Frequency modulation method for large-scale new energy grid-connected system based on neural network

Abstract

The invention provides a frequency modulation method for a large-scale new energy grid-connected system based on a neural network. The frequency modulation method comprises the following steps: S1, installing an auxiliary energy storage unit and a prediction and control unit in the new energy grid-connected system; s2, the prediction and control unit performs prediction-based data acquisition on the energy storage unit, the auxiliary energy storage unit, the grid-connected unit and the distributed power generation unit; s3, on the basis of a neural network, predicting wind power, optical power and grid-connected bus frequency at future moments; s4, optimizing the energy storage SOC, the charging and discharging efficiency and the charging and discharging power; and S5, performing cooperative control of charging and discharging according to the real-time operation states of the energy storage unit and the auxiliary energy storage unit. The frequency modulation requirement under large-scale new energy access is met, comprehensive cooperation of the energy storage system and efficient utilization of stored energy are achieved, the overall performance and intelligence of the new energy grid-connected system can be effectively improved, and meanwhile the service life of the energy storage system is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of renewable energy power generation, and in particular to a frequency modulation method for a large-scale renewable energy grid-connected system based on a neural network. Background Art

[0002] In the process of accelerating energy transformation around the world, although renewable energy power generation is booming, the inherent characteristics of distributed power generation devices (wind turbines, photovoltaics, diesel generators, etc.) have led to severe challenges in the frequency regulation of power systems. Current frequency regulation methods have many limitations, such as the low efficiency of energy storage systems caused by the charging and discharging mode of traditional energy storage batteries, the complex structure of power compensation devices, slow response and easy to cause over-compensation, and prominent aging problems of energy storage systems. Although some relevant technical explorations have been made, none of them have been able to comprehensively and systematically solve the diverse and complex needs of frequency regulation of renewable energy grid-connected systems, and have been unable to effectively deal with frequency stability issues in large-scale renewable energy grid-connected scenarios.

[0003] The current mainstream independent energy storage power station that uses lithium batteries as energy storage application batteries and combines high-voltage cascade technology for frequency modulation scenarios, due to the unbalanced charging and discharging characteristics of lithium batteries themselves, high-voltage cascade technology further amplifies the problem, and frequent daily charging and discharging exacerbates the extremely unbalanced state of energy storage SOC. The passive balancing method used by traditional energy storage manufacturers has little effect on improving this problem, which seriously affects the performance and life of the energy storage system and the frequency modulation effect of the entire new energy grid-connected system. Innovative solutions are urgently needed to overcome this key technical bottleneck. Summary of the invention

[0004] In order to solve the current technical problems, the main purpose of the present invention is to provide a frequency modulation method for a large-scale new energy grid-connected system based on a neural network. Aiming at the problem of large-scale new energy access, by introducing auxiliary energy storage units and prediction and control units, not only the comprehensive coordination and efficient utilization of energy storage systems can be achieved, but also the overall performance and intelligence of the new energy grid-connected system can be effectively improved, while extending the service life of the energy storage system.

[0005] The technical solution adopted by the present invention is: a frequency modulation method for a large-scale new energy grid-connected system based on a neural network, comprising the following steps: S1. Install an auxiliary energy storage unit and a prediction and control unit in the new energy grid-connected system, and connect the prediction and control unit with the auxiliary energy storage unit, and the distributed generation unit, energy storage unit and grid-connected unit in the system through control; S2, the prediction and control unit implements prediction-based data collection on the energy storage unit, the auxiliary energy storage unit, the grid-connected unit and the distributed generation unit; S3, the prediction unit predicts the wind, light power and grid-connected bus frequency at future times based on the collected data and neural network; S4. Optimize the energy storage SOC, charging and discharging efficiency, and charging and discharging power according to the prediction results; S5. Perform coordinated control of charging and discharging according to the real-time operating status of the energy storage unit and the auxiliary energy storage unit.

[0006] In S1, the energy storage unit is responsible for providing buffering for the electric energy to be connected to the common bus.

[0007] In S1, the auxiliary energy storage unit is responsible for providing buffering for electric energy to be incorporated into the public bus. The capacity of the auxiliary energy storage unit is smaller than the capacity of the energy storage unit. When the energy storage unit cannot cope with bus fluctuations, it is supplemented by the auxiliary energy storage unit.

[0008] In S1, the grid-connected unit is responsible for connecting to the common bus and providing line support and current path for the common bus when the power system operates normally.

[0009] In S1, the distributed power generation unit includes a wind power generation device and a photovoltaic power generation device, and outputs the power regulation amount of the power grid through the wind power generation device and the photovoltaic power generation device respectively.

[0010] In S2, prediction-based data collection includes: Total power, output voltage, output current, remaining capacity of energy storage unit, charging and discharging efficiency, charging and discharging power and grid connection point frequency.

[0011] In S3, based on the neural network, the prediction methods for wind, light power, and grid-connected bus frequency at future moments include: A neural network model is established and trained through the collected data. The grid connection point frequency collected in real time is used as the feedback signal during training. The wind and light intensity at the next time point are predicted based on the current time point. The wind and light power at the next time point are predicted based on the current time point, as well as the bus frequency at the next time point.

[0012] Methods for training neural network models using collected data include: Set the training time step Td, and use the grid connection point frequency, wind, light intensity, wind, optical power, and energy storage SOC value at the current time points t0, t0+1, and t0+Td as the input vectors Xt0, Xt0+1, and Xt0+Td. Use the grid connection point frequency as the ideal output value, and predict the frequency value Yt0 of the grid connection point at time ti. Compare the frequency value Yt0 of the grid connection point at time ti with the actual value ei collected to obtain the error E. Correct the model parameters according to the error E, update the input vector, increase the training time step by one, and conduct the next training until the error E is less than the set value. After that, use the trained neural network to predict the frequency value Yt0 at time ti and the wind and optical power corresponding to time ti according to the collected data.

[0013] In S4, the optimization method for energy storage SOC, charging and discharging efficiency, and charging and discharging power includes: Based on the wind and light intensity, the predicted value of the grid connection point frequency in the future period is compared with the grid connection set frequency value. If the grid connection point frequency is lower than the set frequency during this period, the energy storage unit releases electricity when the power is sufficient. If the energy storage SOC is higher than the set value, or lower than the set value, the charging and discharging power is reduced to improve the charging and discharging efficiency. If the energy storage SOC is not lower than the set value, the current charging and discharging state is maintained.

[0014] In S5, the method for coordinated control of charging and discharging includes: If the frequency is monitored to be out of range, it is determined whether the charging and discharging power is zero. If the charging and discharging power is zero, the charging and discharging of the energy storage unit is started and the output is increased. If the charging and discharging power is not zero, the power generation power of the distributed generation unit is gradually adjusted to the minimum safe operating level or an orderly shutdown operation is performed, and then the charging and discharging coefficient of the energy storage unit is adjusted.

[0015] The present invention has the following beneficial effects: The present invention meets the frequency regulation requirements under large-scale new energy access, not only realizing the comprehensive coordination of the energy storage system and the efficient utilization of energy storage, but also effectively improving the overall performance and intelligence of the new energy grid-connected system, while extending the service life of the energy storage system. The prediction data is provided by the prediction and control unit, and the fine coordinated control of the frequency regulation response and frequency regulation is realized in combination with the real-time operating status of the system. Under the premise of ensuring that the system has no power limit exceeding, the optimal control strategy is obtained through the system optimization algorithm to minimize the frequency regulation cost, maximize the service life of energy storage, and maximize the output power of distributed power generation devices, which significantly reduces the imbalance of the energy storage system during the frequent frequency regulation cycle adjustment process and the problem of energy storage loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a system diagram of the present invention.

[0018] Figure 2 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] like Figure 1 , 2 As shown, the present invention provides a frequency modulation method for a large-scale new energy grid-connected system based on a neural network, comprising the following steps: S1. Install an auxiliary energy storage unit and a prediction and control unit in the new energy grid-connected system, and connect the prediction and control unit with the auxiliary energy storage unit, and the distributed generation unit, energy storage unit and grid-connected unit in the system by wired or wireless means; S2, the prediction and control unit implements prediction-based data collection on the energy storage unit, the auxiliary energy storage unit, the grid-connected unit and the distributed generation unit; S3, the prediction unit predicts the wind, light power, and grid-connected bus frequency at future times based on the data collected in S2 and based on the neural network; S4. Optimize the energy storage SOC, charging and discharging efficiency, and charging and discharging power according to the prediction results; S5. Perform coordinated control of charging and discharging according to the real-time operating status of the energy storage unit and the auxiliary energy storage unit.

[0021] By introducing auxiliary energy storage units and prediction and control units, the present invention meets the frequency regulation requirements under large-scale renewable energy access, which not only realizes the comprehensive coordination of energy storage systems and efficient utilization of energy storage, but also can effectively improve the overall performance and intelligence of renewable energy grid-connected systems, while extending the service life of energy storage systems.

[0022] In S1, the energy storage unit is responsible for providing buffer for the electric energy to be connected to the public bus. The capacity of the energy storage unit is 100 kWh, or other values, and the charging and discharging efficiency is 90%.

[0023] In S1, the auxiliary energy storage unit is responsible for providing buffering for the electric energy to be incorporated into the public bus. The capacity of the auxiliary energy storage unit is smaller than that of the energy storage unit, the charging and discharging efficiency is 90%, and the charging and discharging coefficient is 20%. It can cope with short-term bus fluctuations. When the energy storage unit cannot cope with bus fluctuations, it is supplemented by the auxiliary energy storage unit.

[0024] In S1, the grid-connected unit is responsible for connecting to the common bus and providing line support and current path for the common bus when the power system is operating normally. Specifically, the grid-connected unit includes a step-up transformer and an inverter.

[0025] The energy storage unit and the auxiliary energy storage unit are provided with corresponding energy storage status monitoring units. The prediction and control unit obtains the charging and discharging status information of the energy storage unit and the auxiliary energy storage unit through the energy storage status monitoring units, and selects the energy storage unit to cooperate in responding to the predicted grid-connected power according to the charging and discharging status information of the energy storage unit and the auxiliary energy storage unit.

[0026] In S1, the distributed power generation unit includes a wind power generation device and a photovoltaic power generation device, and outputs the power regulation amount of the power grid through the wind power generation device and the photovoltaic power generation device respectively.

[0027] In S2, prediction-based data collection includes: Total power, output voltage, output current, remaining capacity of energy storage unit, charging and discharging efficiency, charging and discharging power and grid connection point frequency.

[0028] In S3, based on the neural network, the prediction methods for wind, light power, and grid-connected bus frequency at future moments include: A neural network model is established and trained through the collected data. The grid connection point frequency collected in real time is used as the feedback signal during training. The wind and light intensity at the next time point are predicted based on the current time point. The wind and light power at the next time point are predicted based on the current time point, as well as the bus frequency at the next time point.

[0029] Methods for training neural network models using collected data include: The neural network model adopts BP neural network, sets the training time step Td, and uses the grid connection point frequency, wind, light intensity, wind, optical power, and energy storage SOC value at the current time points t0, t0+1, and t0+Td as input vectors Xt0, Xt0+1, and Xt0+Td. The grid connection point frequency is used as the ideal output value, and the frequency value Yt0 of the grid connection point at time ti is predicted. The frequency value Yt0 of the grid connection point at time ti is compared with the actual value ei collected to obtain the error E. The model parameters are corrected according to the error E, the input vector is updated, the training time step is increased by one, and the next training is performed until the error E is less than the set value. After that, the neural network after the training is used to predict the frequency value Yt0 at time ti and the wind and optical power corresponding to time ti according to the collected data.

[0030] Among them, the prediction time scale is 10 minutes, 30 minutes, and 1 hour.

[0031] In S4, the optimization method for energy storage SOC, charging and discharging efficiency, and charging and discharging power includes: Based on the wind and light intensity, the predicted value of the grid connection point frequency in the future period is compared with the grid connection set frequency value. If the grid connection point frequency is lower than the set frequency during this period, the energy storage unit releases electricity when the power is sufficient. If the energy storage SOC is higher than the set value, or lower than the set value, the charging and discharging power is reduced to improve the charging and discharging efficiency. If the energy storage SOC is not lower than the set value, the current charging and discharging state is maintained.

[0032] SOC is the percentage of the remaining battery power to the rated capacity of the battery.

[0033] Set the frequency to 50Hz±0.2Hz.

[0034] Accurate wind and light intensity monitoring data, using neural network models, deduce the changing trend of grid connection point frequency within a specific time period in the future, and then obtain the corresponding prediction value. This prediction value is then compared with the grid connection standard frequency threshold pre-set by the system.

[0035] Once it is determined that the frequency of the grid connection point is lower than the set standard during the forecast period, the system's attention is immediately focused on the power reserve details of the energy storage unit. First, confirm whether the energy storage unit has sufficient power reserves to implement the discharge operation. The basis for this judgment is the state of charge value of the energy storage unit. Only when the power level represented by the energy storage SOC can meet the current discharge demand while fully taking into account the power stability and continuity of the subsequent operation of the system, and will not cause subsequent power supply shortages or cause system operation risks due to this discharge, will the discharge process of the energy storage unit be started to achieve timely replenishment of power to the power grid, thereby having a positive effect on raising the low power grid frequency.

[0036] After the discharge program is officially started, the system enters the stage of real-time dynamic monitoring and fine-tuning of the energy storage SOC. When it is detected that the energy storage SOC has climbed above a specific high threshold, such as 90%, in view of the series of negative effects that may be caused by the battery continuing to discharge at a high current under a high SOC state, such as accelerating the imbalance of the internal chemical reaction of the battery, aggravating the aging of the battery plate, and shortening the overall cycle life of the battery, the system will quickly activate the control mechanism to automatically and accurately reduce the charge and discharge power. By finely adjusting the charge and discharge current and voltage amplitude, while maintaining the basic frequency modulation function of the system, the energy loss in the charge and discharge conversion process is minimized, the charge and discharge efficiency is significantly improved, and the physical structure and chemical performance integrity of the battery are effectively protected, ensuring its stability and reliability during long-term operation.

[0037] When the energy storage SOC drops below a preset low threshold, such as 20%, the system will decisively start the power limit program to reduce the charge and discharge power, because excessive discharge of the battery at low power can easily cause irreversible sulfidation of the battery plates, a large loss of active substances, and a sharp increase in the battery's internal resistance, which seriously damages the battery's performance and life. With the help of the optimization control algorithm to reasonably adjust the current and voltage, the charging and discharging efficiency is improved, energy loss is reduced, and the battery is prevented from falling into a deep loss state due to excessive discharge, ensuring that the battery can maintain basic performance and capacity levels in subsequent charge and discharge cycles.

[0038] If the energy storage SOC is always maintained above the set low threshold during the discharge process, it indicates that the battery power is in a relatively ideal and stable range. In this case, the system will follow the principle of maintaining stable operation and maintain the current charge and discharge state unchanged. Such an operating strategy effectively avoids the adverse phenomena such as system voltage fluctuations, current shocks, and efficiency impairment caused by frequent and unnecessary switching of charge and discharge states, ensuring the stability, continuity and efficiency of the new energy grid-connected system during the frequency regulation operation, and providing a solid guarantee for the stable and reliable operation of the entire power system.

[0039] In S5, the method for coordinated control of charging and discharging includes: If the frequency is monitored to be out of range, it is determined whether the charging and discharging power is zero. If the charging and discharging power is zero, the charging and discharging of the energy storage unit is started and the output is increased. If the charging and discharging power is not zero, the power generation power of the distributed generation unit is gradually adjusted to the minimum safe operating level or an orderly shutdown operation is performed, and then the charging and discharging coefficient of the energy storage unit is adjusted.

[0040] When the grid frequency is detected to be beyond the preset safety range, the charging and discharging power of the energy storage unit is first accurately monitored and judged. If the charging and discharging power is zero, after comprehensively evaluating the charge state and health state of the energy storage unit and the specific value and change rate of the current grid frequency deviation, if it is determined that the energy storage unit has safe and efficient charging and discharging conditions, the charging and discharging of the energy storage unit is started, and according to the performance characteristic curve of the energy storage unit and the grid frequency recovery requirements, its output is gradually increased at an appropriate rate to ensure that while stabilizing the grid frequency, it will not cause excessive impact or damage to the energy storage unit.

[0041] If the charge and discharge power is not zero, a comprehensive analysis of the operating status of the distributed generation unit is given priority, including its current power generation, power generation efficiency, equipment stability, and degree of coupling with the power grid. According to the analysis results, if the continued operation of the distributed generation unit may have a serious adverse impact on the stability of the power grid frequency, the power generation of the distributed generation unit is gradually adjusted to the minimum safe operating level or an orderly shutdown operation is performed under the premise of following the safety shutdown specifications and optimized control strategies. In this process, the energy storage unit is dynamically evaluated simultaneously, and combined with its charge state, health status, current charge and discharge power, and real-time changes in the power grid frequency, the intelligent control algorithm is used to accurately calculate and reasonably adjust the charge and discharge coefficient to achieve effective regulation of the power storage unit on the power grid frequency, and to ensure that the energy storage unit is always in a safe, stable and efficient operating range, avoiding the performance degradation of the energy storage unit or the increase of the risk of failure due to excessive regulation.

Claims

1. A frequency modulation method for a large-scale new energy grid-connected system based on a neural network, characterized in that: The following steps are involved: S1. Install an auxiliary energy storage unit and a prediction and control unit in the new energy grid-connected system, and control and connect the prediction and control unit with the auxiliary energy storage unit, as well as the distributed generation unit, energy storage unit and grid-connected unit in the system; S2, the prediction and control unit implements prediction-based data collection on the energy storage unit, the auxiliary energy storage unit, the grid-connected unit and the distributed generation unit; S3, the prediction unit predicts the wind, light power and grid-connected bus frequency at future times based on the collected data and neural network; S4. Optimize the energy storage SOC, charging and discharging efficiency, and charging and discharging power according to the prediction results; S5. Perform coordinated control of charging and discharging according to the real-time operating status of the energy storage unit and the auxiliary energy storage unit.

2. According to claim 1, a method for frequency modulation of a large-scale new energy grid-connected system based on a neural network is characterized in that: In S1, the energy storage unit is responsible for providing buffering for the electric energy to be connected to the common bus.

3. According to the frequency modulation method of a large-scale new energy grid-connected system based on a neural network as described in claim 1, it is characterized in that: In S1, the auxiliary energy storage unit is responsible for providing buffering for electric energy to be incorporated into the public bus. The capacity of the auxiliary energy storage unit is smaller than the capacity of the energy storage unit. When the energy storage unit cannot cope with bus fluctuations, it is supplemented by the auxiliary energy storage unit.

4. According to the method of frequency modulation of large-scale new energy grid-connected system based on neural network in claim 1, it is characterized in that: In S1, the grid-connected unit is responsible for connecting to the common bus and providing line support and current path for the common bus when the power system operates normally.

5. The method for frequency modulation of a large-scale new energy grid-connected system based on a neural network according to claim 1, characterized in that: In S1, the distributed power generation unit includes a wind power generation device and a photovoltaic power generation device, and outputs the power regulation amount of the power grid through the wind power generation device and the photovoltaic power generation device respectively.

6. The method for frequency modulation of a large-scale new energy grid-connected system based on a neural network according to claim 1, characterized in that: In S2, prediction-based data collection includes: Total power, output voltage, output current, remaining capacity of energy storage unit, charging and discharging efficiency, charging and discharging power and grid connection point frequency.

7. The method for frequency modulation of a large-scale new energy grid-connected system based on a neural network according to claim 1, characterized in that: In S3, based on the neural network, the prediction methods for wind, light power, and grid-connected bus frequency at future moments include: A neural network model is established and trained through the collected data. The grid connection point frequency collected in real time is used as the feedback signal during training. The wind and light intensity at the next time point are predicted based on the current time point. The wind and light power at the next time point are predicted based on the current time point, as well as the bus frequency at the next time point.

8. A method for frequency modulation of a large-scale new energy grid-connected system based on a neural network according to claim 7, characterized in that: Methods for training neural network models using collected data include: Set the training time step Td, and use the grid connection point frequency, wind, light intensity, wind, optical power, and energy storage SOC value at the current time points t0, t0+1, and t0+Td as the input vectors Xt0, Xt0+1, and Xt0+Td. Use the grid connection point frequency as the ideal output value, and predict the frequency value Yt0 of the grid connection point at time ti. Compare the frequency value Yt0 of the grid connection point at time ti with the actual value ei collected to obtain the error E. Correct the model parameters according to the error E, update the input vector, increase the training time step by one, and conduct the next training until the error E is less than the set value. After that, use the trained neural network to predict the frequency value Yt0 at time ti and the wind and optical power corresponding to time ti according to the collected data.

9. The method for frequency modulation of a large-scale new energy grid-connected system based on a neural network according to claim 1, characterized in that: In S4, the optimization method for energy storage SOC, charging and discharging efficiency, and charging and discharging power includes: Based on the wind and light intensity, the predicted value of the grid connection point frequency in the future period is compared with the grid connection set frequency value. If the grid connection point frequency is lower than the set frequency during this period, the energy storage unit releases electricity when the power is sufficient. If the energy storage SOC is higher than the set value, or lower than the set value, the charging and discharging power is reduced to improve the charging and discharging efficiency. If the energy storage SOC is not lower than the set value, the current charging and discharging state is maintained.

10. The method for frequency modulation of a large-scale new energy grid-connected system based on a neural network according to claim 1, characterized in that: In S5, the method for coordinated control of charging and discharging includes: If the frequency is monitored to be out of range, it is determined whether the charging and discharging power is zero. If the charging and discharging power is zero, the charging and discharging of the energy storage unit is started and the output is increased. If the charging and discharging power is not zero, the power generation power of the distributed generation unit is gradually adjusted to the minimum safe operating level or an orderly shutdown operation is performed, and then the charging and discharging coefficient of the energy storage unit is adjusted.

Images