Power data adjusting method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of power system control, in particular to a power data adjusting method and device, electronic equipment and a storage medium. The electric power data adjusting method comprises the following steps: acquiring real-time data information of an electric power system; inputting the real-time data information into the trained photovoltaic prediction model, and obtaining a photovoltaic power generation prediction value; inputting the real-time data information and the photovoltaic power generation prediction value into the trained energy storage prediction model, and obtaining an energy storage charging and discharging power prediction value; generating a scheduling instruction based on the energy storage charging and discharging power prediction value, the photovoltaic power generation prediction value and the real-time data information; and adjusting the operation efficiency of the target equipment in real time based on the scheduling instruction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system control, and in particular to a power data adjustment method, an adjustment device, an electronic device and a storage medium. BACKGROUND

[0002] A virtual power plant aggregates diversified resources such as distributed power generation, energy storage units and flexible loads to realize coordinated regulation and flexible optimization of the power system, thereby effectively improving the reliability, economy and environmental friendliness of system operation. However, due to the complex energy structure covered by the virtual power plant and the dynamic changes in the operating state, the virtual power plant is also affected by multiple uncertain factors such as fluctuations in power market price signals and meteorological sensitivity of renewable energy output. The traditional scheduling method based on deterministic rules or classical optimization theory has gradually exposed limitations such as response lag and strategy solidification when dealing with multi-dimensional and strong uncertain scheduling scenarios.

[0003] In related technologies, although some research has attempted to introduce AI methods into the operation optimization of virtual power plants, there are still problems such as insufficient fusion processing capability of multi-source heterogeneous data, weak generalization and adaptability of scheduling models in dynamic scenarios, and difficulty in balancing the accuracy and real-time performance of the decision-making process, which restricts its large-scale application in actual engineering. SUMMARY

[0004] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application provides a power data adjustment method that can effectively improve the accuracy and scientificity of virtual power plant scheduling decisions.

[0005] In a first aspect, the present application provides a power data adjustment method, comprising: obtaining real-time data information of a power system; inputting the real-time data information into a trained photovoltaic prediction model and obtaining a photovoltaic power generation prediction value; inputting the real-time data information and the photovoltaic power generation prediction value into a trained energy storage prediction model and obtaining an energy storage charging and discharging power prediction value; generating a scheduling instruction based on the energy storage charging and discharging power prediction value, the photovoltaic power generation prediction value and the real-time data information; and adjusting the operating efficiency of a target device in real time based on the scheduling instruction.

[0006] In some embodiments, before acquiring the real-time data information of the power system, the method further comprises: acquiring historical data information of the power system; performing cleaning processing on the historical data information to obtain cleaned data information; performing interpolation processing on the cleaned data information to obtain interpolated data information; performing normalization processing on the interpolated data information to obtain normalized data information; inputting the normalized data information into a photovoltaic prediction model for training to obtain a trained photovoltaic prediction model; inputting the normalized data information and output information of the trained photovoltaic prediction model into an energy storage prediction model for training to obtain a trained energy storage prediction model; wherein the output information at least includes a photovoltaic power generation prediction value In some embodiments, the normalized data information includes historical battery state of charge and historical load prediction data; inputting the normalized data information and the output information of the trained photovoltaic prediction model into the energy storage prediction model for training to obtain the trained energy storage prediction model comprises: constructing an input feature vector based on the output information, the historical battery state of charge and the historical load prediction data; inputting the input feature vector into a preset machine learning model to obtain an initial prediction error; taking the initial prediction error as a reference, using an improved particle swarm optimization algorithm to perform offline optimization on a penalty factor and a kernel function width of the machine learning model, and obtaining an optimal hyperparameter combination; retraining the machine learning model based on the optimal hyperparameter combination to obtain the trained energy storage prediction model.

[0007] In some embodiments, the real-time data information includes meteorological data information and photovoltaic output information; inputting the real-time data information into the trained photovoltaic prediction model to obtain the photovoltaic power generation prediction value comprises: extracting spatial features of the meteorological data information to obtain meteorological spatial features; determining time series features of photovoltaic output based on the photovoltaic output information, and determining a dependency relationship between the time series features and the meteorological data information; based on the meteorological spatial features, the time series features and the dependency relationship, obtaining the photovoltaic power generation prediction value.

[0008] In some embodiments, the real-time data information further includes battery state of charge and load prediction data; inputting the real-time data information and the photovoltaic power generation prediction value into the trained energy storage prediction model to obtain an energy storage charging and discharging power prediction value comprises: based on the load prediction data and the photovoltaic power generation prediction value, obtaining the energy storage charging and discharging power prediction value, the energy storage charging and discharging power prediction value is expressed as:

[0009] wherein, is the energy storage charging and discharging power prediction value; is a battery SOC at a previous moment; is a time feature; is an output information of the photovoltaic prediction model at time t; is historical load prediction data.

[0010] In some embodiments, after obtaining the energy storage charging and discharging power prediction value, the method further comprises: performing calculation based on the energy storage charging and discharging power prediction value and the battery state of charge, and obtaining an available energy prediction value, which is represented as:

[0011] wherein, is the available energy prediction value; is an initial available energy; is a sequence of charging and discharging power prediction values; is a time interval; is a charging and discharging efficiency.

[0012] In some embodiments, the scheduling instruction is generated based on the energy storage charging and discharging power prediction value, the photovoltaic power generation prediction value and the real-time data information, comprising: generating a first adjustment index based on the available energy prediction value, the energy storage charging and discharging power prediction value and the photovoltaic power generation prediction value; generating a second adjustment index based on the load demand information of the power system and the market electricity price information; constructing a target revenue function according to the first adjustment index and the second adjustment index; solving the target revenue function under the target constraint by using a genetic algorithm to obtain a target operation efficiency; and generating the scheduling instruction based on the target operation efficiency.

[0013] In a second aspect, an embodiment of the present application provides a power data adjustment device, comprising: a data information acquisition module configured to acquire real-time data information of a power system; a photovoltaic data prediction module configured to input the real-time data information into a trained photovoltaic prediction model and obtain a photovoltaic power generation prediction value; an energy storage data prediction module configured to input the real-time data information and the photovoltaic power generation prediction value into a trained energy storage prediction model and obtain an energy storage charging and discharging power prediction value; a scheduling instruction generation module configured to generate a scheduling instruction based on the energy storage charging and discharging power prediction value, the photovoltaic power generation prediction value and the real-time data information; and a data adjustment module configured to adjust the energy storage charging and discharging power and the air conditioner operation efficiency in real time based on the scheduling instruction.

[0014] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor and a memory, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of the power data adjustment method of the first aspect when executing the program.

[0015] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores computer instructions for causing a computer to execute the steps of the power data adjustment method of the first aspect.

[0016] The technical scheme provided by the present application can train photovoltaic prediction models and energy storage prediction models according to various influencing factors, input real-time data information of a power system obtained to the trained prediction models to obtain photovoltaic power generation prediction values and energy storage charging and discharging power prediction values, and further generate corresponding scheduling instructions according to the prediction values, so as to realize real-time adjustment of the operation efficiency of the target equipment according to the scheduling instructions, effectively improve the total operation benefit of the virtual power plant in the case of guaranteeing data security and stability, and improve the reliability and feasibility of the scheduling strategy.

[0017] Additional aspects and advantages of the application will be described in the following description, will become apparent from the following description, or will be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0018] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description, taken in conjunction with the following drawings, in which: Figure 1 A flowchart of the power data adjustment method provided for the embodiment of the present application is shown in FIG. 5; Figure 2 A flowchart of training photovoltaic prediction models and energy storage prediction models provided for the embodiment of the present application is shown in FIG. 6; Figure 3 A flowchart of training energy storage prediction models provided for the embodiment of the present application is shown in FIG. 7; Figure 4 A flowchart of calculating photovoltaic power generation prediction values provided for the embodiment of the present application is shown in FIG. 8; Figure 5 A schematic diagram of the power data adjustment device provided for the embodiment of the present application is shown in FIG. 9; Figure 6 A more specific electronic device hardware structure schematic diagram provided for the embodiment of the present application is shown in FIG. 10.

[0019] Reference signs: 500-power data adjustment device; 510-data information acquisition module; 520-photovoltaic data prediction module; 530-energy storage data prediction module; 540-scheduling instruction generation module; 550-data adjustment module; 610-processor; 620-memory; 630-input / output interface; 640-communication interface; 650-bus. DETAILED DESCRIPTION Embodiments of the present application will be described below in greater detail with reference to the accompanying drawings. While certain embodiments of the present application are shown in the drawings, it is understood that the present application can be embodied in various forms and should not be interpreted as being limited to the embodiments set forth herein; rather, these embodiments are provided so that the present application can be more thoroughly and completely understood. It is to be understood that the drawings of the present application and the embodiments thereof are for illustrative purposes and are not to be construed as limiting the scope of the present application.

[0020] It should be understood that each of the steps recited in the method embodiments of the present application can be performed in different orders and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0021] The power data adjustment method of the embodiments of the present application will be described below with reference to the accompanying drawings.

[0022] Reference is made to Figure 1 The flowchart of the power data adjustment method provided by the embodiments of the present application is shown in FIG. 1.

[0023] In step S101, real-time data information of the power system is obtained. In step S102, the real-time data information is input into the trained photovoltaic prediction model, and a photovoltaic power generation prediction value is obtained. In step S103, the real-time data information and the photovoltaic power generation prediction value are input into the trained energy storage prediction model, and an energy storage charging and discharging power prediction value is obtained. In step S104, a dispatching instruction is generated based on the energy storage charging and discharging power prediction value, the photovoltaic power generation prediction value, and the real-time data information. In step S105, the operating efficiency of the target device is adjusted in real time based on the dispatching instruction.

[0024] Specifically, when implementing power data adjustment, the embodiment of the present application first needs to collect real-time data information of the power system, wherein the real-time data information of the power system includes but is not limited to data information of a photovoltaic system and an energy storage system. Further, the real-time data is processed, and the processed real-time data information is input into a trained photovoltaic prediction model. The trained photovoltaic prediction model is fused with a long short-term memory (LSTM) and a convolutional neural network (CNN). The CNN can extract spatial features in meteorological data, and the LSTM focuses on learning time rules in photovoltaic output historical data and a dependency relationship between photovoltaic output values and meteorological data. Through the cooperation of the LSTM and the CNN, the photovoltaic prediction model can calculate a photovoltaic power generation prediction value according to the real-time data information. Further, the processed real-time data information, the photovoltaic power generation prediction value, and data center load prediction data (i.e., a prediction value of power consumption of the data center in a future period of time) are input into a trained energy storage prediction model to obtain an energy storage charging and discharging power prediction value. An optimization function is constructed according to the photovoltaic power generation prediction value, the charging and discharging power prediction value, and the real-time data information, a scheduling instruction executed on a target device in the power system is determined, and finally the target device is adjusted based on the scheduling instruction to achieve the purpose of adjusting the operation efficiency of the target device. The present application can effectively realize accurate prediction of the power system through the prediction model, provide reliable data support for optimization scheduling decision, and further improve the adaptability of the scheduling strategy.

[0025] Reference Figure 2 The flow chart of training the photovoltaic prediction model and the energy storage prediction model is provided for the embodiment of the present application.

[0026] As an optional embodiment, before acquiring the real-time data information of the power system, the method further includes: Step S201, acquiring historical data information of the power system. Step S202, performing cleaning processing on the historical data information, and obtaining cleaned data information.

[0027] Specifically, before acquiring the real-time data information of the power system, the present application needs to acquire historical data information of the power system to train the photovoltaic prediction model and the energy storage prediction model. Further, the acquired historical data information is cleaned, that is, by setting a reasonable threshold or using a statistical method, data points obviously not conforming to logic or physical rules are identified and removed to ensure the quality of the data.

[0028] Step S203, performing interpolation processing on the cleaned data information, and obtaining interpolated data information.

[0029] Specifically, interpolation processing is performed on the missing points of the historical data information itself or the missing points caused by cleaning, so that the historical data remains complete and continuous, and the interpolation processing can be linear interpolation, spline interpolation, time series interpolation, etc.

[0030] In step S204, the interpolated data information is normalized to obtain normalized data information.

[0031] Specifically, the interpolated historical data information is further normalized, that is, the differences in dimension and numerical range between different historical data information are eliminated, and all historical data information is converted to the same scale, which can effectively accelerate model convergence and improve accuracy.

[0032] In step S205, the normalized data information is input into the photovoltaic prediction model for training, and a trained photovoltaic prediction model is obtained.

[0033] Specifically, the processed data information is input into the photovoltaic prediction model for training, and a trained photovoltaic prediction model is obtained. For the photovoltaic prediction model, the convolutional neural network (CNN) layer of the photovoltaic prediction model is responsible for extracting historical spatial features. In the embodiment of the present application, CNN is preferably used to process two-dimensional historical meteorological data, and effectively captures the correlation and distribution pattern of historical meteorological data in space through convolution operation. Its calculation can be expressed as:

[0034] wherein, is a historical meteorological data spatial feature matrix, is an activation function, is a weight matrix of the convolution kernel, * represents convolution operation, is a bias term, is a historical meteorological data matrix.

[0035] The long short-term memory (LSTM) network layer of the photovoltaic prediction model is responsible for processing historical time series. By inputting the historical meteorological data spatial feature matrix and the time series data of historical photovoltaic output (such as historical power data every 15 minutes in the past few hours), the root mean square error (RMSE) is used as the training loss function to further obtain the photovoltaic output value .

[0036] It should be noted that the output information of the photovoltaic prediction model at least includes the photovoltaic power prediction value, that is, the photovoltaic output value in a future time sequence.

[0037] In step S206, the normalized data information and the output information of the trained photovoltaic prediction model are input into the energy storage prediction model for training, and a trained energy storage prediction model is obtained.

[0038] Specifically, the normalized data information includes historical battery state of charge and historical load prediction data, wherein the historical battery state of charge includes historical battery charging and discharging data and battery state data.

[0039] Reference Figure 3 The flowchart for training the energy storage prediction model is provided for the embodiments of the present application.

[0040] In step S301, an input feature vector is constructed based on the output information, the historical battery state of charge and the historical load prediction data. Specifically, the input feature vector can be represented as:

[0041] wherein, represents the input feature vector at time t; represents the output information (photovoltaic output value) of the photovoltaic prediction model at time t; represents the historical battery state of charge at time t; represents the historical load prediction data at time t.

[0042] In step S302, the input feature vector is input into a preset machine learning model to obtain an initial prediction error.

[0043] In step S303, based on the initial prediction error, an improved particle swarm optimization algorithm is used to perform offline optimization on the penalty factor and the kernel function width of the machine learning model, and an optimal hyperparameter combination is obtained.

[0044] Specifically, the input feature vector is input into the preset machine learning model SVM model to calculate the initial prediction error. Further, based on the initial prediction error, the hyperparameters of the SVM regression model are optimized by using the improved particle swarm optimization algorithm IPSO to improve the generalization ability and prediction accuracy.

[0045] For the particle swarm optimization algorithm, each particle represents a set of hyperparameter candidate values (penalty factor and kernel function width) in the SVM. For the th particle, the velocity and position update formula of the (t+1)th iteration in the dth dimension (i.e., the dth parameter) is:

[0046]

[0047] wherein, represents the velocity of the particle in the dth dimension at the tth iteration; represents the position of the particle Position of the t-th iteration in the d-th dimension; is the inertia weight, which is used to control the influence of the current velocity on the next generation velocity; , is the learning factor, which respectively controls the influence of individual experience and group experience on particle velocity; , is a random number uniformly distributed in the range of [0, 1], which is used to introduce randomness; is the particle is the individual optimal position in the d-th dimension (i.e. the particle finds the optimal position so far); is the global optimal position of the entire group in the d-th dimension. Further, the inertia weight decreases linearly with the iteration number, and the linear decrease formula is:

[0048] wherein, is the initial inertia weight; is the final inertia weight; is the iteration number; is the maximum iteration number.

[0049] Further, the root mean square error of the SVM model is used as the fitness function of IPSO, and the RMSE is more sensitive to large errors, which can effectively punish the prediction deviating seriously to realize the evaluation of the good and bad of the position (i.e. candidate solution) of each particle, and the formula is:

[0050] wherein, is the number of training samples; is the true value of the i-th sample, which is the actual historical charge and discharge value in the embodiment of the present application; indicates the predicted value of the i-th sample using the current SVM model.

[0051] The improved particle swarm optimization algorithm initializes the particles, trains the SVM with the hyperparameters of each particle, calculates the RMSE, further updates the pbest and gbest to record the individual and global optimal positions, updates the velocity and position of the particles according to the formula to realize the iterative optimization, stops to realize the convergence judgment when the maximum iteration number is reached or the accuracy requirement is met, and finally outputs the optimal particle, i.e. the penalty factor and the kernel function width, which can more efficiently find the hyperparameter combination that makes the prediction accuracy of the SVM model highest, thereby improving the accuracy of the energy charge and discharge power prediction.

[0052] ​​Step S304, retraining the machine learning model based on the optimal hyperparameter combination to obtain a trained energy storage prediction model.

[0053] Specifically, after obtaining the optimal hyperparameter combination of the machine learning model determined by the improved particle swarm optimization algorithm, the model is retrained to obtain a trained energy storage prediction model, ensuring that the final prediction model put into use has high optimal generalization ability and prediction accuracy.

[0054] Reference Figure 4 The flowchart for calculating the prediction value of the photovoltaic power generation capacity provided by the embodiment of the present application.

[0055] Step S401, extracting the spatial features of the meteorological data information to obtain meteorological spatial features.

[0056] Specifically, in the actual application of the photovoltaic power generation model, the actual data includes real-time meteorological data information and real-time photovoltaic output information, and further, the real-time meteorological data matrix is input into the trained photovoltaic prediction model to extract real-time meteorological spatial features through convolution operation.

[0057] Step S402, determining the time series features of the photovoltaic output based on the photovoltaic output information, and determining the dependency relationship between the time series features and the meteorological data information.

[0058] Further, the real-time photovoltaic output information is constructed as time series data, matched with the real-time meteorological spatial features to obtain a fusion feature vector, and the fusion feature vector is input into the long short-term memory network for time series modeling. Through the collaborative operation of the forget gate, the input gate and the output gate of the LSTM unit, the dynamic change law of the photovoltaic output in the time dimension is captured, and the nonlinear mapping relationship and the dependency relationship between the time series features and the meteorological spatial features are established. The forget gate is used to filter the importance of historical information, the input gate is used to control the update degree of new features, and the output gate is used to determine the output of the hidden state at the current time.

[0059] Step S403, obtaining the prediction value of the photovoltaic power generation capacity based on the meteorological spatial features, the time series features and the dependency relationship.

[0060] Specifically, the hidden state finally output by the LSTM network is dimensionally transformed and nonlinearly mapped through the fully connected layer, and the distributed photovoltaic power generation capacity prediction value at a specific time scale in the future is directly output, which comprehensively reflects the joint action of the meteorological spatial features and the output time series law, and provides accurate power generation side prediction data support for the optimization and scheduling of the virtual power plant.

[0061] As an optional embodiment, the real-time data information further comprises a battery state of charge and load prediction data; the real-time data information and the photovoltaic power generation prediction value are input into the trained energy storage prediction model, and an energy storage charging and discharging power prediction value is obtained, which is represented as:

[0062] wherein, is the energy storage charging and discharging power prediction value; is a previous time battery SOC; is a time feature; is output information of the photovoltaic prediction model at time t; is historical load prediction data.

[0063] Specifically, the real-time data information further comprises a real-time battery state of charge and real-time load prediction data, on the basis of obtaining the photovoltaic power generation prediction value, an input feature vector is constructed in combination with the real-time state of charge and the real-time load prediction data, and the input feature vector is input into the trained energy storage prediction model, and through a nonlinear mapping relationship established inside the model, an energy storage charging and discharging power prediction value of the energy storage system at any future time is calculated, wherein the energy storage prediction model determines an optimal hyperparameter combination based on an improved particle swarm optimization algorithm, can accurately capture the complex coupling relationship between photovoltaic power generation, load demand and energy storage state, and the energy storage charging and discharging power prediction value is represented as:

[0064] wherein, is the energy storage charging and discharging power prediction value; is a previous time battery SOC; is a time feature; is output information of the photovoltaic prediction model at time t; is historical load prediction data.

[0065] As an optional embodiment, after obtaining the energy storage charging and discharging power prediction value, the method further comprises: calculating based on the energy storage charging and discharging power prediction value and the battery state of charge, and obtaining an available energy prediction value, which is represented as:

[0066] wherein, is the available energy prediction value; is an initial available energy; is a charging and discharging power prediction value sequence; is a time interval; is a charging and discharging efficiency.

[0067] Specifically, based on the above-mentioned integrated charge and discharge power prediction value sequence of the charge and discharge power prediction value , and the current battery state of charge, further through energy accumulation calculation, the available energy prediction value of the energy storage system at future k time is obtained, and the available energy prediction value is represented as:

[0068] , wherein, The initial available energy can be calculated by multiplying the battery rated capacity and the current battery state of charge; The time interval, i.e. the prediction time step, is consistent with the time resolution of the charge and discharge power prediction value; The charge and discharge efficiency is <1, representing the energy conversion loss existing in the charging process; when discharging, >1, representing the additional energy loss that needs to be compensated in the discharging process.

[0069] The available energy prediction value reflects the theoretical maximum dischargeable amount of the energy storage system at future k time, and provides key energy boundary constraints for aspects such as daily scheduling planning of the virtual power plant or backup capacity assessment. Based on the prediction value, the system can dynamically evaluate the continuous power supply capacity of the energy storage system, optimize the energy distribution scheme of multiple time scales, and ensure the reliability and economy of the virtual power plant operation.

[0070] As an optional embodiment, a scheduling instruction is generated based on the energy storage charge and discharge power prediction value, the photovoltaic power generation prediction value and real-time data information, including: A first adjustment index is generated based on the available energy prediction value, the energy storage charge and discharge power prediction value and the photovoltaic power generation prediction value; A second adjustment index is generated based on the load demand information of the power system and the market electricity price information; A target income function is constructed according to the first adjustment index and the second adjustment index; A genetic algorithm is used to solve the target income function under the target constraint to obtain a target operation efficiency; and a scheduling instruction is generated based on the target operation efficiency.

[0071] Specifically, a first adjustment index is generated based on the available energy prediction value, the energy storage charge and discharge power prediction value and the photovoltaic power generation prediction value, which is used to evaluate the internal energy supply and demand balance state of the virtual power plant; a second adjustment index is generated based on the load demand information of the power system and the market electricity price information, which is used to evaluate the external market environment and grid demand; further, a target income function is constructed according to the first adjustment index and the second adjustment index, and the specific form of the target income function is:

[0072] wherein, is the data center net load at time t, In the embodiment of the present application, the load reference value at the same time of the previous five days of the data center is preferred; is the unit price of the demand side response peak shaving subsidy participated by the virtual power plant.

[0073] In the embodiment of the present application, the target constraint condition preferably considers that the data center IT equipment and heating equipment load of the data center accounts for 80-90% of the total load, so the influence of the IT equipment load and the air conditioning load on the data center net load is mainly considered, according to the recommended operation temperature range of the data center set by the TIA-942 standard, which is 20-27℃, the air conditioning working power range is determined by measuring the air conditioning power load of the data center at 20℃ and 27℃. wherein, corresponds to the power at the operating condition of 27℃, corresponds to the power at the operating condition of 20℃; and the data center net load calculation formula is:

[0074] wherein, is the air conditioning working power; is the current operating power of the data center IT equipment, which is preferred to be a non-adjustable load in the embodiment of the present application; The value range of is [0, +α], wherein α is the maximum photovoltaic power generation power; the energy storage output power is positive when discharging, and the value range is [0, +β1], β1 is the maximum discharging power of the energy storage, and the energy storage output power is negative when charging, and the value range is [-β2, 0], -β2 is the maximum charging power of the energy storage.

[0075] It should be noted that the net load value is positive when the data center is powered by the power grid, and the net load value is negative when the data center can send power back to the power grid.

[0076] Further, the system operation also needs to consider the distributed photovoltaic power generation constraint, the energy storage system charging and discharging constraint, and the data center response capability constraint, which specifically includes: Photovoltaic power generation power constraint: ; Energy storage energy balance constraint: ; Air conditioning load power adjustment range constraint: ; wherein, is the output information of the photovoltaic prediction model at time t; α is the maximum photovoltaic power generation power; is the remaining energy of the energy storage system at time t; is the residual energy of the energy storage system at the previous time point of time t; is the flexible regulation power of the energy storage system at time t, and when the value is positive, it indicates that the energy storage system is in the discharging state, and when the value is negative, it indicates that the energy storage system is in the charging state; is a time interval; is a charging and discharging efficiency loss coefficient, for example, when 0.01 represents a loss of 1%; is the working power of the air conditioner; is the power of the air conditioner in the operating condition of 27 DEG C; is the power of the air conditioner in the operating condition of 20 DEG C.

[0077] Further, a genetic algorithm is used to solve the target benefit function under the target constraint to obtain a target operation efficiency that maximizes the benefit of the virtual power plant, and then a scheduling instruction is generated based on the target operation efficiency, so that the operation efficiency when the target operation efficiency is the maximum benefit, for example, corresponding control instructions are sent to the energy storage device and the air conditioning system in the virtual power plant, to realize the optimized scheduling of the power; the energy storage system operates according to the charging and discharging power plan, and the air conditioning system operates according to the set temperature parameter.

[0078] As an optional embodiment, the present application also proposes to monitor the execution effect of the scheduling strategy in real time, collect actual power generation data, load data, power grid operation data, etc., and compare and analyze them with the predicted data and the expected scheduling effect; feedback information is input into the prediction model to update and optimize the model online, and a rolling optimization strategy is used to improve the prediction accuracy of the model and the adaptability of the scheduling strategy.

[0079] As an optional embodiment, the present application also proposes to establish an evaluation index system to evaluate the scheduling strategy from multiple dimensions such as economy and reliability.

[0080] Specifically, at the economic level, the market profitability of the virtual power plant is significantly improved by accurately assessing the demand side response benefit and energy cost saving; at the reliability level, the continuous and reliable power supply of important loads such as data centers is effectively guaranteed by monitoring key indicators such as power supply interruption probability and load satisfaction rate. Through the evaluation system, scientific basis can be further provided for the optimization of the scheduling strategy, and systematic protection can be provided for the efficient, reliable and sustainable development of the virtual power plant in the power market environment.

[0081] The power data adjustment method provided by the embodiment of the present application can train the photovoltaic prediction model and the energy storage prediction model according to various influencing factors, input the real-time data information of the power system obtained into the trained prediction model to obtain the photovoltaic power generation prediction value and the energy storage charging and discharging power prediction value, and further generate the corresponding scheduling instruction according to the prediction value, so as to realize the real-time adjustment of the operation efficiency of the target device according to the scheduling instruction, effectively improve the total operation benefit of the virtual power plant in the case of ensuring the data security and stability, and improve the reliability and feasibility of the scheduling strategy.

[0082] Reference Figure 5 The schematic diagram of the power data adjustment device provided by the embodiment of the present application is shown.

[0083] Based on the same inventive concept, the present application further provides a power data adjustment device 500 corresponding to the power data adjustment method of any of the above embodiments, which comprises a data information acquisition module 510, a photovoltaic data prediction module 520, an energy storage data prediction module 530, a scheduling instruction generation module 540 and a data adjustment module 550.

[0084] The data information acquisition module 510 is configured to acquire real-time data information of a power system; the photovoltaic data prediction module 520 is configured to input the real-time data information into a trained photovoltaic prediction model and obtain a photovoltaic power generation prediction value; the energy storage data prediction module 530 is configured to input the real-time data information and the photovoltaic power generation prediction value into a trained energy storage prediction model and obtain an energy storage charging and discharging power prediction value; the scheduling instruction generation module 540 is configured to generate a scheduling instruction based on the energy storage charging and discharging power prediction value, the photovoltaic power generation prediction value and the real-time data information; and the data adjustment module 550 is configured to adjust the energy storage charging and discharging power and the air conditioner operation efficiency in real time based on the scheduling instruction.

[0085] In some embodiments, the power data adjustment device 500 is further configured to, before acquiring the real-time data information of the power system, acquire historical data information of the power system; perform cleaning processing on the historical data information and obtain cleaned data information; perform interpolation processing on the cleaned data information and obtain interpolated data information; perform normalization processing on the interpolated data information and obtain normalized data information; input the normalized data information into the photovoltaic prediction model for training and obtain a trained photovoltaic prediction model; and input the normalized data information and the output information of the trained photovoltaic prediction model into the energy storage prediction model for training and obtain a trained energy storage prediction model; wherein the output information at least includes the photovoltaic power generation prediction value.

[0086] In some embodiments, the normalized data information comprises historical battery state of charge and historical load prediction data; the power data conditioning device 500 is further configured to: input the normalized data information and output information of the trained photovoltaic prediction model into an energy storage prediction model for training, and obtain a trained energy storage prediction model, construct an input feature vector based on the output information, the historical battery state of charge and the historical load prediction data; input the input feature vector into a preset machine learning model to obtain an initial prediction error; take the initial prediction error as a reference, and use an improved particle swarm optimization algorithm to perform offline optimization on a penalty factor and a kernel function width of the machine learning model, and obtain an optimal hyperparameter combination; retrain the machine learning model based on the optimal hyperparameter combination to obtain the trained energy storage prediction model.

[0087] In some embodiments, the real-time data information comprises meteorological data information and photovoltaic output information; the photovoltaic data prediction module 520 is further configured to: extract spatial features of the meteorological data information to obtain meteorological spatial features; determine time sequence features of photovoltaic output based on the photovoltaic output information, and determine a dependency relationship between the time sequence features and the meteorological data information; based on the meteorological spatial features, the time sequence features and the dependency relationship, obtain a photovoltaic power generation prediction value.

[0088] In some embodiments, the real-time data information further comprises a battery state of charge and load prediction data; the energy storage data prediction module 530 is further configured to: based on the load prediction data and the photovoltaic power generation prediction value, obtain an energy storage charging and discharging power prediction value, which is expressed as:

[0089] wherein, is the energy storage charging and discharging power prediction value; is a battery SOC at a previous moment; is a time feature; is output information of the photovoltaic prediction model at time t; is historical load prediction data.

[0090] In some embodiments, after obtaining the energy storage charging and discharging power prediction value, the energy storage data prediction module 530 is further configured to:

[0091] wherein, is an available energy prediction value; is an initial available energy; is a charging and discharging power prediction value sequence; is a time interval; is a charging and discharging efficiency.

[0092] In some embodiments, the scheduling instruction generation module 540 is further configured to generate a first adjustment index based on the available energy prediction value, the energy storage charge-discharge power prediction value, and the photovoltaic power generation prediction value; generate a second adjustment index based on load demand information of the power system and market electricity price information; construct a target benefit function according to the first adjustment index and the second adjustment index; solve the target benefit function under the target constraint by using a genetic algorithm to obtain a target operation efficiency; and generate the scheduling instruction based on the target operation efficiency.

[0093] The power data adjustment device provided by the embodiment of the present application can train the photovoltaic prediction model and the energy storage prediction model according to various influencing factors, input the real-time data information of the power system obtained to the trained prediction model to obtain the photovoltaic power generation prediction value and the energy storage charge-discharge power prediction value, and further generate the corresponding scheduling instruction according to the prediction values, so as to realize the real-time adjustment of the operation efficiency of the target equipment according to the scheduling instruction, effectively improve the total operation benefit of the virtual power plant in the case of guaranteeing the data security and stability, and improve the reliability and feasibility of the scheduling strategy.

[0094] Based on the same concept, the present application further provides an electronic device, which comprises a processor and a memory, the memory stores programs or instructions that can be run on the processor, and the programs or instructions are executed by the processor to realize the power data adjustment method.

[0095] Figure 6 A more specific electronic device hardware structure schematic diagram provided by the embodiment of the present application is shown, which can comprise a processor 610, a memory 620, an input / output interface 630, a communication interface 640, and a bus 650. The processor 610, the memory 620, the input / output interface 630, and the communication interface 640 are connected to each other through the bus 650 for internal communication.

[0096] The processor 610 can be implemented by a general CPU (Central Processing Unit, central processor), a microprocessor, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, etc., for executing related programs to realize the technical solutions provided by the embodiments of the present application.

[0097] The memory 620 can be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 620 can store an operating system and other application programs, and when the technical solutions provided by the embodiments of the present specification are implemented by software or firmware, the related program codes are stored in the memory 620 and are called and executed by the processor 610.

[0098] The input / output interface 630 is configured to connect an input / output module to realize information input and output. The input / output module can be configured as a component in the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. The input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0099] The communication interface 640 is configured to connect a communication module (not shown in the figure) to realize the communication interaction between the device and other devices. The communication module can realize communication through a wired manner (such as USB, network cable, etc.) or through a wireless manner (such as mobile network, WIFI, Bluetooth, etc.).

[0100] The bus 650 includes a channel for transmitting information between various components (such as the processor 610, the memory 620, the input / output interface 630, and the communication interface 640) of the device.

[0101] It should be noted that although the above device only shows the processor 610, the memory 620, the input / output interface 630, the communication interface 640, and the bus 650, in the specific implementation process, the device can also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device can also only contain the components necessary to implement the embodiments of the present specification, and does not have to contain all the components shown in the figure.

[0102] The electronic device of the above embodiments is used to implement the corresponding power data adjustment method in any of the preceding embodiments, and has the beneficial effects of the corresponding power data adjustment method embodiments, which are not described here.

[0103] Based on the same concept, corresponding to the power data adjustment method provided by any of the above embodiments, the present application also provides a computer readable storage medium, and the above readable storage medium stores programs or instructions, and the programs or instructions are executed by the processor to realize the steps of the above power data adjustment method.

[0104] The computer readable storage medium described above can be any available medium or data storage device that can be accessed by a computer, including but not limited to a magnetic storage (such as a floppy disk, a hard disk, a magnetic tape, a magneto-optical disk (MO), etc.), an optical storage (such as a CD, a DVD, a BD, a HVD, etc.), and a semiconductor storage (such as a ROM, an EPROM, an EEPROM, a NAND FLASH, a solid state disk (SSD)), etc.

[0105] The computer instructions stored in the storage medium of the above-described embodiments are used to make the computer execute the corresponding power data adjustment method in any of the foregoing embodiments, and have the beneficial effects of the corresponding power data adjustment method embodiment, which are not described herein again.

[0106] It should be noted that, in this document, the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article, or device. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article, or device that includes the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, but can also include performing functions in a substantially simultaneous manner or in reverse order, for example, the described method can be performed in an order different from that described, and various steps can also be added, omitted, or combined. In addition, the features described with reference to certain examples can be combined in other examples.

[0107] From the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of computer software product and general hardware platform as necessary, of course, it can also be realized by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.), including a plurality of instructions, used to make the terminal or network side device execute the method described in each embodiment of the present application.

[0108] The embodiments of the present application are described above in combination with the drawings, but the present application is not limited to the above-described specific embodiments, which are only illustrative and not limiting. Those skilled in the art can make many forms of embodiments under the inspiration of the present application without departing from the purpose of the present application and the scope protected by the claims, and these embodiments all belong to the protection of the present application.

Claims

1. A method for regulating power data, characterized in that, include: Obtain real-time data information from the power system; The real-time data information is input into the trained photovoltaic prediction model to obtain the predicted value of photovoltaic power generation; The real-time data information and the predicted photovoltaic power generation are input into the trained energy storage prediction model to obtain the predicted energy storage charging and discharging power. Dispatch instructions are generated based on the predicted energy storage charging and discharging power, the predicted photovoltaic power generation, and the real-time data information. The operating efficiency of the target equipment is adjusted in real time based on the scheduling instructions.

2. The power data adjustment method according to claim 1, characterized in that, Before acquiring real-time data information of the power system, the method further includes: Obtain historical data information of the power system; The historical data information is cleaned to obtain cleaned data information; The cleaned data is then interpolated to obtain the interpolated data. The interpolated data is then normalized to obtain normalized data. The normalized data information is input into the photovoltaic prediction model for training, and the trained photovoltaic prediction model is obtained. The normalized data and the output of the trained photovoltaic prediction model are input into the energy storage prediction model for training, and the trained energy storage prediction model is obtained. The output information includes at least the predicted value of photovoltaic power generation.

3. The power data adjustment method according to claim 2, characterized in that, The normalized data information includes historical battery state of charge and historical load prediction data. The step of inputting the normalized data information and the output information of the trained photovoltaic prediction model into the energy storage prediction model for training, and obtaining the trained energy storage prediction model, includes: Based on the output information, the historical battery state of charge, and the historical load prediction data, an input feature vector is constructed. The input feature vector is input into a preset machine learning model to obtain the initial prediction error; Based on the initial prediction error, an improved particle swarm optimization algorithm is used to optimize the penalty factor and kernel function width of the machine learning model offline, and the optimal hyperparameter combination is obtained. The machine learning model is retrained based on the optimal hyperparameter combination to obtain the trained energy storage prediction model.

4. The power data adjustment method according to claim 1, characterized in that, The real-time data information includes meteorological data and photovoltaic power output information; The real-time data information is input into the trained photovoltaic prediction model to obtain the predicted photovoltaic power generation value, including: Spatial features are extracted from the meteorological data to obtain meteorological spatial features; Based on the photovoltaic output information, the time series characteristics of photovoltaic output are determined, and the dependency relationship between the time series characteristics and the meteorological data information is determined. Based on the meteorological spatial characteristics, the time series characteristics, and the dependency relationship, the predicted value of photovoltaic power generation is obtained.

5. The power data adjustment method according to claim 4, characterized in that, The real-time data information also includes battery state of charge and load forecast data; The step of inputting the real-time data information and the photovoltaic power generation prediction value into the trained energy storage prediction model to obtain the energy storage charging and discharging power prediction value includes: Based on the load forecast data and the photovoltaic power generation forecast, the energy storage charging and discharging power forecast is obtained, and the energy storage charging and discharging power forecast is expressed as follows: in, The predicted value of the energy storage charging and discharging power; This represents the battery's state of charge (SOC) at the previous moment. It is a time-related feature; This refers to the output information of the photovoltaic prediction model at time t. This is historical load forecast data.

6. The power data adjustment method according to claim 5, characterized in that, After obtaining the predicted energy storage charging and discharging power value, the method further includes: Based on the predicted energy storage charge / discharge power and the battery state of charge, a predicted value of available energy is obtained, which is expressed as follows: in, The predicted value of available energy; Initial available energy; This is a sequence of predicted charge and discharge power values; For time intervals; This refers to the charge / discharge efficiency.

7. The power data adjustment method according to claim 6, characterized in that, The process of generating scheduling instructions based on the predicted energy storage charging and discharging power, the predicted photovoltaic power generation, and the real-time data information includes: A first adjustment index is generated based on the predicted value of available energy, the predicted value of energy storage charging and discharging power, and the predicted value of photovoltaic power generation. A second regulation indicator is generated based on the load demand information of the power system and the electricity price information in the market. Construct a target return function based on the first adjustment index and the second adjustment index; The target reward function is solved using a genetic algorithm under the target constraints to obtain the target operating efficiency. Scheduling instructions are generated based on the target operating efficiency.

8. A power data regulation device, characterized in that, include: The data acquisition module is configured to acquire real-time data information from the power system. The photovoltaic data prediction module is configured to: input the real-time data information into the trained photovoltaic prediction model and obtain the predicted value of photovoltaic power generation; The energy storage data prediction module is configured to input the real-time data information and the photovoltaic power generation prediction value into the trained energy storage prediction model, and obtain the energy storage charging and discharging power prediction value. The scheduling instruction generation module is configured to generate scheduling instructions based on the predicted energy storage charging and discharging power, the predicted photovoltaic power generation, and the real-time data information. The data adjustment module is configured to adjust the energy storage charging and discharging power and the air conditioning operating efficiency in real time based on the scheduling instructions.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the power data regulation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing computer instructions for causing a computer to perform the power data regulation method according to any one of claims 1 to 7.

Images