Smart grid optimal dispatching method and system based on multi-element energy storage collaborative dispatching

Abstract

The application discloses a multi-element energy storage collaborative scheduling-based intelligent power grid optimal scheduling method and system, relates to the technical field of power grid optimal scheduling, and comprises the following steps: constructing a prediction model based on first data, generating prediction data, coupling energy storage characteristic parameters of different types of energy storage devices with the prediction data, and establishing a multi-energy collaborative scheduling model; dynamically screening an energy storage scheduling strategy set based on preset real-time efficiency evaluation indexes, and generating an optimal strategy subset; performing differentiated charging and discharging control instructions on an energy storage device cluster according to the optimal strategy subset; collecting second data in a charging and discharging control process, calculating deviation values of the second data and the prediction data, converting the deviation values into feature vectors, inputting the feature vectors into a preset incremental learning algorithm, and optimizing parameters of the multi-energy collaborative scheduling model. The application implements hierarchical screening in combination with real-time efficiency evaluation indexes, and ensures that optimal scheduling schemes can be quickly selected under different time periods and uncertain disturbance conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid optimal scheduling, and more particularly to an intelligent power grid optimal scheduling method and system based on multi-element energy storage collaborative scheduling. BACKGROUND

[0002] With the continuous increase in the proportion of renewable energy and the increasing complexity of power grid load characteristics, traditional power system scheduling methods are facing unprecedented challenges. Multi-element energy storage systems (including electrochemical energy storage, flywheel energy storage, pumped storage, hydrogen energy storage, and thermal energy storage) have different time scales, energy densities, and response speeds, and are becoming an important support for improving the flexibility of power grids and enhancing the resilience of the system. The intelligent power grid optimization method based on multi-element energy storage collaborative scheduling is becoming an important direction of current intelligent power grid research and engineering application.

[0003] In practical applications, a single type of energy storage often cannot meet the dual needs of large-scale energy regulation and rapid dynamic response. Therefore, by complementing and cooperating multiple energy storage resources, and according to their respective characteristics, hierarchical, time-sharing, and zoned scheduling can effectively improve the overall performance of the system. At the same time, the output of new energy sources such as wind and solar energy has high uncertainty, requiring scheduling strategies to have good robustness and adaptive ability. In addition, to achieve real-time scheduling, edge computing, 5G communication, and Internet of Things technologies are often combined to build a cloud-edge-end collaborative control system.

[0004] Existing multi-element energy storage collaborative optimization scheduling methods mainly use multi-objective optimization, rolling time domain control, distributed optimization algorithms (such as ADMM), and reinforcement learning techniques, taking into account economic efficiency, reliability, environmental friendliness, and other indicators. At the same time, emerging technologies such as digital twinning and artificial intelligence are widely used in energy storage system state prediction, fault warning, and adaptive optimization, improving the intelligent level of scheduling decisions.

[0005] For example, the invention patent with the announcement number CN118539521B announces a source network load storage collaborative method and system, which relates to the technical field of data processing. By collecting real-time power generation data of renewable energy power stations, real-time operation state data of power grids, and power load demand data of terminal users from multiple sources, a preliminary data set is formed. A prediction model is trained based on the data set to output prediction results of renewable energy generation and user load change trends in the future. The prediction results are combined with the real-time operation state data of the power grid to construct a multi-objective optimization model, which includes multiple control strategies. The optimal scheduling strategy is obtained by solving the multi-objective optimization model based on the multiple control strategies to adjust the operation state of the power grid. By collecting necessary data from multiple sources in real time and introducing a multi-objective optimization framework, various constraint conditions and objective functions can be considered to achieve a globally optimal scheduling strategy.

[0006] For example, the invention patent announcement CN118263908A, which describes a method and system for improving energy storage efficiency in conjunction with energy management, relates to the field of energy storage system technology. The method includes: reading the installed capacity and grid connection status of energy storage systems in a preset control area, and monitoring the distribution status of energy storage; performing energy-side capacity forecasting and demand-side absorption forecasting, fitting the forecast time series, and determining energy storage trends; constructing an energy storage management module, including an energy management block and a battery management block with a staggered network structure, capable of lateral data interaction and configured with a penalty function; performing electrochemical conversion decisions and energy dispatch decisions to determine an energy storage dispatch scheme; and transmitting the data to the energy management system to assist the energy storage converter in performing energy storage management based on energy dispatch. This invention solves the technical problems of traditional energy storage system management, which often lacks comprehensive data support and intelligent decision-making, and cannot accurately predict energy supply and demand, resulting in inflexible and unintelligent energy storage management.

[0007] The aforementioned publicly disclosed technical solutions suffer from at least the following technical problems: Most microgrid projects often adopt a simple configuration method based on maximum load demand or design peak electricity consumption when configuring energy storage capacity. This approach ignores the synergistic effects between various renewable energy sources (such as wind power and photovoltaics) and energy storage systems, and fails to consider the utilization efficiency of energy storage devices, resulting in excessively large energy storage capacity, increased investment costs, and low energy storage utilization.

[0008] To address the above problems, this invention proposes a solution. Summary of the Invention

[0009] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a smart grid optimization scheduling method and system based on multi-energy storage collaborative scheduling. By deeply integrating the dynamic characteristics of energy storage devices with load forecasting and renewable energy output forecasting, a collaborative scheduling model oriented towards fluctuations at different time scales is constructed. This addresses the problems of neglecting the synergistic effects between various renewable energy sources (such as wind power and photovoltaics) and energy storage systems, failing to consider the utilization efficiency of energy storage devices, resulting in excessive energy storage capacity, increased investment costs, and low energy storage utilization.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] The intelligent power grid optimal scheduling method based on multi-element energy storage collaborative scheduling comprises the following steps: constructing a prediction model based on first data to generate prediction data, wherein the first data comprises historical load data of power grid nodes, meteorological characteristic parameters and physical parameters of renewable energy power generation equipment; coupling energy storage characteristic parameters of different types of energy storage equipment with the prediction data to establish a multi-energy collaborative scheduling model and output a number of hierarchical energy storage scheduling strategy sets; dynamically screening the energy storage scheduling strategy sets based on preset real-time performance evaluation indicators to generate an optimal strategy subset; differentiating the charging and discharging control instructions for the energy storage equipment cluster according to the optimal strategy subset; collecting second data in the charging and discharging control process, calculating the deviation value of the second data and the prediction data, converting the deviation value into a feature vector and inputting the feature vector into a preset incremental learning algorithm to optimize the parameters of the multi-energy collaborative scheduling model.

[0012] In a preferred embodiment, the coupling of the energy storage characteristic parameters of different types of energy storage equipment with the prediction data specifically comprises: establishing corresponding parameter models according to the types of the energy storage equipment;

[0013] Fusing the parameter models with the prediction data to establish differentiated scheduling constraint conditions for different types of energy storage equipment; the prediction data comprises load prediction data and renewable energy output prediction data; based on the fusion result, analyzing the operation capability boundary and available resource space-time distribution of each type of energy storage equipment in the future scheduling period; according to the device type and the current state of charge and health state parameters, matching different time scale load fluctuation characteristics to determine the role range and scheduling priority of each type of energy storage equipment in the multi-level scheduling system.

[0014] In a preferred embodiment, the coupling of the energy storage characteristic parameters of different types of energy storage equipment with the prediction data to establish a multi-energy collaborative scheduling model and output a number of hierarchical energy storage scheduling strategy sets specifically comprises: obtaining load prediction data and renewable energy output prediction data in a target scheduling period; establishing parameterized characteristic models of each type of energy storage equipment according to the types of the energy storage equipment; performing time-synchronous discrete processing on the parameterized characteristic models and the prediction data to generate dynamic operation boundaries of each energy storage unit in each scheduling period; based on power grid power balance constraints, energy storage equipment state transition constraints and multi-device response capability coupling relationships, constructing a collaborative scheduling constraint set; taking the minimization of comprehensive operation cost as the target, fusing the dynamic operation boundaries and the coordinated scheduling constraint set to establish a multi-energy collaborative scheduling model; calling a preset optimization algorithm to solve the scheduling optimization model to output optimal charging and discharging power instructions of each energy storage equipment in each scheduling period, thereby constituting a hierarchical energy storage scheduling strategy set.

[0015] In a preferred embodiment, the preset optimization algorithm is called to solve the scheduling optimization model to output optimal charging and discharging power instructions of each energy storage device at each scheduling period, thereby forming a hierarchical energy storage scheduling strategy set, specifically: based on a multi-energy collaborative scheduling model, an optimization objective function is constructed; a multi-objective optimization method is used to solve the multi-energy collaborative scheduling model to generate a scheduling strategy set; in view of the uncertainty of the prediction data, a number of disturbance scenarios based on historical error distribution are constructed, and multi-scenario performance simulation evaluation is performed on each scheduling strategy; the strategy score is calculated according to the real-time performance evaluation index, and the optimal energy storage scheduling strategy at the current time is selected and executed according to the score.

[0016] In a preferred embodiment, the preset real-time performance evaluation index is used to dynamically screen the energy storage scheduling strategy set, specifically: an efficiency evaluation index system including safety index, economy index and equipment health index is constructed; first layer evaluation: the scheduling strategy set is hard-constrained and filtered based on grid constraint parameters; second layer evaluation: the remaining strategies passing the first layer evaluation are calculated for a weighted comprehensive score of the performance evaluation index; the top k strategies with the highest comprehensive score are selected, and the optimal strategy is taken as the execution scheme, and the remaining strategies are stored in a standby strategy pool.

[0017] In a preferred embodiment, the preset real-time performance evaluation index is used to dynamically screen the energy storage scheduling strategy set, further comprising: when it is detected that there is a power instruction logic conflict between the strategies in the standby strategy pool, performing: calling a strategy fusion algorithm to identify a conflict segment set, calculating a conflict power difference, constructing a conflict matrix, and constructing a compromise scheduling power instruction.

[0018] In a preferred embodiment, the parameters of the optimized multi-energy collaborative scheduling model further include a judgment of the optimization result, specifically: the voltage value and system frequency of each grid node are obtained in real time; the system frequency deviation is calculated, and the disturbance level is divided according to a preset frequency threshold; for a key node set, the voltage fluctuation range and instantaneous voltage offset are calculated; if the instantaneous voltage offset is greater than a preset voltage offset threshold, the node is marked as unstable, and an energy storage priority scheduling instruction is generated; when the deviation is greater than a preset second deviation threshold and there are multiple nodes with instantaneous voltage offset greater than a preset voltage offset threshold, it is determined that there is a large power disturbance or a sudden change in renewable energy.

[0019] In a preferred embodiment, the bias value is converted into a feature vector, specifically: based on a preset time sliding window, the actual load value and the predicted load value, the actual output and the predicted output data of the renewable energy at each time in the window are obtained, and a multi-dimensional disturbance observation matrix is constructed; the real-time operation state parameters of the energy storage device cluster are fused into the observation matrix to form an extended state matrix; the covariance matrix of the extended state matrix is calculated, and the eigenvalue decomposition is performed on the covariance matrix to obtain a group of feature vectors arranged in descending order according to the eigenvalue; a principal component retention threshold is set, the first g feature vectors with a cumulative contribution rate meeting the threshold are selected, and the key disturbance response feature vector is generated by splicing.

[0020] In a preferred embodiment, the bias value is converted into a feature vector input into a preset incremental learning algorithm, specifically: an incremental extreme learning machine is used to build a strategy parameter learning model, and the current power system coordinated scheduling parameter set is used as the output layer target value of the model to complete the model initialization; the key disturbance response feature vector is input into the initialized strategy parameter learning model, and the nonlinear mapping relationship between the disturbance feature and the scheduling parameter is constructed combined with the feedback data; the feedback data includes the charge and discharge efficiency of the energy storage system, the power grid frequency recovery time, and the voltage stability coefficient; the bias value of the model output parameter and the current operation strategy parameter is calculated; the hidden layer-output layer weight matrix of the strategy parameter learning model is iteratively updated according to the preset learning rate.

[0021] The intelligent power grid optimization scheduling system based on multi-element energy storage coordinated scheduling includes the following modules: a prediction modeling module for obtaining first data and constructing a prediction model to generate prediction data; a scheduling modeling module for coupling the energy storage characteristics of different types of energy storage devices with the prediction data, establishing a multi-energy coordinated scheduling model, and outputting a set of energy storage scheduling strategies at several levels; a strategy screening module for dynamically screening the set of energy storage scheduling strategies based on real-time efficiency evaluation indicators; an execution control module for performing differential charge and discharge control on the energy storage device cluster based on the screened energy storage scheduling strategy; and an adaptive optimization module for optimizing the coordinated scheduling parameters according to second data in the charge and discharge control process, wherein the optimization is performed by converting the bias value between the actual data and the prediction data into a feature vector and inputting it into a preset incremental learning algorithm for optimization.

[0022] The intelligent power grid optimization scheduling method and system based on multi-element energy storage coordinated scheduling have the following technical effects and advantages:

[0023] 1.The application fully considers the differences in response rate, efficiency, capacity attenuation characteristics, etc. of energy storage types such as lithium-ion batteries, supercapacitors, and flow batteries, and constructs differentiated parameter models and scheduling constraints. By deeply integrating the dynamic characteristics of energy storage devices with load forecasting and renewable energy output forecasting, a collaborative scheduling model is constructed for different time scale fluctuations, so that fast energy storage can quickly respond to short-term disturbances, and energy storage can undertake medium and long-term energy balance tasks, significantly improving the overall response capability and scheduling accuracy of the system, and avoiding the waste of "one-size-fits-all" resources.

[0024] 2.The application generates three types of strategies including economic scheduling, emergency response and balance regulation by constructing a multi-objective collaborative optimization model, and implements hierarchical screening combined with real-time performance evaluation indexes to ensure that the optimal scheduling scheme can be quickly selected under different time periods and uncertain disturbance conditions. This mechanism not only avoids the local optimal problem brought by a single objective function, but also enhances the robustness of strategy switching through the standby strategy pool and conflict fusion mechanism, so as to realize the global optimization of scheduling economy and device life under the premise of considering power grid safety constraints. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 Fig. 1 is a flowchart of the intelligent power grid optimization scheduling method based on multi-element energy storage collaborative scheduling of the application;

[0026] Figure 2 Fig. 2 is a structural diagram of the intelligent power grid optimization scheduling system based on multi-element energy storage collaborative scheduling of the application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the application will be clearly and completely described 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 the embodiments. 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 the application.

[0028] Embodiment 1, Figure 1 The intelligent power grid optimization scheduling method based on multi-element energy storage collaborative scheduling of the application is given, including the following steps:

[0029] S1, a prediction model is constructed based on first data to generate prediction data.

[0030] The first data includes historical load data of power grid nodes, meteorological characteristic parameters, and physical parameters of renewable energy power generation devices.

[0031] The prediction data includes predicted electricity load values and renewable energy power generation prediction intervals.

[0032] Obtain historical power consumption data of key grid nodes in the target micro-grid system, including daily, hourly, and minute-level time scales;

[0033] Synchronize the collection of meteorological feature data in the target area, including but not limited to environmental temperature, wind speed, wind direction, solar irradiance, humidity, etc.

[0034] Obtain physical and operational parameters of deployed renewable energy equipment such as photovoltaic modules and wind power equipment, such as photovoltaic panel area, conversion efficiency, fan blade diameter, rated power, etc.

[0035] The first data is used to construct a prediction model and generate prediction data, specifically:

[0036] Obtain first data and extract long-term fluctuation trends and short-term change characteristics based on a multi-model structure combining long and short-term data.

[0037] The long-term prediction part uses time series modeling methods to capture seasonal and trend characteristics.

[0038] The short-term prediction part introduces machine learning or deep learning models to predict short-term load fluctuations and renewable energy output changes in the future time steps.

[0039] Output the power consumption prediction value sequence in the future dispatching period, including prediction values and confidence intervals.

[0040] Output the power generation prediction interval of renewable energy such as photovoltaic and wind power, including minimum-maximum range or quantile prediction results.

[0041] S2, couple the energy storage characteristics parameters of different types of energy storage devices with the prediction data, establish a multi-energy collaborative scheduling model, and output several levels of energy storage scheduling strategy sets;

[0042] The energy storage devices include lithium ion batteries, super capacitors, and flow batteries.

[0043] The coupling of energy storage characteristics parameters of different types of energy storage devices with prediction data is specifically:

[0044] According to the type of energy storage device, a corresponding parameter model is established, including but not limited to charging and discharging efficiency, capacity attenuation coefficient, state response time, maximum and minimum charging and discharging power limit, current state of charge, and health status.

[0045] Fuse the parameter model with the load prediction data and renewable energy output prediction data, and establish differentiated scheduling constraint conditions for different types of energy storage;

[0046] Based on the parameter model and the prediction data fusion result, the operation ability boundary and the available resource distribution of each type of energy storage device in the future scheduling period are analyzed.

[0047] Further, according to the device type (such as fast response type and energy type) and the current state parameter (such as SOC, SOH, response delay), the load fluctuation characteristics in different time scales are matched to determine the role range and scheduling priority of each type of energy storage device in the multi-level scheduling system.

[0048] Among them, the fast response type energy storage device is matched to suppress short-time power disturbance of high-frequency fluctuation, and the energy type energy storage device is used to support medium and long-term energy balance control.

[0049] The parameter model is fused with the load prediction data and the renewable energy output prediction data, specifically:

[0050] The predicted values of the grid load and the renewable energy output in the future multiple time periods generated by the prediction model are obtained to form a scheduling input data set;

[0051] According to each type of energy storage device, the corresponding parameterized mathematical model is called, and the parameterized mathematical model includes a charge and discharge efficiency function, a health state updating model, and a response rate model;

[0052] The prediction data are aligned with the state parameters (such as SOC, SOH, and available capacity) of the energy storage model according to the time stamp, to realize consistency in the time dimension;

[0053] Based on the aligned data, in combination with the models of various devices, the maximum charge and discharge power range, the efficiency variation trend, and the health state boundary that can be supported in each time period are calculated, and differentiated device operation capability constraint conditions are extracted and formed;

[0054] The prediction values are fused with the device capability constraints to construct a structured multi-source collaborative scheduling constraint matrix.

[0055] The corresponding parameter model is established, specifically:

[0056] Nonlinear charge and discharge efficiency function: for lithium batteries, the efficiency variation under the influence of SOC is modeled as:

[0057]

[0058] Device health state updating model: considering the influence of temperature, DOD, and cycle number, the SOH updating expression is:

[0059]

[0060] Super capacitor response rate modeling: define the maximum response slope:

[0061]

[0062] wherein, is the charge and discharge efficiency at different power, is the rated efficiency, is the current power, is the current state of charge, is the power attenuation coefficient, is the SOC offset penalty coefficient, is the device health state at t+1, is the device health state at t, is the battery aging loss function, the input is the depth of discharge (DOD), temperature (T) and cycle number (N), is the maximum response rate of super capacitor, is the super capacitor power change rate, is the current output of super capacitor.

[0063] The establishment of differentiated scheduling constraint conditions for different energy storage types is specifically:

[0064] SOC update equation:

[0065]

[0066] State-dependent schedulable power range:

[0067]

[0068] Frequency support device scheduling response trigger condition:

[0069]

[0070] wherein, is the state of charge of the energy storage device at t+1, is the state of charge of the energy storage device at t, and is the charge and discharge efficiency at different power, is the charging power of the energy storage device at t, is the discharging power of the energy storage device at t, is the rated capacity of the energy storage device, is the scheduling time step, is the maximum discharging power limit of the i-th device, is the frequency response threshold, is the response power of the super capacitor at t, is the grid frequency deviation at t, The response power is proportional to the frequency deviation.

[0071] The establishment of the multi-energy coordinated scheduling model is specifically as follows:

[0072] Acquire load forecast data and renewable energy output forecast data within the target scheduling period;

[0073] Parametric characteristic models for each type of energy storage device were established, including charge / discharge efficiency functions, capacity decay functions, and response rate constraints.

[0074] By performing time-synchronized discretization processing on the parameterized characteristic model and the prediction data, the dynamic operating boundary of each energy storage unit in each scheduling period is obtained.

[0075] A set of collaborative scheduling constraints is constructed based on the coupling relationship between power balance, state constraints, and equipment response capabilities.

[0076] A multi-energy coordinated dispatch model is constructed based on dynamic operating boundaries and constraint sets, with the goal of minimizing the overall operating cost. The operating cost includes grid power purchase cost, energy storage aging cost, and dispatch ramp-up cost.

[0077] The preset optimization algorithm is invoked to solve the multi-energy collaborative scheduling model, and the optimal charging and discharging power command of each energy storage device in each scheduling period is output, forming a hierarchical energy storage scheduling strategy set.

[0078] The multi-energy coordinated scheduling model is specifically as follows:

[0079]

[0080] in, For the total scheduling cycle, Let t be the cost of purchasing electricity from the main grid. Let be the weighting coefficient for the aging cost of the i-th device. Let be the aging loss cost of the i-th energy storage device at time t. Let be the ramp-up cost weighting coefficient for the j-th device. Let be the ramp-up cost of the j-th energy storage device at time t. The weighting coefficients for device aging costs and ramp-up costs are obtained using the analytic hierarchy process (AHP).

[0081] The output consists of a set of energy storage scheduling strategies at several levels, specifically:

[0082] Based on the multi-energy coordinated scheduling model, optimization objective functions are constructed respectively, corresponding to the economic optimization objective, the safety and stability objective, and the balance objective;

[0083] Adopting the sub-target optimization method, a scheduling strategy is generated; the scheduling strategy includes an economic scheduling strategy, an emergency response strategy and a balance regulation strategy;

[0084] In view of the uncertainty of the prediction data, a plurality of disturbance scenarios are constructed, and the efficiency simulation evaluation of the scheduling strategy is carried out under various scenarios;

[0085] According to the set real-time efficiency evaluation index, the multiple scheduling strategies are scored and sorted, and the optimal energy storage scheduling strategy at the current time is selected for actual execution.

[0086] The emergency response strategy is triggered when the node voltage deviation exceeds 7% or the grid frequency fluctuation exceeds 0.5Hz, and the preset safe charging and discharging template is used to ensure system stability;

[0087] The economic scheduling strategy: combined with the trough and peak period division of electricity price, the charging and discharging plan of preferentially using low-cost electric energy is made, and the device aging loss model is introduced for cost conversion;

[0088] The balance regulation strategy: based on the minute-level photovoltaic or wind power output fluctuation detection (amplitude exceeding ±10%), the smoothing scheduling algorithm is triggered to suppress the impact of power drastic change on the system.

[0089] The real-time efficiency evaluation index is specifically:

[0090] S3, based on the preset real-time efficiency evaluation index, the optimal strategy subset is generated by dynamically screening the energy storage scheduling strategy set;

[0091] The optimal strategy subset is generated by dynamically screening the energy storage scheduling strategy set based on the preset real-time efficiency evaluation index, specifically:

[0092] The performance index including safety index, economy index and device health index is constructed; the safety index includes voltage stability coefficient; the economy index includes peak-valley arbitrage net income and device loss cost; the device health index includes storage unit aging distribution standard deviation;

[0093] The first layer evaluation excludes the strategy scheme that violates the safety hard constraint in the simulation scheduling process according to the multi-energy collaborative scheduling model and the grid constraint parameters (voltage upper and lower limit, frequency deviation threshold, power balance condition);

[0094] The second layer evaluation calculates the weighted comprehensive score according to the performance evaluation index of the remaining strategy;

[0095] The first k strategies with the highest comprehensive score are selected as the execution scheme of the current period, and the suboptimal strategies are stored in the standby strategy pool for subsequent switching;

[0096] When there is a logical conflict between the standby strategies (such as requiring charging and discharging at the same time in the same period), a strategy fusion algorithm is called to calculate the conflict power difference and construct a compromise scheduling power;

[0097] Actual operation data is continuously collected within the strategy execution period, and the safety, economy and energy storage utilization rate scores are updated in real time. When the current strategy performance deviates from the predicted score by more than a preset threshold, the strategy switching mechanism is triggered to replace the suboptimal scheme in the standby strategy pool.

[0098] The voltage stability coefficient, specifically:

[0099]

[0100] The conflict power difference, specifically:

[0101]

[0102] The compromise scheduling power is:

[0103]

[0104] The weighted comprehensive score, specifically:

[0105]

[0106]

[0107]

[0108]

[0109] wherein, is a voltage stability coefficient, is a frequency recovery time, is a voltage deviation, is a normal voltage, is a conflict power difference, is an expected power of the i-th scheduling object, is an average power currently allocable by the system, is a total number of scheduling objects, is a compromise scheduling power, is a priority weight of the i-th scheduling object, is a total power that can be provided by the system, is a preset adjustment coefficient, is a weighted comprehensive score, is a safety score, is an economy score, is a health score, , , is the weight coefficient corresponding to three indicators, which is set according to historical experience, is the net income = peak-valley arbitrage income - equipment wear cost, is the maximum net income, is the minimum net income, is the standard deviation of the aging distribution of the energy storage unit, the smaller the better, is the maximum standard deviation, is the minimum standard deviation.

[0110] S4, according to the optimal strategy subset, the differential charging and discharging control instructions of the energy storage device cluster are carried out;

[0111] The differential charging and discharging control instructions of the energy storage device cluster according to the optimal strategy subset are specifically:

[0112] For lithium battery equipment, according to the current state of charge, the health state and the target power value set in the strategy, a nonlinear power control function is called to calculate the actual execution power, so as to avoid overcharging and overdischarging and inhibit life loss;

[0113] For supercapacitor equipment, according to the frequency response type control logic, when the grid frequency deviation exceeds the threshold, the fast response mode is triggered, and according to the current SOC and the response rate upper limit, the maximum output power is set;

[0114] For each type of energy storage device, a scheduling execution window is set, and combined with the time priority label in the scheduling strategy, the power instruction is dynamically updated in each window period;

[0115] In the whole control cycle, the running state data of each device in the energy storage cluster is continuously monitored, and when parameter abnormalities (such as temperature overrun, voltage out of bounds) are detected, the charging and discharging power of the device is automatically adjusted or participation in the current cycle control is suspended.

[0116] S5, collect the second data in the charging and discharging control process, calculate the deviation value of the second data and the predicted data, convert the deviation value into a feature vector, input the preset incremental learning algorithm, and optimize the parameters of the multi-energy collaborative scheduling model.

[0117] The second data includes grid frequency deviation and node voltage fluctuation;

[0118] The optimization of the parameters of the multi-energy collaborative scheduling model also includes the judgment of the optimization result, which is specifically:

[0119] The voltage value and system frequency of each grid node are obtained in real time;

[0120] The deviation amount of the system frequency and the reference frequency is calculated, and the frequency disturbance level is judged;

[0121] For each key node, the voltage fluctuation range is counted, and the instantaneous voltage offset is calculated:

[0122] If the instantaneous voltage offset is greater than the preset voltage offset, it is marked as an unstable node, and priority scheduling energy storage support is performed;

[0123] When the deviation is greater than the preset second deviation threshold and there are multiple nodes with instantaneous voltage offset greater than the preset voltage offset, it is determined that there is a large power disturbance or a sudden change in renewable energy.

[0124] The frequency disturbance level is determined, specifically:

[0125] If the deviation is less than the preset first deviation threshold, it is considered to be normal fluctuation;

[0126] If the deviation is greater than the preset first deviation threshold and less than the preset second deviation threshold, a mild adjustment strategy is triggered;

[0127] If the deviation is greater than the preset second deviation threshold, an emergency strategy is started and the output power of the energy storage fast response device is adjusted.

[0128] The deviation value is converted into a feature vector, specifically:

[0129] Based on the preset time sliding window, the actual load value and the predicted load value, the actual output and the predicted output of the renewable energy at the corresponding time are obtained, and a multi-dimensional disturbance observation matrix is constructed;

[0130] Based on the observation matrix, the real-time running state parameters of the current energy storage device cluster are fused to form an extended state matrix; the real-time running state parameters include average state of charge, health state, frequency deviation and node voltage variance;

[0131] The covariance matrix of the extended state matrix is calculated and eigenvalue decomposition is performed to obtain a set of feature vectors and corresponding eigenvalues;

[0132] Set the principal component retention threshold, and splice the first g eigenvalues that meet the cumulative contribution rate to obtain the key disturbance response feature vector.

[0133] The input preset incremental learning algorithm optimizes the parameters of the multi-energy collaborative scheduling model, specifically:

[0134] Based on the incremental extreme learning machine initialization strategy parameter learning model, and taking the current collaborative scheduling parameter set as the output target;

[0135] The key disturbance response feature vector is input into the strategy parameter learning model, and the feedback data is combined to construct a nonlinear mapping relationship between the disturbance and the scheduling parameters; the feedback data includes the charging and discharging efficiency of the energy storage system, the power grid frequency recovery time, and the voltage stability coefficient.

[0136] The deviation of the calculation model output and the current operation strategy parameter is calculated, and the parameter is updated with a preset learning rate.

[0137] Embodiment 2, Figure 2 The application discloses an intelligent power grid optimal scheduling system based on multi-element energy storage collaborative scheduling, and has the characteristics that the system comprises the following modules:

[0138] A prediction modeling module is configured to construct a prediction model based on first data, and generate prediction data, wherein the first data comprises historical load data of a power grid node, meteorological characteristic parameters, and physical parameters of a renewable energy power generation device.

[0139] A scheduling modeling module is configured to couple energy storage characteristic parameters of different types of energy storage devices with the prediction data, establish a multi-energy collaborative scheduling model, and output a plurality of levels of energy storage scheduling strategy sets.

[0140] A strategy screening module is configured to dynamically screen the energy storage scheduling strategy sets based on preset real-time performance evaluation indexes, and generate an optimal strategy subset.

[0141] An execution control module is configured to perform differentiated charging and discharging control instructions on an energy storage device cluster according to the optimal strategy subset.

[0142] An adaptive optimization module is configured to collect second data in a charging and discharging control process, calculate a deviation value of the second data and the prediction data, convert the deviation value into a feature vector, input the feature vector into a preset incremental learning algorithm, and optimize parameters of the multi-energy collaborative scheduling model.

[0143] The above formulas are all dimensionless numerical calculations, 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 formulas are set by a person skilled in the art according to actual conditions.

[0144] 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.

[0145] 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.

[0146] In addition, each function module in each embodiment of the present application can be integrated in one processing module, or each module can be physically present alone, or two or more modules can be integrated in one module.

[0147] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, 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.

[0148] Finally: the above is only a preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An intelligent power grid optimal scheduling method based on multi-element energy storage collaborative scheduling, characterized in that, The method comprises the following steps: Based on the first data, a prediction model is constructed, and prediction data is generated, wherein the first data comprises historical load data of a power grid node, meteorological characteristic parameters, and physical parameters of a renewable energy power generation device; The energy storage characteristics parameters of different types of energy storage devices are coupled with the prediction data to establish a multi-energy collaborative scheduling model, load prediction data and renewable energy output prediction data in a target scheduling period are obtained, a parameterized characteristic model is established according to the types of energy storage devices, time synchronization and discrete processing are performed on the prediction data to generate a dynamic operation boundary, a collaborative scheduling constraint set is constructed based on power grid power balance constraints, energy storage device state transition constraints, and coupling relationships of multi-device response capabilities, and the multi-energy collaborative scheduling model is established by fusing the dynamic operation boundary and the collaborative scheduling constraint set with the minimum comprehensive operation cost as the target; A preset optimization algorithm is called to solve the multi-energy collaborative scheduling model, a plurality of disturbance scenarios based on historical error distribution are constructed for the uncertainty of the prediction data, multi-scenario performance simulation evaluation is performed on each scheduling strategy, and a plurality of hierarchical energy storage scheduling strategy sets are output; Based on preset real-time performance evaluation indexes, the energy storage scheduling strategy set is dynamically screened, strategy scores are calculated according to the real-time performance evaluation indexes, and an optimal strategy subset is generated by sorting according to the scores; According to the optimal strategy subset, differential charging and discharging control instructions are generated for the energy storage device cluster. Second data in the charging and discharging control process is collected, deviation values of the second data and the prediction data are calculated, the deviation values are converted into feature vectors, a preset incremental learning algorithm is input, and parameters of the multi-energy collaborative scheduling model are optimized. 2.The multi-energy storage collaborative scheduling based smart grid optimal scheduling method according to claim 1, characterized in that, The coupling of the energy storage characteristic parameters of different types of energy storage devices with the prediction data is specifically as follows: Parameter models corresponding to the types of energy storage devices are respectively established; The parameter models are fused with the prediction data to establish differential scheduling constraint conditions for different types of energy storage devices; the prediction data comprises load prediction data and renewable energy output prediction data; Based on the fusion result, the operation capability boundary and available resource space-time distribution of each type of energy storage device in the future scheduling period are analyzed; According to the types of devices, the current state of charge, and the health state parameters, the time scale load fluctuation characteristics are matched, the role range and scheduling priority of each type of energy storage device in the multi-level scheduling system are determined. 3.The multi-energy storage collaborative scheduling based smart grid optimal scheduling method according to claim 2, characterized in that, The dynamic screening of the energy storage scheduling strategy set based on the preset real-time performance evaluation indexes is specifically as follows: An efficiency evaluation index system comprising safety indexes, economic indexes, and device health indexes is constructed: First layer evaluation: the scheduling strategy set is subjected to hard constraint filtering based on grid constraint parameters; Second layer evaluation: the remaining strategies that pass the first layer evaluation are calculated for weighted comprehensive scores of the efficiency evaluation indexes; The first k strategies with the highest comprehensive scores are selected, the optimal strategy is taken as an execution scheme, and the remaining strategies are stored in a standby strategy pool. 4.The multi-energy storage collaborative scheduling based smart grid optimal scheduling method of claim 3, wherein, The dynamic screening of the energy storage scheduling strategy set based on the preset real-time performance evaluation indexes further comprises: When a power instruction logic conflict between strategies in the backup strategy pool is detected, the following is performed: a strategy fusion algorithm is called to identify a conflict segment set, a conflict power difference value is calculated, a conflict matrix is constructed, and a compromise scheduling power instruction is constructed.

5. The multi-energy storage collaborative scheduling based smart grid optimal scheduling method according to claim 4, characterized in that, The parameters of the optimized multi-energy collaborative scheduling model also include a judgment of the optimization result, specifically: Real-time acquisition of voltage values and system frequency of each power grid node; Calculate the system frequency deviation, and divide the disturbance level according to the preset frequency threshold; For the key node set, calculate the voltage fluctuation range and instantaneous voltage offset: If the instantaneous voltage offset is greater than the preset voltage offset threshold, mark it as an unstable node and generate a storage priority scheduling instruction; When the deviation is greater than the preset second deviation threshold and there are multiple nodes with instantaneous voltage offset greater than the preset voltage offset threshold, it is determined that there is a large power disturbance or a sudden change in renewable energy. 6.The multi-energy storage collaborative scheduling based smart grid optimal scheduling method according to claim 5, characterized in that, The deviation value is converted into a feature vector, specifically: Based on the preset time sliding window, the actual load value and the predicted load value, the actual output and the predicted output data of the renewable energy at each time in the window are obtained, and a multi-dimensional disturbance observation matrix is constructed; Fuse the real-time running state parameters of the energy storage device cluster into the observation matrix to form an extended state matrix; Calculate the covariance matrix of the extended state matrix, and perform eigenvalue decomposition on the covariance matrix to obtain a set of feature vectors arranged in descending order of eigenvalue size; Set a principal component retention threshold, select the first g feature vectors whose cumulative contribution rate meets the threshold, and splice to generate a key disturbance response feature vector. 7.The multi-energy storage collaborative scheduling based smart grid optimal scheduling method of claim 6, wherein, The deviation value is converted into a feature vector input into a preset incremental learning algorithm, specifically: Based on the incremental extreme learning machine, a strategy parameter learning model is built, and the current power system collaborative scheduling parameter set is used as the output layer target value of the model to complete model initialization; Input the key disturbance response feature vector into the initialized strategy parameter learning model, and construct the nonlinear mapping relationship between the disturbance feature and the scheduling parameter based on the feedback data; the feedback data includes the charge and discharge efficiency of the energy storage system, the power grid frequency recovery time, and the voltage stability coefficient; Calculate the deviation value of the model output parameter and the current running strategy parameter; Iteratively update the hidden layer-output layer weight matrix of the strategy parameter learning model according to the preset learning rate.

8. A system using the multi-element energy storage collaborative scheduling based smart grid optimal scheduling method according to any one of claims 1-7, characterized in that, It includes the following modules: A prediction modeling module is used to obtain first data and build a prediction model to generate prediction data; A scheduling modeling module is used to couple the energy storage characteristics of different types of energy storage devices with the prediction data, build a multi-energy collaborative scheduling model, and output a set of energy storage scheduling strategies at several levels; A strategy screening module is used to dynamically screen the energy storage scheduling strategy set based on real-time performance evaluation indicators; An execution control module is used to perform differential charge and discharge control on the energy storage device cluster based on the screened energy storage scheduling strategy; An adaptive optimization module is used to optimize the collaborative scheduling parameters according to the second data in the charge and discharge control process. The optimization is performed by converting the deviation value of the actual data and the prediction data into a feature vector and inputting it into a preset incremental learning algorithm for optimization.

Images