A distributed energy power control method and system

Abstract

The application discloses a kind of distributed energy power control method and system, it is related to power control technical field, including the following steps: obtaining and based on first data constructs multilayer frequency deviation index, constructs frequency response urgency feature;Based on reinforcement learning model dynamic adjustment charge-discharge parameter, according to frequency response urgency feature obtains power adjustment period;According to power adjustment period, in combination with the remaining power demand of electric vehicle and the urgency of charging cutoff time, quantize user demand constraint as charging priority;According to power grid frequency stability, charging priority and energy cost establishes distributed optimization objective function, and obtains the optimal power distribution scheme by genetic algorithm solution;Based on the optimal power distribution scheme, the state of each node is synchronized, and the charging and discharging power of each electric vehicle in the region is distributed.The application can quickly and reasonably allocate electric vehicle charging and discharging power, effectively suppress frequency deviation, and significantly improve the stability and reliability of power grid.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power control, more particularly, the present application relates to a distributed energy power control method and system. BACKGROUND

[0002] With the transformation of global energy structure and the rapid development of renewable energy, distributed energy power control methods play a crucial role in modern energy systems. With the widespread application of distributed energy such as solar energy, wind energy and other renewable energy, as well as the development of new power grid forms such as microgrid and smart grid, the proportion of distributed energy (such as solar photovoltaic, wind power, energy storage system, etc.) in the power system is gradually increasing. Distributed energy has the characteristics of decentralization, flexibility and renewable energy, which brings new vitality and opportunities to the power system. However, the intermittency and volatility of distributed energy also bring challenges to the stable operation of the power grid.

[0003] In order to ensure that distributed energy can be safely, stably and efficiently connected to the power grid, and realize the coordinated operation with traditional energy, power control technology becomes the key. Power control technology can accurately control the output power of distributed energy, so that it can meet the demand of power grid while maintaining the stability and reliability of the system.

[0004] Distributed energy power control technology involves many aspects, including the selection of control strategy, the design of control algorithm, the implementation of control system, etc. Different types of distributed energy and application scenarios have different requirements for power control technology. Therefore, researchers need to conduct in-depth research according to the specific situation, and develop power control technology suitable for specific application scenarios.

[0005] Under the background of existing technology, distributed energy power control faces many technical challenges. The output of distributed energy has significant volatility and uncertainty, such as wind power affected by wind speed changes, photovoltaic power affected by light intensity and weather conditions, which leads to the instability of energy supply. At the same time, with the increasing penetration rate of distributed energy in the power grid, the direction of power flow becomes more complex and variable, and the traditional one-way power flow control strategy cannot meet the actual demand.

[0006] For example, the invention patent announcement No. CN112394704B announces a power control strategy diagnosis method and device for new energy power station, which comprises: testing each power control parameter of the new energy power station in each test period; using the test results to diagnose the power control strategy of the new energy power station, and warning based on the diagnosis results. The power control strategy diagnosis method and device for new energy power station provided by the present application extracts the power control parameters of the new energy power station from the massive and complex new energy power station operation data, and tests them based on a set of calculation system. The test results are used to diagnose the power control strategy of the new energy power station, which can realize the early warning of the new energy power station, and provide technical support for the safe and stable operation of the power grid and the priority scheduling of the new energy power station.

[0007] For example, the invention patent announcement No. CN116736708A announces a power control method for photovoltaic hydrogen production system based on fuzzy logic system, which comprises the following steps: S1, obtaining the battery charge state at the previous moment and the current photovoltaic output, inputting them into the fuzzy logic system, determining the operation mode of the photovoltaic hydrogen production system, and obtaining the current electrolytic cell and battery power command through the operation mode of the photovoltaic hydrogen production system; S2, based on the electrolytic cell and battery power command obtained in step S1, a first-order low-pass filter algorithm is used to determine the power values of the electrolytic cell, battery and super capacitor. This method considers the power control calculation time, and also needs to consider avoiding the high start-stop frequency of the electrolytic unit module to ensure the safety and service life of the electrolytic cell, and realizes real-time power control of the photovoltaic hydrogen production system.

[0008] The above disclosed technical solutions have at least the following technical problems: centralized control depends on central node communication, has single point failure risk and slow response speed, lacks real-time compensation mechanism for power grid frequency fluctuation, resulting in high frequency response delay and inability to provide fast frequency support for power grid. In view of the above problems, the present application provides a solution. SUMMARY

[0009] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a distributed energy power control method and system, which establishes a frequency response urgency feature by using multiple frequency deviation indexes, and acquires a power adjustment period according to the frequency response urgency feature, so as to solve the problem of high frequency response delay and inability to provide fast frequency support for power grid.

[0010] In order to achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0011] The application discloses a distributed energy power control method, comprising the following steps: acquiring and constructing a multi-layer frequency deviation index based on first data, the multi-layer frequency deviation index comprising an instantaneous frequency fluctuation rate, a minute-level average deviation and a regional integral deviation, and adopting a sliding window to adaptively adjust the weights of the three types of indexes to construct a frequency response urgency feature; dynamically adjusting charging and discharging parameters based on a reinforcement learning model, acquiring a power adjustment period according to the frequency response urgency feature; quantifying user demand constraints into charging priorities according to the power adjustment period, the residual power demand of an electric vehicle and the urgency of a charging cutoff time; establishing a distributed optimization objective function according to the power frequency stability, the charging priorities and the energy cost, and obtaining an optimal power distribution scheme by solving the objective function through a genetic algorithm, the genetic algorithm taking the initial value of the charging and discharging parameters output by the reinforcement learning model as an iteration starting point; and synchronizing the states of all nodes based on the optimal power distribution scheme and distributing the charging and discharging power of all electric vehicles in the region.

[0012] In a preferred embodiment, the multi-layer frequency deviation index comprises an instantaneous frequency fluctuation rate; and the method for acquiring the instantaneous frequency fluctuation rate comprises: taking a dynamic time interval linked with the frequency response urgency feature as a sampling period, and acquiring the power grid frequency value at each sampling period; and calculating the instantaneous frequency fluctuation rate based on the power grid frequency value, the calculation method of the instantaneous frequency fluctuation rate being the ratio of the absolute value of the difference between the frequency of the current sampling period and the frequency of the previous period to the frequency of the previous period.

[0013] In a preferred embodiment, the construction of the frequency response urgency feature comprises: analyzing historical power grid operation data to determine the influence degrees of the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation on the stability of the power grid in different value ranges; dividing the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation into three levels of low, medium and high according to the influence degrees; giving the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation initial weights according to the levels and adaptively adjusting the initial weights based on a sliding window; and calculating the frequency response urgency feature according to the calculation results of the indexes and the given weights.

[0014] In a preferred embodiment, the adaptive adjustment based on the sliding window comprises: inversely proportionally adjusting the weight of the instantaneous frequency fluctuation rate according to the current regional inertia level; the weight of the minute-level average deviation is positively correlated with the new energy penetration rate in a segmented linear manner; and when the regional integral deviation exceeds a preset threshold, the weight of the regional integral deviation is corrected by using an exponential decay.

[0015] In a preferred embodiment, the reinforcement learning model-based dynamic adjustment of the charging and discharging parameters specifically comprises: constructing a state space of the reinforcement learning model, defining a set of charging and discharging parameters of the electric vehicles as an action space, and setting a value range of the power; at each time step, the reinforcement learning model executes an action according to the current state, and stores the current state, the action, and the output of the reinforcement learning model in an experience replay buffer; random data are obtained from the experience replay buffer, the parameters of the reinforcement learning model are adjusted by minimizing a loss function, and a trained reinforcement learning model is obtained; before the start of each power adjustment period, the edge computing node obtains real-time state information, inputs the real-time state information into the trained reinforcement learning model, and outputs the charging and discharging parameters of each electric vehicle.

[0016] In a preferred embodiment, the power adjustment period is obtained according to the frequency response urgency characteristic, specifically comprising: establishing a mapping relationship between the frequency response urgency characteristic and the power adjustment period based on regression analysis; and obtaining the corresponding power adjustment period through the established mapping relationship according to the current frequency response urgency characteristic value.

[0017] In a preferred embodiment, the user demand constraint is quantified as a charging priority according to the power adjustment period, the urgency of the remaining power demand of the electric vehicle, and the charging cutoff time, specifically comprising: for each electric vehicle, the product of the difference between the target SOC and the current SOC and the battery capacity is taken as the remaining power demand of the electric vehicle; the charging cutoff time of each electric vehicle is obtained, and the charging priority is calculated based on the power adjustment period in combination with the remaining power demand of the electric vehicle.

[0018] In a preferred embodiment, the distributed optimization objective function is established according to the grid frequency stability, the charging priority, and the energy cost, and the optimal power distribution scheme is obtained by solving the genetic algorithm, specifically comprising: establishing a distributed optimization objective function according to the grid frequency stability, the charging priority, and the energy cost; generating a plurality of random power distribution schemes that satisfy the constraint conditions; dividing the solution space of the power distribution scheme into a short-term power rapid regulation layer and a long-term energy cost optimization layer, and constructing a first population and a second population, respectively; in the first population, a genetic operator with a high mutation probability is used to improve the rapid response ability to frequency disturbance; in the second population, a genetic operator with a low mutation probability and an elite preservation genetic operator are used to ensure the economic convergence performance; exchanging elite individuals at a preset generation interval and sharing information between different optimization layers;

[0019] When the convergence condition is reached, the optimal power distribution scheme is selected based on the comprehensive fitness value of the double population.

[0020] The technical effects and advantages of the distributed energy power control method and system of the present application are:

[0021] 1.The application can comprehensively and meticulously evaluate the power grid frequency condition by constructing multi-layer frequency deviation indicators. Based on these indicators, a frequency response urgency feature is established, and the weights are adaptively adjusted according to factors such as regional inertia level and new energy penetration rate, which can accurately judge the urgency of power grid frequency stability. Then, the power adjustment period is dynamically adjusted according to the urgency feature, so that the system can quickly and reasonably allocate the charging and discharging power of electric vehicles when facing different degrees of frequency fluctuation, effectively suppress the frequency deviation, significantly improve the stability and reliability of the power grid, and ensure the safe and stable operation of the power grid under complex working conditions.

[0022] 2.The application quantifies user demand constraints as charging priority through power adjustment period, and establishes a distributed optimization objective function based on this. The optimal power distribution scheme is solved by optimization methods such as genetic algorithm. In the execution process, each node makes collaborative decisions based on synchronization signals and local information, which can reasonably allocate charging and discharging power under the constraints of meeting power balance and battery capacity limitations, etc., not only meeting user charging demand, but also reducing energy cost, achieving efficient optimization of distributed energy, and improving the overall efficiency of the entire energy system. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A flowchart of a distributed energy power control method according to the application is shown.

[0024] Figure 2 A structure diagram of a distributed energy power control system according to the application is shown. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0026] Embodiment 1, Figure 1 A distributed energy power control method according to the application is given, including the following steps:

[0027] S1, acquire and construct multi-layer frequency deviation indicators based on first data, the multi-layer frequency deviation indicators include instantaneous frequency fluctuation rate, minute-level average deviation and regional integral deviation, and the weights of the three types of indicators are adaptively adjusted using a sliding window to construct a frequency response urgency feature;

[0028] The first data includes power grid frequency deviation dynamic price signal, state of charge (SOC) of each electric vehicle in the region, and charging cutoff time;

[0029] The first data is collected in real time by an edge computing node, specifically:

[0030] According to the coverage area of the power grid and the distribution of electric vehicles, edge computing nodes are deployed. The edge computing nodes need to have sufficient computing power, storage capacity, and communication capabilities to process and store the collected data in real time. The edge computing nodes are connected to power grid monitoring devices (such as frequency sensors, electricity meters, etc.) and electric vehicle charging facilities to ensure smooth data transmission.

[0031] According to the pre-set collection period (for example, once every second), the edge computing node obtains the power grid frequency data from the frequency sensor.

[0032] The collected frequency data is processed in real time to calculate the frequency deviation. The calculation method of the frequency deviation is the difference between the actual measured frequency and the rated frequency of the power grid (such as 50Hz or 60Hz).

[0033] The calculated frequency deviation data is filtered to remove noise and interference, improving the quality of the data.

[0034] The dynamic electricity price signal contains electricity price information for different time periods.

[0035] The edge computing node sends a data request instruction to the charging facility, requesting the SOC and charging cutoff time of the electric vehicle.

[0036] After receiving the request, the charging facility reads the SOC and charging cutoff time from the battery management system of the electric vehicle and returns this information to the edge computing node.

[0037] The SOC (State of Charge) of an electric vehicle refers to the state of charge of the battery, which is the remaining amount of power in the vehicle's power battery. It is the percentage of the remaining available power relative to the total power, usually presented in percentage form, ranging from 0% to 100%.

[0038] The multi-layer frequency deviation index includes instantaneous frequency fluctuation rate, minute-level average deviation, and regional integral deviation.

[0039] The method for obtaining the instantaneous frequency fluctuation rate is as follows:

[0040] The dynamic time interval linked with the frequency response urgency feature is used as the sampling period (the higher the urgency, the shorter the sampling period). The power grid frequency value at each sampling period is obtained. Based on the power grid frequency value, the instantaneous frequency fluctuation rate is calculated. The calculation method of the instantaneous frequency fluctuation rate is the ratio of the absolute value of the difference between the current sampling period frequency and the previous period frequency to the previous period frequency.

[0041] The method for obtaining the minute-level average deviation is specifically:

[0042] The minute-level average deviation is calculated by obtaining the average deviation of the frequency and the rated frequency in the past 1 minute.

[0043] The regional integral deviation is specifically:

[0044] The time integral of the frequency deviation in the past 1 hour reflects the long-term cumulative frequency deviation.

[0045] The instantaneous frequency fluctuation rate is specifically:

[0046]

[0047] The minute-level average deviation is specifically:

[0048]

[0049] The regional integral deviation is specifically:

[0050]

[0051] wherein, is the instantaneous frequency fluctuation rate, is the absolute value of the frequency difference between the current sampling period and the previous period, reflecting the absolute fluctuation amplitude of the frequency, is the frequency value of the previous period, is the minute-level average deviation, is the power grid frequency value at the i-th moment, is the rated frequency of the power grid, is the regional integral deviation, is the power grid frequency in one hour, and x is the time variable.

[0052] The frequency response urgency feature is specifically established as:

[0053] The historical power grid operation data is analyzed to determine the influence degree of the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation on the power grid stability in different value ranges;

[0054] According to the influence degree, the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation are divided into three levels of low, medium and high;

[0055] The initial weights are assigned to the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation according to the levels, and the sliding window is adaptively adjusted;

[0056] According to the calculation results of each index and the assigned weights, the frequency response urgency feature is calculated.

[0057] The adaptive adjustment based on the sliding window is specifically:

[0058] The weight of the instantaneous frequency fluctuation rate is inversely adjusted according to the current regional inertia level;

[0059] The weight of the minute-level average deviation is positively related to the new energy penetration rate in a segmented linear manner;

[0060] For the weight of the regional integral deviation, when the regional integral deviation exceeds a preset threshold, an exponential decay correction is enabled.

[0061] For example, the instantaneous frequency fluctuation rate is set to be less than 0.05 Hz / s for a low level, 0.05-0.1 Hz / s for a medium level, and greater than 0.1 Hz / s for a high level.

[0062] Minute-level average deviation: Based on historical data, the minute-level average deviation is evaluated to grade the impact on short-term operation of the power grid, such as less than 0.1 Hz for a low level, 0.1-0.5 Hz for a medium level, and greater than 0.5 Hz for a high level.

[0063] Regional integral deviation: The long-term cumulative deviation of the power grid frequency is studied to grade the impact on the overall performance of the power grid, such as less than 0.5 Hz / h for a low level, 0.5-2 Hz / h for a medium level, and greater than 2 Hz / h for a high level.

[0064] The sliding window adaptive adjustment mechanism satisfies the following constraint conditions:

[0065]

[0066] wherein, , and are the adjusted weights, , and are the initial weights of the instantaneous frequency fluctuation rate, the minute-level average deviation, and the regional integral deviation, is the regional reference inertia value, is the current regional inertia level, is the new energy influence factor, is the new energy penetration rate, is the maximum design value of the new energy penetration rate, is the preset penetration rate threshold, is the decay rate coefficient, is the regional integral deviation, is the integral deviation activation threshold, which is generally 0.15 Hz / min.

[0067] The frequency response urgency feature is specifically:

[0068]

[0069] wherein, is a frequency response urgency feature, , and are respectively an instantaneous frequency fluctuation rate, a minute-level average deviation and a regional integral deviation adjusted weight, is an instantaneous frequency fluctuation rate, is a minute-level average deviation, is a regional integral deviation.

[0070] S2, dynamically adjusting the charging and discharging parameters based on the reinforcement learning model, and acquiring a power adjustment period according to the frequency response urgency feature;

[0071] The dynamically adjusting the charging and discharging parameters based on the reinforcement learning model is specifically:

[0072] A state space of the reinforcement learning model is constructed, and a multi-layer frequency deviation index, a frequency response urgency feature, a state of charge (SOC) of each electric vehicle, a charging cutoff time and a dynamic price signal information of the power grid are taken as elements of the state space;

[0073] A charging and discharging parameter set of the electric vehicle is defined as an action space, and a value range of the power is set;

[0074] A reward function is designed to comprehensively consider the power grid frequency stability, the user charging demand satisfaction degree and the energy cost;

[0075] At each time step, the reinforcement learning model selects an action to execute according to the current state, and stores the current state and the action and the output of the reinforcement learning model in an experience replay buffer;

[0076] Random data is obtained from the experience replay buffer, and the parameters of the reinforcement learning model are adjusted by minimizing a loss function to obtain a trained reinforcement learning model;

[0077] Before the start of each power adjustment period, the edge computing node obtains real-time current state information, inputs the current state information into the trained reinforcement learning model, and outputs the charging and discharging parameters of each electric vehicle.

[0078] The acquiring a power adjustment period according to the frequency response urgency feature is specifically:

[0079] A mapping relationship between the frequency response urgency feature and the power adjustment period is established based on regression analysis;

[0080] According to a current frequency response urgency feature value, a corresponding power adjustment period is obtained through the established mapping relationship.

[0081] the power adjustment period, in particular,

[0082]

[0083]

[0084] wherein, is the adjusted power period, is the initial power period, is the relaxation factor, is the frequency response urgency feature, is the average electricity price of the power adjustment period, is the dynamic electricity price signal.

[0085] S3, according to the power adjustment period, combining the remaining power demand of the electric vehicle and the urgency of the charging cutoff time, quantifying the user demand constraint into a charging priority;

[0086] The charging priority is quantified according to the power adjustment period, combining the remaining power demand of the electric vehicle and the urgency of the charging cutoff time, in particular,

[0087] For each electric vehicle, the product of the difference between the target SOC and the current SOC and the battery capacity is taken as the remaining power demand of the electric vehicle;

[0088] The charging cutoff time of each electric vehicle is obtained in combination with the remaining power demand of the electric vehicle to calculate the charging priority based on the power adjustment period.

[0089] The charging priority, in particular,

[0090]

[0091] wherein, is the charging priority, is the charging cutoff time, is the adjusted power period, is the target state of charge of the electric vehicle i, is the current state of charge of the electric vehicle i, is the battery capacity of the electric vehicle i, is the largest remaining charging demand among all electric vehicles.

[0092] S4, according to the power grid frequency stability, the charging priority and the energy cost to establish a distributed optimization objective function, and through a genetic algorithm to obtain the optimal power distribution scheme, the genetic algorithm takes the initial value of the charge and discharge parameters output by the reinforcement learning model as the iteration starting point;

[0093] The distributed optimization objective function is established according to grid frequency stability, charging priority and energy cost, and an optimal power distribution scheme is obtained by solving the genetic algorithm, specifically:

[0094] A distributed optimization objective function is established according to grid frequency stability, charging priority and energy cost;

[0095] A plurality of random power distribution schemes satisfying the constraint condition are generated;

[0096] The solution space of the power distribution scheme is divided into a short-term power fast regulation layer and a long-term energy cost optimization layer, and a first population and a second population are constructed respectively;

[0097] In the first population, a genetic operator with high mutation probability is used to improve the rapid response ability to frequency disturbance;

[0098] In the second population, a genetic operator with low mutation probability and an elite reservation genetic operator are used to ensure economic convergence performance;

[0099] Exchange elite individuals within a preset generation interval, and share information between different optimization layers;

[0100] When the convergence condition is reached, the optimal power distribution scheme is selected based on the comprehensive fitness value of the double population.

[0101] The distributed optimization objective function is specifically:

[0102]

[0103] The constraint condition is specifically:

[0104]

[0105] Power limit:

[0106] Battery capacity limit:

[0107] Grid power balance:

[0108] Wherein, is the distributed optimization objective function, is a preset grid frequency stability weight, is a frequency response urgency feature, is a preset charging priority weight, is a charging priority, is a preset energy cost weight, is a dynamic electricity price, is the charging and discharging power of the electric vehicle i at time t, is the minimum grid power, is the allowed maximum grid power, , are the minimum and maximum safe state of charge of the battery respectively, is the current state of charge of the electric vehicle i, is the battery capacity of the electric vehicle i, is the maximum total power provided by the grid, is the adjusted power cycle.

[0109] S5, synchronizing the states of each node based on the optimal power allocation scheme and allocating the charging and discharging power of each electric vehicle in the region.

[0110] The optimal power allocation scheme is based on the synchronization of the states of each node and the allocation of the charging and discharging power of each electric vehicle in the region, specifically:

[0111] A power synchronization signal source is set, and each node adjusts its charging and discharging power by detecting and tracking the synchronization signal, so that the change of power is consistent with the synchronization signal;

[0112] When the amplitude of the synchronization signal reaches the preset amplitude threshold, the node adjusts the power according to the optimal power allocation scheme to obtain the synchronized node state;

[0113] Based on the synchronized node state, each electric vehicle's charging device makes local decisions according to its own SOC and estimated stay time, received information of adjacent vehicles and public information of the grid, and the rules of the optimal power allocation scheme;

[0114] Each electric vehicle's charging device exchanges information and cooperatively adjusts through communication, and when the SOC of the electric vehicle reaches the target value in advance, the power allocation is adjusted in time to distribute the excess power to other vehicles that need to be charged;

[0115] If the available power of the grid changes, the charging and discharging power of each electric vehicle is recalculated and adjusted to ensure that the system is always in an optimal operating state.

[0116] Embodiment 2, Figure 2 A distributed energy power control system is given, which comprises the following modules:

[0117] The frequency urgency feature acquisition module is used to acquire and construct a multi-layer frequency deviation index based on the first data, the multi-layer frequency deviation index includes instantaneous frequency fluctuation rate, minute-level average deviation and regional integral deviation, and a sliding window is used to adaptively adjust the weights of the three types of indexes to construct the frequency response urgency feature;

[0118] The reinforcement learning regulation module is configured to dynamically adjust the charging and discharging parameters based on a reinforcement learning model, and obtain a power adjustment period according to a frequency response urgency characteristic;

[0119] The charging priority quantification module is configured to quantize the user demand constraint into a charging priority according to the power adjustment period, in combination with a residual power demand of the electric vehicle and an urgency of a charging cutoff time;

[0120] The power optimization module is configured to establish a distributed optimization objective function according to a power grid frequency stability, the charging priority and an energy cost, and obtain an optimal power distribution scheme by solving the objective function through a genetic algorithm, wherein the genetic algorithm takes the initial value of the charging and discharging parameters output by the reinforcement learning model as an iteration starting point.

[0121] The node cooperative distribution module is configured to synchronize states of all nodes based on the optimal power distribution scheme, and distribute charging and discharging powers of all electric vehicles in the region.

[0122] The above formulas are all dimensionless values, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation, and preset parameters in the formula are set by a person skilled in the art according to actual conditions.

[0123] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized in the form of a computer program product wholly or partially.

[0124] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solutions. A person skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0125] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module.

[0126] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0127] Finally: the above only for the preferred embodiments of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the scope of protection of the present application.

Claims

1. A method of distributed energy power control, the method comprising: The method comprises the following steps: obtaining and constructing a multi-layer frequency deviation index based on the first data, the multi-layer frequency deviation index comprising an instantaneous frequency fluctuation rate, a minute-level average deviation and a regional integral deviation, and adaptively adjusting the weights of the three types of indexes by using a sliding window to construct a frequency response urgency feature, specifically: analyzing historical power grid operation data to determine the influence degree of the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation on the stability of the power grid under different value ranges; dividing the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation into three levels of low, medium and high according to the influence degree; assigning initial weights to the instantaneous frequency fluctuation rate, the minute-level average deviation and the regional integral deviation according to the levels and adaptively adjusting the initial weights based on the sliding window; calculating the frequency response urgency feature according to the calculation results of the indexes and the assigned weights; dynamically adjusting the charging and discharging parameters based on the reinforcement learning model, obtaining a power adjustment period according to the frequency response urgency feature; quantifying user demand constraints into charging priorities according to the power adjustment period, the remaining power demand of the electric vehicle and the urgency of the charging cutoff time; establishing a distributed optimization objective function according to the frequency response urgency feature, the charging priorities and the energy cost, and solving the optimal power distribution scheme by using a genetic algorithm, wherein the genetic algorithm takes the initial values of the charging and discharging parameters output by the reinforcement learning model as the iteration starting point; synchronizing the states of the nodes based on the optimal power distribution scheme, and distributing the charging and discharging power of the electric vehicles in the region.

2. The method of claim 1, wherein, The multi-layer frequency deviation index comprises an instantaneous frequency fluctuation rate. The method for obtaining the instantaneous frequency fluctuation rate comprises the following steps: taking a dynamic time interval linked with the frequency response urgency feature as a sampling period, and obtaining the power grid frequency value at each sampling period; calculating the instantaneous frequency fluctuation rate based on the power grid frequency value, wherein the calculation method for the instantaneous frequency fluctuation rate is the ratio of the absolute value of the difference between the current sampling period frequency and the previous period frequency to the previous period frequency.

3. The method of claim 2, wherein, The adaptive adjustment based on the sliding window comprises the following steps: for the weight of the instantaneous frequency fluctuation rate, inversely proportionally adjusting the weight according to the current regional inertia level; the weight of the minute-level average deviation is positively correlated with the new energy penetration rate in a segmented linear manner; for the weight of the regional integral deviation, enabling exponential decay correction when the regional integral deviation exceeds a preset threshold.

4. The method of claim 3, wherein, The dynamic adjustment of the charging and discharging parameters based on the reinforcement learning model comprises the following steps: constructing a state space of the reinforcement learning model, defining a set of charging and discharging parameters of the electric vehicle as an action space, and setting the value range of the power; at each time step, the reinforcement learning model executes an action according to the current state, and stores the current state and the action and the output of the reinforcement learning model in an experience replay buffer; randomly obtaining data from the experience replay buffer, adjusting the parameters of the reinforcement learning model by minimizing a loss function, and obtaining a trained reinforcement learning model; before the start of each power adjustment period, an edge computing node obtains real-time state information, inputs the real-time state information into the trained reinforcement learning model, and outputs the charging and discharging parameters of the electric vehicles.

5. The method of claim 4, wherein, The power adjustment period is obtained according to the frequency response urgency feature, specifically as follows: A mapping relationship between the frequency response urgency feature and the power adjustment period is established based on regression analysis; According to the current frequency response urgency feature value, the corresponding power adjustment period is obtained through the established mapping relationship.

6. The method of claim 5, wherein, The charging priority is quantified according to the power adjustment period, combined with the urgency of the remaining power demand and the charging cutoff time of the electric vehicle, specifically as follows: For each electric vehicle, the product of the difference between the target SOC and the current SOC and the battery capacity is taken as the remaining power demand of the electric vehicle; The charging priority is calculated based on the power adjustment period, combined with the charging cutoff time of each electric vehicle and the remaining power demand of the electric vehicle.

7. The method of claim 6, wherein, The distributed optimization objective function is established according to the frequency response urgency feature, the charging priority and the energy cost, and the optimal power allocation scheme is obtained by solving the genetic algorithm, specifically as follows: The distributed optimization objective function is established according to the frequency response urgency feature, the charging priority and the energy cost; A number of random power allocation schemes that meet the constraint conditions are generated; The solution space of the power allocation scheme is divided into a short-term power rapid regulation layer and a long-term energy cost optimization layer, and the first population and the second population are constructed respectively; In the first population, a genetic operator with high mutation probability is used to improve the rapid response ability to frequency disturbance; In the second population, a genetic operator with low mutation probability and an elite preservation genetic operator are used to ensure economic convergence performance; Elite individuals are exchanged within a preset generation interval, and information is shared between different optimization layers; When the convergence condition is reached, the optimal power allocation scheme is selected based on the comprehensive fitness value of the double population.

8. The method of claim 7, wherein, The distributed optimization objective function is specifically as follows: The constraint condition is specifically as follows: Power limits: Battery capacity limitations: Grid power balancing: wherein, is a distributed optimization objective function, is a preset grid frequency stability weight, is a frequency response urgency feature, is a preset charging priority weight, is a charging priority, is a preset energy cost weight, is a dynamic electricity price, is a charging / discharging power of an electric vehicle i at time t, is a minimum grid power, is an allowed maximum grid power, , are minimum and maximum safe state-of-charge of the battery, respectively, is a current state-of-charge of the electric vehicle i, is a battery capacity of the electric vehicle i, is a maximum total power provided by the grid, is an adjusted power cycle.

9. A system using a distributed energy power control method according to any one of claims 1 to 8, characterized in that, The following modules are included: A frequency urgency feature acquisition module is used to acquire and construct a multi-layer frequency deviation index based on first data, the multi-layer frequency deviation index includes instantaneous frequency fluctuation rate, minute-level average deviation and regional integral deviation, and a sliding window is used to adaptively adjust the weights of the three types of indexes to construct the frequency response urgency feature; A reinforcement learning regulation module is used to dynamically adjust the charging and discharging parameters based on a reinforcement learning model, and to obtain the power adjustment period according to the frequency response urgency feature; A charging priority quantification module is used to quantify the user demand constraint as a charging priority according to the power adjustment period, combined with the urgency of the remaining power demand and the charging cutoff time of the electric vehicle; A power optimization module is used to establish a distributed optimization objective function according to the frequency response urgency feature, the charging priority and the energy cost, and to obtain the optimal power allocation scheme by solving the genetic algorithm, the genetic algorithm takes the initial value of the charging and discharging parameters output by the reinforcement learning model as the iteration starting point; A node cooperative allocation module is used to synchronize the states of each node based on the optimal power allocation scheme, and to allocate the charging and discharging power of each electric vehicle in the region.

Images