A hybrid mode home full-scene energy coordination optimization system and method

Abstract

The application discloses a mixed mode home full-scene energy coordination optimization system and method, and particularly relates to the technical field of home energy management, the method comprising: collecting photovoltaic, energy storage, load and electricity price data in real time, performing quality control and short-time prediction; through calculating the data freshness and credibility comprehensive quantity, gating fusion is performed on key measuring points to construct a credible observation set; on this basis, a double-target optimization model with the minimum total home electricity cost and the maximum on-grid income as the targets is established, and an improved multi-objective particle swarm algorithm is adopted to perform rolling solution, and output a Pareto optimal solution set; according to the selected scheme, execution instructions are generated to realize dynamic switching of charging pile energy consumption and on-grid mode and physical isolation of the on-grid loop, and model parameters are updated rolling according to the actual operation state. The application improves the stability, economy and on-grid income of the home energy system, and is suitable for the coordinated management of home energy in the scenario where time-of-use electricity price and on-grid electricity price coexist.

Description

Technical Field

[0001] This invention relates to the field of home energy management technology, and more specifically, to a hybrid-mode home energy collaborative optimization system and method for all scenarios. Background Technology

[0002] Home energy collaborative control technology is mainly used for the automated collaborative scheduling of energy-consuming devices such as distributed photovoltaic systems, energy storage, electric vehicle charging stations, heat pumps, and electric underfloor heating in homes. This enables peak shaving and valley filling, reducing electricity purchase costs, and balancing grid connection benefits while meeting safety and comfort constraints. This technology is widely used in residential energy management and microgrid control scenarios, especially in situations where time-of-use pricing and grid connection pricing coexist. It can improve energy efficiency and reduce the risks associated with energy consumption fluctuations through rolling optimization.

[0003] The existing technology has the following shortcomings: Traditional home energy management systems rely on data reported periodically by smart meters, home gateways, and various devices. Within each control cycle, photovoltaic (PV) power, household load power, energy storage state of charge (SOC), electricity price, and charging power are directly combined into a single input, and charging / discharging and grid-connected control strategies are generated based on this. However, in real-world home scenarios, brief router outages and discrepancies in reporting cycles between different devices are common. This means that seemingly up-to-date data may be outdated, and the timing of events at different measurement points may differ. Furthermore, the charging power of electric vehicles can increase from 0 to 7kW / 11kW within seconds after plugging in, and rooftop PV power can rapidly decrease from high output within tens of seconds under cloud cover. If these data are still directly combined and used as the starting point for optimization within a single control cycle, it's easy to misjudge excess PV power, triggering grid connection or incorrectly allocating power, leading to increased instantaneous electricity purchases; or misjudge SOC margin, causing continuous discharge and triggering energy storage protection; or incorrectly reduce the power of heat pumps and underfloor heating to meet economic requirements, resulting in a significant drop in room temperature. This can lead to frequent switching, infeasible strategies, and closed-loop oscillations.

[0004] To address the aforementioned issues, this invention proposes a hybrid-mode home energy collaborative optimization system and method for all scenarios. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a hybrid mode home full-scene energy collaborative optimization system and method to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A hybrid-mode home energy synergy optimization method includes the following steps: Key operational quantities of household-side power generation, energy storage, energy consumption, and grid connection are collected and formed into a unified input; and the key operational quantities are uniformly formatted and time-stamped to obtain a set of household-side status inputs. For key state variables entering the optimization solution, a comprehensive measure of data freshness and reliability is determined. This comprehensive measure is determined by at least one or more of the following: arrival delay, number of consecutive missing measurements, event time rollback or rearrangement depth, short-term fluctuation intensity factor, and arrival jitter. Gating processing is performed on the observations according to a preset threshold to obtain a reliable set of observations for optimization solution. The short-term fluctuation intensity factor is used to characterize the intensity of change or ramp-up characteristics of key operational variables within a short time window. A dual-objective optimization model for energy consumption and internet access is constructed based on a reliable observation set. State consistency is verified on the household-side state input set, and a comprehensive state consistency anomaly is determined. This comprehensive state consistency anomaly is determined by at least one or more of the following: power balance residual, residual persistence index, and key quantity time dispersion. When the comprehensive state consistency anomaly exceeds a preset threshold, state reassessment and bias correction are triggered to reconstruct the state packet. Observations with lower comprehensive data freshness and reliability are prioritized for replacement or fusion to ensure the reconstructed state packet satisfies energy conservation constraints. Based on the state packet that passes the state consistency verification, an improved multi-objective particle swarm optimization algorithm is used to dynamically solve the dual-objective optimization model for energy consumption and internet access, outputting a Pareto optimal solution set and forming executable scheduling candidate schemes. The scheduling scheme is selected from the Pareto optimal solution set and transformed into a set of executable instructions. The central control unit performs power allocation and state switching for each subsystem according to the set of executable instructions, realizing dynamic switching between the charging pile's energy consumption mode and the Internet access mode, and rolling updates the subsequent scheduling strategy based on the execution deviation.

[0007] In a preferred embodiment, the key operating quantities include at least the real-time output of the dual photovoltaic system, the state of charge (SOC) of the hybrid energy storage unit and its charge / discharge related quantities, the power of each household load, indoor and outdoor temperatures, hot water tank temperature, electric vehicle charging status, real-time electricity price, and weather forecast data; and based on historical data and weather forecasts, a photovoltaic output prediction sequence and a load demand prediction sequence are generated to form a rolling prior input.

[0008] In a preferred embodiment, the unified formatting and time stamping includes binding a unified time identifier to key operational quantities and unifying the dimensional expression to form a set of home-side status inputs aligned with a preset control cycle.

[0009] In a preferred embodiment, the unified formatting and time stamping also includes performing quality control on key operational quantities. The quality control includes at least boundary rationality verification and processing of outliers, completion processing of missing data, and statistics on the arrival delay of key data; and writing missing data markers or completion markers into the home-side status input set to distinguish between measured data and completed data.

[0010] In a preferred embodiment, the gating process includes selecting the observation processing method based on the comparison result between the comprehensive data freshness confidence level and a preset threshold. When the comprehensive data freshness confidence level reaches the high confidence threshold, the observation is used as a valid input. When the comprehensive data freshness confidence level is between the high confidence threshold and the low confidence threshold, the observation and short-term predicted value are fused. When the comprehensive data freshness confidence level is lower than the low confidence threshold, the observation is replaced by the short-term predicted value or the consistency inference value to obtain a reliable observation set.

[0011] In a preferred embodiment, the dual-objective optimization model for energy consumption and grid connection takes the total cost of household electricity consumption and the revenue from grid connection as optimization objectives, and sets at least the following constraints in the feasible domain: state of charge (SOC) safety constraints of hybrid energy storage units, charging and discharging power constraints of hybrid energy storage units, upper limit constraints of grid-connected output power, household energy consumption guarantee constraints, and indoor thermal comfort constraints.

[0012] In a preferred embodiment, the state consistency verification includes at least calculating the power balance residual based on the household-side state input set, and determining the state consistency anomaly comprehensive quantity by combining the residual persistence index and the time dispersion of key quantities; when the state consistency anomaly comprehensive quantity exceeds a preset threshold, priority is given to replacing or merging observations with low data freshness and reliability comprehensive quantities, and bias correction is performed on at least one of the purchased power, grid-connected output power or total household load power to reconstruct the state package and make it satisfy the energy conservation constraint.

[0013] In a preferred embodiment, the improved multi-objective particle swarm optimization algorithm includes using an external archive to store candidate non-dominated solution sets and filtering them based on Pareto dominance relations, introducing dynamic inertial weights to update particle velocity and position, and suppressing infeasible solutions through a constraint violation penalty mechanism; the scheduling candidate schemes include at least the decision quantities of hybrid energy storage unit charging and discharging power, adjustable load start / stop or power setting, electric vehicle charging power, and grid-connected output power.

[0014] In a preferred embodiment, the execution instruction set includes at least power setting values ​​and state switching instructions for the hybrid energy storage unit, various household loads, electric vehicle charging unit, and grid-connected output; the dynamic switching between the charging pile's energy consumption mode and grid connection mode is achieved by the central control unit controlling intelligent relays or equivalent isolation devices, and setting switching interlock conditions to avoid simultaneous grid connection and back-feeding conflicts; the rolling update corrects the scheduling strategy for the next control cycle based on the execution deviation.

[0015] In a preferred embodiment, the following modules are included: The real-time data acquisition and forecasting module collects key operational quantities related to the coordinated operation of household-side power generation, energy storage, energy consumption, and grid connection, and forms a unified input; it also performs unified formatting and time stamping on the key operational quantities to obtain a set of household-side status inputs. The trusted gating modeling module determines a comprehensive measure of data freshness and trustworthiness for key state variables entering the optimization solution. This comprehensive measure is determined by at least one or more of the following: arrival delay, number of consecutive missing measurements, event time rollback or rearrangement depth, short-term fluctuation intensity factor, and arrival jitter. Gating processing is then applied to the observations based on a preset threshold to obtain a trusted set of observations for the optimization solution. The short-term fluctuation intensity factor characterizes the intensity of change or ramp-up characteristics of key operational variables within a short time window. The multi-objective particle swarm optimization module constructs a dual-objective optimization model for energy consumption and internet access based on a set of reliable observations. It performs state consistency verification on the household-side state input set and determines the comprehensive quantity of state consistency anomalies. This comprehensive quantity is determined by at least one or more of the following: power balance residual, residual persistence index, and time dispersion of key quantities. When the comprehensive quantity of state consistency anomalies exceeds a preset threshold, state reassessment and bias correction are triggered to reconstruct the state packet. Observations with lower comprehensive quantities of data freshness and reliability are prioritized for replacement or fusion to ensure that the reconstructed state packet satisfies energy conservation constraints. Based on the state packets that pass the state consistency verification, an improved multi-objective particle swarm optimization algorithm is used to dynamically solve the dual-objective optimization model for energy consumption and internet access, outputting a Pareto optimal solution set and forming executable scheduling candidate schemes. The rolling closed-loop control module selects a scheduling scheme from the Pareto optimal solution set and converts it into a set of executable instructions that can be issued. The central control unit performs power allocation and state switching for each subsystem according to the set of executable instructions, realizing dynamic switching between the charging pile's energy consumption mode and the grid connection mode, and rolling updates the subsequent scheduling strategy based on the execution deviation.

[0016] The technical effects and advantages of this invention are as follows: The technical effects and advantages of this invention are as follows: This invention addresses the problem of misaligned splicing caused by the inconsistency between the reporting cycles of household-side supplementary reporting or delays and multiple devices. By calculating the comprehensive quantity of data freshness and reliability and setting high and low reliability thresholds, it implements gating processing on key observation values, including direct adoption, fusion with short-term predictions, and substitution with prediction or consistency inference, to form a reliable observation set. This reduces the risks of increased instantaneous power purchase, triggering of energy storage protection, and incorrect allocation of load power caused by misjudging surplus photovoltaic power and misjudging the SOC margin of energy storage from the source.

[0017] Furthermore, before entering the optimization solution, the present invention performs state consistency verification. Based on the power balance residual, residual persistence index and key quantity time dispersion, a state consistency anomaly comprehensive quantity is constructed. When the anomaly exceeds the threshold, state re-estimation and bias correction are triggered. Low-confidence observations are replaced or fused first, so that the reconstructed state package satisfies the energy conservation constraint, thereby suppressing the infeasibility of the strategy and closed-loop oscillation caused by state inconsistency.

[0018] Based on this, the present invention constructs a dual-objective optimization model with the total cost of household electricity and the revenue from grid connection as objectives. Under constraints such as SOC safety, charging and discharging power, grid-connected output power limit, household energy security and thermal comfort, an improved multi-objective particle swarm optimization algorithm with the introduction of external files, dynamic inertia weights and constraint violation penalty mechanisms is used for rolling solution, outputting an executable Pareto optimal solution set, taking into account the feasibility and diversity of solutions, and facilitating the rapid selection of implementation solutions under different electricity prices and user preferences.

[0019] At the execution level, this invention transforms the selected scheme into a set of executable instructions that can be issued, enabling dynamic switching between the charging pile's energy consumption mode and the grid connection mode. It also implements physical isolation of the grid connection circuit through intelligent relays, ensuring that the grid connection power is limited to the surplus portion of the charging pile. Simultaneously, it performs rolling updates based on execution deviations, enabling the scheduling strategy to adaptively correct itself with photovoltaic fluctuations, load changes, and state deviations, thereby improving the system's operational stability, economy, and grid connection benefits. Attached Figure Description

[0020] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flowchart illustrating a hybrid-mode home energy synergy optimization method according to the present invention. Figure 2 To improve the flowchart of the multi-objective particle swarm optimization algorithm; Figure 3 This is a schematic diagram of the structure of a hybrid-mode home energy collaborative optimization system according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0022] Example 1: A hybrid-mode home energy synergy optimization method according to the present invention, such as... Figure 1 As shown, it includes the following steps: Step 1: Real-time data collection and prediction; In this embodiment, for the status information required for the coordinated operation of household power generation, energy storage, energy consumption, and grid connection, key operational quantities are first collected in real time within each control cycle and formed into a unified input. The collected data includes at least: real-time output of dual photovoltaic systems, the state of charge (SOC) of the hybrid energy storage unit and its charge / discharge related quantities, the power of each household load, indoor and outdoor temperatures, hot water tank temperature, electric vehicle charging status, real-time electricity prices, and weather forecast data. The real-time output of dual photovoltaic systems includes the output power of the rooftop photovoltaic sub-unit and the balcony photovoltaic sub-unit. The power of each household load includes the energy consumption power of rigid and adjustable loads. The electric vehicle charging status includes whether it is connected, the charging power, and the user-set necessary charging requirements. To ensure the availability of data for subsequent rolling optimization, the data collection is preferably refreshed continuously at a cycle of no more than 30 seconds, and the collected data is uniformly formatted within each control cycle to give it a unified time identifier and dimensional expression, facilitating the subsequent use of the dual-objective optimization model for energy consumption and grid connection.

[0023] During data preprocessing, considering the possibility of communication jitter on the home side, inconsistent sampling frequencies, and transient noise affecting state construction, basic quality control and time-stamping were performed on the collected data: boundary rationality verification and anomaly removal were conducted for power, temperature, and SOC data; short-term missing values ​​that may occur within the acquisition period were preserved or interpolated for completion, preferably retaining the missing and interpolation markers in the missing and interpolation markers to distinguish between measured and completed values ​​in subsequent optimization. After the above processing, a home-side state input set identical to the control period is generated, providing a stable data foundation for prediction and subsequent collaborative optimization.

[0024] After real-time data acquisition and preprocessing, a rolling solution-based prediction sequence is further generated to meet the changing needs of photovoltaic power output and load demand. A CNN-BiLSTM-Attention hybrid model is used to predict photovoltaic power output for the next 4 hours, outputting a photovoltaic power output prediction sequence. The CNN-BiLSTM-Attention hybrid model takes as input historical photovoltaic power output sequences from the past 24 hours and related weather factor time series. The CNN layer extracts local features, the BiLSTM layer captures forward and backward temporal dependencies, and the Attention layer focuses on key time steps, ultimately outputting a photovoltaic power output prediction sequence for the next 4 hours. The input of historical photovoltaic power output sequences and weather forecast-related factors provides adaptability to short-term fluctuations such as cloud cover. To meet engineering usability requirements, this embodiment provides constraints on prediction errors, preferably using the Mean Absolute Percentage Error (MAPE) to evaluate prediction accuracy, defined as: ; Where N represents the number of sample points within the evaluation window; This represents the measured photovoltaic output at the nth sample point; This represents the predicted photovoltaic output corresponding to the nth sample point. Preferably, the model is trained using historical weather data and photovoltaic output data, and the MAPE is controlled to be no more than 5% to ensure that the predicted sequence can be used for subsequent optimization decisions.

[0025] Meanwhile, to ensure stable system scheduling regardless of user energy consumption habits and load changes, it is also necessary to predict load demand sequences based on historical energy consumption data and user habits. User energy consumption habits can be represented by historical time-of-use electricity curves, electric vehicle access time distribution, and typical energy consumption periods for underfloor heating and domestic hot water, and can be updated as the system operates. By organizing the photovoltaic output prediction sequence, load demand prediction sequence, and state variables such as real-time electricity price, SOC, and temperature, a rolling prior input is obtained for the next step of establishing a dual-objective optimization model for energy consumption and grid connection. This makes it possible to prepare feasible data and predictions to ensure household energy needs are met and to improve grid connection benefits.

[0026] Step 2: Trusted Gating Modeling; After completing the real-time data acquisition and short-term prediction in Step 1, this step, within each control cycle k, first performs credibility processing on the key state variables entering the optimization solution, and then establishes a dual-objective optimization model for energy consumption and grid connection based on the credibility input. Specifically, for key measurement points such as real-time output of dual photovoltaic systems, total household load power, energy storage SOC and its charging and discharging power, purchased electricity power, charging pile charging power and grid connection power, indoor temperature and hot water-related temperatures, and real-time electricity price, considering network jitter, gateway buffer retransmission, and sampling frequency differences that may occur on the household side, this step calculates a comprehensive quantity of data freshness and credibility for the observed values ​​of each measurement point i. This is to characterize whether the observation can serve as a reliable input for the starting point of optimization in this cycle. Therefore, let... This represents the event timestamp at measurement point i. This indicates the arrival timestamp of the observation at the control station. If the unified timescale corresponding to the current control cycle is represented, then the arrival delay is defined as follows: The event time rollback or reordering depth is defined as Number of consecutive missing tests This indicates the number of consecutive cycles in which the measurement point failed to arrive on time or was deemed unusable before the current cycle; arrival jitter. Used to characterize the degree of fluctuation in arrival intervals, it can be derived from the sequence of the most recent N arrival intervals. The ratio of the standard deviation to the mean is used to obtain the value. ,in To prevent zero-volume output, and further, to adapt to highly volatile scenarios such as the rapid increase in power output after electric vehicles are plugged in and the rapid decrease in photovoltaic output due to cloud cover, a short-term fluctuation intensity factor is introduced. Its power output ramp-up rate within the short window Household load ramp rate Common characterization; whereby the gradient rate is calculated using discrete difference: , ,here To control the cycle.

[0027] And through the scale parameter , Normalization, the optimal definition is: ; in, It can be taken as 10%-20% of the rated power of photovoltaic power. You can use 10%-20% of the maximum load power of the household.

[0028] To standardize the units and facilitate comprehensive calculations, , , , Normalized according to preset scale parameters respectively to obtain , , , ;in , , , ; , , , , where is the normalized scaling constant; The maximum allowable arrival delay is typically set to 0.5 to 1 times the communication protocol timeout, such as 60 seconds. The maximum allowed event time rollback depth is typically set to 2 to 4 control cycle durations. The maximum number of consecutive missing tests allowed is typically set to 3 to 5. The maximum allowable arrival jitter coefficient is typically set between 0.5 and 1.0. Based on this value, the overall data freshness and reliability score is calculated: ; in For bias terms, , , , , The weighting coefficients are determined based on historical data and expert experience. An example set of values ​​is as follows: , , , , , ; The value range is (0,1), with a larger value indicating a more reliable observation. A high-confidence threshold is then set based on this. With low confidence threshold For example, high confidence threshold Low confidence threshold :when When, the observation is directly used as valid input; when At that time, compare the observation with its short-term forecast. according to By fusion Enter after; when In such cases, the observation is not used directly, but a short-term predicted value or a consistent inferred value that satisfies the energy balance constraint is used instead. Through the above gating and substitution, this step obtains a reliable set of observations for optimization, suppressing the optimization starting point shift caused by reporting old values ​​or time-stitching from the source.

[0029] A dual-objective optimization model for energy consumption and grid connection is established based on a reliable observation set. The model aims to minimize the total cost of household electricity consumption and maximize the benefits of grid connection, and the feasible region is defined by a set of implementable constraints. The objective of total household electricity cost can be expressed as: ; Where t is the time within a day, and the integration interval [0,24] represents the optimization time domain with 24 hours as the objective. The electricity purchase price at time t, To purchase power from the power grid; The on-grid electricity price at time t. This refers to the grid-connected output power. The revenue target from grid connection can be expressed as... ; in For internet access revenue, This refers to the power output to the internet generated by the charging station.

[0030] To ensure that household energy needs are met first, a full guarantee constraint for household energy consumption is set: ; in Total household load power, This represents the total output power of the dual photovoltaic systems. For energy storage discharge power, This refers to the power purchase capacity. To ensure that essential charging is prioritized and surplus power is only fed into the grid, a power grid connection for charging stations with surplus power is established: ; in The available charging power for the charging station. The equivalent power corresponding to the user's required charging needs. Further setting of upper limits for internet power: ; in This represents the maximum allowable grid connection power for the charging pile. To ensure the safe operation of energy storage, state of charge constraints are set: ; in In the state of energy storage charge, and These are the lower and upper safety limits, respectively. To ensure thermal comfort, indoor temperature constraints are set: ; in Indoor temperature, Set the temperature for the user. Allowable deviation bandwidth.

[0031] Through the above processing, this step completes the construction of a dual-objective optimization model based on a reliable observation set. The model simultaneously reflects the household electricity purchase cost and grid connection benefits, and forms a computable optimization feasible region under constraints such as household energy security, surplus grid connection of charging piles, grid connection power limit, energy storage SOC safety and thermal comfort. This provides a stable and implementable model foundation for subsequent rolling solution and execution strategy generation.

[0032] Step 3: Solving for multi-objective particle swarm optimization; After obtaining the reliable observation set and bi-objective optimization model output from step two, this step first performs a consistency check on the current state packet within each control cycle k. Then, based on the state that passes the check or is re-estimated, an improved multi-objective particle swarm optimization algorithm is used to dynamically solve the bi-objective optimization model, outputting a Pareto optimal solution set and forming an executable scheduling candidate scheme. The optimization variables for the dynamic solution include energy storage charging and discharging power, underfloor heating power, hot water heating power, electric vehicle charging power, grid connection power, and flexible load start-up and shutdown time. A fitness function is constructed with bi-objectives and a penalty term. Dynamic inertia weights, Pareto front screening, and constraint violation penalty mechanisms are introduced to improve solution efficiency and feasibility convergence capability.

[0033] To avoid state inconsistencies that arise from seemingly complete but out-of-sync or drifting data entering the solution process, this step first calculates the comprehensive quantity of state consistency anomalies. .make The power balance residual represents the control period k, which is used to characterize the conservation deviation between the supply and demand sides at the same time scale, and is defined as: ; in, The total output power of the dual photovoltaic system during period k is given. To purchase electricity from the grid, and These are the energy storage discharge power and the charging power, respectively. Total household load power, Charging power for electric vehicles, This refers to the grid-connected output power, including the surplus power from charging piles fed into the grid. To distinguish between instantaneous noise and persistent misalignment, a residual persistence index is introduced. It is preferable to use exponential smoothing to accumulate out-of-limit events: ; in, For smoothing coefficients, For indicator functions, This represents the residual amplitude threshold. Furthermore, to directly characterize the time-discrete features of the splicing state, a key quantity, time-discreteness, is introduced. It can be defined as the set of timestamps of key measurement point events that participate in the construction of the state packet in this cycle. Discrete range: ; in, Let be the event timestamp of measurement point i within period k. Based on the above three quantities, construct the comprehensive quantity of state consistency anomalies: ; in, , , For the weighting coefficients, a set of example values ​​are: =0.5, =0.3, =0.2; R0 is the reference value for power balance, such as the average household load power or the rated photovoltaic power; D0 is the reference value for time dispersion, such as the duration of a control cycle. If Exceeding the preset threshold ,For example If the current state packet shows an inconsistency anomaly, this step triggers state reassessment and bias correction: observations at measurement points with low overall data freshness and reliability are prioritized for replacement, and their short-term predicted values ​​or consistency inferences are used to reconstruct the state packet, minimizing the bias. The online bias of the purchased power, grid-connected power, or load power is corrected to ensure that the re-estimated state packet satisfies the energy conservation constraint and... Pull back within the threshold; the online bias correction can adopt a proportional correction method, for example, by slightly lowering the power purchase setting value for the current cycle proportionally. until the power balance residual The absolute value is below the threshold ;when If the current state packet is considered to meet the consistency requirements, proceed directly to the subsequent solution.

[0034] After completing the consistency check, this step uses an improved multi-objective particle swarm optimization algorithm to dynamically solve the bi-objective optimization model. The improved multi-objective particle swarm optimization algorithm process is as follows: Figure 2 As shown. For ease of description, the position vector of a particle is denoted as... This corresponds to a set of executable scheduling decision variables, including energy storage charging and discharging power, underfloor heating power, hot water heating power, electric vehicle charging power, grid connection power, and flexible load start-up and shutdown time; the particle's velocity vector is denoted as... Dynamic inertia weights are introduced during the iterative solution process. And it decreases linearly with the number of iterations: ; in, and These are the upper and lower limits of the inertia weight, with typical values ​​being... , Where t is the current iteration number and T is the maximum iteration number. Based on the inertia weight, the particle velocity and position are updated: , ; in, , This is the learning factor, typically taking values ​​in the range [1.5, 2.0]. , A random number in the interval [0,1]. This represents the optimal position in the history of an individual particle. The global boot solution is selected from external archives; the global boot solution The solution is selected from the current Pareto non-dominated solution set stored in the external archive using a roulette wheel selection method, with the reciprocal of the fitness function value as the selection probability to maintain solution diversity; after updating... Perform boundary processing to ensure it falls within the feasible region.

[0035] Simultaneously, this step constructs a fitness function to achieve a trade-off between the lowest cost and highest benefit, and incorporates constraint violations into the fitness function as penalty terms: ; in, Total household electricity cost For revenue generated from charging piles connecting to the grid, , This is a weighting coefficient used to adjust the relative importance of the cost minimization and revenue maximization objectives. Its value can be dynamically set according to user preferences or time-of-use pricing, for example, set during peak electricity price periods. Larger, a set of static example values =1.0, =0.8, To constrain the penalty terms for violations, the penalty terms are preferably obtained by weighted summation based on the severity of the violation of each constraint: ; in, This represents the violation amount of the q-th constraint, which is usually normalized by dividing it by the corresponding constraint limit, such as for the SOC upper limit constraint. ; Here, Q represents the number of constraints and the penalty weight. This method encourages particles to search for feasible solutions that satisfy the constraints during the optimization process. Based on the Pareto dominance relation, this step performs Pareto front screening on the non-dominated solutions of the particle swarm and maintains an external archive to store candidate non-dominated solution sets. This results in the Pareto optimal solution set at the end of the iteration, which serves as input for subsequent scheme selection and execution.

[0036] The Pareto optimal solution set output in this step corresponds to multiple executable scheduling schemes. Each scheme includes at least the energy storage charging and discharging power, the underfloor heating operating power, the hot water heating power, the electric vehicle charging power, the flexible load start and stop time, and the grid power decision quantity related to the charging pile mode switching. These are used in the next step to implement the schemes in conjunction with the mode switching logic and execution priority.

[0037] Step 4: Rolling closed-loop control; After selecting the scheduling scheme for the current control cycle from the Pareto optimal solution set obtained in step three, this step transforms the scheme into a set of directly executable instructions. During execution, it dynamically switches between the charging pile's energy consumption mode and grid connection mode to ensure that household energy needs are met first, while selectively connecting surplus electricity from the charging pile to the grid to generate revenue. The set of executable instructions includes at least: energy storage charging and discharging power settings, underfloor heating operating power settings, hot water heating power settings, electric vehicle charging power settings, grid connection output power settings, and flexible load start / stop time or operating window. The central control unit allocates power and switches states for each subsystem according to this set of instructions, and uses intelligent relays to physically isolate the grid connection circuit during the switching action, ensuring that grid connection power comes only from the surplus portion of the charging pile, thus avoiding adverse effects on household electricity safety.

[0038] To achieve real-time switching control between energy consumption and grid connection, this step constructs a mode judgment criterion based on photovoltaic output, energy storage SOC, and total household load demand within each control cycle, and switches the charging pile mode accordingly. Preferably, photovoltaic output is denoted as... The total household load demand is denoted as The energy storage state of charge is denoted as When the photovoltaic output is sufficient and the energy storage SOC is high, the charging pile switches to grid connection mode; when the total household load demand is close to or exceeds the photovoltaic output, the charging pile switches back to energy consumption mode. The criterion can be expressed in the following form: when... And enter grid connection mode when photovoltaic output is in sufficient range; when The system then returns to the energy consumption mode. Here, 0.8 represents the SOC threshold, corresponding to 80%; 0.9 represents the load-to-PV output ratio threshold, corresponding to 90%; and the optimal PV output is determined by combining the current PV output with the available margin of the predicted output to avoid frequent switching due to short-term fluctuations such as cloud cover.

[0039] In addition to power and SOC conditions, this step also incorporates electricity price factors into the switching and power allocation decisions to achieve a dynamic trade-off between economic efficiency and profitability. Let the real-time electricity price be... The on-grid electricity price is ,when Prioritize household use, meaning that while meeting the total household load requirements, thermal comfort, and necessary charging needs, minimize the grid-connected output power; when At the same time, without affecting the guarantee of household energy use, the proportion of surplus electricity connected to the grid should be increased to increase grid connection revenue.

[0040] To avoid power supply conflicts or decreased comfort due to multiple loads competing for power, this step constrains power through a clear execution priority: power supply to rigid loads first ensures the basic needs of underfloor heating and hot water, then the necessary electric vehicle charging needs. After these needs are met, energy storage is charged to a safe limit before power supply is given to flexible loads. Finally, surplus power from the charging piles is fed into the grid. Through this priority constraint, when electricity prices drop or photovoltaic output decreases temporarily, it avoids the household side from suddenly having to purchase large amounts of electricity or experiencing a significant decrease in thermal comfort due to grid connection priority.

[0041] To achieve closed-loop adaptive operation, rolling updates are performed after execution. The scheduling strategy is updated every 5 minutes, taking into account the discrepancy between actual and predicted photovoltaic output, fluctuations in household load, and user feedback. The weight parameters and related penalty coefficients in the multi-objective optimization are dynamically updated to make the solution of the next rolling cycle closer to the actual operating state. At the same time, the load demand forecasting and scheduling rules are updated according to user habits to ensure that the electricity cost is minimized and the grid connection benefit is maximized under different electricity price periods, different weather conditions, and different energy preferences.

[0042] Example 2: The design of a hybrid-mode home full-scene energy collaborative optimization system of the present invention is based on the method in Example 1, specifically as follows... Figure 3 The following modules are shown: The real-time data acquisition and prediction module is used to collect, format, and time-stamp key operational quantities required for the coordinated operation of household power generation, energy storage, energy consumption, and grid connection in real time during each control cycle, generating a set of household-side status inputs. The data collected includes at least: real-time output of dual photovoltaic systems, the state of charge (SOC) of the hybrid energy storage unit and its charge / discharge related quantities, the power of various household loads, indoor and outdoor temperatures, hot water tank temperature, electric vehicle charging status, real-time electricity prices, and weather forecast data. Based on this, the module uses a CNN-BiLSTM-Attention hybrid model to predict photovoltaic output for the next four hours, generating a photovoltaic output prediction sequence; simultaneously, it combines historical energy consumption data and user habits to predict load demand sequences, forming rolling prior inputs for subsequent modeling.

[0043] The trusted gating modeling module performs trusted processing on key state variables entering the optimization solution, calculates the combined quantity of data freshness and trustworthiness, and gates, fuses, or replaces observations based on high and low trust thresholds to obtain a trusted observation set for optimization. Based on this trusted observation set, the module constructs a dual-objective optimization model for energy consumption and grid connection. This model aims to minimize the total cost of household electricity consumption and maximize the benefits of grid connection, and sets constraints such as household energy security, surplus grid connection of charging piles, upper limit of grid connection power, and safety and thermal comfort of energy storage SOC, forming a computable feasible optimization region.

[0044] The multi-objective particle swarm optimization module dynamically solves the bi-objective optimization model within each control cycle. It first performs a state consistency check; if the check fails, it triggers state reestimation and bias correction. Subsequently, it employs an improved multi-objective particle swarm optimization algorithm to solve the model, outputting a Pareto optimal solution set, and forming executable scheduling candidate schemes based on this set. The candidate schemes include at least the energy storage charging and discharging power, underfloor heating operating power, hot water heating power, electric vehicle charging power, flexible load start-up and shutdown time, and grid connection power decisions related to charging pile mode switching.

[0045] The rolling closed-loop control module selects the scheduling scheme for the current control cycle from the Pareto optimal solution set and converts it into a set of executable instructions. The central control unit uses this instruction set to allocate power and switch states for each subsystem. During execution, it dynamically switches between the charging pile's energy consumption mode and grid connection mode, and uses intelligent relays to physically isolate the grid connection loop. After execution, the module updates the scheduling strategy according to a rolling update mechanism, using the deviation between actual and predicted photovoltaic output, household load fluctuations, and user feedback as correction criteria. It dynamically updates the weight parameters and related penalty coefficients in the multi-objective optimization, making the solution for the next rolling cycle more closely reflect the actual operating conditions.

[0046] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0047] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A hybrid-mode home energy synergy optimization method, characterized in that, Includes the following steps: Key operational quantities of household-side power generation, energy storage, energy consumption, and grid connection are collected and formed into a unified input; and the key operational quantities are uniformly formatted and time-stamped to obtain a set of household-side status inputs. For the key state variables entering the optimization solution, a comprehensive measure of data freshness and reliability is determined. The comprehensive measure of data freshness and reliability is determined by at least one or more of the following: arrival delay, number of consecutive missing measurements, event time backtracking or rearrangement depth, short-term fluctuation intensity factor, and arrival jitter. Gating processing is performed on the observations according to a preset threshold to obtain a reliable set of observations for optimization solution. The short-term fluctuation intensity factor is used to characterize the intensity of change or ramp-up characteristics of key operating quantities within a short time window; A dual-objective optimization model for energy consumption and internet access is constructed based on a set of reliable observations. The state consistency of the household-side state input set is verified and the comprehensive quantity of state consistency anomalies is determined. The comprehensive quantity of state consistency anomalies is determined by at least one or more of the following: power balance residual, residual persistence index, and time dispersion of key quantities. When the comprehensive quantity of state consistency anomalies exceeds a preset threshold, state re-estimation and bias correction are triggered to reconstruct the state package. Observations with lower comprehensive quantities of data freshness and reliability are preferentially replaced or fused to ensure that the reconstructed state package meets the energy conservation constraint. Based on the state packets that have passed the state consistency verification, an improved multi-objective particle swarm optimization algorithm is used to dynamically solve the dual-objective optimization model of energy consumption and internet access, output the Pareto optimal solution set and form an executable scheduling candidate scheme; The scheduling scheme is selected from the Pareto optimal solution set and transformed into a set of executable instructions. The central control unit performs power allocation and state switching for each subsystem according to the set of executable instructions, realizing dynamic switching between the charging pile's energy consumption mode and the Internet access mode, and rolling updates the subsequent scheduling strategy based on the execution deviation.

2. The hybrid-mode home energy synergy optimization method according to claim 1, characterized in that: Key operational parameters include at least the real-time output of dual photovoltaic systems, the state of charge (SOC) of the hybrid energy storage unit and its charge / discharge related parameters, the power of each household load, indoor and outdoor temperatures, hot water tank temperature, electric vehicle charging status, real-time electricity prices, and weather forecast data; and based on historical data and weather forecasts, photovoltaic output prediction sequences and load demand prediction sequences are generated to form rolling prior inputs.

3. The hybrid-mode home energy synergy optimization method according to claim 1, characterized in that: Unified formatting and time stamping include binding a unified time identifier to key operational quantities and unifying the dimensional expression to form a set of home-side status inputs aligned with the preset control cycle.

4. The hybrid-mode home energy synergy optimization method according to claim 1, characterized in that: Unified formatting and time stamping also include quality control of key operational quantities. Quality control includes at least boundary rationality verification and handling of outliers, completion of missing data, and statistics on arrival delays of key data; and writing missing or completion flags into the home-side status input set to distinguish between measured data and completed data.

5. The hybrid-mode home energy synergy optimization method according to claim 1, characterized in that: Gating processing includes selecting the observation processing method based on the comparison result between the comprehensive data freshness and confidence level and the preset threshold. When the comprehensive data freshness and confidence level reaches the high confidence threshold, the observation is used as a valid input; when the comprehensive data freshness and confidence level is between the high confidence threshold and the low confidence threshold, the observation and short-term prediction are fused. When the combined freshness and reliability of data is below the low reliability threshold, short-term predicted values ​​or consistent inference values ​​are used to replace the observed values ​​to obtain a reliable observation set.

6. The hybrid-mode home energy synergy optimization method according to claim 1, characterized in that: The dual-objective optimization model for energy consumption and grid connection takes the total cost of household electricity consumption and the benefits of grid connection as optimization objectives. In the feasible domain, it sets at least the following constraints: the state of charge (SOC) safety constraint of the hybrid energy storage unit, the charging and discharging power constraint of the hybrid energy storage unit, the upper limit constraint of the grid-connected output power, the household energy consumption guarantee constraint, and the indoor thermal comfort constraint.

7. The hybrid-mode home energy synergy optimization method according to claim 1, characterized in that: State consistency verification includes at least calculating the power balance residual based on the household-side state input set, and determining the state consistency anomaly comprehensive quantity by combining the residual persistence index and the time dispersion of key quantities. When the state consistency anomaly comprehensive quantity exceeds a preset threshold, observations with lower data freshness and reliability comprehensive quantities are preferentially replaced or fused, and at least one of the purchased power, grid-connected output power, or total household load power is biased to reconstruct the state package and make it satisfy the energy conservation constraint.

8. The hybrid-mode home energy synergy optimization method according to claim 1, characterized in that: The improved multi-objective particle swarm optimization algorithm includes using an external archive to store candidate non-dominated solution sets and filtering them based on Pareto dominance relations, introducing dynamic inertial weights to update particle velocity and position, and suppressing infeasible solutions through a constraint violation penalty mechanism; the scheduling candidate schemes include at least the decision quantities of hybrid energy storage unit charging and discharging power, adjustable load start / stop or power setting, electric vehicle charging power, and grid-connected output power.

9. The hybrid-mode home energy synergy optimization method according to claim 1, characterized in that: The execution instruction set includes at least the power setting values ​​and state switching instructions for the hybrid energy storage unit, various household loads, electric vehicle charging unit, and grid-connected output; the dynamic switching between the charging pile's energy consumption mode and grid connection mode is achieved by the central control unit controlling intelligent relays or equivalent isolation devices, and setting switching interlock conditions to avoid conflicts between simultaneous grid connection and reverse power supply. Rolling updates adjust the scheduling strategy for the next control cycle based on execution deviations.

10. A hybrid-mode home energy collaborative optimization system, characterized in that, The optimization system is used to implement the method according to any one of claims 1-9, and includes the following modules: The real-time data acquisition and forecasting module collects key operational quantities related to the coordinated operation of household-side power generation, energy storage, energy consumption, and grid connection, and forms a unified input; it also performs unified formatting and time stamping on the key operational quantities to obtain a set of household-side status inputs. The trusted gating modeling module determines the comprehensive value of data freshness and trustworthiness for the key state variables entering the optimization solution. The comprehensive value of data freshness and trustworthiness is determined by at least one or more of the following: arrival delay, number of consecutive missing measurements, event time backtracking or rearrangement depth, short-term fluctuation intensity factor, and arrival jitter. The module also performs gating processing on the observations according to a preset threshold to obtain a trusted observation set for the optimization solution. The short-term fluctuation intensity factor is used to characterize the intensity of change or ramp-up characteristics of key operating quantities within a short time window; The multi-objective particle swarm optimization module constructs a dual-objective optimization model for energy consumption and internet access based on a set of reliable observations. It performs state consistency verification on the household-side state input set and determines the comprehensive quantity of state consistency anomalies. The comprehensive quantity of state consistency anomalies is determined by at least one or more of the following: power balance residual, residual persistence index, and time dispersion of key quantities. When the comprehensive quantity of state consistency anomalies exceeds a preset threshold, it triggers state reassessment and bias correction to reconstruct the state package. Observations with lower comprehensive quantities of data freshness and reliability are given priority for replacement or fusion to ensure that the reconstructed state package meets the energy conservation constraint. Based on the state packets that have passed the state consistency verification, an improved multi-objective particle swarm optimization algorithm is used to dynamically solve the dual-objective optimization model of energy consumption and internet access, output the Pareto optimal solution set and form an executable scheduling candidate scheme; The rolling closed-loop control module selects a scheduling scheme from the Pareto optimal solution set and converts it into a set of executable instructions that can be issued. The central control unit performs power allocation and state switching for each subsystem according to the set of executable instructions, realizing dynamic switching between the charging pile's energy consumption mode and the grid connection mode, and rolling updates the subsequent scheduling strategy based on the execution deviation.

Images