A multi-element coexistence fusion method and system of adaptive photovoltaic energy storage equipment

Abstract

The application discloses a kind of self-adapting photovoltaic energy storage equipment multi-element coexistence fusion method and system, it is related to new energy power generation and energy storage system collaborative control technical field.The method comprises: collecting photovoltaic, wind energy and power input data such as electricity, combine grid stability data and environmental parameter calculation energy contribution ratio;On the basis of historical trend and real-time fluctuation mode, build deep sequence prediction model to generate energy supply prediction value, and according to the prediction uncertainty dynamic adjustment power supply stability threshold to form adjustment priority sequence;According to energy input matching energy storage capacity, voltage and health state generates initial energy allocation scheme, and determines basic power transmission path under multiple constraint conditions using fast model predictive control algorithm;When photovoltaic power generation appears positive deviation, adjust photovoltaic side power distribution to generate charging input flow.Thereby realize the adaptive coordination of multi-source energy and the stable operation of energy storage system.

Description

Technical Field

[0001] This invention relates to the field of collaborative control technology for new energy power generation and energy storage systems, and in particular to a method and system for the multi-element coexistence and integration of adaptive photovoltaic energy storage equipment. Background Technology

[0002] With the large-scale integration of renewable energy sources such as photovoltaic and wind power, the coexistence of multiple energy sources in power systems is becoming increasingly common. However, renewable energy has significant volatility, intermittency, and unpredictability, placing higher demands on power systems in terms of stable operation, load balancing, and energy storage dispatch. To alleviate these problems, energy storage devices are gradually becoming important equipment for improving the absorption capacity of new energy sources and ensuring the stable operation of the power grid.

[0003] Existing photovoltaic (PV) energy storage integrated control methods typically rely on static power allocation strategies or energy estimation data based solely on a single prediction model for scheduling. These methods generally suffer from the following shortcomings: First, existing solutions often focus only on instantaneous power information, lacking quantitative analysis of the contribution of different energy types and failing to reasonably reflect the differentiated value of multiple energy sources in complex environments. Second, traditional prediction methods cannot simultaneously integrate long-term trends and short-term fluctuations, resulting in limited prediction accuracy for fluctuating energy sources, especially with significant prediction deviations in scenarios involving rapid changes in sunlight and wind speed. With increasingly stringent grid stability requirements, fixed power supply stability thresholds cannot adapt to dynamic operating conditions and struggle to respond to energy prediction uncertainties, easily leading to redundant energy storage regulation or insufficient power supply. Furthermore, in multi-energy collaborative operation scenarios, the lack of a mechanism to adjust energy allocation ratios promptly based on prediction changes makes it difficult to achieve optimal power transmission paths, impacting energy storage utilization efficiency and grid operational stability.

[0004] Therefore, there is an urgent need for an adaptive photovoltaic energy storage integration method that can comprehensively consider the characteristics of multiple energy types, integrate deep prediction results, and have dynamic adjustment capabilities, so as to improve energy utilization efficiency while ensuring grid stability and provide an feasible technical solution for multi-energy synergy in complex operating scenarios. Summary of the Invention

[0005] To address the problems of insufficient energy prediction depth, inadequate threshold adjustment capability, and lack of dynamic optimization capability in existing photovoltaic energy storage integration control methods, this invention proposes an adaptive photovoltaic energy storage equipment multi-element coexistence integration method and system.

[0006] The present invention achieves the above objectives through the following technical solutions:

[0007] A method for the coexistence and integration of multiple adaptive photovoltaic energy storage devices, the method comprising:

[0008] Acquire energy input data, power grid stability data, and environmental parameters, and calculate the contribution ratio of the energy input data, wherein the energy input data includes photovoltaic power generation data, wind power generation data, and grid power data; the power grid stability data includes grid voltage deviation, frequency deviation, and related stability indicators; and the environmental parameters include light intensity and ambient temperature.

[0009] Based on the contribution ratio, and by integrating historical data trends and real-time fluctuation patterns, a deep sequence prediction model is used to generate energy supply forecasts, while extracting the corresponding forecast uncertainty indicators.

[0010] Determine whether the predicted energy supply value is lower than a preset power supply stability threshold, and dynamically adjust the power supply stability threshold based on the prediction uncertainty index to generate an adjustment priority sequence;

[0011] Based on the adjustment priority sequence, the capacity status and voltage status of the energy storage device are obtained, and the energy input data is matched to form an initial energy allocation scheme that includes the target allocation ratio of each energy type.

[0012] Based on the initial energy allocation scheme, an optimization algorithm is used to allocate resources and generate a basic power transmission path. During the optimization process, when a positive deviation is detected between the photovoltaic power generation data and the predicted energy supply value, the photovoltaic power allocation is adjusted according to the basic power transmission path to determine the charging input flow.

[0013] As a preferred embodiment of the present invention, calculating the contribution ratio of the energy input data includes:

[0014] Based on the instantaneous photovoltaic power output value corresponding to photovoltaic power generation data Instantaneous wind energy function output value corresponding to wind power generation data Instantaneous mains power output value corresponding to mains power data Establish a power vector group for energy input data. ;

[0015] The power output ratio is calculated based on the power vector set, using the following formula: ,in Indicates energy type The corresponding power output ratio Indicates energy type The corresponding instantaneous power output value;

[0016] Based on grid voltage deviation Frequency deviation and related stability indicators Generate stability correction weights for power percentage correction The formula is: ,in Indicates the correction factor;

[0017] Based on light intensity and ambient temperature Constructing environmental impact factors for power ratio correction The formula is: ,in Indicates environmental parameter coefficients;

[0018] The power output ratio, stability correction weight, and environmental impact factor are fused together, and the contribution ratio of the energy input data is generated through dynamic normalization calculation, as shown in the formula:

[0019] ;

[0020] In the formula, Energy type The corresponding contribution ratio , Indicates the fusion weight factor. Indicates energy type The corresponding power output ratio For the number of energy types.

[0021] As a preferred embodiment of the present invention, the input to the deep sequence prediction model is constructed through the following steps:

[0022] Historical data trend sequences were constructed based on photovoltaic power generation data, wind power generation data, and grid power data. Real-time fluctuation pattern sequences were constructed based on solar irradiance and ambient temperature.

[0023] A multi-scale time window is constructed based on the historical data trend sequence and real-time fluctuation pattern sequence, and a time series feature tensor is established within the multi-scale time window;

[0024] A weight vector is constructed based on the contribution ratio of energy input data, and the time series feature tensor and the weight vector are combined according to the feature concatenation method to generate the input feature matrix.

[0025] A structured feature set is constructed based on the input feature matrix to serve as the input data for the deep sequence prediction model.

[0026] As a preferred embodiment of the present invention, the step of generating energy supply forecast values ​​and simultaneously extracting corresponding forecast uncertainty indicators includes:

[0027] The structured feature set is input into a multivariate variational mode decomposition model, based on the set modality number parameter. Output There are 1 modal component sequences, and each modal component sequence corresponds to a decomposed subsequence;

[0028] The modal component sequences are input into a deep sequence sub-model consisting of several layers of long short-term memory networks. The deep sequence sub-model generates a corresponding prediction sub-sequence for each modal component sequence.

[0029] The main prediction sequence is constructed based on all predicted subsequences in a modal superposition manner, which is the predicted value of energy supply.

[0030] Using the confidence interval expression Generate a predictive uncertainty index, in which For at any time The confidence interval for the energy supply forecast. For at any time The average of the energy supply forecasts, For at any time The standard deviation of energy supply forecasts This is a quantification factor.

[0031] As a preferred embodiment of the present invention, dynamically adjusting the power supply stability threshold based on the predicted uncertainty index includes:

[0032] Based on energy supply forecasts Compared with the preset power supply stability threshold Constructing the initial stability difference The formula is: ;

[0033] Based on the mean of energy supply forecasts Standard deviation of energy supply forecasts and quantification factor Constructing the prediction confidence interval bias : ;

[0034] A threshold correction coefficient is constructed based on the initial stability difference and the predicted confidence interval deviation. : ,in, , To adjust the weighting coefficients;

[0035] Update the power supply stability threshold based on the threshold correction coefficient: ,in This is the updated power supply stability threshold.

[0036] As a preferred embodiment of the present invention, the generation of the adjustment priority sequence includes:

[0037] A multidimensional priority factor is constructed based on the energy supply forecast and the updated power supply stability threshold. The multidimensional priority factor includes an energy supply forecast deviation factor, an energy available power factor, and an energy response speed factor.

[0038] An adjusted priority sequence is generated by weighted sorting based on multidimensional priority factors, and the adjusted priority sequence is smoothed by exponential smoothing within the time window.

[0039] An allocation weight vector is constructed based on the adjusted priority sequence. The allocation weight vector is generated in a normalized manner and used for subsequent calculations of the initial energy allocation scheme.

[0040] As a preferred embodiment of the present invention, the formation of an initial energy allocation scheme comprising target allocation ratios for each energy type includes:

[0041] Obtain the current remaining capacity, maximum rechargeable capacity, real-time voltage, and health status information of the energy storage device to construct energy storage availability information;

[0042] Based on the adjustment priority sequence, a target matching order for multiple energy types is constructed. Energy input data is matched with energy storage availability information according to the target matching order. The matching content includes the available power of energy input data, the contribution ratio of energy input data, the time series trend factor corresponding to energy input data, and the instantaneous characteristics of energy input data.

[0043] Based on the matching results of energy storage availability information and energy input data, an energy storage allocation factor is constructed. The energy storage allocation factor includes the allowable energy storage charging amount, the allowable energy storage voltage range, the allowable energy storage charging rate, and the energy storage response speed parameter.

[0044] An energy allocation search space is constructed based on energy storage allocation factors and adjustment priority sequences. The energy allocation search space includes energy storage protection interval constraints, energy storage capacity constraints, and available energy input power constraints.

[0045] An initial energy allocation scheme is generated based on the energy allocation search space, combined with the branch power flow constraints of the power grid, the power grid stability conditions, and the feasibility domain pruning method. The initial energy allocation scheme includes the target allocation ratio corresponding to each energy type.

[0046] As a preferred embodiment of the present invention, the generation of the basic power transmission path includes:

[0047] Based on the initial energy allocation scheme, a resource allocation input set is constructed, which includes the target allocation ratio of each energy type, the available power of energy input data, the energy storage allocation factor, and the real-time status observation of the energy storage device.

[0048] Based on the resource allocation input set, combined with the power grid branch flow constraints, power grid stability conditions, allowable charging rate of energy storage, and energy storage protection interval, a resource allocation constraint set is constructed. The resource allocation constraint set includes inviolable constraints and adjustable constraints.

[0049] Within the rolling time domain, a fast model predictive control algorithm is used to search for feasible resource allocation action sequences based on the resource allocation constraint set. The resource allocation action sequences correspond to the feasible power output of each energy type at multiple prediction times.

[0050] Based on the range of resource allocation action changes between adjacent prediction times, a rule for limiting the change of allocation ratio is constructed, and the change magnitude of resource allocation action is controlled according to the rule for limiting the change during the resource allocation action search process.

[0051] The basic power transmission path is determined based on the resource allocation action sequence. The basic power transmission path is a set of paths that represent the power transmission relationship from each energy type to the energy storage device.

[0052] As a preferred embodiment of the present invention, the method for determining the charging input current includes:

[0053] The deviation information between the actual power value of photovoltaic power generation data at the current moment and the predicted energy supply value is obtained, and multi-level deviation judgment conditions are constructed based on the deviation information. The multi-level deviation judgment conditions include slight deviation, moderate deviation and severe deviation.

[0054] The photovoltaic side deviation level is determined based on multi-level deviation judgment conditions. The photovoltaic side deviation level corresponds to a preset photovoltaic power adjustment strategy. The adjustment strategy includes adjusting the distribution relationship between photovoltaic output power and energy storage allowable charging amount.

[0055] Within the photovoltaic power allocation section corresponding to the basic power transmission path, a power adjustment step is performed based on the photovoltaic side deviation level. The power adjustment step is limited by the maximum allowable output power of photovoltaic, the maximum allowable charging power of energy storage, and the energy storage protection range.

[0056] The charging input stream is constructed based on the photovoltaic power adjustment steps, and a smoothing rule for the charging input stream is constructed based on the historical charging input stream. The final charging input stream is generated through the smoothing rule.

[0057] A multi-element coexistence and integration system for adaptive photovoltaic energy storage equipment, the system comprising:

[0058] The energy data acquisition module is used to acquire energy input data, power grid stability data, and environmental parameters, and to calculate the contribution ratio of the energy input data.

[0059] The energy forecasting module is used to generate energy supply forecasts based on the contribution ratio of energy input data through a deep sequence forecasting model, by integrating historical data trends and real-time fluctuation patterns, and to extract the corresponding forecast uncertainty indicators.

[0060] The threshold adjustment module is used to determine whether the predicted energy supply value is lower than the preset power supply stability threshold, and to dynamically adjust the power supply stability threshold according to the prediction uncertainty index to generate an adjustment priority sequence.

[0061] The matching processing module is used to obtain the capacity status and voltage status of the energy storage device according to the adjustment priority sequence, and to perform matching processing on the energy input data to form an initial energy allocation scheme that includes the target allocation ratio of each energy type.

[0062] The resource allocation module is used to allocate resources using an optimization algorithm based on the initial energy allocation scheme in order to generate a basic power transmission path;

[0063] The deviation adjustment module is used to adjust the photovoltaic power allocation according to the basic power transmission path to determine the charging input flow when a positive deviation is detected in the photovoltaic power generation data relative to the energy supply forecast during the resource allocation optimization process.

[0064] The beneficial effects of this invention are as follows: By constructing contribution ratios for photovoltaic power generation data, wind power generation data, and grid power data respectively, and correcting these contribution ratios by combining grid stability data and environmental parameters, the system can accurately identify the comprehensive weight of different energy types at the current moment, providing a reliable basis for subsequent prediction and allocation. By generating energy supply forecast values ​​through a deep sequence prediction model that integrates historical trend sequences and real-time fluctuation patterns, and extracting prediction uncertainty indicators, the prediction results can simultaneously reflect long-term trends, real-time fluctuations, and model uncertainties, improving the accuracy and adaptability of energy prediction in renewable energy scenarios. By constructing a dynamic adjustment mechanism between energy supply forecast values ​​and power supply stability thresholds, the thresholds can adaptively change according to prediction deviations and uncertainties, thereby improving the robustness of grid operation and the flexibility of energy dispatch. By generating initial energy allocation schemes based on adjusting priority sequences to match the status of energy storage devices, energy allocation can reflect the energy storage capacity status, voltage status, and real-time availability of different energy types, enhancing the refinement and security of energy storage regulation. By constructing a resource allocation input set and a resource allocation constraint set, and employing a fast model predictive control algorithm in the rolling time domain to generate the basic power transmission path, the system can achieve optimal power transmission path selection under multiple constraints, thereby improving energy dispatch efficiency. By detecting the deviation between photovoltaic power generation data and energy supply forecasts and implementing tiered photovoltaic-side power adjustment strategies, the system can effectively cope with sudden increases in photovoltaic power generation, enhancing the energy storage system's adaptability to new energy fluctuations and ensuring the stability and continuity of the charging input flow. Attached Figure Description

[0065] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the modular structure of the system in an embodiment of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0067] like Figure 1 As shown, this is an embodiment of the present invention. This embodiment provides a method for the multi-source coexistence and integration of adaptive photovoltaic energy storage equipment. Through steps such as energy data acquisition, predictive calculation, dynamic threshold adjustment, energy matching processing, and power transmission path optimization, adaptive multi-source integration under different energy input conditions is achieved. It is applicable to energy storage systems that are simultaneously connected to photovoltaic power generation devices, wind power generation devices, and the mains power grid.

[0068] This method mainly includes the following steps:

[0069] S1: Acquire energy input data, power grid stability data, and environmental parameters, and calculate the contribution ratio of energy input data.

[0070] In practical systems, photovoltaic inverters, wind turbine controllers, and grid-connected devices can all output corresponding power data in real time. This embodiment obtains data in the following way:

[0071] Energy input data should include at least:

[0072] Photovoltaic power generation data: Real-time output of instantaneous photovoltaic power from photovoltaic array inverters; Wind power generation data: Instantaneous output of wind turbine power from the converter of wind power generation device; Grid power data: Real-time collection of grid power input to the energy storage system from grid-connected monitoring nodes.

[0073] Power grid stability data should include at least:

[0074] Grid voltage deviation: The current voltage value is obtained from the field meter and compared with the rated voltage to obtain the deviation, which is used to indicate the degree of voltage deviation from the rated value; Grid frequency deviation: The real-time frequency is obtained from the grid synchronization module and compared with the rated frequency to obtain the deviation, which is used to indicate the degree of grid frequency deviation from the rated frequency; Stability indicators: These can be calculated by the grid monitoring terminal, such as voltage fluctuation amplitude, harmonic content, etc., to indicate the overall stability of the grid at the current moment; These data are used for subsequent adjustment contribution ratios.

[0075] Environmental parameters include at least: light intensity: collected by a light sensor; ambient temperature: collected by a temperature sensor.

[0076] Based on the above energy input data, a power vector is constructed that includes power values ​​from photovoltaic (PV), wind, and grid power. The base power proportion is determined according to the relative power magnitudes of the three. Then, stability correction weights are constructed using voltage deviation, frequency deviation, and stability indices to adjust the power proportions. Furthermore, an environmental impact factor is constructed based on irradiance and ambient temperature to correct the contribution bias of PV and wind power generation. The power proportions, stability correction weights, and environmental impact factors are then fused and normalized to obtain the final contribution ratio of the energy input data, which serves as the key input for subsequent deep sequence prediction models.

[0077] In one specific embodiment, calculating the contribution ratio of energy input data includes:

[0078] The instantaneous power output values, including the instantaneous photovoltaic power output value, are obtained from the photovoltaic power generation device, the wind power generation device, and the mains power interface. Instantaneous wind energy output value and instantaneous mains power output value The above three power values ​​constitute the power vector group of the energy input data. This is used to represent the real-time output of the three energy types at the current moment, and is the basis for subsequent percentage calculations.

[0079] The formula for calculating the power output ratio of each energy type based on the power vector group is as follows: ,in Indicates energy type The corresponding power output ratio Indicates energy type The corresponding instantaneous power output value; the sum of the three terms in the denominator represents the total instantaneous power output of all energy sources, and the basic proportion of each energy type is determined by the ratio of its own instantaneous power output to the total power output;

[0080] In order to ensure that the contribution ratio reflects the impact of grid operating conditions on grid power availability, this embodiment calculates a stability correction weight based on the following grid stability data to reflect the impact of grid conditions on the energy contribution ratio:

[0081] Based on grid voltage deviation Frequency deviation and related stability indicators Generate stability correction weights for power percentage correction The formula is: ,in The correction coefficient is set by the system based on operational experience or preset parameters. In this embodiment, the stability correction weight is added as an additional correction amount to the contribution ratio calculation, so that the contribution ratio can reflect the changes in power grid stability.

[0082] Photovoltaic power generation and wind power generation are closely related to environmental conditions, based on sunlight intensity. and ambient temperature Constructing environmental impact factors for power ratio correction The formula is: ,in The environmental parameter coefficient is used to adjust the impact of sunlight and temperature on the output power; the environmental impact factor is used to correct the contribution of photovoltaic and wind power generation, so that the contribution ratio of new energy types can be reasonably increased when sunlight is enhanced or wind energy resources are abundant.

[0083] The power output ratio, stability correction weight, and environmental impact factor are integrated, and the contribution ratio of energy input data is calculated through dynamic normalization. The formula is as follows:

[0084] ;

[0085] In the formula, Energy type The corresponding contribution ratio , This represents the fusion weighting factor, used to adjust the strength of the stability correction weight and the environmental impact factor. Indicates energy type The corresponding power output ratio The number of energy types is 3 in this embodiment. The denominator is the sum of the corrected proportions of all energy sources, which is normalized to ensure that the total contribution ratio is 1.

[0086] S2: Based on the contribution ratio, and by integrating historical data trends and real-time fluctuation patterns, a deep sequence prediction model is used to generate energy supply forecasts, while extracting the corresponding forecast uncertainty indicators.

[0087] In this embodiment, instantaneous power data corresponding to photovoltaic (PV) power generation, wind power generation, and grid power are continuously collected. The collection method can be timed recording (e.g., updating every 1 second or 5 seconds) or real-time push based on device communication protocols. Since data from different devices may have timestamp discrepancies, this embodiment performs time alignment processing on all input data before it enters the calculation module, ensuring that each type of energy data has a corresponding value on a unified time scale. After time alignment, the historically collected data is arranged chronologically to construct a historical data trend sequence. This trend sequence describes the changing patterns of energy input over a longer time period; for example, PV power generation shows a gradual upward trend during the day, and wind power exhibits significant fluctuations during certain time periods. This trend information serves as the primary source of long-term time-series characteristics for the prediction model.

[0088] Real-time fluctuation patterns are primarily used to reflect changes on short timescales. In this embodiment, illuminance and ambient temperature are periodically collected and aligned in the same timeframe. Illuminance can characterize rapid output changes that may occur in photovoltaic devices, such as a sudden drop caused by cloud cover; ambient temperature is used to depict short-term factors affecting equipment operating characteristics, such as inverter power limitation due to high temperatures. The aligned illuminance and temperature data are combined to form a real-time fluctuation pattern sequence, which, together with the historical trend sequence, is used to capture the comprehensive temporal behavior of energy input.

[0089] To enable the prediction model to learn both short-term and long-term variation characteristics simultaneously, this embodiment constructs time windows using a multi-scale approach. A multi-scale time window refers to the system extracting multiple continuous data segments from historical trend sequences and real-time fluctuation pattern sequences according to preset lengths (e.g., 10 seconds for short-term, 1 minute for medium-term, and 10 minutes for long-term). Each data segment within a time window contains multiple data sequences, including photovoltaic, wind, and mains power sequences, as well as light intensity and temperature sequences. These sequences are stacked along the time dimension to generate a time series feature tensor with a unified format. This tensor is represented by a three-dimensional array, with the three dimensions representing the number of time steps, the number of different features, and different sample windows, respectively. The time series feature tensor plays a core input role in the deep sequence model, providing a complete multi-time-scale information representation for subsequent prediction tasks.

[0090] After generating the time-series feature tensor, a weight vector is generated based on the contribution ratios calculated from the energy input data. The contribution ratios reflect the actual impact of different energy types at the current moment. For example, during the midday hours when sunlight is strong, photovoltaic (PV) power generation typically contributes a higher percentage and has a larger weight; while at night or on cloudy days, wind power or grid power may become the primary energy source, and the weight vector will change accordingly. The structure of the weight vector is consistent with the number of energy types, typically a three-dimensional vector corresponding to PV, wind power, and grid power inputs. This weight vector enhances the model's ability to distinguish between different energy types, enabling the predictive model to dynamically focus on energy types with a greater impact on the system during the learning process.

[0091] The time-series feature tensors and weight vectors are merged according to a feature concatenation method. This concatenation can be achieved by adding weight information to the feature dimensions; that is, adding one or more columns of weight features related to the contribution ratio to the feature dimensions of the input tensor. This allows the deep sequence model to consider the impact of the current energy input structure on the overall prediction task during learning. The concatenated input feature matrix serves as the final structured feature set. This feature set has a uniform format, fixed shape, and complete temporal information, and can be directly input into the deep sequence prediction model. The structured feature set contains data dimensional relationships that are clear and features arranged in a fixed manner; those skilled in the art can directly construct it using conventional data processing tools or deep learning frameworks.

[0092] Through the above processing, the structured feature set now includes time-series feature tensors, real-time environmental fluctuation features, weight vectors corresponding to the contribution ratios of energy input data, and multi-dimensional feature sequences extracted from multi-scale time windows. The structured feature set typically uses a three-dimensional data format, with the following dimensional structure:

[0093] First dimension: Time step length;

[0094] Second dimension: Number of input features (including photovoltaic power sequence, wind power sequence, mains power sequence, irradiance sequence, temperature sequence, etc.);

[0095] The third dimension: the sample batch dimension.

[0096] To ensure consistency in model input, the system performs normalization on the structured feature set during the data input phase, ensuring that all input features have a uniform numerical range. This step can be achieved using linear scaling or mean-variance normalization methods, which can be directly implemented by those skilled in the art.

[0097] To enhance the ability of deep sequence models to identify complex time series data, this embodiment employs a multivariate variational mode decomposition model (MVMD model) to decompose multi-source time series in a structured feature set into several modal components with different frequency characteristics.

[0098] The system decomposes the input sequence according to a preset modality number parameter M (e.g., M=4 or M=6). The decomposition process can be understood as splitting the input time series into M subsequences based on different frequencies or oscillation characteristics:

[0099] Low-frequency modes are used to characterize long-term trends;

[0100] Mid-frequency modes are used to characterize intraday variations or general fluctuations;

[0101] High-frequency modes are used to capture rapid changes, such as photovoltaic power jumps caused by sudden changes in illumination.

[0102] Each modal component sequence is output as a fixed-length time series and used as input for subsequent depth sequence models.

[0103] The deep prediction model in this embodiment uses LSTM (Long Short-Term Memory) network as the temporal modeling structure. For each modal component sequence, the following model operations are performed sequentially:

[0104] (1) Input layer

[0105] The input layer accepts a sequence of modal components in the form of a two-dimensional array, which contains the time step dimension and the feature dimension (generally, each modal component contains only one feature dimension).

[0106] (2) First-layer LSTM network

[0107] This layer is used to identify short-term dependencies in modal sequences and contains several LSTM units. This embodiment does not limit the number of units; those skilled in the art can select, for example, 32, 64, or 128 units depending on resource configuration.

[0108] (3) Deep LSTM layer

[0109] Following the first layer, the model may contain 1–3 LSTM hidden layers to capture long-term dependencies in the modal sequences. In this embodiment, a multi-layered structure is optional, but at least one hidden LSTM layer is required to ensure the model has sufficient expressive power.

[0110] (4) Output layer

[0111] The model ultimately outputs the predicted subsequence corresponding to the modal component. The output format is usually a sequence of predicted results that is consistent with the input time step.

[0112] Each modal component is predicted separately using an LSTM model with the same structure but different parameters to ensure that each modal feature is fully learned.

[0113] The model training process can utilize existing deep learning frameworks, with training data including historical feature sets as input and actual energy outputs as labels. Training methods typically employ backpropagation and gradient descent optimization, requiring no additional special steps.

[0114] After obtaining the predicted subsequences corresponding to all modal components, the predicted subsequences are added point-by-point according to the modal superposition method to reconstruct the complete energy supply prediction master sequence. The prediction master sequence is a time-varying sequence, representing the predicted energy supply capacity at each future moment. The sequence's first... The element is the time. The energy supply forecast will be used in subsequent steps for dynamic threshold adjustments and energy allocation strategy formulation.

[0115] To characterize the uncertainty of the prediction results, this embodiment constructs a prediction uncertainty index by calculating the prediction mean and prediction standard deviation of the main prediction sequence. The prediction mean is directly obtained from the values ​​of the corresponding time points of the main prediction sequence.

[0116] The predicted standard deviation is obtained by comparing the error between the model output and the historical true value, or by sampling the model's output multiple times. This embodiment uses the square root of the approximate residual variance as the predicted standard deviation, enabling those skilled in the art to calculate this index using conventional estimation methods.

[0117] Construct a prediction confidence interval using the predicted mean and predicted standard deviation. The upper bound of the interval is given by the following relationship: ;in For at any time The confidence interval for the energy supply forecast. For at any time The average of the energy supply forecasts, For at any time The standard deviation of energy supply forecasts This is a quantification factor, set by the system. The size of k can be set according to task requirements, for example, to 1, 1.5 or 2, to reflect the predicted risk in different time periods.

[0118] S3: Determine whether the predicted energy supply value is lower than the preset power supply stability threshold, and dynamically adjust the power supply stability threshold based on the prediction uncertainty index to generate an adjustment priority sequence.

[0119] In one preferred embodiment, the system first obtains the energy supply forecast value output by the energy supply forecasting model, that is, the energy supply forecast value at a future time. The projected energy supply capacity is denoted as The system also has a preset power supply stability threshold. This threshold represents the minimum energy input capacity that the system needs to guarantee under normal operating conditions. To determine whether future energy shortages are possible, the difference between the predicted energy value and the power supply stability threshold is constructed: ;in, Indicates at time The initial stability difference; if If so, it indicates a potential shortage of energy supply; if If the energy supply forecast meets or exceeds the stability requirements, then the difference reflects the degree to which the forecast deviates from the stability target.

[0120] Two uncertainty indicators are obtained from the prediction model:

[0121] : Predicted mean, i.e., the predicted value of the main sequence at time t. The corresponding value;

[0122] The predicted standard deviation is obtained by analyzing the historical error of the model or the sampling distribution.

[0123] Constructing the prediction confidence interval bias : ;

[0124] A threshold correction coefficient is constructed based on the initial stability difference and the predicted confidence interval deviation. : ,in, , The weighting coefficients are set by the system and can be adjusted according to application requirements, for example... , This indicates a greater emphasis on the impact of inadequate forecasting;

[0125] Update the power supply stability threshold based on the threshold correction coefficient: ,in The updated power supply stability threshold is used to replace the fixed threshold. Execute subsequent priority generation actions. The system will continuously update as the prediction results change. This dynamic threshold adjustment mechanism can be implemented in software within an energy storage controller, energy management system (EMS), or industrial automation control unit.

[0126] Based on energy supply forecasts Compared with the updated power supply stability threshold Construct a multidimensional priority factor, including:

[0127] The energy supply forecasting bias factor measures the potential of different energy types to compensate for system supply imbalances in the future.

[0128] Available power factor: Calculated in real time based on the output power of different energy types, such as instantaneous power of photovoltaics, instantaneous power of wind power, and power available from the mains.

[0129] Energy response speed factor: used to distinguish the dynamic adjustability of different energy types. For example, photovoltaics responds quickly and can adjust its output rapidly; wind energy is greatly affected by natural conditions; grid electricity is stable but may be limited by the grid.

[0130] To derive the ranking results for different energy types, the system assigns corresponding weights to the three factors mentioned above. For example, the deviation factor weight highlights the compensation capability when forecasts are insufficient, the available power factor weight reflects the actual power that an energy type can currently provide, and the response speed factor weight determines the applicability of an energy type in dynamic adjustments. The three factors are weighted and summed to obtain a comprehensive priority score for each energy type. The system then ranks these scores to obtain an adjustment priority sequence, for example: photovoltaic energy > wind energy > grid power. This sequence is updated in each scheduling cycle to adapt to changes in energy input.

[0131] To avoid frequent and drastic changes in priority within a short period, the system introduces an exponential smoothing method to smooth the priority sequence within several consecutive time windows. The smoothing process uses a decay coefficient to increase the weight of recent priorities and decrease the weight of older records, thus achieving a more stable priority trend. An allocation weight vector is constructed based on the smoothed priority sequence to reflect the target proportion of different energy types in the initial energy allocation scheme.

[0132] The weight vector is normalized so that the sum of all weight values ​​is 1, thus meeting the requirements for subsequent power allocation. For example: photovoltaic: 0.50, wind power: 0.35, grid power: 0.15. This allocation weight vector will be directly used to construct the initial energy allocation scheme.

[0133] S4: Based on the adjustment priority sequence, obtain the capacity status and voltage status of the energy storage device, perform matching processing on the energy input data, and form an initial energy allocation scheme that includes the target allocation ratio of each energy type.

[0134] In one embodiment, the key operating status of the energy storage device is first obtained, including:

[0135] Remaining capacity: This represents the energy available for charging the energy storage device, which is measured in real time by the energy storage management system (BMS).

[0136] Maximum rechargeable capacity: This indicates the maximum charging capacity that the energy storage device can accept under current operating conditions. It is usually calculated in real time by the BMS based on battery health status, temperature, and lifespan management strategies.

[0137] Real-time voltage: Represents the current terminal voltage of the energy storage device, which can be used to determine whether further charging is permitted;

[0138] Status of Health (SOH): This represents the overall performance of an energy storage unit after it degrades over time, and is usually expressed as a percentage.

[0139] The above four types of information are combined to form energy storage availability information, which is used to guide the subsequent energy matching and allocation process. Energy storage availability information reflects the ability of an energy storage system to accept different energy inputs and provides safety and reliability boundaries for allocation schemes.

[0140] Based on the generated adjustment priority sequence, photovoltaic energy, wind energy, and grid power are arranged from high to low according to their comprehensive priority, constructing a target matching order for multiple energy types, and obtaining the input data corresponding to each energy type, including:

[0141] Available power of energy input data: This represents the maximum output power that each energy type can provide at the current moment;

[0142] Contribution ratio of energy input data: This represents the proportion of each energy type in the total energy input;

[0143] Time series trend factor corresponding to energy input data: calculated from historical data trend series, used to characterize the stability trend of energy in future moments;

[0144] Instantaneous characteristics of energy input data: such as the impact of instantaneous solar radiation on photovoltaics, the impact of instantaneous temperature on wind power equipment, and the instantaneous voltage fluctuations of the mains power.

[0145] Based on the target matching order, the above energy input data is sequentially matched with energy storage availability information to determine whether different energy inputs are suitable for the current energy storage state. For example, when the energy storage voltage is close to its upper limit, the system will reduce the charging ratio of the corresponding energy type; while when the energy storage capacity is sufficient, it allows the allocation of higher energy inputs.

[0146] Based on the matching results between energy input data and energy storage availability information, an energy storage allocation factor is constructed. The energy storage allocation factor includes the following:

[0147] Allowable charging capacity for energy storage: This represents the amount of energy that can be allocated to the energy storage device at the current moment. This amount is determined by the remaining capacity and the health status of the energy storage.

[0148] Allowable voltage range for energy storage: Used to determine whether the input of a certain type of energy may cause the voltage to exceed the limit, thereby limiting the allocation ratio;

[0149] Allowable charging rate for energy storage: Determined by the safe charging rate set by the battery manufacturer and real-time temperature conditions, to ensure that the charging rate does not exceed the safe range of the equipment;

[0150] Energy storage response speed parameter: This indicates the ability of an energy storage device to handle power changes per unit time. For example, some energy storage systems can accept high-power charging instantaneously, while some systems respond more slowly.

[0151] The energy storage allocation factor comprehensively reflects the real-time charging capability of the energy storage system.

[0152] To ensure that the final energy allocation scheme meets physical, security, and grid constraints, this embodiment constructs an energy allocation search space. This search space defines the possible range of values ​​for the initial energy allocation scheme and includes the following three types of constraints:

[0153] Energy storage protection range constraint: Limit the input power so that the voltage, temperature or SOC of the energy storage system does not exceed the allowable range.

[0154] Energy storage capacity constraint: The input power allocated to energy storage is limited to the maximum available charging capacity.

[0155] Energy input available power constraint: restricts the allocation ratio of a certain energy type from exceeding its real-time available power.

[0156] By constructing the above constraints into a multidimensional feasible region, the subsequent allocation algorithm will not produce outputs that violate the operating constraints during the search process.

[0157] After constructing the energy allocation search space, further considerations are needed:

[0158] Power flow constraints in power grid branches: used to limit the maximum power that can be transmitted in power grid lines to prevent line overload;

[0159] Power grid stability conditions: such as allowable voltage deviation range and frequency deviation range, to ensure that the distribution scheme does not affect power grid stability;

[0160] Feasibility domain pruning method: used to reduce the search space and improve the efficiency of energy allocation solution.

[0161] By analyzing all constraints, the target allocation ratio for each energy type is determined. For example, if photovoltaic power generation currently has sufficient energy and energy storage capacity, the target allocation ratio for the photovoltaic side will be appropriately increased; if wind power fluctuates significantly, the allocation ratio may be reduced; and when the external power grid is unstable, grid power input may be severely limited.

[0162] S5: Based on the initial energy allocation scheme, an optimization algorithm is used to allocate resources and generate a basic power transmission path. During the optimization process, when a positive deviation is detected between the photovoltaic power generation data and the predicted energy supply value, the photovoltaic power allocation is adjusted according to the basic power transmission path to determine the charging input flow.

[0163] In this embodiment, the system first constructs a resource configuration input set based on the initial energy allocation scheme. The initial energy allocation scheme includes the target allocation ratios for different energy types (photovoltaic, wind, and grid power) in the current scheduling cycle. To ensure the information integrity of the optimization process, the system includes the following content in the resource configuration input set:

[0164] The target allocation ratio for each energy type: used to represent the initial reference point of the optimization problem, giving the optimization process a clear direction;

[0165] Available power of energy input data: provided by real-time monitoring, reflecting whether each energy type can actually meet the requirements of the initial allocation plan;

[0166] Energy storage allocation factor: constructed by step S4, including allowable energy storage charging amount, allowable voltage range, allowable charging rate and energy storage response speed;

[0167] Real-time status observations of the energy storage device include current voltage, SOC (state of charge), temperature, etc., to ensure that the optimization results meet the operating conditions of the energy storage device.

[0168] The resource configuration input set provides the necessary initial data structure for subsequent modeling. The system uses the data management module inside the controller to uniformly number all the above parameters and repackage them into the input format required by the optimization algorithm.

[0169] To ensure that the optimization results meet system safety requirements and power grid stability conditions, this embodiment constructs a resource configuration constraint set based on the resource configuration input set. The resource configuration constraint set is divided into two categories:

[0170] (1) Unbreakable constraints: These are used to ensure that optimization actions do not cause the system to enter an unsafe state, and include the following:

[0171] Branch power flow constraint: Used to limit the maximum power transmission capacity of power grid branches to prevent line overload;

[0172] Power grid stability conditions include allowable voltage deviation range and frequency stability range, which must not be exceeded.

[0173] Allowable charging rate for energy storage: Used to limit the maximum charging power that energy storage can accept per unit time;

[0174] Energy storage protection range: such as the upper and lower limits of permissible voltage, the safe operating range of SOC, etc.

[0175] If any resource allocation action violates the above constraints, the system will directly determine it as an infeasible solution during the optimization search phase.

[0176] (2) Adjustable constraints: used to adjust the optimization direction of the objective function, including: soft constraints on the allocation ratio deviating from the initial target, soft constraints on the power adjustment speed, and preference settings for resource utilization efficiency.

[0177] During the optimization process, the system adjusts the degree of influence of such constraints through a penalty factor to ensure that the final configuration changes continuously within a reasonable range.

[0178] In this embodiment, a fast model predictive control algorithm is used to perform resource allocation action search within the rolling time domain. Its principle is as follows:

[0179] First, establish the optimization problem based on the current resource allocation input set and constraint set;

[0180] Set a prediction time window (e.g., the next 10 time steps) and search for feasible power output trajectories for different energy types within that window;

[0181] The MPC controller determines the feasibility of each candidate action based on the resource configuration constraint set and directly excludes actions that violate the inviolable constraints.

[0182] The system uses a solver (which can be a quadratic programming QP or a nonlinear programming NLP solver) to find the sequence of actions that optimizes the objective function within the current prediction window in the feasible solution space.

[0183] The resulting action sequence is the resource configuration action sequence.

[0184] Each action sequence contains multiple feasible power output values ​​at consecutive moments. For example: photovoltaic power at moment... , , The target power output; the allowable input power of wind energy at future moments; and the compensation power capacity of the mains power in the time domain.

[0185] MPC employs a rolling optimization approach, meaning that after each action is executed, the system moves to the next time step and recalculates the resource configuration action.

[0186] To avoid drastic fluctuations in energy input between adjacent prediction times, the system establishes rules to limit changes in the allocation ratio. In this embodiment, the rules include:

[0187] Limit the variation in power output of a certain energy type between two consecutive moments, for example, by not exceeding the maximum permissible variation;

[0188] By constraining the rate of change, the power change process is ensured to be smooth, avoiding shocks to the equipment caused by sudden and large adjustments;

[0189] More stringent restrictions should be imposed on energy types with high volatility (such as wind power) to enhance overall dispatch stability.

[0190] The system applies the above rules in real time when searching for action sequences, so that the selected resource configuration actions are continuous and stable in terms of power changes.

[0191] After the resource allocation action sequence is determined, the system converts this sequence into a basic power transmission path. The basic power transmission path is a set of paths describing the power transmission relationship between different energy types and the energy storage device. The path structure records the target power input direction and magnitude for each energy source at future times, thus providing a clear data foundation for photovoltaic offset adjustment and charging input flow generation.

[0192] In this embodiment, the basic power transmission path can be formed into a data structure within the energy storage equipment controller, implemented as a graph structure or a time series structure. Those skilled in the art can complete data storage and retrieval according to the specific hardware platform.

[0193] Furthermore, methods for determining the charging input current include:

[0194] Obtain the actual power value of photovoltaic power generation at the current moment. Compared with energy supply forecasts Deviation information between , is represented as: ;

[0195] The deviation value reflects whether actual photovoltaic power generation exceeds prediction, i.e., whether there is additional usable power on the photovoltaic side. To classify and handle deviations of different degrees, this embodiment establishes multi-level deviation judgment conditions, including:

[0196] Mild deviation level: Above zero but below a slightly preset threshold;

[0197] Moderate deviation level: Above the mild threshold but not exceeding the moderate threshold;

[0198] Severe deviation level: A value above the moderate threshold indicates a significant surge in photovoltaic power.

[0199] The threshold can be set by system management parameters. For example, the threshold for mild cases is 5% to 10% of the extra power, and the threshold for moderate cases is 10% to 20%. The specific value can be configured according to the actual situation of the project and adjusted according to the characteristics of different equipment.

[0200] After obtaining the deviation level, the system selects the corresponding preset photovoltaic power adjustment strategy based on the deviation level. The purpose of the adjustment strategy is to: rationally utilize the extra power when there is a sudden increase in photovoltaic power generation; prevent the energy storage system from entering an unsafe state due to overcharging; and maintain a stable and continuous power transmission process.

[0201] Preset strategies include, but are not limited to, the following:

[0202] Slight deviation strategy: Slightly increase photovoltaic power allocation while keeping the allowable charging capacity of energy storage unchanged or slightly increased; prioritize the continued energy flow of the current base path.

[0203] Moderate deviation strategy: moderately increase the proportion of photovoltaic power input to energy storage; if the energy storage allows insufficient charging capacity, some photovoltaic power will be limited; adjust the allocation to avoid voltage or SOC approaching the upper limit.

[0204] Severe deviation strategy: Significantly increase photovoltaic input power but not exceed the maximum allowable charging power of energy storage; if energy storage can no longer absorb, implement power trimming or feed back to the inverter for derating operation; prevent safety hazards such as energy storage overcharging and voltage overshoot.

[0205] The strategy is triggered by the photovoltaic side deviation level. Different levels correspond to different incremental coefficients, response parameters and constraint rules. The system can implement this through tabular parameters or logical operations.

[0206] The adjustment of photovoltaic power must be carried out within the photovoltaic power allocation segment corresponding to the base power transmission path. During the adjustment process, the system must simultaneously meet the following constraints:

[0207] Photovoltaic maximum allowable output power constraint: Photovoltaic inverters have a rated output limit that cannot be exceeded by hardware limitations;

[0208] Maximum allowable charging power constraint for energy storage: determined by the allowable charging rate and health status of the energy storage system;

[0209] Energy storage protection range constraints include voltage range, SOC range, and temperature limits.

[0210] Based on the above constraints, the system makes restrictive adjustments to the photovoltaic input power according to the deviation level. For example:

[0211] like If the energy storage capacity is large enough and there is still room for further energy storage, then more input power can be added;

[0212] If the energy storage capacity is already close to full, the increase will be limited and excess photovoltaic power may be reduced.

[0213] The final adjusted power input value on the photovoltaic side is thus formed.

[0214] In this embodiment, the charging input stream represents the actual charging power sequence ultimately executed by the energy storage device, and is the target input sent by the system control module to the energy storage converter and charging controller. The final charging input stream is generated through the following two steps:

[0215] (1) Generate candidate charging input streams based on adjustment steps: After performing photovoltaic power adjustment, the system combines the adjustment results with the basic power paths of other energy types to form the original charging input value at the current moment. This value may contain instantaneous fluctuations, so it cannot be directly used as the final control signal.

[0216] (2) Construct smoothing rules based on historical charging input flow: In order to avoid drastic changes in power at continuous moments (which would cause mechanical and electrical shocks to energy storage devices), the system constructs smoothing rules, such as exponential smoothing, moving average, and linear smoothing with a limit on the rate of change.

[0217] Smoothing rules can be implemented in the following form: ,in The smoothing coefficient can be configured according to the system response speed, such as 0.2 to 0.5, resulting in the final... This is the final charging input stream.

[0218] like Figure 2 As shown, another embodiment of the present invention provides a multi-element coexistence and fusion system for adaptive photovoltaic energy storage equipment, comprising:

[0219] The energy data acquisition module is used to acquire energy input data, power grid stability data, and environmental parameters, and to calculate the contribution ratio of the energy input data.

[0220] The energy forecasting module is used to generate energy supply forecasts based on the contribution ratio of energy input data through a deep sequence forecasting model, by integrating historical data trends and real-time fluctuation patterns, and to extract the corresponding forecast uncertainty indicators.

[0221] The threshold adjustment module is used to determine whether the predicted energy supply value is lower than the preset power supply stability threshold, and to dynamically adjust the power supply stability threshold according to the prediction uncertainty index to generate an adjustment priority sequence.

[0222] The matching processing module is used to obtain the capacity status and voltage status of the energy storage device according to the adjustment priority sequence, and to perform matching processing on the energy input data to form an initial energy allocation scheme that includes the target allocation ratio of each energy type.

[0223] The resource allocation module is used to allocate resources using an optimization algorithm based on the initial energy allocation scheme in order to generate a basic power transmission path;

[0224] The deviation adjustment module is used to adjust the photovoltaic power allocation according to the basic power transmission path to determine the charging input flow when a positive deviation is detected in the photovoltaic power generation data relative to the energy supply forecast during the resource allocation optimization process.

[0225] In summary, this invention achieves a dynamic and adaptive energy fusion control system for multi-energy input scenarios by constructing an energy input contribution ratio, integrating deep sequence prediction, dynamically adjusting power supply thresholds, matching energy storage status, optimizing resource allocation, and adjusting charging input flow based on deviation triggering. This invention can improve the energy storage system's ability to withstand new energy fluctuations, optimize energy utilization efficiency, and provide a highly reliable and flexible solution for multi-energy collaborative operation, while ensuring grid stability requirements.

[0226] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all 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 method for the coexistence and integration of multiple adaptive photovoltaic energy storage devices, characterized in that, The method includes: Acquire energy input data, power grid stability data, and environmental parameters, and calculate the contribution ratio of the energy input data, wherein the energy input data includes photovoltaic power generation data, wind power generation data, and grid power data; the power grid stability data includes grid voltage deviation, frequency deviation, and related stability indicators; and the environmental parameters include light intensity and ambient temperature. Based on the contribution ratio, and by integrating historical data trends and real-time fluctuation patterns, a deep sequence prediction model is used to generate energy supply forecasts, while extracting the corresponding forecast uncertainty indicators. Determine whether the predicted energy supply value is lower than a preset power supply stability threshold, and dynamically adjust the power supply stability threshold based on the prediction uncertainty index to generate an adjustment priority sequence; Based on the adjustment priority sequence, the capacity status and voltage status of the energy storage device are obtained, and the energy input data is matched to form an initial energy allocation scheme that includes the target allocation ratio of each energy type. Based on the initial energy allocation scheme, an optimization algorithm is used to allocate resources and generate a basic power transmission path. During the optimization process, when a positive deviation is detected between the photovoltaic power generation data and the predicted energy supply value, the photovoltaic power allocation is adjusted according to the basic power transmission path to determine the charging input flow. Calculating the contribution ratio of the energy input data includes: Based on the instantaneous photovoltaic power output value corresponding to photovoltaic power generation data Instantaneous wind power output value corresponding to wind power generation data Instantaneous mains power output value corresponding to mains power data Establish a power vector group for energy input data. ; The power output ratio is calculated based on the power vector set, using the following formula: ,in Indicates energy type The corresponding power output ratio Indicates energy type The corresponding instantaneous power output value; Based on grid voltage deviation Frequency deviation and related stability indicators Generate stability correction weights for power percentage correction The formula is: ,in Indicates the correction factor; Based on light intensity and ambient temperature Constructing environmental impact factors for power ratio correction The formula is: ,in Indicates environmental parameter coefficients; The power output ratio, stability correction weight, and environmental impact factor are fused together, and the contribution ratio of the energy input data is generated through dynamic normalization calculation, as shown in the formula: ； In the formula, Energy type The corresponding contribution ratio , Indicates the fusion weight factor. Indicates energy type The corresponding power output ratio For the number of energy types.

2. The method for multi-element coexistence and integration of adaptive photovoltaic energy storage equipment according to claim 1, characterized in that, The input to the deep sequence prediction model is constructed through the following steps: Historical data trend sequences were constructed based on photovoltaic power generation data, wind power generation data, and grid power data; and real-time fluctuation pattern sequences were constructed based on light intensity and ambient temperature. A multi-scale time window is constructed based on the historical data trend sequence and real-time fluctuation pattern sequence, and a time series feature tensor is established within the multi-scale time window; A weight vector is constructed based on the contribution ratio of energy input data, and the time series feature tensor and the weight vector are combined according to the feature concatenation method to generate the input feature matrix. A structured feature set is constructed based on the input feature matrix to serve as the input data for the deep sequence prediction model.

3. The method for multi-element coexistence and integration of adaptive photovoltaic energy storage equipment according to claim 2, characterized in that, The generated energy supply forecast value is simultaneously extracted, along with corresponding forecast uncertainty indicators, including: The structured feature set is input into a multivariate variational mode decomposition model, based on the set modality number parameter. Output There are 1 modal component sequences, and each modal component sequence corresponds to a decomposed subsequence; The modal component sequences are input into a deep sequence sub-model consisting of several layers of long short-term memory networks. The deep sequence sub-model generates a corresponding prediction sub-sequence for each modal component sequence. The main prediction sequence is constructed based on all predicted subsequences in a modal superposition manner, which is the predicted value of energy supply. Using the confidence interval expression Generate a predictive uncertainty index, in which For at any time The confidence interval for the energy supply forecast. For at any time The average of the energy supply forecasts, For at any time The standard deviation of energy supply forecasts This is a quantification factor.

4. The method for multi-element coexistence and integration of adaptive photovoltaic energy storage equipment according to claim 1, characterized in that, Dynamically adjusting the power supply stability threshold based on the predicted uncertainty index includes: Based on energy supply forecasts Compared with the preset power supply stability threshold Constructing the initial stability difference The formula is: ; Based on the mean of energy supply forecasts Standard deviation of energy supply forecasts and quantification factor Constructing the prediction confidence interval bias : ; A threshold correction coefficient is constructed based on the initial stability difference and the predicted confidence interval deviation. : ,in, , To adjust the weighting coefficients; Update the power supply stability threshold based on the threshold correction coefficient: ,in This is the updated power supply stability threshold.

5. The method for multi-element coexistence and integration of adaptive photovoltaic energy storage equipment according to claim 4, characterized in that, The generation of the adjustment priority sequence includes: A multidimensional priority factor is constructed based on the energy supply forecast and the updated power supply stability threshold. The multidimensional priority factor includes an energy supply forecast deviation factor, an energy available power factor, and an energy response speed factor. An adjusted priority sequence is generated by weighted sorting based on multidimensional priority factors, and the adjusted priority sequence is smoothed by exponential smoothing within the time window. An allocation weight vector is constructed based on the adjusted priority sequence. The allocation weight vector is generated in a normalized manner and used for subsequent calculations of the initial energy allocation scheme.

6. The method for multi-element coexistence and integration of adaptive photovoltaic energy storage equipment according to claim 1, characterized in that, The formation of the initial energy allocation scheme, which includes the target allocation ratio for each energy type, includes: Obtain the current remaining capacity, maximum rechargeable capacity, real-time voltage, and health status information of the energy storage device to construct energy storage availability information; Based on the adjustment priority sequence, a target matching order for multiple energy types is constructed. Energy input data is matched with energy storage availability information according to the target matching order. The matching content includes the available power of energy input data, the contribution ratio of energy input data, the time series trend factor corresponding to energy input data, and the instantaneous characteristics of energy input data. Based on the matching results of energy storage availability information and energy input data, an energy storage allocation factor is constructed. The energy storage allocation factor includes the allowable energy storage charging amount, the allowable energy storage voltage range, the allowable energy storage charging rate, and the energy storage response speed parameter. An energy allocation search space is constructed based on energy storage allocation factors and adjustment priority sequences. The energy allocation search space includes energy storage protection interval constraints, energy storage capacity constraints, and available energy input power constraints. An initial energy allocation scheme is generated based on the energy allocation search space, combined with the branch power flow constraints of the power grid, the power grid stability conditions, and the feasibility domain pruning method. The initial energy allocation scheme includes the target allocation ratio corresponding to each energy type.

7. The method for multi-element coexistence and integration of adaptive photovoltaic energy storage equipment according to claim 1, characterized in that, The generation of the basic power transmission path includes: Based on the initial energy allocation scheme, a resource allocation input set is constructed, which includes the target allocation ratio of each energy type, the available power of energy input data, the energy storage allocation factor, and the real-time status observation of the energy storage device. Based on the resource allocation input set, combined with the power grid branch flow constraints, power grid stability conditions, allowable charging rate of energy storage, and energy storage protection interval, a resource allocation constraint set is constructed. The resource allocation constraint set includes inviolable constraints and adjustable constraints. Within the rolling time domain, a fast model predictive control algorithm is used to search for feasible resource allocation action sequences based on the resource allocation constraint set. The resource allocation action sequences correspond to the feasible power output of each energy type at multiple prediction times. Based on the range of resource allocation action changes between adjacent prediction times, a rule for limiting the change of allocation ratio is constructed, and the change magnitude of resource allocation action is controlled according to the rule for limiting the change during the resource allocation action search process. The basic power transmission path is determined based on the resource allocation action sequence. The basic power transmission path is a set of paths that represent the power transmission relationship from each energy type to the energy storage device.

8. The method for multi-element coexistence and integration of adaptive photovoltaic energy storage equipment according to claim 1, characterized in that, Methods for determining the charging input current include: The deviation information between the actual power value of photovoltaic power generation data at the current moment and the predicted energy supply value is obtained, and multi-level deviation judgment conditions are constructed based on the deviation information. The multi-level deviation judgment conditions include slight deviation, moderate deviation and severe deviation. The photovoltaic side deviation level is determined based on multi-level deviation judgment conditions. The photovoltaic side deviation level corresponds to a preset photovoltaic power adjustment strategy. The adjustment strategy includes adjusting the distribution relationship between photovoltaic output power and energy storage allowable charging amount. Within the photovoltaic power allocation section corresponding to the basic power transmission path, a power adjustment step is performed based on the photovoltaic side deviation level. The power adjustment step is limited by the maximum allowable output power of photovoltaic, the maximum allowable charging power of energy storage, and the energy storage protection range. The charging input stream is constructed based on the photovoltaic power adjustment steps, and a smoothing rule for the charging input stream is constructed based on the historical charging input stream. The final charging input stream is generated through the smoothing rule.

9. A multi-element coexistence fusion system for adaptive photovoltaic energy storage equipment, applied to the multi-element coexistence fusion method for adaptive photovoltaic energy storage equipment as described in claim 1, characterized in that, The system includes: The energy data acquisition module is used to acquire energy input data, power grid stability data, and environmental parameters, and to calculate the contribution ratio of the energy input data. The energy forecasting module is used to generate energy supply forecasts based on the contribution ratio of energy input data through a deep sequence forecasting model, by integrating historical data trends and real-time fluctuation patterns, and to extract the corresponding forecast uncertainty indicators. The threshold adjustment module is used to determine whether the predicted energy supply value is lower than the preset power supply stability threshold, and to dynamically adjust the power supply stability threshold according to the prediction uncertainty index to generate an adjustment priority sequence. The matching processing module is used to obtain the capacity status and voltage status of the energy storage device according to the adjustment priority sequence, and to perform matching processing on the energy input data to form an initial energy allocation scheme that includes the target allocation ratio of each energy type. The resource allocation module is used to allocate resources using an optimization algorithm based on the initial energy allocation scheme in order to generate a basic power transmission path; The deviation adjustment module is used to adjust the photovoltaic power allocation according to the basic power transmission path to determine the charging input flow when a positive deviation is detected in the photovoltaic power generation data relative to the energy supply forecast during the resource allocation optimization process.

Images