A wind power photovoltaic energy storage matching intelligent coordination optimization system

Abstract

This invention discloses an intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratios, relating to the field of energy storage system technology. The system achieves optimization through multi-module collaboration: it constructs a four-entity non-cooperative game model involving wind, photovoltaic, energy storage, and flexible loads, incorporating flexible load demand into the ratio decision; it employs a CNN-LSTM algorithm to monitor the health status of wind, photovoltaic, and energy storage equipment throughout their entire lifecycle, using associated loss costs as optimization constraints; it dynamically adjusts the optimization target weights based on grid dispatch instructions, achieving closed-loop linkage between instructions and ratios; it adaptively allocates tiered batteries and newly added energy storage power based on the predicted confidence level of wind and photovoltaic output; and it sets up a closed-loop correction module, combining regular cycle and emergency trigger correction schemes, and avoids power surges through linear transition. This system can improve the renewable energy absorption rate, extend equipment lifespan, enhance grid connection compliance and energy storage utilization, and adapt to the integrated operation requirements of source-grid-load-storage in high-penetration new energy scenarios.

Description

Technical Field

[0001] This invention relates to the field of energy storage system technology, specifically to an intelligent coordination and optimization system for wind power and photovoltaic energy storage ratio. Background Technology

[0002] With the acceleration of the global energy transition, the installed capacity of renewable energy sources such as wind power and photovoltaics continues to grow. Wind power, photovoltaics, and energy storage joint operation systems have become a core solution to address the randomness and volatility of wind and solar power output. Currently, wind-solar-storage systems generally adopt a "source-side optimization-oriented" operation mode, which uses traditional algorithms (such as particle swarm optimization and LSTM prediction) to achieve power distribution and basic grid connection control among wind power, photovoltaics, and energy storage. For example, the charging and discharging sequence of energy storage is adjusted according to the predicted wind and solar power output, or grid reverse current is avoided through a simple power balance model. Such systems can meet basic operational needs in scenarios with low renewable energy penetration.

[0003] However, existing wind power-solar energy storage ratio optimization systems generally suffer from the core technical problem of insufficient multi-dimensional coordination between source, storage, load, grid, and equipment. On the one hand, the system only focuses on the internal ratio optimization of wind, solar, and energy storage, without incorporating the adjustable capacity and response delay of flexible loads such as industrial adjustable loads and solar-energy storage charging piles into the collaborative decision-making. This leads to a disconnect between "source-side optimization" and "load-side demand." During the peak output of solar power at noon, the energy storage is already full, but the flexible loads do not respond in time to absorb the energy, and solar power still needs to be curtailed. During peak load periods, the energy storage discharge is insufficient, requiring reliance on the grid to purchase electricity at high prices. On the other hand, the system does not use the full life cycle health parameters such as wind power equipment bearing wear, solar module degradation, and energy storage battery SOH (state of health) as optimization constraints, and excessive use of equipment leads to rapid lifespan degradation. At the same time, the system lags in responding to dynamic instructions from the grid dispatch, such as real-time electricity prices, frequency support, and reserve capacity. The grid-connected power fluctuations are prone to exceeding the standards and triggering grid assessments. Ultimately, this results in low renewable energy absorption rate, high equipment operation and maintenance costs, and poor grid compliance, making it difficult to adapt to the "integrated source-grid-load-energy storage" operation requirements in high-penetration new energy scenarios.

[0004] In view of the above, this application is hereby submitted. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent coordination and optimization system for the ratio of wind power, photovoltaic power and energy storage, so as to solve the problems mentioned in the background art.

[0006] To address the aforementioned technical problems, this invention provides an intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratios, comprising a data acquisition module, a grid command receiving module, a device lifecycle health status monitoring module, a wind and solar power output prediction confidence calculation module, a wind, solar, and energy storage flexible load multi-entity game optimization module, and an energy storage ratio execution module.

[0007] The data acquisition module collects operating data from wind power equipment, photovoltaic equipment, energy storage systems, and flexible loads.

[0008] The power grid instruction receiving module receives dynamic instructions from the power grid dispatch center;

[0009] The equipment lifecycle health status monitoring module acquires health status parameters of wind power equipment, photovoltaic equipment, and energy storage systems.

[0010] The wind and solar power output prediction confidence calculation module generates wind power output prediction results, photovoltaic power output prediction results, and corresponding prediction confidence levels;

[0011] The wind-solar-storage flexible load multi-entity game optimization module takes the collected data from the data acquisition module, the dynamic instructions from the grid instruction receiving module, the health status parameters from the equipment life cycle health status monitoring module, and the prediction results from the wind and solar output prediction confidence calculation module as inputs to construct a non-cooperative game model of wind power entity, photovoltaic entity, energy storage entity and flexible load entity, and solves the initial wind-solar-storage ratio scheme.

[0012] The energy storage allocation execution module allocates power to the cascaded batteries and new energy storage in the energy storage system based on the prediction confidence of the wind and solar power output prediction confidence calculation module and the initial wind, solar and energy storage allocation scheme, thereby realizing dynamic allocation optimization of wind power and solar energy storage.

[0013] Energy storage systems belong to electrical energy storage systems. The data acquisition module, grid command receiving module, wind-solar-storage flexible load multi-entity game optimization module, and energy storage allocation ratio execution module together constitute the control core of the power supply or distribution circuit device. By constructing a wind-solar-storage flexible load multi-entity game optimization module, flexible loads are incorporated as independent game entities into the wind power-solar-storage allocation ratio optimization system. This breaks through the technical limitations of existing technologies that only focus on the internal allocation ratio of wind, solar, and storage. At the same time, it integrates grid command receiving, equipment health status monitoring, and prediction confidence calculation functions to form a complete power supply or distribution circuit device control system and energy storage system collaborative mechanism. This realizes the transformation of allocation ratio optimization from "one-way adaptation of source and storage" to "multi-entity collaboration", significantly improving the renewable energy consumption rate and system operation economy.

[0014] Furthermore, the power grid command receiving module communicates with the power grid dispatch center using the IEC 61850 protocol. The received dynamic commands include real-time electricity prices, frequency support requirements, and reserve capacity requirements. The power grid command receiving module also includes a command parsing unit, which converts dynamic commands into target weights for the wind, solar, and energy storage flexible load multi-entity game optimization module. When receiving frequency support requirement commands, the grid connection stability weight is increased; when receiving real-time peak price commands, the cost optimization weight is increased. By clarifying the communication protocol and command parsing logic of the power grid command receiving module, it is ensured that the power grid dynamic commands can be accurately converted into target guidance for allocation optimization, enabling the power supply or distribution circuit devices to proactively respond to the power grid dispatch requirements. This avoids grid connection compliance issues caused by delayed command response in existing systems, while also improving the benefits of energy storage systems participating in power grid ancillary services.

[0015] Furthermore, the equipment lifecycle health status monitoring module includes a multi-device health parameter acquisition unit and a health status assessment unit. The multi-device health parameter acquisition unit collects bearing vibration data of wind power equipment, component operation data of photovoltaic equipment, and cell voltage and temperature data of energy storage system. The health status assessment unit uses the CNN-LSTM algorithm to process the collected data and generate health status parameters for wind power equipment, photovoltaic equipment, and energy storage system. These health status parameters include health level and remaining life prediction values. The health status assessment unit also constructs an equipment loss cost model, correlates health status parameters with equipment loss costs, and outputs them to the wind-solar-storage flexible load multi-entity game optimization module. By refining the composition and function of the equipment lifecycle health status monitoring module, the health status of the energy storage system and wind and solar equipment can be integrated into the ratio optimization in real time, avoiding the lifespan degradation problem caused by neglecting equipment aging in the existing system. At the same time, the loss cost model realizes the synergy between "short-term ratio optimization" and "long-term equipment maintenance", reducing the lifecycle operating cost of power supply or distribution circuit devices.

[0016] Furthermore, the wind and solar power output prediction confidence calculation module includes a prediction data input unit and a confidence calculation unit. The prediction data input unit acquires numerical weather forecast data and historical operating data of wind and solar equipment. The confidence calculation unit processes the input data using a random forest algorithm to generate wind power output prediction results and photovoltaic power output prediction results, and calculates the corresponding prediction confidence. This prediction confidence reflects the reliability of the prediction results and is output to the wind-solar-storage flexible load multi-agent game optimization module and the energy storage allocation execution module. By clarifying the implementation method of the wind and solar power output prediction confidence calculation module, a reliable prediction reliability basis is provided for energy storage allocation, avoiding the power allocation inaccuracy problem caused by prediction errors in existing energy storage systems. At the same time, it improves the adaptability of power supply or distribution circuit devices to wind and solar power output fluctuations and ensures system operation stability.

[0017] Further, the wind-solar-storage flexible load multi-agent game optimization module performs the following steps: S1, determining the game agents, including wind power agents, photovoltaic agents, energy storage agents, and flexible load agents, wherein flexible load agents include industrial adjustable loads, charging pile loads, and building HVAC loads; S2, acquiring input data, including data collected by the data acquisition module, target weights converted by the grid command receiving module, health status parameters and loss cost model data from the equipment life cycle health status monitoring module, and prediction results from the wind and solar power output prediction confidence calculation module; S3, constructing the game objective function, which aims to maximize total revenue, including... This includes revenue from green electricity, revenue from grid ancillary services, deduction of electricity purchase costs, and deduction of equipment loss costs; S4, using a Nash equilibrium solver to solve the objective function of the game, generating an initial wind-solar-storage allocation scheme. This scheme includes the wind power output allocation ratio, the photovoltaic power output allocation ratio, and the total charging and discharging power of the energy storage system; by refining the execution steps of the multi-agent game, clarifying the logic of the game subjects and the construction of the objective function, the allocation optimization of the power supply or distribution circuit devices can fully consider the demand for flexible loads, avoiding the problem of "disconnection between source-storage optimization and load-side demand" in the existing system. At the same time, the Nash equilibrium solution ensures the optimal interests of multiple subjects and improves the coordination efficiency of the energy storage system with wind and solar equipment and flexible loads.

[0018] Furthermore, in S3, when constructing the game objective function, the health status parameters output by the equipment lifecycle health status monitoring module are used as constraints. When the health status of the energy storage system is lower than a preset threshold, the charging and discharging power ratio of the energy storage system is reduced. When the health status of wind power equipment or photovoltaic equipment is lower than a preset threshold, the output allocation ratio of the corresponding equipment is adjusted. By introducing equipment health status constraints into the game objective function, the health protection of the energy storage system and wind and solar equipment is further strengthened, avoiding the risk of failure caused by excessive use of equipment in the existing power supply or distribution circuit devices. At the same time, it ensures that the allocation optimization scheme is balanced between economy and equipment safety, and extends the overall service life of the system.

[0019] Furthermore, the energy storage allocation execution module includes an energy storage type identification unit and a power allocation unit. The energy storage type identification unit identifies the secondary batteries and newly added energy storage in the energy storage system. The power allocation unit allocates the total charging and discharging power of the energy storage system in the initial wind-solar-storage allocation scheme based on the prediction confidence level output by the wind and solar power output prediction confidence level calculation module. When the prediction confidence level is high, the power allocation ratio of the secondary batteries is increased; when the prediction confidence level is low, the power allocation ratio of the newly added energy storage is increased. By clarifying the power allocation logic of the energy storage allocation execution module, differentiated utilization of secondary batteries and newly added energy storage is achieved, avoiding the problem of excessively rapid degradation of secondary batteries or insufficient utilization of newly added energy storage caused by the fixed power allocation in existing energy storage systems. At the same time, it improves the ability of power supply or distribution circuit devices to cope with fluctuations in wind and solar power output and reduces the operating cost of the energy storage system.

[0020] Furthermore, it also includes a closed-loop correction module. This module acquires real-time data from the data acquisition module, the latest dynamic commands from the grid command receiving module, the latest health status parameters from the equipment lifecycle health status monitoring module, and the latest prediction results from the wind and solar power output prediction confidence calculation module according to a preset cycle. This data is then input into the wind-solar-storage flexible load multi-entity game optimization module to regenerate the initial wind-solar-storage allocation scheme. The power allocation is then adjusted through the energy storage allocation execution module to achieve dynamic correction of the wind power-solar-storage allocation scheme. By adding the closed-loop correction module, the allocation optimization of the power supply or distribution circuit devices can adapt to changes in operating conditions in real time, avoiding the lag problem caused by static optimization in the existing system. At the same time, it ensures the long-term collaborative stability of the energy storage system with wind and solar equipment, flexible loads, and the grid, further improving the renewable energy absorption rate and system operational reliability.

[0021] Furthermore, the system is applied to grid-connected microgrids in industrial and commercial parks. The power supply or distribution circuit devices in this microgrid include a microgrid controller and a grid-connected interface device. The energy storage system includes cascaded battery packs and newly added energy storage battery packs. Both the cascaded battery packs and the newly added energy storage battery packs are connected to the microgrid controller through energy storage converters and receive power allocation commands from the energy storage ratio execution module. By limiting the application scenarios and hardware connection relationships of the system, the collaborative logic between the power supply or distribution circuit devices and the energy storage system is made more in line with actual application needs. This ensures that the system can effectively play a multi-entity collaborative optimization role in the industrial and commercial park microgrid scenario, significantly improve the park's green electricity self-sufficiency rate, reduce the park's electricity costs, and provide technical support for the stable collaboration between the microgrid and the main grid.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] 1. By constructing a four-entity non-cooperative game optimization mechanism of "wind, solar, and energy storage - flexible load", the adjustable capacity and response delay of industrial adjustable loads, solar and energy storage charging piles and other flexible loads are incorporated into collaborative decision-making. This breaks through the limitation of the existing focus on only the internal ratio of wind, solar and energy storage, and achieves deep coupling between "source-side optimization" and "load-side demand". This not only avoids the phenomenon of wind and solar curtailment during peak wind and solar output, but also reduces the dependence on high-priced electricity purchases from the grid during peak load, significantly improves the renewable energy consumption rate and system operation economy, and reduces the adjustment cost of flexible loads.

[0024] 2. Based on the coupled optimization model of the health status of wind, solar and energy storage equipment throughout its entire life cycle, the CNN-LSTM algorithm is used to identify health parameters such as bearing vibration of wind power equipment, degradation of photovoltaic modules and SOH of energy storage batteries in real time. An equipment loss cost model is constructed to associate the health status with the optimization goal. This breaks through the existing optimization logic that ignores equipment health, avoids the lifespan degradation caused by excessive use of equipment, extends the overall service life of wind, solar and energy storage equipment, reduces equipment operation and maintenance costs, and achieves the synergistic unity of "short-term ratio optimization" and "long-term equipment maintenance", solving the problem of balancing economy and equipment safety.

[0025] 3. The dynamic allocation ratio adaptive adjustment system driven by power grid dispatching instructions receives dynamic power grid instructions in real time through the IEC 61850 protocol and transforms them into dynamic weights of optimization targets. This achieves closed-loop linkage between power grid instructions and allocation ratio strategies, breaking through the existing static optimization mode with fixed weights. It ensures that the system can quickly respond to the needs of power grid frequency support, reserve capacity, and real-time electricity prices, significantly improving grid connection compliance. At the same time, it increases the benefits of participating in power grid ancillary services and avoids triggering power grid assessments due to delayed instruction response.

[0026] 4. By leveraging the adaptive matching strategy of predicted confidence levels for cascaded batteries and new energy storage, the power allocation ratio of the two types of energy storage is dynamically adjusted based on the predicted confidence levels of wind and solar power output. In high-confidence scenarios, cascaded batteries are prioritized to reduce costs, while in low-confidence scenarios, new energy storage is prioritized to improve response speed. This breaks through the limitations of the existing fixed allocation ratio, fully leverages the technical advantages of both types of energy storage, improves the overall utilization rate of the energy storage system, reduces the total life cycle cost of hybrid energy storage, and avoids the problems of excessively rapid degradation of cascaded batteries or insufficient utilization of new energy storage.

[0027] 5. By using a closed-loop correction module, the system combines regular periodic correction with emergency trigger correction. A linear smooth transition algorithm is designed for large deviation schemes to overcome the lag of existing static optimization schemes. This ensures that the system can adapt to dynamic scenarios such as wind and solar power output fluctuations, grid command changes, and equipment health status updates in real time, maintaining long-term stable operation of the system. It further improves the accuracy of power fluctuation smoothing and power supply reliability, meeting the operational requirements of "integrated generation, grid, load, and storage" in high-penetration new energy scenarios. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of a wind power, photovoltaic, and energy storage intelligent coordination and optimization system. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Please see Figure 1 This invention provides a technical solution: a wind power, photovoltaic, and energy storage ratio intelligent coordination and optimization system, mainly applied to grid-connected microgrid scenarios in industrial and commercial parks. It achieves dynamic ratio optimization of wind power, photovoltaic, energy storage, and flexible loads through multi-module collaboration, solving problems such as "disconnection between source, storage, and load coordination, conflict between equipment lifespan and optimization, and lag in grid command response" in existing technologies.

[0031] I. Application Scenario: This embodiment uses a grid-connected microgrid in a provincial industrial park as its carrier. The load of this park is mainly industrial production, commercial electricity consumption, and charging piles. Specific parameters are as follows: The wind and solar equipment includes a 10MW distributed photovoltaic array and a 5MW distributed wind farm. The distributed photovoltaic array uses monocrystalline silicon modules with a rated efficiency of 23%. The distributed wind farm has a single unit capacity of 2.5MW, a cut-in wind speed of 3m / s, and a rated wind speed of 12m / s. The energy storage system is a 20MWh hybrid energy storage system, including a 12MWh cascaded power battery and an 8MWh newly added lithium iron phosphate battery. The cascaded power battery is sourced from retired new energy vehicle batteries, with ternary lithium cells and an initial capacity decay to 80%. The newly added lithium iron phosphate battery... The cycle life is 6000 cycles, and the rated voltage is 3.2V. The flexible load includes a 5MW adjustable industrial motor, a 3MW photovoltaic-storage-charging pile, and a 2MW building HVAC system. The adjustable industrial motor is a continuous production load with an adjustment range of 0.5-5MW and a response delay of ≤200ms. The photovoltaic-storage-charging pile includes 50 DC fast charging piles with a single pile power of 60kW, supporting dynamic power allocation. The building HVAC system has an adjustment range of 0.8-2MW, with peak electricity consumption from 18:00 to 22:00 daily. The grid connection conditions are to connect to a 110kV distribution network, which must meet the requirements of frequency fluctuation ≤±0.2Hz, voltage fluctuation ≤±2%, and must respond to the real-time electricity price, frequency support, and reserve capacity instructions from the power grid dispatch center.

[0032] The system adopts a "layered control + modular collaboration" architecture. The hardware includes a microgrid controller, a multi-channel data acquisition terminal, an energy storage converter, and a grid connection interface device. The core chip of the microgrid controller is the TI TMS320F28379D, the energy storage converter has a rated efficiency of 96%, and the grid connection interface device includes an SVG dynamic reactive power compensation module. The software includes a data acquisition layer, a command parsing layer, a health monitoring layer, a predictive calculation layer, a game theory optimization layer, an execution control layer, and a closed-loop correction layer. The logical connections of each module are as follows: The data acquisition layer, i.e., the data acquisition module, collects real-time operating data of wind, solar, energy storage, and flexible loads through sensors and a communication bus; the command parsing layer, i.e., the grid command receiving module, receives data through IEC... The 61850 protocol interfaces with the power grid dispatch center, parsing dynamic commands and converting them into optimized target weights. The health monitoring layer, i.e., the equipment's full lifecycle health status monitoring module, calculates the health and remaining lifespan of wind, solar, and energy storage equipment based on the CNN-LSTM algorithm. The prediction calculation layer, i.e., the wind and solar power output prediction confidence calculation module, uses the random forest algorithm to generate predicted wind and solar power output values ​​and confidence levels. The game optimization layer, i.e., the wind, solar, and energy storage flexible load multi-agent game optimization module, constructs a four-agent non-cooperative game model to solve the initial allocation scheme. The execution control layer, i.e., the energy storage allocation execution module, allocates the power of cascaded batteries and newly added energy storage according to the prediction confidence level. The closed-loop correction layer, i.e., the closed-loop correction module, updates data and corrects the allocation scheme according to a preset cycle to ensure dynamic system adaptation.

[0033] II. Data Acquisition Module: Data acquisition is the foundation of system optimization. It requires obtaining multi-dimensional data on "wind and solar power output, energy storage status, flexible load, and equipment health" to address the problem of "optimization inaccuracies caused by single data dimensions" in existing technologies. This module achieves data acquisition through distributed sensors and an industrial bus, including Modbus-RTU and EtherCAT. The acquisition frequency is set differently based on the data type: high-frequency data such as energy storage cell voltage and wind turbine speed are acquired every 1 second; medium-frequency data such as photovoltaic power output and load power are acquired every 10 seconds; and low-frequency data such as equipment health parameters and meteorological data are acquired every minute. Specific technical methods are as follows:

[0034] The data acquisition module collects wind power equipment data including wind speed, wind direction, turbine speed, pitch angle, and output, which are acquired using ultrasonic anemometers and encoders. The ultrasonic anemometers support RS485 output, and the acquisition frequency is 1 second per acquisition. Photovoltaic equipment data includes irradiance, module temperature, and array output, acquired using irradiance sensors and thermocouples. The irradiance sensor has a measurement range of 0-2000 W / m². 2The data collection frequency is 10 seconds per data point; the energy storage system data includes cell voltage, current, temperature, SOC and SOH, which are collected through the battery management system with a sampling accuracy of ±0.01V and a collection frequency of 1 second per data point; the flexible load data includes industrial motor power, charging pile current and HVAC load, which are collected using power sensors and smart meters with an accuracy of 0.5 class, and a collection frequency of 10 seconds per data point; the meteorological auxiliary data includes precipitation, humidity and cloud coverage, which are collected through an automatic weather station with wireless transmission, and a collection frequency of 1 minute per data point.

[0035] The collected data needs to be filtered and outlier removed to avoid noise interference with the optimization results. This embodiment uses Kalman filtering and the 3σ criterion for preprocessing.

[0036] 1. Taking wind power output data filtering as an example, the state equation of Kalman filtering is: ,in Let k be the actual wind power output at time k. This is the state transition matrix, with a value of 0.98 to reflect the continuity of force. for Input wind speed at all times. This is the input coefficient, with a value of 0.05, calibrated based on the fan characteristics. For process noise, it follows the rule N(0,0.01). 2 Distribution; Observation equation is ,in The wind power output observation value collected at time k is... The observation matrix takes the value 1. To observe the noise, it follows the rule N(0,0.02). 2 The filtering update process includes prior estimation, prior covariance calculation, Kalman gain calculation, posterior estimation, and posterior covariance calculation. The prior estimation formula is... The formula for prior covariance is: ,in The process noise covariance is set to 0.01. 2 The Kalman gain formula is ,in To observe the noise covariance, a value of 0.02 was used. 2 The posterior estimation formula is: The formula for posterior covariance is: ,in It is an identity matrix.

[0037] 2. The specific process for outlier removal based on criteria is as follows: Calculate the mean of the preprocessed data. with standard deviation If a certain data point satisfy If the value is not found, it is considered an outlier and is replaced by the mean of the data from the previous three time points.

[0038] Example: Taking the data analysis of a weekday from 14:00 to 14:05 as an example:

[0039] 1. In the original data collection, the observed wind power output values ​​were 4.2MW, 4.5MW, 3.1MW, 4.3MW, and 4.4MW. Among them, the 3.1MW value at 14:02 was an outlier, corresponding to wind speeds of 8.5m / s, 8.8m / s, 7.0m / s, 8.6m / s, and 8.7m / s.

[0040] 2. In the Kalman filtering stage, the initial value is set to... MW, The prior estimate at 14:00 is calculated as follows: MW, the prior covariance is calculated as The Kalman gain is approximately The posterior estimate is MW; similarly, the filtered output at each time point from 14:00 to 14:05 is calculated to be 4.213MW, 4.485MW, 3.892MW, 4.291MW, and 4.387MW;

[0041] 3. In the outlier removal phase, the mean of the filtered data is calculated. MW, standard deviation MW, The value is approximately 0.654MW; at 14:02, the difference between the filtered value of 3.892MW and the mean is 0.361MW, which is less than 0.654MW, so it is judged to be normal, and the original outlier has been corrected by filtering;

[0042] 4. Final output of preprocessed data: Wind power output is 4.21MW, 4.49MW, 3.89MW, 4.29MW, and 4.39MW; photovoltaic power output is 8.12MW, 8.08MW, 8.05MW, 8.09MW, and 8.10MW; industrial motor load is 4.50MW, 4.52MW, 4.48MW, 4.51MW, and 4.53MW.

[0043] Existing publicly available documents only collect power output and SOC data for wind, solar, and energy storage, omitting detailed data such as the adjustable capacity of flexible loads and equipment vibration. Furthermore, they employ simple moving average filtering, resulting in low preprocessing accuracy. This module's unique technical approach is primarily reflected in three aspects: 1. Data Dimension Expansion: Including the "adjustable range + response delay" of flexible loads and the "vibration / temperature" of equipment in the data collection scope, providing data support for subsequent multi-agent game theory and health monitoring; 2. Preprocessing Algorithm Optimization: Employing Kalman filtering combined with... The criteria are that the error is reduced compared to the moving average filtering. In this embodiment, the output data error is ≤2%, ensuring the reliability of the data input to the optimization model; 3. The data preprocessing accuracy is improved, which improves the adaptability of the subsequent ratio scheme, avoids overcharging and discharging of energy storage due to data noise, and extends the energy storage life.

[0044] III. Grid Command Receiving Module: Grid commands are the core constraints for the grid-connected operation of the system. Dynamic commands from the dispatch center, such as electricity prices and frequency support, need to be transformed into weights for optimization objectives, addressing the problem of "disconnect between grid commands and allocation optimization" in existing technologies. This module achieves real-time command reception via the IEC 61850 protocol, and then transforms the commands into quantifiable optimization objectives through a weight allocation model, ensuring the system proactively responds to grid demands. The specific technical solution is as follows:

[0045] This module adopts the IEC 61850-8-1 standard protocol and communicates with the SCADA system of the power grid dispatch center via Ethernet at a communication rate of 100Mbps and a transmission latency of ≤50ms. Command types include real-time electricity price commands, frequency support commands, and reserve capacity commands. The parsing rules for each type of command are as follows: Real-time electricity price commands are formatted as "time period - electricity price," for example, "14:00-18:00-1.2 yuan / kWh," and are parsed as the basis for cost optimization weight adjustment; Frequency support commands are formatted as "current frequency - power to be released," for example, "49.8Hz-2MW," and are parsed as the basis for grid connection stability weight adjustment; Reserve capacity commands are formatted as "reserved capacity - duration," for example, "3MWh-2h," and are parsed as energy storage capacity constraints.

[0046] The system's optimization objectives include "economic efficiency ( ), grid connection stability ( Equipment safety The total weight satisfies the following conditions. The weight allocation formula is as follows: The baseline state is defined by the absence of specific instructions. , , When receiving a peak price instruction, i.e., the electricity price is ≥ 1.0 yuan / kWh, , , , must meet , When receiving frequency support commands, i.e., frequency ≤ 49.9Hz, , , , must meet , When receiving a reserve capacity instruction, the weights are not adjusted, but the constraint condition "energy storage reserved capacity ≥ instruction value" is added to the game model.

[0047] Example: Taking the power grid command processing at 14:00 on a certain workday as an example:

[0048] 1. During the instruction receiving phase, the dispatch center sends three instructions: the first is the real-time electricity price instruction, with the price set at 1.2 yuan / kWh from 14:00 to 18:00, which is a peak price; the second is the frequency instruction, with the current frequency at 49.85Hz, requiring the release of 1.5MW of power; the third is the reserve capacity instruction, reserving 3MWh of capacity for 2 hours, from 14:00 to 16:00.

[0049] 2. During the instruction parsing phase, peak price instructions trigger an increase in economic weight, while frequency instructions trigger an increase in stability weight; these weights need to be calculated cumulatively. First, the baseline weight is determined as follows: , , Based on the adjustment of the peak price order, , , Then adjust the commands based on frequency support. , , ,because ≥0.2 is required, therefore the final adjustment is as follows: , , ;

[0050] 3. In the constraint setting stage, the constraint "energy storage reserved capacity ≥ 3MWh" is added to the game model;

[0051] 4. The output result is the target weight. (Economic efficiency) (stability), (Safety) Energy storage reserved capacity constraint of 3MWh.

[0052] Existing publicly available documents use fixed weights, such as the fixed allocation method proposed in a journal article with an economic weight of 0.5 and a stability weight of 0.5. This method cannot respond to dynamic grid commands, resulting in low grid connection compliance rates. The unique technical means of this module are reflected in three aspects: 1. Command-weight linkage mechanism: Real-time electricity price, frequency support and other commands are quantified into target weights to achieve a closed loop of "command change → dynamic weight adjustment", rather than static preset; 2. Multi-command superposition processing: For scenarios where multiple commands are received simultaneously, a weight superposition algorithm is designed and a minimum weight threshold is set to avoid system imbalance caused by excessive priority of a single target; 3. The grid command response latency is shortened to 180ms, the grid connection compliance rate is improved, and the annual income from participating in grid ancillary services such as frequency regulation and reserve is increased. In this embodiment, the income from a single frequency support is approximately 200 yuan / h.

[0053] IV. Equipment Lifecycle Health Status Monitoring Module: Equipment health status, or SOH, is a crucial constraint for ratio optimization. It requires real-time monitoring of the aging degree and remaining lifespan of wind, solar, and energy storage equipment to address the problem of "neglecting equipment health leading to excessively rapid lifespan degradation" in existing technologies. This module employs a CNN-LSTM fusion algorithm to calculate SOH in two stages: "feature extraction" and "time series prediction." Simultaneously, it constructs a loss cost model to link health status with economic costs. The specific technical solution is as follows:

[0054] Equipment health parameters include core parameters for wind power equipment, photovoltaic equipment, and energy storage systems: the health parameter for wind power equipment is bearing vibration acceleration. With gearbox oil temperature bearing vibration acceleration The larger the value, the more severe the mechanical wear and the higher the gearbox oil temperature. Excessive levels will accelerate equipment aging; the health parameter of photovoltaic equipment is the module degradation rate. With open circuit voltage Component attenuation rate A smaller value indicates a more severe degradation in photoelectric conversion efficiency and an open-circuit voltage. A decrease in voltage will directly affect photovoltaic output; the health parameter of an energy storage system is the cell voltage difference. With the number of loops Cell voltage difference The larger the number of cycles, the more severe the battery capacity degradation. The more you have, the shorter your remaining lifespan.

[0055] The CNN feature extraction uses a three-layer convolutional neural network to extract spatial features of health parameters, such as the frequency components of vibration signals and the fluctuation features of voltage curves. The input layer dimension is... ,in The time series length is 100 time points. For the number of parameters, such as energy storage systems Corresponding cell voltage difference With the number of loops Convolutional layer 1 has 32 filters, a kernel size of 3×1, uses ReLU activation, and has an output dimension of [missing information]. Pooling layer 1 uses max-pooling, with a kernel size of 2×1 and an output dimension of... Convolutional layer 2 has 64 filters, a kernel size of 3×1, uses ReLU as the activation function, and has an output dimension of [missing information]. Pooling layer 2 uses max-pooling, kernelsize=2×1, and the output dimension is... The output dimension of the fully connected layer is , which serves as the feature vector input to the LSTM.

[0056] The feature vectors extracted by the CNN are input into the LSTM network for predicting the state of interest (SOH). The LSTM cell state update formula includes the calculation of the forget gate, input gate, candidate cell state, and cell state and output gate: the forget gate formula is as follows. The input gate formula is The candidate cell state formula is: The cell state formula is: The output gate formula is The hidden state formula is ,in The feature vector output by the CNN. The hidden state at time t-1 , , , This is the weight matrix. , , , For bias terms, For the sigmoid function, It is an element-wise product. The SOH output formula is: ,in To output weights, For output bias, The value range is [0,1], where 1 represents a brand new state and 0 represents a scrapped state.

[0057] The equipment loss cost model links SOH (Solar Energy Utilization) with equipment loss cost, with the specific formula as follows: 1. The formula for the loss cost of an energy storage system is: ,in The initial cost per unit capacity for energy storage is 500 yuan / kWh for secondary batteries and 1200 yuan / kWh for new energy storage. 1. Energy storage capacity, unit: kWh; 2. Wind power equipment loss cost formula: ,in The initial cost per unit power of wind power is taken as 3000 yuan / kW. 3. The formula for photovoltaic equipment loss cost is: (Wind power rated power, unit: kW) ,in The initial cost per unit power of photovoltaic power is taken as 2000 yuan / kW. 4. The formula for total equipment loss cost is: (The formula is missing from the original text.) .

[0058] Example: Taking the health monitoring of energy storage systems and wind power equipment at 14:00 as an example:

[0059] 1. During the health parameter collection phase, the maximum voltage difference between cells in the cascaded battery. Number of loops Next; the maximum voltage difference between the newly added energy storage cells. Number of loops Second; bearing vibration acceleration of the fan Gearbox oil temperature ;

[0060] 2. In the CNN feature extraction stage, the length of the input sequence is... The data corresponding to the 100 seconds before 14:00 includes the number of energy storage system parameters. Number of wind turbine parameters Convolutional layer 1 outputs 32-dimensional features, which are reduced to 16-dimensional features after processing by pooling layer 1; Convolutional layer 2 outputs 64-dimensional features, which are reduced to 32-dimensional features after processing by pooling layer 2; The fully connected layer outputs a 128-dimensional feature vector.

[0061] 3. LSTM prediction of the SOH stage and the LSTM hidden state of the secondary battery. Calculations yielded That is, 78%; the hidden state of the newly added energy storage LSTM Calculations yielded That is, 95%; the LSTM hidden state of the wind turbine. Calculations yielded That is, 85%; photovoltaic equipment after pretreatment That is, 92%;

[0062] 4. Loss cost calculation stage: tiered battery loss cost Yuan; additional energy storage loss cost Yuan; Wind power equipment loss cost Yuan; Photovoltaic equipment loss cost Yuan; Total equipment depreciation cost Yuan, the average daily loss is calculated by dividing the annual loss by 365, approximately 15,500 yuan / day;

[0063] 5. The output results are: SOH=78% for the secondary battery, SOH=95% for the newly added energy storage, SOH=85% for the wind turbine, SOH=92% for the photovoltaic system, and the average daily equipment loss cost is 15,500 yuan.

[0064] Existing publicly available documents only estimate the SOH of energy storage using SOC, with an error of ≥8%, and do not correlate it with the health status of wind and solar power equipment, nor do they include a loss cost model. Existing documents only monitor the SOC of energy storage batteries, without addressing the correlation between equipment health and cost. This module's unique technical approach is reflected in four aspects: 1. Multi-device health monitoring fusion: Incorporating the health parameters of wind power, solar power, and energy storage into the same monitoring model, rather than monitoring them separately; 2. CNN-LSTM algorithm innovation: Using CNN to extract spatial features and LSTM to capture temporal changes, significantly reducing the SOH error compared to traditional SOC estimation methods; 3. Health-cost correlation: Constructing a loss cost model, transforming the abstract SOH into quantifiable economic costs, providing a quantitative basis for "equipment safety" in game theory optimization; 4. Precise monitoring of equipment health status reduces the number of overcharge / discharge cycles in energy storage, extends the lifespan of cascaded batteries, extends wind turbine maintenance cycles, and significantly reduces annual equipment operation and maintenance costs.

[0065] V. Wind and Solar Output Prediction Confidence Calculation Module: The randomness of wind and solar output is the core challenge of energy storage allocation optimization. It is necessary to predict the output value and quantify the reliability of the prediction, i.e., the confidence level, to solve the problem in existing technologies where "only the output value is predicted, ignoring the confidence level, leading to inaccurate energy storage allocation." This module uses the Random Forest algorithm to achieve prediction from two aspects: "multi-feature input - probability output," and defines a confidence level index to provide a basis for energy storage allocation. Specific technical methods are as follows:

[0066] Twelve characteristics affecting wind and solar power output were selected and categorized into three types: meteorological characteristics, historical characteristics, and equipment characteristics. Meteorological characteristics include wind speed. Light intensity ,temperature ,humidity These are environmental factors that directly affect the power output of wind and solar power; historical characteristics include the power output in the first hour. , the same period of output in the first 24 hours This is used to reflect the time-dependent nature of force; equipment characteristics include the wind turbine pitch angle. Photovoltaic module temperature It is used to reflect the impact of equipment operating status on output.

[0067] Random forests consist of 100 decision trees, outputting the final predicted value through a voting method, while simultaneously calculating the prediction variance to reflect the confidence level. The decision tree construction steps are as follows: 1. Sample sampling: Randomly select 70% of the historical data as the training set and 30% as the test set. The historical data consists of 8760 time points over one year, using Bootstrap sampling (sampling with replacement); 2. Feature selection: Randomly select 6 features for each decision tree node, representing 50% of the total number of features, and calculate the information gain ratio to select the optimal splitting feature; 3. Tree pruning: Use post-pruning to avoid overfitting, setting the minimum number of samples per leaf node to ≥5.

[0068] Taking photovoltaic power output as an example, the output formula for a single decision tree is as follows: ,in This corresponds to 100 decision trees; the random forest output uses a voting method, and the formula is... The prediction variance is used to reflect the degree of volatility, and the formula is: .

[0069] Prediction confidence Reflecting the reliability of the predicted values, it is negatively correlated with the predicted variance, as shown in the formula: ,in This represents the current prediction variance. The largest historical forecast variance was obtained through one year of data statistics for photovoltaics. MW 2 Wind power MW 2 The value of C is in the range of [0,1]. Defined as high confidence level, Defined as medium confidence level, Defined as low confidence level.

[0070] Example: Taking the wind and solar power output forecast from 14:00 to 15:00 as an example:

[0071] 1. During the input feature data stage, the meteorological feature at 14:00 is wind speed. m / s, light intensity W / m 2 ,temperature ,humidity Historical characteristics include wind power output in the first hour. MW, wind power output in the first 24 hours MW, photovoltaic power output in the first hour MW, photovoltaic power output in the first 24 hours MW; Equipment characteristics include wind turbine pitch angle Photovoltaic module temperature ;

[0072] 2. In the random forest prediction phase, the output of 100 decision trees in wind power output prediction. The range is 3.8MW-4.6MW, calculated as follows: MW, Prediction Variance MW 2 Output of 100 decision trees in photovoltaic power output prediction The range is 7.8MW-8.4MW, calculated as follows: MW, Prediction Variance MW 2 ;

[0073] 3. Confidence Calculation Stage: Wind Power Confidence It belongs to the high confidence level; photovoltaic confidence level This indicates a high level of confidence.

[0074] 4. The output results are as follows: wind power predicted output of 4.2MW with confidence level of 0.93 and photovoltaic power predicted output of 8.1MW with confidence level of 0.95, both of which are high confidence levels.

[0075] Existing publicly available documents using the LSTM algorithm only output power prediction values ​​without confidence indices, leading to an inability to adapt energy storage allocation to prediction reliability. For example, a dissertation only achieved point prediction of wind and solar power output without quantifying prediction uncertainty. Even with low confidence, a fixed allocation ratio was still used, resulting in a fluctuation smoothing error ≥12%. This module's unique technical approach lies in three aspects: 1. Confidence quantification: Defining a confidence index based on prediction variance transforms "prediction reliability" into a quantifiable value, rather than a subjective judgment; 2. Multi-feature fusion prediction: Incorporating equipment features such as propeller pitch angle reduces the prediction error by 50% compared to LSTM predictions using only meteorological features. In this embodiment, the power prediction error is ≤3%; 3. Confidence-guided energy storage allocation reduces power fluctuation smoothing errors, avoids excessive energy storage response in low-confidence scenarios, and reduces the number of charge / discharge cycles for cascaded batteries.

[0076] VI. Multi-Agent Game Theory Optimization Module for Wind, Solar, and Energy Storage Flexible Loads: Multi-agent game theory is the core of system optimization. It requires treating wind power, solar power, energy storage, and flexible loads as independent players, and using Nash equilibrium to find the optimal allocation scheme, addressing the problem in existing technologies that "only optimize the internal aspects of wind, solar, and energy storage, neglecting load-side demand." This module constructs a four-agent non-cooperative game model with the goal of maximizing total revenue, while incorporating grid command weights and equipment health constraints to ensure the scheme balances economy, stability, and security. Specific technical methods are as follows:

[0077] The game involves wind power entities, photovoltaic entities, energy storage entities, and flexible load entities. The decision variables, i.e., strategies, and strategy space constraints for each entity are as follows:

[0078] 1. The decision variable for wind power entities is the power output allocation ratio. , used to represent the ratio allocated to load and energy storage, with a policy space constraint of . and ,in To contribute to actual wind power, To meet the maximum load demand, in this embodiment MW;

[0079] 2. The decision variable for photovoltaic entities is the power output allocation ratio. , used to represent the ratio allocated to load and energy storage, with a policy space constraint of . and ,in To contribute to the actual development of photovoltaics;

[0080] 3. The decision variable for energy storage entities is charging and discharging power. Positive sign represents charging, negative sign represents discharging, and the strategy space constraint is: and ,in To achieve the maximum charge and discharge power for energy storage, in this embodiment... MW, , ;

[0081] 4. The decision variable for the flexible load subject is the adjustable load power. , used to represent the power of load increase or decrease, with policy space constraints as follows: and ,in To minimize load requirements, in this embodiment... MW, This represents the current load power. In this embodiment, the maximum load adjustment is... MW.

[0082] The revenue functions of each entity need to be combined with the weights of the power grid command. , , Total revenue The weighted sum of the gains of each entity is calculated using the following formula:

[0083] 1. The main revenue from wind power includes revenue from green electricity and wind curtailment losses, as shown in the formula: ,in Yuan / kWh is the green electricity price. Yuan / kWh represents wind curtailment losses. For wind power loss costs, refer to the calculation results of the equipment life cycle health status monitoring module;

[0084] 2. The main revenue of photovoltaic power includes green electricity revenue and curtailment losses, as shown in the formula: ,in The photovoltaic loss cost is calculated based on the results of the equipment lifecycle health status monitoring module.

[0085] 3. The revenue of energy storage entities includes arbitrage profits and charging / discharging losses, as shown in the formula: ,in The discharge power, i.e. Take the absolute value. The charging power, i.e. The value at time, The peak price is 1 yuan / kWh. Yuan / kWh is the off-peak price. The cost per kWh is the fluctuation penalty cost. The energy storage loss cost is calculated based on the results of the equipment's full life-cycle health status monitoring module.

[0086] 4. The main benefits of flexible loads include savings in electricity costs and adjustment costs, as shown in the formula: ,in Yuan / kWh represents the load regulation cost;

[0087] 5. The total revenue function is: .

[0088] Nash equilibrium refers to a state where each agent's strategy maximizes its own payoff, and without changing its strategy, the payoffs of other agents cannot be increased. This module uses an improved particle swarm optimization algorithm, IPSO, to solve for Nash equilibrium. The specific steps are as follows: 1. Initialization: Set the number of particles to 50, the dimension to 4, and the corresponding... , , , Four decision variables, maximum number of iterations 100, inertia weights w; 2. Fitness Function: Set the fitness function to the total reward. The objective is to maximize total revenue; 3. Speed ​​update: The speed update formula is... ,in As a learning factor, , A random number in the range [0,1]. For the optimal position of an individual, 4. Position Update: The position update formula is: After the update, it is necessary to verify whether the policy space constraints of each subject are met. If the constraints are exceeded, the policy space constraints should be adjusted to the constraint boundary. 5. Convergence judgment: When the number of iterations reaches 100 or the fitness function changes by ≤0.1% for 10 consecutive iterations, the iteration is stopped and the global optimal solution, i.e., the Nash equilibrium point, is output.

[0089] Example: Taking a game optimization exercise at 14:00 as an example, the input data is as follows: wind power output in the wind and solar power generation. MW, photovoltaic MW; Current load demand MW, including 4.5MW for industrial use, 3.0MW for charging piles, and 1.5MW for HVAC. MW, MW; in energy storage status , MW; target weights , , Cost parameters Yuan / kWh Yuan / kWh Yuan / kWh Yuan / kWh.

[0090] The IPSO solution process is as follows:

[0091] 1. During the particle initialization phase, the initial strategy of a certain particle is set to... , , MW stands for charging. MW stands for load increase;

[0092] 2. In the fitness calculation stage, the wind power main revenue Yuan (Note: unit conversion is required here; actual calculation is based on hours and adjusted accordingly). Yuan / h; similarly, the revenue of the photovoltaic main body is calculated. Yuan / h, revenue of energy storage entity Yuan / h represents the charging cost, while the revenue of the flexible load entity is also a factor. Yuan / hour; Total Revenue Yuan / h;

[0093] 3. In the iterative optimization phase, after 100 iterations, the global optimal solution is: That is, 95% of wind power is allocated to the load, and 0.05% is curtailed. This means that 100% of the photovoltaic power is allocated to the load, with no curtailment. MW instant charging, with a reserved capacity of 3MWh. MW means the load increased by 0.3MW, bringing the total load to 9.3MW;

[0094] 4. During the Nash equilibrium verification phase, if any entity changes its strategy, for example, the wind power entity will... Adjusting the value to 0.9 and recalculating the total revenue to 128 yuan / h proves that this solution is a Nash equilibrium point.

[0095] 5. The initial output allocation scheme is as follows: wind power output of 4.0MW (4.2×0.95) is allocated to the load, photovoltaic output of 8.1MW is allocated to the load, energy storage and charging of 0.8MW is allocated, and the flexible load is increased to 9.3MW. There is no wind or solar curtailment, which meets the reserved capacity constraints.

[0096] Existing publicly available documents employ single-objective optimization, focusing only on economic efficiency and failing to treat flexible loads as independent entities. Furthermore, they lack Nash equilibrium solutions. For example, one patent document only optimizes the output allocation of wind, solar, and energy storage, neglecting load-side demand, resulting in a curtailment rate of ≥8%. This module's unique technical approach is reflected in four aspects: 1. Four-entity game architecture: For the first time, flexible loads are incorporated into the game system, breaking through the traditional logic of "one-way load adaptation by source and storage," achieving "source-storage-load" synergy; 2. Multi-objective revenue function: Integrating grid weights and equipment loss costs, the total revenue is increased by 40% compared to single-objective optimization, increasing from 94 yuan / h to 132 yuan / h in this embodiment; 3. IPSO Nash equilibrium solution: Convergence speed is improved by 30% compared to the traditional particle swarm optimization algorithm, avoiding local optima and ensuring the solution is globally optimal; 4. Increased renewable energy absorption rate, reducing wind and solar curtailment rates to below 4%, and lowering flexible load adjustment costs.

[0097] VII. Energy Storage Allocation Module: Energy storage allocation is crucial for the successful implementation of the optimization plan. It requires allocating the power of secondary batteries and new energy storage based on the predicted confidence level, addressing the problem in existing technologies where "fixed allocation ratios lead to excessively rapid degradation of secondary batteries or insufficient utilization of new energy storage." This module dynamically adjusts the allocation coefficient based on the confidence level, achieving differentiated utilization where "high confidence levels utilize secondary batteries for low cost, while low confidence levels utilize new energy storage for high response." Specific technical methods are as follows:

[0098] Energy storage systems include two categories: secondary batteries and new energy storage. The rated power of secondary batteries is... MW, response latency ms, unit cost Yuan / kWh, suitable for high-confidence scenarios with low fluctuations and low costs; newly added energy storage rated power MW, response latency ms, unit cost Priced at RMB / kWh, it is suitable for scenarios with low confidence levels, i.e., large fluctuations and high response requirements.

[0099] Define allocation coefficients This indicates the proportion of secondary battery power to the total energy storage power, and is related to the confidence level of wind and solar power predictions. Positive correlation, the specific formula is as follows:

[0100] 1. When That is, at high confidence level, ,For example hour , hour ;

[0101] 2. When That is, at medium confidence level, A fixed 50% allocation ratio is adopted;

[0102] 3. When That is, at low confidence levels, ,For example hour , hour After taking the absolute value, the newly added energy storage ratio is 150%. At this point, it is necessary to verify whether the rated power constraint is met. If it is exceeded, the rated power limit shall apply.

[0103] The power allocation formula is as follows: power of secondary battery ,in The total energy storage power output in the game is represented by positive values ​​for charging and negative values ​​for discharging; the newly added energy storage power... Power allocation must meet the rated power constraint, i.e. and If the constraints are exceeded, the corresponding energy storage power will be adjusted to the rated power, and the power of another type of energy storage will be recalculated to ensure that the total power conforms to the game theory scheme.

[0104] During the control phase, power distribution commands are sent to the energy storage converter (PCS) via the Modbus-RTU protocol. The PCS uses a droop control algorithm to adjust the output power. The droop control formula is as follows: ,in For output frequency, Hz is the rated frequency. This is the droop factor, with a value of 0.02Hz / MW, to ensure that the output frequency remains stable within the range required by the power grid.

[0105] Example: Taking the energy storage ratio at 14:00 as an example:

[0106] 1. Input data phase: Game theory outputs total energy storage power. MW stands for charging; wind power in wind and solar forecast confidence. Photovoltaics Take the average value This indicates a high level of confidence.

[0107] 2. In the allocation coefficient calculation stage, based on the high confidence formula... That is, the proportion of recycled batteries is 91%;

[0108] 3. Power distribution stage, tiered battery power MW, approximately 0.73MW, which is 0.73MW of secondary battery charging; additional energy storage capacity. MW, approximately 0.07MW, meaning an additional 0.07MW of energy storage charging; verifying rated power constraints, , The constraints are satisfied.

[0109] 4. During the control execution phase, power allocation commands are sent to the energy storage converters. The energy storage converters corresponding to the secondary batteries are adjusted according to a charging power of 0.73MW, and the energy storage converters corresponding to the newly added energy storage are adjusted according to a charging power of 0.07MW. Based on the droop control algorithm, the frequency decreases slightly due to the increase in charging power; the output frequency is calculated. Hz, which meets the power grid frequency requirements of 49.8-50.2Hz;

[0110] 5. The output results are 0.73MW of secondary battery charging, 0.07MW of new energy storage charging, and 0.8MW of total energy storage charging, which meets the game theory scheme and grid frequency constraints.

[0111] Existing publicly available documents use a fixed allocation ratio. One document proposes a fixed allocation method of 50% for secondary batteries and 50% for new energy storage. This cannot adapt to the predicted confidence level, resulting in a 30% reduction in the cycle life of secondary batteries or a utilization rate of less than 40% for new energy storage. The unique technical means of this module are reflected in three aspects: 1. Confidence-allocation linkage: The allocation coefficient is dynamically adjusted based on the predicted confidence level, rather than a fixed ratio, to achieve a balance between reliability and cost; 2. Differentiated utilization strategy: Secondary batteries are prioritized in high-confidence scenarios to reduce costs, while new energy storage is prioritized in low-confidence scenarios to improve response speed. Compared with a fixed allocation, the lifespan of secondary batteries is extended by 45%, and the utilization rate of new energy storage is increased by 60%, from 40% to 64%; 3. The total life cycle cost of the hybrid energy storage system is reduced, and the power response speed meets the grid requirements of ≤100ms.

[0112] 8. Closed-Loop Correction Module: Closed-loop correction ensures dynamic system adaptation. It updates data and corrects the energy storage ratio scheme according to a preset cycle, solving the problem of "static optimization leading to scheme lag" in existing technologies. This module adopts a 5-minute correction cycle, updating data acquisition, command parsing, health monitoring, and prediction calculation results in real time, re-triggering game optimization and energy storage ratio to ensure the system always operates in an optimal state. Specific technical means are as follows:

[0113] Correction triggers include two categories: regular corrections and emergency corrections. Regular corrections use a fixed cycle, set to 5 minutes (300 seconds), triggering a full-process correction every 5 minutes, covering all stages of data acquisition, command parsing, health monitoring, predictive calculation, game theory optimization, and energy storage allocation. Emergency corrections are set with trigger conditions, and are triggered immediately when any of the following conditions are met, without waiting for the regular cycle: changes in grid commands, such as a frequency drop from 49.85Hz to 49.7Hz; abnormal equipment health, such as a sudden drop of 5% in the energy storage SOH; and excessive output fluctuations, such as a deviation of ≥20% between actual output and predicted value.

[0114] The revision process includes three steps: data update, scheme comparison, and smooth transition.

[0115] 1. Data update: Re-collect real-time data within 5 minutes, including wind power output, photovoltaic power output, load power, energy storage SOC, etc., and update grid commands, equipment SOH, and prediction confidence to ensure the real-time nature of input data;

[0116] 2. Scheme Comparison: Calculate the deviation between the current implementation scheme and the new optimized scheme. Taking the energy storage power deviation as an example, the formula is: ,in For the energy storage capacity of the new scheme, This represents the current energy storage capacity.

[0117] 3. Smooth transition: If deviation MW means small deviation; switch directly to the new solution. If the deviation is small... MW stands for large deviation, which uses a linear transition method. The transition formula is as follows: ,in The transition time, with a value ranging from 0 to T, The system will smoothly switch to the new scheme within 60 seconds to avoid power fluctuations caused by sudden power changes.

[0118] Example: Taking the regular correction at 14:05 as an example:

[0119] 1. During the data update phase, real-time data is collected from 14:00 to 14:05, including wind power in the actual wind and solar power output. MW, a deviation of 7.1% from the predicted 4.2MW, photovoltaic MW, a deviation of 2.5% from the predicted 8.1MW; the frequency in the grid instruction dropped to 49.8Hz, increasing the required power output from 1.5MW to 2.0MW; the SOH of the secondary battery in the equipment is 77.9%, without a sudden drop, while the SOH of the newly added energy storage is 95%; prediction confidence level It remains at a high confidence level;

[0120] 2. During the optimization phase of the new plan, the target weights are updated to... From 0.5 to 0.6, , The new solution obtained through game theory is: MW means a discharge of 0.5MW, which meets the frequency support requirements. That is, all wind power is allocated. That is, all photovoltaic power is allocated. That is, the load is not adjusted;

[0121] 3. Solution comparison phase: Old solution MW charging, a new solution MW stands for discharge, deviation. For MW values ​​greater than 0.5MW, a smooth transition is required; based on the transition formula. ,in The duration is 0-60 seconds.

[0122] 4. During the correction phase, 14:05-14:06 is the transition period. At 14:05, the energy storage power is 0.8MW, and at 14:05:30, the calculated energy storage power is... MW, energy storage capacity reached at 14:06 MW; After 14:06, the new scheme stabilized and discharged 0.5MW of energy storage, and the frequency rose back to 49.85Hz, meeting the grid requirements;

[0123] 5. The output result is the revised scheme: 3.9MW of wind power and 7.9MW of photovoltaic power are all allocated to the load, 0.5MW of energy storage is discharged, the load is maintained at 9.3MW, there is no wind curtailment or solar curtailment, and the frequency is stable.

[0124] Existing publicly available documents lack a closed-loop correction mechanism, and the optimization scheme remains fixed. Publicly available documents update the scheme every 24 hours, resulting in a lag and a deviation of ≥15%. This module's unique technical approach is mainly reflected in three aspects: 1. Dual-trigger correction mechanism: Combining regular periodic correction and emergency trigger correction, the lag time is reduced by 80% compared to fixed-period correction schemes, from 120 minutes to 24 minutes; 2. Smooth transition strategy: Avoiding power surges through linear transition, frequency fluctuations are reduced from ±0.15Hz to ±0.05Hz, ensuring grid stability; 3. Improved system operating optimality rate, grid-connected power fluctuation amplitude controlled within ±3%, and response speed improved to 100ms in emergency scenarios, meeting the dynamic requirements of the power grid.

[0125] The core technical advantages of this system are as follows: 1. Multi-entity collaborative breakthrough: Flexible loads are incorporated into the game theory system, constructing a four-entity game model of "source-storage-load" to solve the problem of load-side demand disconnect in existing technologies and achieve optimal benefits for multiple entities; 2. Health-optimization fusion breakthrough: CNN-LSTM health monitoring and loss cost model are embedded into the optimization process to solve the problem of conflict between equipment lifespan and optimization objectives in existing technologies and achieve synergy between short-term optimization and long-term maintenance; 3. Command-weight linkage breakthrough: Dynamic grid commands are quantified into optimization target weights to solve the problem of grid connection command response lag in existing technologies and achieve proactive coordination between the system and the grid; 4. Confidence-energy storage adaptation breakthrough: Confidence is defined based on prediction variance, and tiered batteries and new energy storage power are dynamically allocated to solve the problem of energy storage utilization imbalance in existing technologies and achieve differentiated cost control; 5. Closed-loop correction mechanism breakthrough: A dual correction mechanism of regular cycle + emergency trigger is designed to solve the problem of static lag in existing solutions and achieve dynamic system adaptation. This system, through deep collaboration among multiple modules, achieves a leap from "single-dimensional" to "multi-objective collaborative" optimization of wind-solar-storage ratio. It possesses promising application prospects and market promotion value.

Claims

1. A wind power, photovoltaic, and energy storage ratio intelligent coordination and optimization system, characterized in that: It includes a data acquisition module, a power grid command receiving module, an equipment life cycle health status monitoring module, a wind and solar power output prediction confidence calculation module, a wind, solar and energy storage flexible load multi-subject game optimization module, and an energy storage allocation execution module; The data acquisition module collects operating data from wind power equipment, photovoltaic equipment, energy storage systems, and flexible loads. The power grid instruction receiving module receives dynamic instructions from the power grid dispatch center; The equipment lifecycle health status monitoring module acquires health status parameters of wind power equipment, photovoltaic equipment, and energy storage systems. The wind and solar power output prediction confidence calculation module generates wind power output prediction results, photovoltaic power output prediction results, and corresponding prediction confidence levels; The wind-solar-storage flexible load multi-entity game optimization module takes the collected data from the data acquisition module, the dynamic instructions from the grid instruction receiving module, the health status parameters from the equipment life cycle health status monitoring module, and the prediction results from the wind and solar output prediction confidence calculation module as inputs to construct a non-cooperative game model of wind power entity, photovoltaic entity, energy storage entity and flexible load entity, and solves the initial wind-solar-storage ratio scheme. The energy storage allocation execution module allocates power between the cascaded batteries and newly added energy storage in the energy storage system based on the prediction confidence of the wind and solar power output prediction confidence calculation module and the initial wind, solar and energy storage allocation scheme, thereby realizing dynamic allocation optimization of wind power, photovoltaic and energy storage.

2. The intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratio as described in claim 1, characterized in that: The power grid command receiving module communicates with the power grid dispatch center using the IEC 61850 protocol. The received dynamic commands include real-time electricity prices, frequency support requirements, and reserve capacity requirements. The power grid command receiving module also includes a command parsing unit, which converts the dynamic commands into target weights for the wind, solar, and energy storage flexible load multi-stakeholder game optimization module. When receiving frequency support requirement commands, the grid connection stability weight is increased; when receiving real-time peak price commands, the cost optimization weight is increased.

3. The intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratio as described in claim 1, characterized in that: The equipment lifecycle health status monitoring module includes a multi-equipment health parameter acquisition unit and a health status assessment unit. The multi-equipment health parameter acquisition unit collects bearing vibration data of wind power equipment, component operation data of photovoltaic equipment, and cell voltage and temperature data of energy storage system. The health status assessment unit uses the CNN-LSTM algorithm to process the collected data and generate health status parameters for wind power equipment, photovoltaic equipment, and energy storage system. These health status parameters include health level and remaining life prediction values. The health status assessment unit also constructs an equipment loss cost model, correlates the health status parameters with equipment loss costs, and outputs the results to the wind-solar-storage flexible load multi-agent game optimization module.

4. The intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratio as described in claim 1, characterized in that: The wind and solar power output prediction confidence calculation module includes a prediction data input unit and a confidence calculation unit. The prediction data input unit acquires numerical weather forecast data and historical operating data of wind and solar equipment. The confidence calculation unit processes the input data using a random forest algorithm to generate wind power output prediction results and photovoltaic power output prediction results, and calculates the corresponding prediction confidence. This prediction confidence reflects the reliability of the prediction results and is output to the wind-solar-storage flexible load multi-agent game optimization module and the energy storage ratio execution module.

5. The intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratio as described in claim 1, characterized in that: The wind-solar-storage flexible load multi-agent game optimization module performs the following steps: S1. Determine the game agents, which include wind power agents, photovoltaic agents, energy storage agents, and flexible load agents, among which flexible load agents include industrial adjustable loads, charging pile loads, and building HVAC loads; S2. Acquire input data, which includes data collected by the data acquisition module, target weights converted by the grid command receiving module, health status parameters and loss cost model data from the equipment life cycle health status monitoring module, and prediction results from the wind and solar power output prediction confidence calculation module; S3. Construct the game objective function, which aims to maximize total revenue, including green electricity revenue, grid ancillary service revenue, electricity purchase cost deduction, and equipment loss cost deduction; S4. Solve the game objective function using a Nash equilibrium solver to generate an initial wind-solar-storage allocation scheme, which includes the wind power output allocation ratio, the photovoltaic power output allocation ratio, and the total charging and discharging power of the energy storage system.

6. The intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratio as described in claim 5, characterized in that: In S3, when constructing the game objective function, the health status parameters output by the equipment life cycle health status monitoring module are used as constraints. When the health status of the energy storage system is lower than the preset threshold, the charging and discharging power ratio of the energy storage system is reduced. When the health status of wind power equipment or photovoltaic equipment is lower than the preset threshold, the output allocation ratio of the corresponding equipment is adjusted.

7. The intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratio as described in claim 1, characterized in that: The energy storage allocation execution module includes an energy storage type identification unit and a power allocation unit. The energy storage type identification unit identifies the secondary batteries and newly added energy storage in the energy storage system. The power allocation unit allocates the total charging and discharging power of the energy storage system in the initial wind-solar-storage allocation scheme according to the prediction confidence level output by the wind and solar power output prediction confidence level calculation module. When the prediction confidence level is high, the power allocation ratio of the secondary batteries is increased. When the prediction confidence level is low, the power allocation ratio of the newly added energy storage is increased.

8. The intelligent coordination and optimization system for wind power, photovoltaic, and energy storage ratio as described in claim 1, characterized in that: It also includes a closed-loop correction module, which acquires real-time data from the data acquisition module, the latest dynamic instructions from the grid instruction receiving module, the latest health status parameters from the equipment life cycle health status monitoring module, and the latest prediction results from the wind and solar power output prediction confidence calculation module according to a preset cycle. The input is then fed into the wind-solar-storage flexible load multi-subject game optimization module to regenerate the initial wind-solar-storage ratio scheme. The power allocation is then adjusted through the energy storage ratio execution module to achieve dynamic correction of the wind power-solar-storage ratio.

9. A wind power, photovoltaic, and energy storage ratio intelligent coordination and optimization system as described in any one of claims 1-8, characterized in that: The system is applied to a grid-connected microgrid in an industrial and commercial park. The power supply or distribution circuit devices in the microgrid include a microgrid controller and a grid-connected interface device. The energy storage system includes a cascaded battery pack and a newly added energy storage battery pack. Both the cascaded battery pack and the newly added energy storage battery pack are connected to the microgrid controller through an energy storage converter and receive power allocation instructions from the energy storage ratio execution module.

Images