024] The status data includes photovoltaic power generation. Energy storage state of charge Total power demand of charging piles Time-of-use pricing with the power grid Numerical weather forecast data are predicted sequences of future meteorological elements obtained by numerically solving the atmospheric dynamic-thermodynamic equations.

[0025] It should be noted that the status data must include photovoltaic power generation. Energy storage state of charge Total power demand of charging piles and time-of-use electricity pricing These four factors together constitute the complete initial conditions and boundary parameters for intraday rolling optimization: measured values ​​of photovoltaic power and load reflect the current supply-demand balance; SOC determines the dispatchable margin of energy storage; and time-of-use pricing provides external signals for economic optimization. The absence of any one of these factors will render the optimization model infeasible or suboptimal.

[0026] The photovoltaic power output prediction sequence and the charging load prediction sequence are not based on simple statistical extrapolation, but are generated through a pre-set deep time series prediction model (such as LSTM or Transformer with encoder-decoder structure). The input integrates two types of key information: (1) Numerical weather forecast (NWP) data, which is the future meteorological element prediction sequence (including total surface radiation, ambient temperature and cloud cover index, etc.) obtained by numerically solving the atmospheric dynamic-thermodynamic equations, which has physical consistency and medium- to long-term reliability; (2) Historical operating data, especially historical photovoltaic power output sequence, which is used to capture localized features not covered by NWP, such as equipment aging and shading. This physical model and data-driven fusion prediction mechanism significantly improves the robustness of short-term predictions under complex weather conditions and provides high-confidence input for subsequent optimization.

[0027] S202. Based on state data, photovoltaic output prediction sequence and charging load prediction sequence, a multi-timescale collaborative optimization model containing economic objectives and grid smoothing constraints is constructed, and the model is solved to generate the optimal power allocation command sequence in the future control time domain.

[0028] The instruction sequence includes energy storage charging and discharging power instructions. Used for peak shaving and valley filling; power grid interactive power commands Used to balance long-term energy shortages; charging pile power adjustment commands It participates in short-term power regulation as a flexible load resource.

[0029] It should be noted that the economic objective not only includes minimizing the traditional grid purchase cost, but also explicitly introduces the electricity sales revenue term (in the case of grid connection permission), whose weight is dynamically driven by the real-time time-of-use electricity price to ensure that the system operation always responds to market signals; the grid smoothing constraint is not just used as a post-processing filtering method, but is directly incorporated into the objective function in the form of an embedded grid interaction power change rate penalty term.

[0030] S203. Extract the instruction corresponding to the current control cycle from the optimal power allocation instruction sequence, and regulate the energy flow between the photovoltaic power generation unit, the energy storage battery unit, the power grid and the charging pile unit through the instruction.

[0031] Based on the above technical solutions, this application provides a collaborative control method for an integrated photovoltaic, energy storage, and charging system. Existing photovoltaic, energy storage, and charging systems generally suffer from problems such as extensive energy dispatching, insufficient operational economy, and significant impact on the power distribution network. The root cause lies in low prediction accuracy, static control strategies, and a lack of closed-loop execution guarantees. This solution significantly improves the accuracy of photovoltaic and charging load predictions by integrating numerical weather prediction and historical data into a deep time-series prediction model. Based on this, a day-ahead and intraday dual-layer collaborative optimization model is constructed, using time-of-use pricing to drive economic dispatch and introducing a grid power change rate penalty term to actively suppress drastic fluctuations in interactive power. Simultaneously, a dynamic pointer mechanism intelligently extracts current instructions, ensuring the continuity of high-priority fast charging while adapting to communication delays and state deviations. This method not only maximizes the self-consumption rate and economic benefits of photovoltaic power generation but also fundamentally reduces the system's power impact on the grid, achieving refined, efficient, and low-disturbance energy collaborative management.

[0032] In one possible implementation of the embodiments of this application, such as Figure 3 As shown, the above S201 can be specifically implemented through the following S301, S302 and S303, which are explained in detail below: S301. Collect numerical weather prediction data, which includes total surface radiation at each time step within a preset future period. Ambient temperature and cloud cover index ; S302. Obtain historical operating data, which includes photovoltaic power generation, charging load power, date type identifier, and time-of-use electricity price sequence for the same time period within the past N days; wherein, in one embodiment... ; S303. Input numerical weather prediction data and historical operational data into the pre-trained deep time series prediction model respectively; Among them, the photovoltaic output prediction branch adopts a long short-term memory network with an encoder-decoder structure, and its input is... The output is a photovoltaic power output prediction sequence. ;in, For future time variables, For photovoltaic installed capacity, The irradiance-power conversion function is based on component temperature correction. The nonlinear deviation compensated by the cloud cover dynamic disturbance term is used to capture power fluctuations caused by sudden shading. To indicate the number within the history window The measured photovoltaic output value at each moment, , The length of the history window; The charging load forecasting branch employs a time-fusion transformer model. Its inputs include historical load sequences, current day type, future time-of-use electricity price sequences, and the number of online charging piles. The output is the charging load forecast sequence. Among them, the relationship between electricity price sensitivity and user behavior patterns is automatically learned through an attention mechanism.

[0033] Numerical weather forecast data includes total surface radiation at each time step within a preset future period. Ambient temperature and cloud cover index Historical operating data includes past Within a day, the photovoltaic power generation sequence, charging load power sequence, date type identifier, and time-of-use electricity price sequence for the same time period are collected. These two types of data are then input into a pre-trained deep time series prediction model. The deep time series prediction model structure includes a photovoltaic output prediction branch and a charging load prediction branch, which process the input features in parallel and output the prediction results for future time periods independently.

[0034] It should be noted that the dual-branch prediction architecture adopted in this application is not simply a combination of two independent models. Instead, it ensures that photovoltaic and load forecasts are generated collaboratively on the same time scale through a shared time axis and a unified time step alignment mechanism, thereby providing a consistent input basis for subsequent multi-objective optimization. Traditional methods often use a single model or sequential prediction, which can easily lead to spatiotemporal misalignment and error accumulation. This application, however, combines physical process modeling with data-driven approaches: introducing... As a dynamic disturbance compensation term, it significantly improves the forecast robustness under complex weather conditions; by introducing coupled modeling of electricity price and user behavior on the load side, it overcomes the limitation that static load forecasting cannot respond to changes in electricity price.

[0035] For example, suppose the prediction time domain is the next 24 hours, and the time step is... minutes, then (That is, take historical data from the past 24 hours); Numerical weather forecast data comes from the high-resolution NWP product released by the Meteorological Bureau, containing data every 15 minutes. , and ; Historical operational data is collected from the station's SCADA system and stored in a local database, where The days cover different seasons, weather types, and peak electricity consumption scenarios; The dual-branch model is trained end-to-end on historical datasets during the offline phase. The loss function is a weighted combination of mean squared error (MSE) and absolute error (MAE) to balance accuracy and stability. During the online inference phase, a prediction task is triggered every 15 minutes, and the output is... and This is for use by the intraday rolling optimization layer.

[0036] Based on the above technical solutions, in integrated photovoltaic-storage-charging systems, photovoltaic output exhibits strong volatility, and charging load is highly uncertain due to user behavior. Relying solely on historical averages or simple extrapolation for scheduling will lead to frequent energy storage switching, drastic fluctuations in grid interaction power, and even voltage exceeding limits or fast charging interruptions, severely impacting system economy and power supply reliability. Therefore, high-precision, collaborative short-term forecasting capabilities are urgently needed as a prerequisite for optimization decisions. This application employs a dual-branch deep time-series forecasting model, constructing dedicated forecasting paths for both photovoltaic and charging loads: the photovoltaic branch integrates numerical weather prediction (NWP) and historical output data, improving robustness under complex weather conditions through physical model correction and cloud disturbance compensation; the load branch introduces multi-dimensional features such as electricity price, date type, and charging pile status, using an attention mechanism to capture user response behavior. Both share a unified time axis, ensuring spatiotemporal alignment of forecast results. This technical solution retains the physical consistency of NWPs while using data-driven approaches to compensate for unmodeled dynamics such as local shading and equipment aging; simultaneously, it transforms the originally passive "uncontrollable load" into a predictable "quasi-flexible resource," providing reliable input for subsequent multi-objective optimization. Therefore, the system can achieve multiple goals such as extending energy storage life, smoothing grid interaction, and reducing operating costs while ensuring the user's charging experience, fundamentally solving the control instability problem caused by inaccurate prediction in traditional methods.

[0037] In one possible implementation of this application embodiment, the above-mentioned S202 can be specifically described as follows: The multi-timescale collaborative optimization model includes a day-ahead planning layer and an intraday rolling optimization layer; The current planning layer uses a preset long-period period as the optimized time domain and the first time resolution as the discrete step length, based on the photovoltaic power output prediction sequence. and charging load prediction sequence Solve the following economic scheduling problem: ; in, For a moment The grid purchase price of electricity, For a moment The power purchase capacity, This is the discount factor for the on-grid electricity price. For a moment The on-grid electricity price For a moment Internet power, For a moment Energy storage daytime charging and discharging power, For a moment Power grid interaction power; t represents the current time, and T represents the preset optimized time domain; The intraday rolling optimization layer performs periodic optimization with a rolling time domain shorter than the long-cycle period and a second time resolution; wherein, the second time resolution is greater than the first time resolution; Based on current state data, updated photovoltaic output forecast sequences, and short-term charging load forecast sequences, an objective function with grid smoothing penalty is constructed: ; in, For the k-th discrete control time, , To control the time step, To control variables, For a moment Energy storage charging and discharging power, For a moment Grid interaction power, This represents the total number of steps in the rolling time domain. For a moment Electricity purchase price from the grid For a moment The power purchase capacity, For a moment The on-grid electricity price For a moment Internet power, For a moment Net interaction power with the power grid, For a moment Net interaction power with the power grid, For a moment Penalty coefficient for the rate of change of power grid.

[0038] The constraints of the multi-timescale collaborative optimization model are as follows: Energy balance constraint: satisfies at any time t ; Dynamic constraints on the state of charge of energy storage: ;in, For a moment The state of charge of energy storage, For the rated capacity of energy storage, The charge / discharge efficiency function; State of charge safety boundary constraints: ;in, At minimum state of charge, It is at its maximum state of charge; Equipment power limit constraints: , ;in, The maximum permissible power for energy storage charging and discharging. This is the maximum allowed power for internet access; The intraday rolling optimization layer also includes soft constraints for tracking day-ahead plans: ;in, This is the reference power output of the day-ahead planning layer. This is the allowable deviation threshold.

[0039] The day-ahead planning layer uses a long-term (e.g., 24 hours) optimization time domain and a coarse time resolution (e.g., 1 hour) for economic dispatch, aiming to minimize electricity purchase costs and maximize electricity sales revenue. The intraday rolling optimization layer uses a short rolling time domain (e.g., 6 hours) and a high time resolution (e.g., 15 minutes) for periodic re-optimization. The objective function introduces a squared penalty term for the rate of change of grid interaction power on the basis of economic efficiency. To suppress drastic power fluctuations; the two-layer model shares the photovoltaic output prediction sequence. With charging load prediction sequence And achieve upper and lower layer collaboration by tracking soft constraints through the current plan.

[0040] In some implementations, the optimization model satisfies the following physical and operational constraints: energy balance constraints ensure that power generation and consumption match at any given time; and the dynamic constraints on the state of charge of energy storage employ an efficiency function. Accurate modeling of charging and discharging nonlinear losses; SOC safety boundaries and equipment power limits ensure safe system operation; in particular, a day-ahead plan tracking soft constraint is introduced at the intraday layer. This allows for moderate deviations from the day-ahead benchmark under disturbances, avoiding excessive corrections that could lead to control oscillations, while maintaining long-term scheduling intentions.

[0041] It should be noted that this application effectively solves the contradiction between economic efficiency, robustness and grid friendliness that is difficult to balance in single-time-scale optimization by adopting a two-layer architecture of "day-ahead coarse adjustment + intraday fine adjustment": the day-ahead layer uses long-term forecasts to grasp the electricity price arbitrage window, and the intraday layer responds quickly to uncertainties based on updated forecasts and real-time status; and the joint design of grid smoothing penalty term and day-ahead tracking constraint not only prevents the impact of frequent power jumps on the distribution network, but also avoids suboptimal solutions caused by strict tracking. Using only single-layer optimization or ignoring smoothing / tracking mechanism will make it difficult to achieve stable, efficient and compliant coordinated operation in actual fluctuation environment.

[0042] Based on the above technical solutions, integrated photovoltaic-storage-charging systems face the dual challenges of the randomness of photovoltaic output and the uncertainty of charging load. If a single timescale optimization is adopted, it is difficult to balance long-term economic dispatch and short-term fluctuation response: day-ahead planning is prone to execution failure due to forecast deviations, while pure intraday rolling lacks a global cost perspective, easily leading to frequent energy storage switching, drastic fluctuations in grid interaction power, and even violations of distribution network access specifications. Therefore, a collaborative control architecture that can both capture electricity price arbitrage windows and smooth out real-time disturbances is urgently needed. This application proposes a day-ahead-intraday dual-layer optimization model, solving the above problems through layered decoupling and complementary advantages: the day-ahead layer formulates an economic benchmark strategy based on long-term forecasts, locking in off-peak charging and peak discharge / sales opportunities; the intraday layer uses high-resolution rolling correction, introduces a grid power change rate penalty term to suppress fluctuations, and achieves a balance between "following the plan" and "adapting to disturbances" through day-ahead tracking soft constraints. Simultaneously, complete physical constraints (energy balance, SOC dynamics, equipment limits) ensure the feasibility of the solution. This design not only improves operational economy but also significantly enhances the system's robustness to prediction errors and sudden changes in user behavior, effectively balancing the three core objectives of economy, safety, and grid friendliness.

[0043] And, the method for obtaining the penalty coefficient for the rate of change of power grid includes the following steps: Obtain the past grid connection point Calculate the normalized standard deviation of voltage amplitude fluctuation and the mean of normalized voltage change rate; sum the standard deviation of fluctuation and the mean of change rate with weighted coefficients to obtain the voltage fluctuation level. This is used to characterize the current power quality status of the power grid; Obtain the rate of change of grid interaction power within the previous control step. This reflects the degree of dynamic intensity of power; The current penalty coefficient is calculated based on the voltage fluctuation level and power change rate using the following formula: ; in, As the baseline penalty coefficient, This is an adjustable sensitivity coefficient.

[0044] It should be noted that this application achieves a "grid-state-aware" smooth control strategy by coupling voltage fluctuation levels with power change rate to construct a dynamic penalty coefficient. Traditional methods often use fixed or time-based penalty terms, which cannot reflect the actual operating pressure of the power grid. This scheme uses voltage disturbances as a key input, automatically increasing the penalty intensity with grid vulnerability—when voltage fluctuations intensify, the system actively suppresses power jumps to avoid triggering relay protection actions or harmonic amplification. Simultaneously, through a power change rate self-feedback mechanism, it achieves "the faster the fluctuation, the greater the penalty," effectively suppressing frequent equipment start-ups and shutdowns. This design overcomes the limitations of "static smoothing," possesses significant creativity, and is a key innovation for achieving "grid-friendly" photovoltaic-storage-charging synergy.

[0045] For example, suppose The sampling frequency is 10Hz. , , ; If the standard deviation of voltage fluctuation is detected at a certain moment (i.e., 3%), the rate of change of power is ,but This indicates that the penalty for power jumps should be significantly increased at this time to guide the system to prioritize smooth scheduling paths. This mechanism has been verified in field tests of microgrids in multiple parks and can reduce the rate of change of power interaction in the power grid, significantly improving the system's power quality adaptability.

[0046] Based on the above technical solutions, in integrated photovoltaic-storage-charging systems, severe fluctuations in grid interaction power can trigger voltage flicker, harmonic amplification, and relay protection malfunctions, seriously threatening the safe and stable operation of the distribution network. Traditional control methods often employ fixed or static smoothing penalty coefficients, which cannot dynamically adjust the control intensity according to the actual power quality status of the grid. This leads to either "over-smoothing" reducing economic efficiency or "insufficient smoothing" exacerbating grid impacts. Therefore, a dynamic penalty mechanism capable of sensing the real-time grid status and adaptively adjusting the smoothing intensity is urgently needed. This application constructs a voltage fluctuation level index by acquiring grid connection point voltage samples, calculating the normalized voltage amplitude fluctuation standard deviation and the mean rate of change, to characterize the current power quality status of the grid. Simultaneously, it combines the rate of change of grid interaction power within the previous control step to reflect the degree of power dynamic intensity. Furthermore, a dynamic penalty coefficient is designed based on both, which increases synchronously with voltage fluctuations and power changes. This method achieves an intelligent response where "the more vulnerable the grid, the more stringent the smoothing," avoiding both blindly pursuing economic efficiency at the expense of grid stability and excessively suppressing the flexibility of energy storage dispatch to meet power quality requirements. Compared to a fixed penalty strategy, this scheme significantly improves the system's adaptability to complex power grid environments, maximizes operational benefits while ensuring power quality, and has outstanding technological advancement and practical value.

[0047] In one possible implementation of this application embodiment, the above-mentioned S202, which solves for the optimal power allocation command sequence in the future control time domain, is specifically explained as follows: The optimal power allocation command sequence in the future control time domain is obtained by numerically solving the convex optimization problem constructed by the aforementioned multi-timescale collaborative optimization model. When the objective function is quadratic and the constraints are linear or convex nonlinear, the optimization problem can be transformed into a standard quadratic programming form. The solution process is executed using the interior point method or a dedicated quadratic programming solver to obtain a global optimal solution or an approximate optimal solution that meets the engineering accuracy requirements within a finite time.

[0048] In some implementations, the objective function of the intraday rolling optimization layer includes the square of the power grid change rate, and after the constraints are linearized, the entire problem can be modeled as a strictly convex quadratic programming problem. This problem can be solved efficiently by calling open-source solvers (such as OSQP, qpOASES) or commercial solvers (such as Gurobi, CPLEX), and the time for a single optimization is usually controlled within 100 milliseconds, which meets the requirements of real-time control.

[0049] It should be noted that the choice of interior-point method or quadratic programming solver in this application is not a simple application of general algorithms, but rather a deep adaptation to the mathematical structure of the optimization problem: due to the introduction of power grid smoothing penalty term and day-ahead tracking soft constraint, the objective function naturally possesses quadratic differentiability and strong convexity, making the QP solution uniquely optimal and convergent quickly; while heuristic algorithms (such as genetic algorithms and particle swarm optimization) can handle non-convex problems, they suffer from drawbacks such as long solution time, unstable results, and difficulty in guaranteeing real-time performance, making them unsuitable for minute-level rolling control scenarios. Therefore, adopting a deterministic solution method based on convex optimization is a key technical choice that balances computational efficiency, solution quality, and system reliability, demonstrating the advanced nature of this solution at the engineering implementation level.

[0050] For example, in a time domain containing 24 prediction steps ( In the intraday optimization task (with a step size of 15 minutes), the control variable dimension is 48 (including energy storage and grid power). The QP problem generated after modeling contains 48 variables and approximately 150 linear constraints. Using the OSQP solver on the embedded industrial control computer, the average solution time is 65 milliseconds, with a maximum of no more than 90 milliseconds, which fully meets the requirement of a rolling optimization cycle of once every 15 minutes. The obtained optimal instruction sequence can be directly issued to each execution unit, achieving a seamless connection between millisecond-level response and minute-level scheduling.

[0051] In one possible implementation of this application embodiment, the above-mentioned S203 can be specifically described as follows: The future obtained by the solution The optimal instruction sequence is cached in the local instruction queue and marked as such. ;in, , ; Setting a dynamic pointer initial value ; At the start of the current control cycle, read the pointer. The instruction it points to As instructions to be issued; Dynamically update the pointer based on the following conditions : If the instruction from the previous control cycle was successfully executed and the system state deviation is less than the preset threshold, then let... This indicates that the optimization sequence has been restarted; The system state deviation includes the absolute deviation between the actual value and the reference value of the energy storage state of charge. The absolute deviation between the actual and commanded values ​​of the power interaction with the power grid ; like And it was detected that the charging station was performing a high-priority fast charging task, while meeting the requirements. Then maintain the pointer. The instructions from the previous control cycle remain unchanged. Among them, high-priority fast charging tasks are those where the current charging pile's charging power is greater than or equal to the preset charging power, and the battery's state of charge (SOC) is less than the preset SOC, such as charging power ≥ 80% of rated power and battery SOC < 30%. If there is a delay in the communication link Then let The next instruction in the instruction queue is executed sequentially to compensate for timing misalignment; The finalized instructions The data is sent to the energy storage converter, grid-connected inverter, and charging pile controller to complete the energy regulation of this control cycle.

[0052] It should be noted that the dynamic pointer mechanism adopted in this application is not a simple "first-step execution" strategy, but rather a closed-loop execution architecture with state awareness, user protection, and timing fault tolerance capabilities. Traditional model predictive control typically assumes "re-optimization every cycle and execution of only the first step." However, in actual photovoltaic energy storage and charging systems, this assumption often fails due to communication delays, equipment response lags, or high-priority user service demands, leading to frequent command jumps, repeated energy storage switching, or fast charging interruptions. This application introduces dynamic pointers and couples high-priority fast charging task identification, power change judgment, and communication delay compensation logic, enabling the command execution process to maintain control continuity, ensure the quality of service for critical users, and prevent accelerated equipment aging due to command oscillations. Therefore, this mechanism is a crucial bridge connecting upper-level optimization decisions and lower-level physical execution. Its design directly determines the system's robustness and practicality under complex operating conditions, and is by no means a conventional choice for those skilled in the art.

[0053] For example, let the control cycle be... minute, Step (6 hours), a communication delay was detected at a certain moment. minutes (> (minutes), then the pointer will automatically start from Postponed to Execute the second step of the original plan to align with the actual timeline; If the charging station is currently fast charging (120kW power, 25% SOC), and the current If the power suddenly drops from 30kW to 0kW, the freeze mechanism is triggered, maintaining [the freeze state]. The command remains unchanged; continue using the instructions from the previous cycle to ensure uninterrupted charging. In the next cycle, if the system returns to normal and there is no communication delay, then reset. Then, we restarted optimizing the sequence.

[0054] This mechanism has been verified in real-world tests of microgrids in multiple industrial parks, and it can effectively reduce fast charging interruption rates and improve system operational stability.

[0055] Based on the above technical solutions, in integrated photovoltaic-storage-charging systems, although model predictive control can generate the optimal sequence of future instructions, directly executing the first instruction without considering actual operational disturbances can easily lead to control instability due to prediction bias, communication delays, or sudden changes in user behavior. For example, during fast charging, charging may be interrupted due to sudden changes in power instructions, or network delays may cause misalignment between instructions and actual timing, leading to risks such as overcharging of energy storage and exceeding grid power limits. Therefore, there is an urgent need for an instruction execution mechanism that can dynamically adapt to actual operating conditions. This application proposes an intelligent instruction extraction method based on dynamic pointers. By introducing a triple judgment logic of state deviation verification, high-priority fast charging protection, and communication delay compensation, it achieves adaptive reading of the instruction queue: the pointer is reset only when the system is stable to enable the latest optimization results; when critical user services are detected and power changes are too large, the instructions are frozen to ensure charging continuity; and the pointer is extended when communication is abnormal to maintain timing consistency. This design upgrades the traditional "open-loop execution" to "closed-loop perception-decision-execution," retaining the optimization advantages of MPC while significantly improving the robustness of the system and user experience under non-ideal conditions. Compared to fixed strategies, this solution effectively avoids frequent device switching and service interruptions. In engineering practice, it combines security, economy, and user-friendliness, demonstrating outstanding practical value and innovation.

[0056] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

Claims

1. A collaborative control method for an integrated photovoltaic-storage-charging system, characterized in that, include: Real-time acquisition of status data of the photovoltaic-storage system, including photovoltaic power generation, energy storage status of charge, total power demand of charging piles, and time-of-use electricity price of the grid. Furthermore, based on numerical weather forecast data and historical operational data, a pre-defined deep time series prediction model is used to generate a photovoltaic power output prediction sequence and a charging load prediction sequence for a pre-defined period in the future. The numerical weather prediction data is a future meteorological element prediction sequence obtained by numerically solving the atmospheric dynamic-thermodynamic equations. Based on the state data, the photovoltaic output prediction sequence, and the charging load prediction sequence, a multi-timescale collaborative optimization model is constructed, which includes economic objectives and grid smoothing constraints. The model is then solved to generate the optimal power allocation command sequence in the future control time domain. The command sequence includes energy storage charging and discharging power commands, grid interaction power commands, and charging pile power adjustment commands. Extract the instruction corresponding to the current control cycle from the optimal power allocation instruction sequence, and regulate the energy flow between the photovoltaic power generation unit, the energy storage battery unit, the power grid and the charging pile unit through the instruction.

2. The collaborative control method for an integrated photovoltaic-storage-charging system according to claim 1, characterized in that, The method for obtaining the photovoltaic power output prediction sequence and charging load prediction sequence within the preset future time period includes: Numerical weather forecast data is collected, including total surface radiation at time steps within a preset future period. Ambient temperature and cloud cover index ; Acquire historical operating data, which includes photovoltaic power generation, charging load power, date type identifier, and time-of-use electricity price sequence for the same time period in the past N days; The numerical weather forecast data and historical operational data are respectively input into the pre-trained deep time series prediction model; Among them, the photovoltaic output prediction branch adopts a long short-term memory network with an encoder-decoder structure, and its input is... The output is a photovoltaic power output prediction sequence. ; For future time variables, For photovoltaic installed capacity, The irradiance-power conversion function is based on component temperature correction. This is the nonlinear deviation compensated by the cloud cover dynamic disturbance term; To indicate the number within the history window The measured photovoltaic output value at each moment, , The length of the history window; The charging load forecasting branch employs a time-fusion transformer model. Its inputs include historical load sequences, current day type, future time-of-use electricity price sequences, and the number of online charging piles. The output is the charging load forecast sequence. .

3. The collaborative control method for an integrated photovoltaic-storage-charging system according to claim 1, characterized in that, The multi-timescale collaborative optimization model includes a day-ahead planning layer and an intraday rolling optimization layer; The day-ahead planning layer uses a preset long-period time period as the optimized time domain and the first time resolution as the discrete step length, based on the photovoltaic power output prediction sequence. and charging load prediction sequence Solve the following economic scheduling problem: ； in, For a moment The grid purchase price of electricity, For a moment The power purchase capacity, This is the discount factor for the on-grid electricity price. For a moment The on-grid electricity price For a moment Internet power, For a moment Energy storage daytime charging and discharging power, For a moment Power grid interaction power; t represents the current time, and T represents the preset optimized time domain; The intraday rolling optimization layer performs periodic optimization with a rolling time domain shorter than the long period and a second time resolution; wherein the second time resolution is greater than the first time resolution; Based on current state data, updated photovoltaic output forecast sequences, and short-term charging load forecast sequences, an objective function with grid smoothing penalty is constructed: ； in, For the k-th discrete control time, , To control the time step, To control variables, For a moment Energy storage charging and discharging power, For a moment Grid interaction power, This represents the total number of steps in the rolling time domain. For a moment Electricity purchase price from the grid For a moment The power purchase capacity, For a moment The on-grid electricity price For a moment Internet power, For a moment Net interaction power with the power grid, For a moment Net interaction power with the power grid, For a moment Penalty coefficient for the rate of change of power grid.

4. The collaborative control method for an integrated photovoltaic-storage-charging system according to claim 3, characterized in that, The method for obtaining the penalty coefficient for the rate of change of power grid includes: Real-time acquisition of grid voltage signals at grid connection points Calculate the current time voltage fluctuation level ; Obtain the rate of change of grid interaction power within the previous control step. ; Based on the voltage fluctuation level and power change rate, the current penalty coefficient is calculated using the following formula: ； in, As the baseline penalty coefficient, This is an adjustable sensitivity coefficient.

5. The collaborative control method for an integrated photovoltaic-storage-charging system according to claim 3, characterized in that, The voltage fluctuation level Methods for obtaining [the information] include: Obtain the past grid connection point Calculate the normalized standard deviation of voltage amplitude fluctuation and the mean of normalized voltage change rate; sum the standard deviation of fluctuation and the mean of change rate with weighted coefficients to obtain the voltage fluctuation level. .

6. The collaborative control method for an integrated photovoltaic-storage-charging system according to claim 3, characterized in that, The constraints of the multi-timescale collaborative optimization model include: Energy balance constraint: satisfies at any time t ; Dynamic constraints on the state of charge of energy storage: ;in, For a moment The state of charge of energy storage, For the rated capacity of energy storage, The charge / discharge efficiency function; State of charge safety boundary constraints: ;in, At minimum state of charge, It is at its maximum state of charge; Equipment power limit constraints: , ;in, The maximum permissible power for energy storage charging and discharging. This is the maximum allowed power for internet access; The intraday rolling optimization layer also includes a soft constraint for tracking day-ahead plans: ;in, This is the reference power output of the day-ahead planning layer. This is the allowable deviation threshold.

7. The collaborative control method for an integrated photovoltaic-storage-charging system according to claim 1, characterized in that, The optimal power allocation command sequence in the future control time domain is obtained by solving the interior point method or a quadratic programming solver.

8. The collaborative control method for an integrated photovoltaic-storage-charging system according to claim 1, characterized in that, Extracting the instruction corresponding to the current control cycle from the optimal power allocation instruction sequence includes: The future obtained by the solution The optimal instruction sequence is cached in the local instruction queue and marked as such. ;in, , ; Setting a dynamic pointer initial value ; At the start of the current control cycle, read the pointer. The instruction it points to As instructions to be issued; Dynamically update the pointer based on the following conditions : If the instruction from the previous control cycle was successfully executed and the system state deviation is less than the preset threshold, then let... The system state deviation includes the absolute deviation between the actual value and the reference value of the energy storage state of charge and the absolute deviation between the actual value and the command value of the grid interaction power, both of which are less than a preset threshold. and ; like And it was detected that the charging station was performing a high-priority fast charging task, while meeting the requirements. Then maintain the pointer. The instructions from the previous control cycle remain unchanged. If there is a delay in the communication link Then let The next instruction in the instruction queue is executed sequentially to compensate for timing misalignment; The finalized instructions The data is sent to the energy storage converter, grid-connected inverter, and charging pile controller to complete the energy regulation of this control cycle.

9. The collaborative control method for an integrated photovoltaic-storage-charging system according to claim 8, characterized in that, The high-priority fast charging task is defined as the charging power of the current charging pile being greater than or equal to the preset charging power, and the battery state of charge being less than the preset state of charge.

10. A collaborative control system for an integrated photovoltaic-storage-charging system, operating based on the collaborative control method for an integrated photovoltaic-storage-charging system according to any one of claims 1-9, characterized in that, It includes an acquisition module, an optimization module, and an adjustment module; The acquisition module is used to collect the status data of the photovoltaic and energy storage system in real time, and, based on numerical weather forecast data and historical operating data, generate a photovoltaic output prediction sequence and a charging load prediction sequence for a preset period of time through a preset deep time series prediction model. The optimization module is used to construct a multi-timescale collaborative optimization model that includes economic objectives and grid smoothing constraints based on the state data, the photovoltaic output prediction sequence, and the charging load prediction sequence, and solve the model to generate the optimal power allocation command sequence in the future control time domain. The adjustment module is used to extract the instruction corresponding to the current control cycle from the optimal power allocation instruction sequence, and regulate the energy flow between the photovoltaic power generation unit, the energy storage battery unit, the power grid and the charging pile unit through the instruction.

Images