A control method for coordinating voltage fluctuation suppression of distributed energy storage and renewable energy in an agricultural power distribution network

Abstract

The application discloses a kind of agricultural distribution network distributed energy storage and renewable energy collaborative control method of voltage fluctuation, it is related to distribution network voltage stability control technical field, after high proportion distributed photovoltaic is accessed to agricultural distribution network, photovoltaic output and agricultural load strong volatility cause node voltage large fluctuation problem, the method constructs " perception-decision-execution-feedback-optimization " full closed loop control system, through wavelet analysis, attribution analysis and the multi-feature algorithm of attention mechanism fusion, accurately identify voltage fluctuation type, leading factor and complete grade determination;Dynamic weight multi-objective optimization control model is constructed based on LSTM network, and source network load storage collaborative suppression is realized using energy storage SOC layered regulation and control strategy, and model and strategy parameters are corrected through real-time operation data feedback. Realize agricultural distribution network voltage accurate stability control, while giving consideration to photovoltaic high proportion of consumption, energy storage life protection and agricultural production power supply reliability.

Description

Technical Field

[0001] This invention relates to the field of voltage stability control technology for power distribution networks, specifically a control method for synergistically mitigating voltage fluctuations through distributed energy storage and renewable energy in agricultural power distribution networks. Background Technology

[0002] Agricultural production loads (such as irrigation pumps, greenhouse temperature control equipment, and aquaculture oxygen supply and ventilation equipment) exhibit typical characteristics of strong seasonality, strong time-varying nature, and strong abrupt changes: scenarios such as short-term surges in load during irrigation, large fluctuations in photovoltaic output under sunny and rainy weather, and concentrated start-up of temperature control loads at night can easily lead to large fluctuations in voltage at agricultural power distribution network nodes, such as sudden increases or decreases.

[0003] Voltage fluctuations can range from minor issues like malfunctions and increased energy consumption in sensitive agricultural equipment to serious consequences such as unexpected shutdowns of irrigation, seedling temperature control, and livestock oxygen supply systems. This directly leads to reduced crop yields and livestock losses, severely impacting the stability of agricultural production. Therefore, stable voltage control has become a core technical challenge for ensuring high-proportion photovoltaic power grid integration and safe, stable operation in agricultural power distribution networks.

[0004] Existing voltage fluctuation mitigation technologies for power distribution networks have the following core shortcomings:

[0005] Poor scenario adaptability: Existing solutions are mostly designed for urban and industrial power distribution networks, without taking into account the distributed photovoltaic characteristics of agricultural power distribution networks, the operation patterns of agricultural loads and the power supply reliability requirements of agricultural production. They cannot balance agricultural production safety, high photovoltaic consumption ratio and equipment operation economy.

[0006] Insufficient accuracy in fluctuation identification: It often uses a single time-domain or frequency-domain analysis method, which cannot accurately distinguish the type of voltage fluctuation or quantify the contribution of different factors. It is easily affected by abnormal data acquisition, instantaneous equipment failure, and extreme weather, resulting in frequent false fluctuation identification and misjudgment, leading to malfunctions in control strategies.

[0007] Insufficient flexibility of control objectives: The collaborative control model mostly adopts multi-objective optimization with fixed weights, which cannot dynamically adjust the control priority according to the dominant factors of voltage fluctuations. This easily leads to problems such as low photovoltaic absorption rate, excessive energy storage charging and discharging losses, and substandard voltage control effect, making it difficult to adapt to the differentiated scenarios of photovoltaic fluctuations and load changes.

[0008] The energy storage control strategy is unreasonable: the lack of a hierarchical control strategy that matches the voltage fluctuation level and the energy storage SOC state can easily lead to overcharging and over-discharging of energy storage, which can significantly shorten the energy storage cycle life. At the same time, there is no emergency regulation mechanism when the SOC exceeds the limit, and the voltage smoothing capability is completely ineffective in extreme scenarios.

[0009] Lack of closed-loop adaptive optimization capability: Without a complete feedback optimization mechanism, it is unable to adaptively correct model parameters and control strategies based on actual control effects. When the operating conditions of the power distribution network, meteorological conditions, and agricultural production cycles change, the control robustness and adaptability are severely insufficient. Summary of the Invention

[0010] The purpose of this invention is to provide a control method for synergistically mitigating voltage fluctuations in agricultural power distribution networks by combining distributed energy storage and renewable energy. By providing a control method for synergistically mitigating voltage fluctuations in agricultural power distribution networks by combining distributed energy storage and photovoltaics, this invention addresses the core pain points of existing technologies, such as low accuracy in identifying voltage fluctuations in agricultural power distribution networks, poor adaptability to different scenarios, fixed control objectives, large energy storage lifespan losses, and lack of closed-loop optimization. This method achieves accurate, stable, and economical control of voltage in agricultural power distribution networks, while also taking into account high photovoltaic uptake, energy storage cycle life protection, and reliable power supply for agricultural production.

[0011] To achieve the above objectives, the present invention provides the following technical solution:

[0012] A control method for synergistically mitigating voltage fluctuations through distributed energy storage and photovoltaic power in agricultural distribution networks includes the following steps:

[0013] S1: Collect photovoltaic operation parameters, distributed energy storage operation parameters, distribution network operation parameters, agricultural load operation parameters and environmental operation parameters, and integrate multiple types of parameters to establish a collaborative voltage fluctuation control database;

[0014] S2: Based on the monitoring data of the collaborative voltage fluctuation control database, a multi-feature fusion fluctuation identification algorithm is used to identify the type, amplitude and dominant factors of voltage fluctuation in the distribution network, and the voltage fluctuation level is determined by combining the photovoltaic distribution characteristics and agricultural load operation patterns in the area where the agricultural distribution network is located.

[0015] S3: Construct a distributed energy storage and photovoltaic coordinated damping control model. Based on the identified dominant factors of voltage fluctuation, dynamically adjust the control target weights of maximizing photovoltaic absorption rate, minimizing distribution network voltage deviation, and minimizing energy storage charging and discharging losses. Combine the rated capacity of distribution network equipment, real-time operating parameters, and agricultural distribution network operation and maintenance safety standards to set voltage stability constraints and coordinated control objectives.

[0016] S4: Based on the determined voltage fluctuation level and the output results of the collaborative suppression control model, a differentiated collaborative suppression control strategy is generated. The energy storage SOC hierarchical regulation strategy is adopted to dynamically adjust the charging and discharging power of distributed energy storage and photovoltaic output, and to link the reactive power compensation equipment of the distribution network to suppress the voltage fluctuation of the distribution network. S5: The voltage status of the distribution network nodes is monitored in real time. The monitoring results are compared with the collaborative control target. Combined with the distribution network operation and maintenance data, historical fluctuation data and prediction data, the parameters of the collaborative suppression control model and the collaborative suppression control strategy are adjusted to correct the voltage fluctuation identification deviation.

[0017] Further, step S1 specifically involves: collecting parameters by deploying monitoring terminals at photovoltaic grid-connected nodes, energy storage access nodes, key distribution network line nodes, concentrated agricultural load areas, and field meteorological monitoring points; the core parameters collected include photovoltaic output, energy storage SOC, distribution network node voltage, agricultural load power, light intensity, and ambient temperature.

[0018] Furthermore, in step S2, the multi-feature fusion wave recognition algorithm is an algorithm that integrates wavelet analysis, attribution analysis, and attention mechanism, specifically including the following steps:

[0019] S21: The db4 wavelet basis is used to perform 3-5 level wavelet decomposition on the voltage time series monitoring data of the distribution network, separating the steady-state low-frequency component and the fluctuating high-frequency component of the voltage signal. The fluctuating component is reconstructed through inverse wavelet transform, and the peak value, fluctuating rate time domain characteristics and frequency peak frequency domain characteristics of the fluctuating component are extracted.

[0020] S22: Attribution analysis was conducted using variance decomposition. Correlation analysis was performed between photovoltaic power output fluctuation data and agricultural load fluctuation data and voltage fluctuation reconstruction components, respectively. The contribution weights of the two types of fluctuation factors to voltage fluctuation were quantified, and the dominant and secondary fluctuation factors were identified.

[0021] S23: An attention mechanism is introduced to construct a feature weight allocation neural network. The time-domain and frequency-domain features extracted by wavelet analysis and the contribution weights obtained by attribution analysis are used as inputs to strengthen the feature weights corresponding to the dominant factors and weaken the feature weights corresponding to the secondary factors.

[0022] S24: Combine environmental monitoring data and equipment operating status data to set judgment thresholds, filter out false fluctuations caused by equipment momentary failures and abnormal data acquisition, and avoid misjudging real voltage fluctuations caused by extreme weather.

[0023] Furthermore, the voltage fluctuation types mentioned in step S2 include voltage surges caused by sudden increases in photovoltaic output, voltage drops caused by sudden decreases in photovoltaic output, and voltage fluctuations caused by sudden changes in agricultural load; the dominant factor of the fluctuation is either photovoltaic output fluctuation or agricultural load fluctuation; the voltage fluctuation level is divided into three levels—slight, moderate, and severe—based on the voltage deviation range and the degree of impact on agricultural production. The specific judgment criteria are as follows: Slight level: voltage deviation does not exceed ±5% (inclusive) of the rated voltage, with no risk of abnormal shutdown of agricultural production equipment; Moderate level: voltage deviation exceeds ±5% of the rated voltage but does not exceed ±10% (inclusive), which may lead to abnormal operation of sensitive agricultural equipment, with no risk of production interruption; Severe level: voltage deviation exceeds ±10% of the rated voltage, posing a risk of shutdown of agricultural production equipment and interruption of irrigation and temperature control systems.

[0024] Furthermore, step S3 specifically includes the following steps:

[0025] S31: Construct a multi-objective optimization distributed energy storage and photovoltaic coordinated damping control model based on LSTM network. The model takes maximizing photovoltaic absorption rate, minimizing distribution network node voltage deviation, and minimizing energy storage charge and discharge cycle loss as the three core optimization objectives.

[0026] S32: Based on the voltage fluctuation dominant factors identified in step S2, dynamically allocate the weight coefficients of each core optimization objective: when the dominant fluctuation factor is photovoltaic output fluctuation, increase the target weights corresponding to minimizing the voltage deviation of the distribution network nodes and maximizing the photovoltaic absorption rate; when the dominant fluctuation factor is agricultural load fluctuation, increase the target weights corresponding to minimizing the voltage deviation of the distribution network nodes and minimizing the energy storage charging and discharging cycle loss.

[0027] S33: Based on the rated capacity of distribution network equipment, real-time operating parameters and agricultural distribution network operation and maintenance safety standards, set hard constraints. The constraints include the allowable deviation of distribution network node voltage, the safe operating range of distributed energy storage SOC, the upper and lower limits of photovoltaic inverter output, the rated capacity of reactive power compensation equipment, and the power supply reliability of key agricultural production loads.

[0028] S34: Using the set hard constraints as the solution boundary and the dynamically weighted multi-objective optimization function as the solution objective, the improved particle swarm optimization algorithm is used to complete the fast optimization solution of the model and output the optimal objective value and control parameter boundary of the cooperative damping control.

[0029] Furthermore, the energy storage SOC hierarchical control strategy described in step S4 is divided into three control levels based on the real-time value of the energy storage SOC, with each level corresponding to different voltage fluctuation levels and power regulation parameters:

[0030] (1) Slight fluctuation control level: corresponding to the slight voltage fluctuation level, the real-time value of energy storage SOC is in the range of 50% (excluding) to 80% (excluding), the grid-connected power command value of photovoltaic inverter is adjusted first, and the adjustment range is ±1% to ±3% of the rated power of photovoltaic inverter. At the same time, the distributed energy storage is controlled to perform charging and discharging operations not exceeding 30% of the rated power of energy storage, and no start command is sent to the reactive power compensation equipment.

[0031] (2) Moderate fluctuation control level: corresponding to the moderate voltage fluctuation level, the real-time value of energy storage SOC is in the range of 30% (excluding) to 50% (including). According to the voltage fluctuation type, the energy storage charging and discharging direction command is generated. When the voltage rises sharply, the charging command is generated and when the voltage drops sharply, the discharging command is generated. The energy storage charging and discharging power is controlled at 30%-70% of the rated power of energy storage. At the same time, the start command is sent to the reactive power compensation equipment, and the output capacity of the reactive power compensation equipment is adjusted to 20%-50% of its rated capacity.

[0032] (3) Severe fluctuation control level: corresponding to severe voltage fluctuation level, the real-time value of energy storage SOC is in the range of 20% (excluding) to 30% (including), the distributed energy storage is controlled to perform charging and discharging operations not exceeding 100% of the rated power of the energy storage, and at the same time, a power limiting instruction is sent to the photovoltaic inverter to limit the photovoltaic output to 60%-80% of the rated power of the photovoltaic inverter, and a full start instruction is sent to the reactive power compensation equipment to adjust the operating period of agricultural non-critical loads, with a delay start time of 15min-30min.

[0033] Furthermore, the energy storage SOC hierarchical control strategy also sets up an emergency control level for SOC exceeding limits: when the real-time value of energy storage SOC is greater than or equal to 80%, the energy storage charging circuit is blocked, and only the energy storage is allowed to perform discharge operations. The voltage fluctuation is responded to by adjusting the output of the photovoltaic inverter and the reactive power compensation equipment. When the real-time value of energy storage SOC is less than or equal to 20%, the energy storage discharge circuit is blocked, and only the energy storage is allowed to perform low-power charging operations not exceeding 10% of the rated power of the energy storage. The voltage fluctuation is smoothed by fully activating the reactive power compensation equipment, adjusting the photovoltaic output, and controlling the peak shifting of non-critical agricultural loads.

[0034] Further, in step S4, a fuzzy adaptive PID control algorithm is used to perform closed-loop regulation of the distributed energy storage charging and discharging power. The input parameters of the PID control algorithm are the voltage fluctuation amplitude, the real-time value of the energy storage SOC, and the photovoltaic output change rate. The charging and discharging power command value is calculated and output through PID parameter tuning. In the charging and discharging power regulation link, a slope limit value is set for the charging and discharging power command value to limit the power change rate to no more than 10% per second of the rated power of the energy storage, thereby suppressing power mutations. At the same time, based on the seasonal and time-series data of agricultural load fluctuations, start-stop timing commands for adjustable agricultural loads such as irrigation and greenhouse temperature control are generated to control the load operation time to avoid peak periods of photovoltaic output fluctuations and periods prone to voltage fluctuations.

[0035] Furthermore, the execution logic of step S5 includes:

[0036] S51: Real-time collection of actual values ​​of voltage at distribution network nodes, actual photovoltaic output, actual value of distributed energy storage operation, and voltage fluctuation mitigation effect value to construct a real-time operation dataset;

[0037] S52: Calculate the difference between the real-time running dataset and the cooperative control target value to obtain the voltage control deviation value, the model prediction deviation value, and the fluctuation identification deviation value;

[0038] S53: When the voltage control deviation exceeds the preset threshold, based on the historical voltage fluctuation dataset of the distribution network, short-term weather forecast parameters and agricultural production cycle time series data, the LSTM network weight value, PID control parameter value and collaborative control strategy coefficient value of the collaborative smoothing control model are corrected through parameter iteration calculation, and the calculation model of the voltage fluctuation level judgment threshold is dynamically adjusted at the same time.

[0039] S54: Substitute the corrected model parameter values ​​and control strategy coefficient values ​​into the collaborative suppression control model to regenerate the voltage fluctuation suppression control command.

[0040] Furthermore, the improved particle swarm optimization algorithm is a particle swarm optimization algorithm that combines adaptive weighting and dynamic mutation. During the algorithm iteration process, the initial value of the inertia weight is set to 0.9, and it adaptively and linearly decreases to 0.4 with the number of iterations. At the same time, a dynamic Gaussian mutation operation is performed on the global optimal particle. When the algorithm fails to update the global optimal solution for three consecutive iterations, the mutation operation is triggered to expand the search range and avoid premature convergence of the algorithm. The termination condition of the algorithm is that the number of iterations reaches a preset upper limit, or the convergence error of the multi-objective optimization function is less than a preset threshold.

[0041] This method addresses the challenges of fragmented and spatiotemporally asynchronous data perception in agricultural power distribution networks by synchronously collecting and standardizing multi-dimensional operational data. Through a multi-feature recognition algorithm integrating wavelet analysis, attribution analysis, and attention mechanisms, it achieves signal decoupling, causal tracing, and accurate level determination of voltage fluctuations, fundamentally resolving the pain points of low fluctuation recognition accuracy, high misjudgment rate, and lack of targeted regulation in traditional solutions. Based on dynamically adjusting the multi-objective optimization weights according to the dominant fluctuation factors, it transforms the three mutually constraining optimization objectives of photovoltaic absorption, voltage stability, and energy storage lifespan into a dynamically weighted optimization problem adaptable to specific scenarios. Within the hard constraints of power distribution network equipment safety and reliable power supply for agricultural production, it rapidly solves the globally optimal control strategy using an improved particle swarm optimization algorithm, overcoming the limitations of poor adaptability and difficulty in coordinating multiple objectives in traditional fixed-weight schemes. Finally, it employs a hierarchical approach linking energy storage SOC with voltage fluctuation levels. With regulation as the core, a collaborative execution system of source-grid-load-storage is constructed, consisting of "photovoltaic output regulation - energy storage charging and discharging response - reactive power compensation equipment linkage - agricultural controllable load timing optimization". This system achieves graded and precise responses under different fluctuation scenarios, ensuring both the speed and effectiveness of voltage stabilization and avoiding overcharging, over-discharging, and frequent operation of energy storage leading to lifespan loss. Ultimately, through feedback on the deviation between real-time operating data and control objectives, the system adaptively corrects control model parameters, strategy coefficients, and fluctuation judgment thresholds, enabling the entire control system to have self-learning and self-optimization capabilities. It can adapt to the time-varying characteristics of agricultural power distribution networks caused by seasonal changes, weather variations, and production cycle adjustments, forming a complete closed loop of "perception-decision-execution-feedback-optimization". Ultimately, it achieves multi-objective synergistic optimization of long-term stable and controllable voltage in agricultural power distribution networks, high proportion of local photovoltaic consumption, economical and efficient operation of energy storage, and safe and reliable power supply for agricultural production.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] 1. This application adopts a multi-feature recognition algorithm that integrates wavelet analysis, attribution analysis and attention mechanism. It can accurately extract the time domain and frequency domain features of voltage fluctuations, quantify the contribution of photovoltaic and agricultural loads to voltage fluctuations, and identify the dominant factors. By strengthening core features and filtering false fluctuations through attention mechanism, it can effectively avoid misjudgment caused by abnormal data collection and extreme weather. At the same time, it combines the distribution of photovoltaics and the operation law of load in agricultural distribution network to complete the level determination, which is fully adapted to the operation characteristics of agricultural distribution network.

[0044] 2. This application constructs a multi-objective optimization control model based on LSTM network, which can dynamically adjust the weights of the three core objectives according to the dominant fluctuation factors. When photovoltaic fluctuations are dominant, priority is given to ensuring absorption and voltage stability. When load fluctuations are dominant, priority is given to ensuring voltage stability and energy storage life. This solves the pain point of poor adaptability of fixed weight schemes and achieves multi-objective synergistic optimization.

[0045] 3. This application adopts a hierarchical control strategy that links energy storage SOC with voltage fluctuation levels, and matches differentiated photovoltaic, energy storage and reactive power compensation collaborative control schemes. At the same time, it sets up an emergency level for SOC over-limit, which not only achieves precise voltage smoothing under different fluctuation scenarios, but also effectively avoids overcharging and over-discharging of energy storage, reduces cycle losses, extends the service life of energy storage, and greatly improves the control reliability under extreme scenarios.

[0046] 4. This application constructs a closed-loop control system for the entire process, which can adaptively adjust model parameters, control strategies and fluctuation judgment thresholds according to the actual control effect. It can still maintain excellent voltage control effect when the power distribution network conditions, meteorological conditions and agricultural production cycle change, and the control robustness is significantly better than the existing schemes.

[0047] 5. This application is designed to meet the power supply needs of agricultural production. Through three-level voltage control, it strictly manages voltage deviations, effectively avoiding the risk of unexpected shutdowns of agricultural production equipment. It reduces distribution network losses and maintenance costs, contributing to the development of clean energy in rural areas and the modernization of agriculture and rural communities. Attached Figure Description

[0048] Figure 1 This is a flowchart of a control method for synergistically mitigating voltage fluctuations in agricultural power distribution networks using distributed energy storage and renewable energy sources, as described in this invention. Detailed Implementation

[0049] The technical solutions of the present invention will now be described in detail with reference to the accompanying drawings.

[0050] like Figure 1 As shown, it includes the following steps:

[0051] S1: Collect photovoltaic operation parameters, distributed energy storage operation parameters, distribution network operation parameters, agricultural load operation parameters and environmental operation parameters, and integrate multiple types of parameters to establish a collaborative voltage fluctuation control database;

[0052] Further, step S1 specifically involves: collecting parameters by deploying monitoring terminals at photovoltaic grid-connected nodes, energy storage access nodes, key distribution network line nodes, concentrated agricultural load areas, and field meteorological monitoring points; the core parameters collected include photovoltaic output, energy storage SOC, distribution network node voltage, agricultural load power, light intensity, and ambient temperature.

[0053] Furthermore, in another embodiment of the invention;

[0054] High-precision synchronous phasor measurement units (PMUs), intelligent power monitoring instruments, and meteorological sensors are deployed at five core locations: photovoltaic grid-connected nodes, energy storage access nodes, key distribution network line nodes, agricultural load concentration areas, and field meteorological monitoring points. Using a 100Hz sampling frequency and a time synchronization error ≤1ms, multi-source operating parameters are synchronously collected. The core parameter collected is: photovoltaic output. Energy storage SOC, distribution network node voltage Agricultural load power Light intensity Ambient temperature .

[0055] To eliminate data noise and acquisition errors, the raw data is processed sequentially by outlier removal, missing value completion, and unit standardization. The corresponding algorithms and formulas are as follows:

[0056] Outlier removal (3σ criterion): for univariate time series data Calculate the mean with standard deviation Eliminate those that meet the requirements The formula for outliers is:

[0057] ;

[0058] In the formula: The number of sampling points within a single control cycle. for The original sampled value at that moment.

[0059] Missing value completion: Linear interpolation is used to complete the blank data points after outliers are removed, ensuring the continuity of data sequence.

[0060] Dimensional standardization (Min-Max normalization): The preprocessed data is standardized from 0 to 1 to eliminate dimensional differences between different parameters. The formula is:

[0061] ;

[0062] In the formula: For standardized data, , This refers to the maximum and minimum values ​​of the parameter based on historical data.

[0063] A three-tiered collaborative mitigation control database is built using the time-series database InfluxDB, with data stored according to its attributes and purpose:

[0064] Real-time operation layer: Stores standardized real-time acquisition data for the current control cycle (1s) to provide real-time monitoring data for step 2;

[0065] Historical data layer: Stores data on distribution network operation, voltage fluctuations, and equipment maintenance over the past 12 months, providing historical support for the attribution analysis in step 2 and the closed-loop optimization in step 5;

[0066] Predictive data layer: Stores short-term weather forecasts, agricultural load predictions, and photovoltaic output forecasts for the next 72 hours, providing a predictive basis for model optimization in step 3 and strategy generation in step 4.

[0067] S2: Based on the monitoring data of the collaborative voltage fluctuation control database, a multi-feature fusion fluctuation identification algorithm is used to identify the type, amplitude and dominant factors of voltage fluctuation in the distribution network, and the voltage fluctuation level is determined by combining the photovoltaic distribution characteristics and agricultural load operation patterns in the area where the agricultural distribution network is located.

[0068] Furthermore, in step S2, the multi-feature fusion wave recognition algorithm is an algorithm that integrates wavelet analysis, attribution analysis, and attention mechanism, specifically including the following steps:

[0069] S21: The db4 wavelet basis is used to perform 3-5 level wavelet decomposition on the voltage time series monitoring data of the distribution network, separating the steady-state low-frequency component and the fluctuating high-frequency component of the voltage signal. The fluctuating component is reconstructed through inverse wavelet transform, and the peak value, fluctuating rate time domain characteristics and frequency peak frequency domain characteristics of the fluctuating component are extracted.

[0070] S22: Attribution analysis was conducted using variance decomposition. Correlation analysis was performed between photovoltaic power output fluctuation data and agricultural load fluctuation data and voltage fluctuation reconstruction components, respectively. The contribution weights of the two types of fluctuation factors to voltage fluctuation were quantified, and the dominant and secondary fluctuation factors were identified.

[0071] S23: An attention mechanism is introduced to construct a feature weight allocation neural network. The time-domain and frequency-domain features extracted by wavelet analysis and the contribution weights obtained by attribution analysis are used as inputs to strengthen the feature weights corresponding to the dominant factors and weaken the feature weights corresponding to the secondary factors.

[0072] S24: Combine environmental monitoring data and equipment operating status data to set judgment thresholds, filter out false fluctuations caused by equipment momentary failures and abnormal data acquisition, and avoid misjudging real voltage fluctuations caused by extreme weather.

[0073] Furthermore, the voltage fluctuation types mentioned in step S2 include voltage surges caused by sudden increases in photovoltaic output, voltage drops caused by sudden decreases in photovoltaic output, and voltage fluctuations caused by sudden changes in agricultural load; the dominant factor of the fluctuation is either photovoltaic output fluctuation or agricultural load fluctuation; the voltage fluctuation level is divided into three levels—slight, moderate, and severe—based on the voltage deviation range and the degree of impact on agricultural production. The specific judgment criteria are as follows: Slight level: voltage deviation does not exceed ±5% (inclusive) of the rated voltage, with no risk of abnormal shutdown of agricultural production equipment; Moderate level: voltage deviation exceeds ±5% of the rated voltage but does not exceed ±10% (inclusive), which may lead to abnormal operation of sensitive agricultural equipment, with no risk of production interruption; Severe level: voltage deviation exceeds ±10% of the rated voltage, posing a risk of shutdown of agricultural production equipment and interruption of irrigation and temperature control systems.

[0074] Furthermore, in another embodiment of the invention;

[0075] Input for this step: the collaborative damping control database output from step 1; Output for this step: voltage fluctuation type, dominant factors, and accurately determined fluctuation level, providing core judgment basis for dynamic weight allocation of the control model in step 3 and generation of differentiated strategies in step 4.

[0076] This step employs a multi-feature fluctuation recognition algorithm that integrates wavelet analysis, attribution analysis, and attention mechanism. It is executed in four sub-steps, which are connected to form a complete chain of "signal decomposition - factor tracing - feature enhancement - false filtering - level determination".

[0077] 2.1 Wavelet Decomposition and Feature Extraction of Voltage Signals

[0078] Using the db4 wavelet basis to analyze the voltage time series data of distribution network nodes Perform 3-5 level wavelet decomposition to separate steady-state low-frequency components from wavelet high-frequency components. Reconstruct the wavelet components using inverse wavelet transform and extract core features. The formulas and steps are as follows:

[0079] Wavelet decomposition: Decomposes a voltage signal into steady-state and fluctuating components, as shown in the formula:

[0080] ;

[0081] In the formula: The number of decomposition layers, For the first Layered steady-state low-frequency components, For the first High-frequency components of layer fluctuations.

[0082] Voltage fluctuation component reconstruction: Reconstructing voltage fluctuation components using inverse wavelet transform .

[0083] Feature extraction: Extracting temporal features (peak value) from the reconstructed wave components. fluctuation rate ) and frequency domain characteristics (frequency peak) This forms the initial feature set.

[0084] 2.2 Attribution Analysis of Fluctuation Factors and Determination of Dominant Factors

[0085] A vector autoregression (VAR) model of voltage fluctuations and influencing factors was constructed using the variance decomposition method. This model quantifies the contributions of photovoltaic power output fluctuations and agricultural load fluctuations to voltage fluctuations, identifying the dominant factors. The formula is as follows:

[0086] ;

[0087] ;

[0088] In the formula: , Weighting of contributions to photovoltaic and agricultural load fluctuations; , These are the impulse response coefficients of the VAR model; , The variance of the impact of the corresponding factors; The mean square error for predicting voltage fluctuation components; The prediction step size is set to 10.

[0089] Judgment rule: If The analysis indicates that fluctuations in photovoltaic power output are the dominant factor; conversely, fluctuations in agricultural load are the dominant factor.

[0090] 2.3 Enhancement of Feature Weights in Attention Mechanism

[0091] A single-layer attention feature weight allocation neural network is constructed, using the time-frequency features extracted in step 2.1 and the contribution weights obtained in step 2.2 as inputs. The network strengthens the feature weights corresponding to the dominant factors and weakens the weights of secondary factors to improve recognition accuracy. The weight calculation and feature weighting formulas are as follows:

[0092] ;

[0093] In the formula: For the first Attention weight coefficients for each feature; , This is the network weight matrix. For bias terms; The original feature vector, This is to optimize the feature vector after weighting.

[0094] 2.4 False Fluctuation Filtering and Voltage Fluctuation Level Determination

[0095] False fluctuation filtering: By combining environmental monitoring data (light intensity, temperature) with equipment operating status data, a threshold for judging fluctuation signals is set to filter out false fluctuations caused by momentary equipment failures and abnormal data acquisition, thus avoiding misjudgment of real fluctuations caused by extreme weather.

[0096] Fluctuation type identification: based on weighted feature vector The voltage fluctuation types were identified as three categories: voltage surge caused by a sudden increase in photovoltaic output, voltage drop caused by a sudden decrease in photovoltaic output, and voltage fluctuation caused by a sudden change in agricultural load.

[0097] Voltage fluctuation level determination: Based on the voltage deviation range and the degree of impact on agricultural production, a trigonometric membership function is used to determine three levels of fluctuation: slight, moderate, and severe. Voltage deviation rate... ( (Rated voltage of the distribution network), the judgment criteria and membership function strictly match the claims, and the core judgment rule is:

[0098] Mild level: (Including 5%), there is no risk of abnormal shutdown of agricultural equipment;

[0099] Moderate level: (Including 10%), may cause abnormal operation of sensitive equipment;

[0100] Severity level: There is a risk of equipment downtime and production system interruption.

[0101] S3: Construct a distributed energy storage and photovoltaic coordinated damping control model. Based on the identified dominant factors of voltage fluctuation, dynamically adjust the control target weights of maximizing photovoltaic absorption rate, minimizing distribution network voltage deviation, and minimizing energy storage charging and discharging losses. Combine the rated capacity of distribution network equipment, real-time operating parameters, and agricultural distribution network operation and maintenance safety standards to set voltage stability constraints and coordinated control objectives.

[0102] Furthermore, step S3 specifically includes the following steps:

[0103] S31: Construct a multi-objective optimization distributed energy storage and photovoltaic coordinated damping control model based on LSTM network. The model takes maximizing photovoltaic absorption rate, minimizing distribution network node voltage deviation, and minimizing energy storage charge and discharge cycle loss as the three core optimization objectives.

[0104] S32: Based on the voltage fluctuation dominant factors identified in step S2, dynamically allocate the weight coefficients of each core optimization objective: when the dominant fluctuation factor is photovoltaic output fluctuation, increase the target weights corresponding to minimizing the voltage deviation of the distribution network nodes and maximizing the photovoltaic absorption rate; when the dominant fluctuation factor is agricultural load fluctuation, increase the target weights corresponding to minimizing the voltage deviation of the distribution network nodes and minimizing the energy storage charging and discharging cycle loss.

[0105] S33: Based on the rated capacity of distribution network equipment, real-time operating parameters and agricultural distribution network operation and maintenance safety standards, set hard constraints. The constraints include the allowable deviation of distribution network node voltage, the safe operating range of distributed energy storage SOC, the upper and lower limits of photovoltaic inverter output, the rated capacity of reactive power compensation equipment, and the power supply reliability of key agricultural production loads.

[0106] S34: Using the set hard constraints as the solution boundary and the dynamically weighted multi-objective optimization function as the solution objective, the improved particle swarm optimization algorithm is used to complete the fast optimization solution of the model and output the optimal objective value and control parameter boundary of the cooperative damping control.

[0107] Furthermore, in another embodiment of the invention;

[0108] Inputs for this step: the dominant factors and volatility levels output from step 2; Outputs for this step: the dynamically weighted multi-objective optimization model and the optimal control parameter boundary, providing model and parameter support for the generation of the differentiated collaborative mitigation strategy in step 4.

[0109] This step involves building a multi-objective optimization control model based on an LSTM network. It is executed in four sub-steps, forming a complete modeling chain of "model building - weight allocation - constraint setting - algorithm optimization", and finally outputting the globally optimal control parameters.

[0110] 3.1 Construction of a Multi-Objective Optimization Model Based on LSTM Network

[0111] An integrated LSTM collaborative smoothing control model consisting of a prediction layer and an optimization layer is constructed. The prediction layer is a two-layer LSTM network that takes historical time-series data as input and outputs predicted values ​​for photovoltaic power output, agricultural load, and node voltage for the next five control cycles. The optimization layer has three core optimization objectives: maximizing photovoltaic absorption rate, minimizing distribution network voltage deviation, and minimizing energy storage charge-discharge cycle losses. The objective function formula is as follows:

[0112] Maximizing photovoltaic power absorption rate:

[0113] ;

[0114] In the formula: To optimize the number of control cycles, for Real-time grid-connected photovoltaic power output for At any given moment, the photovoltaic system reaches its theoretical maximum output.

[0115] Minimize voltage deviation at distribution network nodes:

[0116] ;

[0117] In the formula: This represents the total number of key monitoring nodes in the power distribution network. for Time of the first The actual voltage value of each node.

[0118] Minimize energy storage charge-discharge cycle losses:

[0119] ;

[0120] In the formula: for Real-time energy storage charging and discharging power (charging is positive, discharging is negative); , To improve the charging and discharging efficiency of energy storage; This is the energy storage cycle life loss coefficient; for Real-time energy storage SOC value; This refers to the rated SOC range for energy storage.

[0121] 3.2 Dynamic allocation of target weights based on the dominant factors of fluctuation

[0122] Based on the dominant fluctuation factors identified in step 2, the weight coefficients of the three optimization objectives are dynamically adjusted. , , (satisfy , This forms a dynamically weighted overall objective function. The weight allocation rules and the formula for the overall objective function are as follows:

[0123] Weighting rules:

[0124] Fluctuations in photovoltaic output are a leading factor: improving voltage stability and increasing the weight of photovoltaic power consumption. , , ;

[0125] Agricultural load fluctuations are a leading factor: Improving voltage stability and controlling energy storage losses are given weight. , , .

[0126] Dynamically weighted overall objective function:

[0127] ;

[0128] 3.3 Setting Hard Constraints for Distribution Networks

[0129] Based on the rated capacity of distribution network equipment, real-time operating parameters, and agricultural distribution network operation and maintenance safety standards, six hard constraints are set as the boundaries for model solution. All constraints are quantified by mathematical formulas, specifically:

[0130] Node voltage tolerance constraints: ;

[0131] Energy storage SOC safety operation constraints: ;

[0132] Energy storage charging and discharging power constraints: ;

[0133] Photovoltaic inverter output constraints: ;

[0134] Reactive power compensation equipment capacity constraints: ;

[0135] Reliability constraints for power supply to critical agricultural loads: .

[0136] In the formula: , , These are the rated capacities of energy storage, photovoltaic inverters, and reactive power compensation equipment, respectively. Real-time power for critical agricultural loads. Minimum guaranteed power for critical loads.

[0137] 3.4 Improved Particle Swarm Optimization (IPSO) Algorithm for Model Optimization and Solution

[0138] Using the hard constraints set in step 3.3 as the solution boundary and the dynamically weighted total objective function in step 3.2 as the solution objective, an improved particle swarm optimization algorithm with adaptive weights and dynamic Gaussian mutation is used to quickly find the optimal model, outputting the optimal objective value and control parameter boundary for collaborative suppression control. The core formulas and rules of the algorithm are as follows:

[0139] Iterative formula for particle velocity and position:

[0140] ;

[0141] ;

[0142] In the formula: This represents the number of iterations. , For the first The first particle The speed and position of the next iteration; Inertial weights; For learning factors; , for Random numbers; For each individual particle, the optimal solution is provided. This is the globally optimal solution for the population.

[0143] Adaptive inertia weight update: The initial value of 0.9 decreases linearly to 0.4 with the number of iterations, as shown in the formula:

[0144] ;

[0145] Dynamic Gaussian Mutation: When the algorithm has not updated the global optimal solution for 3 consecutive generations. When this happens, a mutation operation is triggered to expand the search range and avoid premature convergence. The formula is as follows:

[0146] ;

[0147] In the formula: With a mean of 0 and a variance of 0, Gaussian distributed random numbers.

[0148] Algorithm termination condition: The number of iterations reaches the preset upper limit. or convergence error of multi-objective optimization function Stop the iteration and output the optimal solution.

[0149] S4: Based on the determined voltage fluctuation level and the output results of the collaborative smoothing control model, a differentiated collaborative smoothing control strategy is generated. This strategy uses a layered SOC control strategy to dynamically adjust the charging and discharging power of distributed energy storage and photovoltaic output, and coordinates with the reactive power compensation equipment in the distribution network to smooth voltage fluctuations. Further, the layered SOC control strategy described in step S4 is divided into three control levels based on the real-time SOC value, with each level corresponding to different voltage fluctuation levels and power adjustment parameters.

[0150] (1) Slight fluctuation control level: corresponding to the slight voltage fluctuation level, the real-time value of energy storage SOC is in the range of 50% (excluding) to 80% (excluding), the grid-connected power command value of photovoltaic inverter is adjusted first, and the adjustment range is ±1% to ±3% of the rated power of photovoltaic inverter. At the same time, the distributed energy storage is controlled to perform charging and discharging operations not exceeding 30% of the rated power of energy storage, and no start command is sent to the reactive power compensation equipment.

[0151] (2) Moderate fluctuation control level: corresponding to the moderate voltage fluctuation level, the real-time value of energy storage SOC is in the range of 30% (excluding) to 50% (including). According to the voltage fluctuation type, the energy storage charging and discharging direction command is generated. When the voltage rises sharply, the charging command is generated and when the voltage drops sharply, the discharging command is generated. The energy storage charging and discharging power is controlled at 30%-70% of the rated power of energy storage. At the same time, the start command is sent to the reactive power compensation equipment, and the output capacity of the reactive power compensation equipment is adjusted to 20%-50% of its rated capacity.

[0152] (3) Severe fluctuation control level: corresponding to severe voltage fluctuation level, the real-time value of energy storage SOC is in the range of 20% (excluding) to 30% (including), the distributed energy storage is controlled to perform charging and discharging operations not exceeding 100% of the rated power of the energy storage, and at the same time, a power limiting instruction is sent to the photovoltaic inverter to limit the photovoltaic output to 60%-80% of the rated power of the photovoltaic inverter, and a full start instruction is sent to the reactive power compensation equipment to adjust the operating period of agricultural non-critical loads, with a delay start time of 15min-30min.

[0153] Furthermore, the energy storage SOC hierarchical control strategy also sets up an emergency control level for SOC exceeding limits: when the real-time value of energy storage SOC is greater than or equal to 80%, the energy storage charging circuit is blocked, and only the energy storage is allowed to perform discharge operations. The voltage fluctuation is responded to by adjusting the output of the photovoltaic inverter and the reactive power compensation equipment. When the real-time value of energy storage SOC is less than or equal to 20%, the energy storage discharge circuit is blocked, and only the energy storage is allowed to perform low-power charging operations not exceeding 10% of the rated power of the energy storage. The voltage fluctuation is smoothed by fully activating the reactive power compensation equipment, adjusting the photovoltaic output, and controlling the peak shifting of non-critical agricultural loads.

[0154] Further, in step S4, a fuzzy adaptive PID control algorithm is used to perform closed-loop regulation of the distributed energy storage charging and discharging power. The input parameters of the PID control algorithm are the voltage fluctuation amplitude, the real-time value of the energy storage SOC, and the photovoltaic output change rate. The charging and discharging power command value is calculated and output through PID parameter tuning. In the charging and discharging power regulation link, a slope limit value is set for the charging and discharging power command value to limit the power change rate to no more than 10% per second of the rated power of the energy storage, thereby suppressing power mutations. At the same time, based on the seasonal and time-series data of agricultural load fluctuations, start-stop timing commands for adjustable agricultural loads such as irrigation and greenhouse temperature control are generated to control the load operation time to avoid peak periods of photovoltaic output fluctuations and periods prone to voltage fluctuations.

[0155] Furthermore, in another embodiment of the invention;

[0156] The inputs for this step are: the fluctuation level output from step 2 and the optimal control parameter boundary output from step 3; the outputs for this step are: the coordinated smoothing control commands that are implemented and directly act on the photovoltaic inverter, energy storage converter, reactive power compensation equipment, and agricultural load control terminal to achieve voltage fluctuation smoothing; this step is the core execution link, which is executed according to the logic of "layered regulation - power closed loop - load optimization", and the linkage of each link forms a source-grid-load-storage coordinated control system.

[0157] 4.1 Implementation of the Layered Regulation Strategy for Energy Storage SOC

[0158] Based on the real-time SOC value of energy storage and the voltage fluctuation level determined in step 2, a three-level conventional control + two-level emergency control SOC hierarchical control strategy is constructed. Each level is precisely matched with the fluctuation level, and basic control commands are output for photovoltaic, energy storage, and reactive power compensation equipment. The core control rules and mathematical expressions are as follows (all power parameters are based on rated power):

[0159] 4.1.1 Three-level conventional control hierarchy

[0160] Slight fluctuation control level (SOC∈50%~80%, slight level):

[0161] ;

[0162] Moderate fluctuation control level (SOC∈30%~50%, moderate level):

[0163] ;

[0164] In the formula: For sign function, voltage surge ( When charging (+), the voltage drops sharply ( Discharge (-) when ).

[0165] Severe fluctuation control level (SOC∈20%~30%, severe level):

[0166] ;

[0167] In the formula: Delay the start-up time for non-critical agricultural loads.

[0168] 4.1.2 Two-level emergency control system for SOC exceeding limits

[0169] To prevent overcharging and over-discharging of energy storage, an emergency control level is set up as a supplement to the normal level. Upon triggering, the corresponding charging and discharging circuit is locked.

[0170] SOC≥80%: Lock out charging circuit Prioritize adjusting photovoltaic power output and reactive power compensation equipment;

[0171] SOC≤20%: Close the discharge circuit Priority should be given to balancing voltage through full operation of reactive power compensation, photovoltaic adjustments, and peak shifting of non-critical loads.

[0172] 4.2 Fuzzy Adaptive PID Closed-Loop Regulation of Energy Storage Charging and Discharging Power

[0173] Based on the energy storage charging and discharging power output in step 4.1, a fuzzy adaptive PID control algorithm is used for closed-loop regulation to suppress power surges and improve control accuracy. Simultaneously, a power change rate slope limit is set. The core formulas and rules are as follows:

[0174] PID incremental control formula:

[0175] ;

[0176] In the formula: for Power increment at any moment; , , For the PID proportional, integral, and derivative coefficients; This represents the power control deviation (reference value - actual value).

[0177] Fuzzy adaptive parameter tuning: based on voltage fluctuation amplitude Energy storage SOC, photovoltaic output change rate For 3 inputs, , , The correction value is used as the output to construct a 3-input 3-output fuzzy controller, and the PID parameters are tuned in real time using the Mamdani fuzzy inference rule.

[0178] Power change rate slope limit: Limit power change rate ≤ 10% / second, the formula is:

[0179] ;

[0180] In the formula: It is a saturation function, and the boundary value is taken when the limit value is exceeded.

[0181] 4.3 Temporal Optimization Control of Adjustable Agricultural Loads

[0182] Based on the seasonal and temporal fluctuations in agricultural load, start-stop timing commands for adjustable loads such as irrigation and greenhouse temperature control are generated. The control load operation periods avoid peak periods of photovoltaic power output fluctuations and periods prone to voltage fluctuations. The optimization objective is:

[0183] ;

[0184] In the formula: Real-time power for adjustable load; Time-of-use pricing; , The weighting coefficient is used; the constraints are that the total daily electricity consumption of the load remains unchanged and the number of daily start-ups and shutdowns is ≤3 times, to ensure that agricultural production is not affected.

[0185] S5: Monitor the voltage status of distribution network nodes in real time, compare the monitoring results with the cooperative control target, and combine distribution network operation and maintenance data, historical fluctuation data and predictive data to provide feedback to adjust the parameters and cooperative smoothing control strategy of the cooperative smoothing control model and correct the voltage fluctuation identification deviation.

[0186] Furthermore, the execution logic of step S5 includes:

[0187] S51: Real-time collection of actual values ​​of voltage at distribution network nodes, actual photovoltaic output, actual value of distributed energy storage operation, and voltage fluctuation mitigation effect value to construct a real-time operation dataset;

[0188] S52: Calculate the difference between the real-time running dataset and the cooperative control target value to obtain the voltage control deviation value, the model prediction deviation value, and the fluctuation identification deviation value;

[0189] S53: When the voltage control deviation exceeds the preset threshold, based on the historical voltage fluctuation dataset of the distribution network, short-term weather forecast parameters and agricultural production cycle time series data, the LSTM network weight value, PID control parameter value and collaborative control strategy coefficient value of the collaborative smoothing control model are corrected through parameter iteration calculation, and the calculation model of the voltage fluctuation level judgment threshold is dynamically adjusted at the same time.

[0190] S54: Substitute the corrected model parameter values ​​and control strategy coefficient values ​​into the collaborative suppression control model to regenerate the voltage fluctuation suppression control command.

[0191] Furthermore, the improved particle swarm optimization algorithm is a particle swarm optimization algorithm that combines adaptive weighting and dynamic mutation. During the algorithm iteration process, the initial value of the inertia weight is set to 0.9, and it adaptively and linearly decreases to 0.4 with the number of iterations. At the same time, a dynamic Gaussian mutation operation is performed on the global optimal particle. When the algorithm fails to update the global optimal solution for three consecutive iterations, the mutation operation is triggered to expand the search range and avoid premature convergence of the algorithm. The termination condition of the algorithm is that the number of iterations reaches a preset upper limit, or the convergence error of the multi-objective optimization function is less than a preset threshold.

[0192] Furthermore, in another embodiment of the invention;

[0193] Input for this step: Real-time operating data of the distribution network after the control strategy in step 4 is executed; Output for this step: Corrected control model parameters and control strategy coefficients, which are then substituted back into steps 3 and 4 to achieve closed-loop optimization of the entire process; This step is the system's self-learning and self-optimization stage, forming a closed loop of "monitoring-deviation-correction-update" to ensure that the control system adapts to changes in the operating conditions of the distribution network. It is executed in 4 sub-steps, with each step forming a precise connection.

[0194] 5.1 Real-time Dataset Construction

[0195] The actual values ​​of distribution network node voltage, photovoltaic output, distributed energy storage operation, and voltage fluctuation mitigation effect are collected in real time after the strategy is executed in step 4. The data are standardized according to the preprocessing standard in step 1 to construct a real-time operation dataset, providing a data foundation for deviation calculation.

[0196] 5.2 Multidimensional Control Deviation Calculation

[0197] The difference between the real-time running dataset from step 5.1 and the collaborative control target value set in step 3 is calculated to obtain three types of core deviation values, which serve as the basis for determining parameter correction:

[0198] Voltage control deviation value: ;

[0199] Model prediction bias: ;

[0200] Fluctuation identification deviation value: .

[0201] In the formula: the actual value is the real-time monitoring data, and the target value / predicted value / identified value is the output value of the preceding step.

[0202] 5.3 Adaptive Adjustment of Model and Policy Parameters

[0203] Set voltage control deviation threshold ,like This triggers the parameter correction process; based on historical fluctuation datasets of the distribution network, short-term weather forecast parameters, and agricultural production cycle time-series data, the correction of three types of core parameters is completed through parameter iterative calculation and the Adam adaptive gradient descent algorithm, while dynamically adjusting the voltage fluctuation level judgment threshold calculation model:

[0204] The weight values ​​of the LSTM network in the collaborative damping control model are modified to improve the model's prediction accuracy;

[0205] Correct the fuzzy adaptive PID control parameter values ​​in step 4 to improve the power closed-loop control effect;

[0206] The coefficient values ​​of the collaborative control strategy were revised to optimize the power regulation benchmark for SOC hierarchical control.

[0207] Loss function (aimed at minimizing voltage control deviation):

[0208]

[0209] 5.4 Control Command Update and Closed-Loop Iteration

[0210] The model parameter values ​​and control strategy coefficient values ​​corrected in step 5.3 are substituted back into the distributed energy storage and photovoltaic coordinated smoothing control model in step 3 to regenerate the optimized voltage fluctuation smoothing control command, which is then sent to each execution device to complete this closed-loop correction. At the same time, the corrected parameters and operating data are sent back to the coordinated smoothing control database in step 1 to update the historical data layer and enter the next control cycle, realizing continuous closed-loop iterative optimization of the entire process from S1 to S5.

[0211] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. These embodiments are merely descriptions of preferred embodiments and are not intended to limit the scope or concept of the invention. The specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. Such combinations, as long as they do not violate the spirit of the present invention, should also be considered as part of this disclosure. To avoid unnecessary repetition, the present invention will not further describe the various possible combinations.

[0212] This invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this invention and without departing from the design idea of ​​this invention, all modifications and improvements made by those skilled in the art to the technical solutions of this invention should fall within the protection scope of this invention. The technical content for which protection is sought in this invention has been fully described in the claims.

Claims

1. A control method for synergistically mitigating voltage fluctuations through distributed energy storage and renewable energy in agricultural power distribution networks, characterized in that, Includes the following steps: S1: Collect photovoltaic operation parameters, distributed energy storage operation parameters, distribution network operation parameters, agricultural load operation parameters and environmental operation parameters, and integrate multiple types of parameters to establish a collaborative voltage fluctuation control database; S2: Based on the monitoring data of the collaborative voltage fluctuation control database, a multi-feature fusion fluctuation identification algorithm is used to identify the type, amplitude and dominant factors of voltage fluctuation in the distribution network, and the voltage fluctuation level is determined by combining the photovoltaic distribution characteristics and agricultural load operation patterns in the area where the agricultural distribution network is located. S3: Construct a distributed energy storage and photovoltaic coordinated damping control model. Based on the identified dominant factors of voltage fluctuation, dynamically adjust the control target weights of maximizing photovoltaic absorption rate, minimizing distribution network voltage deviation, and minimizing energy storage charging and discharging losses. Combine the rated capacity of distribution network equipment, real-time operating parameters, and agricultural distribution network operation and maintenance safety standards to set voltage stability constraints and coordinated control objectives. S4: Based on the determined voltage fluctuation level and the output results of the collaborative suppression control model, a differentiated collaborative suppression control strategy is generated. The energy storage SOC hierarchical regulation strategy is adopted to dynamically adjust the charging and discharging power of distributed energy storage and photovoltaic output, and to link the reactive power compensation equipment of the distribution network to suppress the voltage fluctuation of the distribution network. S5: Monitor the voltage status of distribution network nodes in real time, compare the monitoring results with the cooperative control target, and combine distribution network operation and maintenance data, historical fluctuation data and predictive data to provide feedback to adjust the parameters and cooperative smoothing control strategy of the cooperative smoothing control model and correct the voltage fluctuation identification deviation.

2. The control method for synergistic voltage fluctuation mitigation through distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 1, characterized in that, Step S1 specifically involves deploying monitoring terminals at photovoltaic grid-connected nodes, energy storage access nodes, key distribution network line nodes, concentrated agricultural load areas, and field meteorological monitoring points to collect parameters. The core parameters collected include photovoltaic output, energy storage SOC, distribution network node voltage, agricultural load power, light intensity, and ambient temperature.

3. The control method for synergistic voltage fluctuation mitigation through distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 1, characterized in that, In step S2, the multi-feature fusion fluctuation recognition algorithm is an algorithm that integrates wavelet analysis, attribution analysis, and attention mechanism, specifically including the following steps: S21: The db4 wavelet basis is used to perform 3-5 level wavelet decomposition on the voltage time series monitoring data of the distribution network, separating the steady-state low-frequency component and the fluctuating high-frequency component of the voltage signal. The fluctuating component is reconstructed through inverse wavelet transform, and the peak value, fluctuating rate time domain characteristics and frequency peak frequency domain characteristics of the fluctuating component are extracted. S22: Attribution analysis was conducted using variance decomposition. Correlation analysis was performed between photovoltaic power output fluctuation data and agricultural load fluctuation data and voltage fluctuation reconstruction components, respectively. The contribution weights of the two types of fluctuation factors to voltage fluctuation were quantified, and the dominant and secondary fluctuation factors were identified. S23: An attention mechanism is introduced to construct a feature weight allocation neural network. The time-domain and frequency-domain features extracted by wavelet analysis and the contribution weights obtained by attribution analysis are used as inputs to strengthen the feature weights corresponding to the dominant factors and weaken the feature weights corresponding to the secondary factors. S24: Combine environmental monitoring data and equipment operating status data to set judgment thresholds, filter out false fluctuations caused by equipment momentary failures and abnormal data acquisition, and avoid misjudging real voltage fluctuations caused by extreme weather.

4. The control method for synergistic voltage fluctuation mitigation through distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 3, characterized in that, The voltage fluctuation types mentioned in step S2 include voltage surges caused by sudden increases in photovoltaic output, voltage drops caused by sudden decreases in photovoltaic output, and voltage fluctuations caused by sudden changes in agricultural load; the dominant factor of the fluctuation is either photovoltaic output fluctuation or agricultural load fluctuation; the voltage fluctuation level is divided into three levels—slight, moderate, and severe—based on the voltage deviation range and the degree of impact on agricultural production, with the specific judgment criteria as follows: Minor level: Voltage deviation does not exceed ±5% (inclusive) of rated voltage, with no risk of abnormal shutdown of agricultural production equipment; Moderate level: Voltage deviation exceeds ±5% but does not exceed ±10% of the rated voltage, which may cause abnormal operation of sensitive agricultural equipment, but there is no risk of production interruption; Severe level: Voltage deviation exceeds ±10% of rated voltage, posing a risk of agricultural production equipment shutdown and irrigation and temperature control system interruption.

5. The control method for synergistic voltage fluctuation mitigation through distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 1, characterized in that, Step S3 specifically includes the following steps: S31: Construct a multi-objective optimization distributed energy storage and photovoltaic coordinated damping control model based on LSTM network. The model takes maximizing photovoltaic absorption rate, minimizing distribution network node voltage deviation, and minimizing energy storage charge and discharge cycle loss as the three core optimization objectives. S32: Based on the voltage fluctuation dominant factors identified in step S2, dynamically allocate the weight coefficients of each core optimization objective: when the dominant fluctuation factor is photovoltaic output fluctuation, increase the target weights corresponding to minimizing the voltage deviation of the distribution network nodes and maximizing the photovoltaic absorption rate; when the dominant fluctuation factor is agricultural load fluctuation, increase the target weights corresponding to minimizing the voltage deviation of the distribution network nodes and minimizing the energy storage charging and discharging cycle loss. S33: Based on the rated capacity of distribution network equipment, real-time operating parameters and agricultural distribution network operation and maintenance safety standards, set hard constraints. The constraints include the allowable deviation of distribution network node voltage, the safe operating range of distributed energy storage SOC, the upper and lower limits of photovoltaic inverter output, the rated capacity of reactive power compensation equipment, and the power supply reliability of key agricultural production loads. S34: Using the set hard constraints as the solution boundary and the dynamically weighted multi-objective optimization function as the solution objective, the improved particle swarm optimization algorithm is used to complete the fast optimization solution of the model and output the optimal objective value and control parameter boundary of the cooperative damping control.

6. The control method for synergistic voltage fluctuation mitigation through distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 1, characterized in that, The energy storage SOC hierarchical control strategy described in step S4 is divided into three control levels based on the real-time value of the energy storage SOC. Each level corresponds to different voltage fluctuation levels and power regulation parameters. (1) Slight fluctuation control level: corresponding to the slight voltage fluctuation level, the real-time value of energy storage SOC is in the range of 50%-80%, the grid-connected power command value of photovoltaic inverter is adjusted first, the adjustment range is ±1%-±3% of the rated power of photovoltaic inverter, and at the same time, the distributed energy storage is controlled to perform charging and discharging operations not exceeding 30% of the rated power of energy storage, and no start command is sent to the reactive power compensation equipment. (2) Moderate fluctuation control level: corresponding to the moderate voltage fluctuation level, the real-time value of energy storage SOC is in the range of 30%-50%. Energy storage charging and discharging direction commands are generated according to the voltage fluctuation type. Charging commands are generated when the voltage rises sharply and discharging commands are generated when the voltage drops sharply. The energy storage charging and discharging power is controlled at 30%-70% of the rated energy storage power. At the same time, a start command is sent to the reactive power compensation equipment to adjust the output capacity of the reactive power compensation equipment to 20%-50% of its rated capacity. (3) Severe fluctuation control level: corresponding to severe voltage fluctuation level, the real-time value of energy storage SOC is in the range of 20%-30%, control the distributed energy storage to perform charging and discharging operations not exceeding 100% of the rated power of energy storage, and send a limiting command to the photovoltaic inverter to limit the photovoltaic output to 60%-80% of the rated power of the photovoltaic inverter. Send a full start command to the reactive power compensation equipment to adjust the operating period of agricultural non-critical loads, and delay the start time by 15min-30min.

7. The control method for synergistic voltage fluctuation mitigation through distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 6, characterized in that, In step S4, a fuzzy adaptive PID control algorithm is used to perform closed-loop regulation of the charging and discharging power of distributed energy storage. The input parameters of the PID control algorithm are the voltage fluctuation amplitude, the real-time value of energy storage SOC, and the photovoltaic output change rate. The charging and discharging power command value is calculated and output through PID parameter tuning. In the charging and discharging power regulation link, a slope limit value is set for the charging and discharging power command value to limit the power change rate to no more than 10% per second of the rated power of energy storage, thereby suppressing power mutations. At the same time, based on the seasonal and time-series data of agricultural load fluctuations, start-stop timing commands for adjustable agricultural loads such as irrigation and greenhouse temperature control are generated to control the load operation time to avoid peak periods of photovoltaic output fluctuations and periods prone to voltage fluctuations.

8. The control method for synergistic voltage fluctuation mitigation through distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 1, characterized in that, The execution logic of step S5 includes: S51: Real-time collection of actual values ​​of voltage at distribution network nodes, actual photovoltaic output, actual value of distributed energy storage operation, and voltage fluctuation mitigation effect value to construct a real-time operation dataset; S52: Calculate the difference between the real-time running dataset and the cooperative control target value to obtain the voltage control deviation value, the model prediction deviation value, and the fluctuation identification deviation value; S53: When the voltage control deviation exceeds the preset threshold, based on the historical voltage fluctuation dataset of the distribution network, short-term weather forecast parameters and agricultural production cycle time series data, the LSTM network weight value, PID control parameter value and collaborative control strategy coefficient value of the collaborative smoothing control model are corrected through parameter iteration calculation, and the calculation model of the voltage fluctuation level judgment threshold is dynamically adjusted at the same time. S54: Substitute the corrected model parameter values ​​and control strategy coefficient values ​​into the collaborative suppression control model to regenerate the voltage fluctuation suppression control command.

9. The control method for synergistic voltage fluctuation mitigation by distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 6, characterized in that, The energy storage SOC hierarchical control strategy also sets up an emergency control level for SOC exceeding limits: when the real-time value of energy storage SOC is greater than or equal to 80%, the energy storage charging circuit is blocked, and only the energy storage is allowed to perform discharge operations. The voltage fluctuation is responded to first by adjusting the output of the photovoltaic inverter and the reactive power compensation equipment; when the real-time value of energy storage SOC is less than or equal to 20%, the energy storage discharge circuit is blocked, and only the energy storage is allowed to perform low-power charging operations not exceeding 10% of the rated power of the energy storage. The voltage fluctuation is smoothed first by fully activating the reactive power compensation equipment, adjusting the photovoltaic output, and controlling the peak shifting of non-critical agricultural loads.

10. The control method for synergistic voltage fluctuation mitigation by distributed energy storage and photovoltaic power in agricultural distribution networks according to claim 5, characterized in that, The improved particle swarm optimization algorithm is a particle swarm optimization algorithm that combines adaptive weighting and dynamic mutation. During the algorithm iteration process, the initial value of the inertia weight is set to 0.9, and it is adaptively and linearly reduced to 0.4 with the number of iterations. At the same time, a dynamic Gaussian mutation operation is performed on the global optimal particle. When the algorithm fails to update the global optimal solution for three consecutive iterations, the mutation operation is triggered to expand the search range and avoid premature convergence of the algorithm. The termination condition of the algorithm is that the number of iterations reaches a preset upper limit, or the convergence error of the multi-objective optimization function is less than a preset threshold.

Images