Source network load storage operation multi-target planning method based on prediction increment updating

Abstract

The invention discloses a source network load storage operation multi-target planning method based on prediction increment updating, and belongs to the technical field of electricity market day-ahead electricity price prediction. And outputting a source network load storage prediction sequence through a source network load storage prediction data set in combination with a long short-term memory network model, then outputting a source network storage control scheme through a multi-target optimization model, and finally inputting source network load storage operation error data into the long short-term memory model, and performing weight updating on the long short-term memory model by using an error data set. Compared with the prior art, the method has the advantages that the stability influence of load fluctuation on the power system is reduced, the problem of operation lag that planning scheduling lags behind the load fluctuation is solved, high fineness and high efficiency of source-grid-load-storage cooperative operation are guaranteed, the defect of control lag in related technologies is effectively overcome, and the method is suitable for large-scale popularization and application. And the intelligent prediction model can be updated according to the real-time operation data, so that the sensitivity of the model to the real-time data is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of day-ahead electricity price prediction in the power market, and specifically is a multi-objective planning method for source-grid-load-storage operation based on prediction increment update. Background Art

[0002] With the continuous advancement of energy technology, human beings have higher and higher requirements for the operation and management of power systems. As an important part of the power industry, the source-grid-load-storage system has extremely strict requirements on operation efficiency and stability. The application of the source-grid-load-storage system in the power industry is reflected in multiple levels, including power generation, transmission, power consumption and energy storage. The coordinated operation of these links is crucial to ensure the safety and efficiency of the power system.

[0003] The operation planning of the source-grid-load-storage system is a key issue. As the guarantor of the stable operation of the power system and the key link in the efficient use of new energy, it is necessary to formulate the most stringent operation standards and continue to explore innovative planning technologies. The effective management of the source-grid-load-storage system not only involves the generation and distribution of electricity, but also includes demand-side management, scheduling of energy storage equipment and other aspects, all of which need to be considered comprehensively to achieve the optimal operation of the entire system. The operation planning of the source-grid-load-storage system needs to comprehensively consider technical, economic, environmental and other factors to achieve the sustainable development of the power system.

[0004] Traditional planning and operation methods for source-grid-load-storage systems usually do not fully utilize the effective information contained in massive data, and cannot effectively extract the time series and nonlinear characteristics in the data. These problems lead to a decrease in the accuracy and reliability of the prediction results, further causing the planning scheme to lag behind load fluctuations, resulting in network instability and reduced production efficiency. Summary of the invention

[0005] The purpose of the present invention is to address the problem that the existing source-grid-load-storage system planning and operation methods cannot effectively extract the time series characteristics and nonlinear characteristics in the data, resulting in the planning scheme lagging behind the load fluctuation, which in turn leads to network instability and reduced production efficiency. A source-grid-load-storage operation multi-objective planning method based on predicted incremental update is proposed. By collecting various data, the data is preprocessed and a source-grid-load-storage prediction data set is constructed. Then, the constructed long-short-term memory network model is combined, and the training boundary is set according to the equipment operating conditions, and the source-grid-load-storage prediction sequence is output. Subsequently, the source-grid-load-storage prediction sequence is input into the multi-objective optimization model, and the source-grid-load-storage control scheme is output to perform source-grid-load-storage coordinated control operation; finally, the source-grid-load-storage operation error data is calculated and an error data set is constructed, which is input into the long-short-term memory model for weight update and continuous iteration of the long-short-term memory model to output the latest source-grid-storage control scheme. Compared with the existing technology, this method reduces the impact of load fluctuations on the stability of the power system, solves the operation lag problem where planning and scheduling lag behind load fluctuations, ensures the high precision and efficiency of the coordinated operation of source, grid, load and storage, effectively overcomes the shortcomings of control lag in related technologies, and can also update the intelligent prediction model according to real-time operation data, ensuring the sensitivity of the model to real-time data.

[0006] In order to solve the above technical problems, the present invention provides a multi-objective planning method for source-grid-load-storage operation based on prediction increment update, comprising:

[0007] S1. Collect historical source-grid-load-storage equipment operation data, wind-solar-temperature data, and equipment performance data, pre-process the data, and construct a source-grid-load-storage prediction data set;

[0008] S2. Build a long short-term memory network model, use the source-grid-load-storage prediction data set for training, and set the training boundary to output the load prediction sequence and the source-grid-storage prediction sequence according to the equipment operating conditions;

[0009] S3, input the load forecast sequence and the source-grid-storage forecast sequence into a multi-objective optimization model, output a source-grid-storage control scheme, and perform source-grid-load-storage coordinated control operation;

[0010] S4. Calculate the source-grid-load-storage operation error data and construct an error data set, input it into the long-short-term memory model for weight update, continuously iterate the long-short-term memory model, and output the latest source-grid-storage control plan.

[0011] Furthermore, in S1, the operation data of the source-grid-load-storage equipment, the wind-solar-air temperature data, and the equipment performance data include: all historical operation data of the power source, the transmission and distribution network, the load, and the energy storage equipment, and also include the wind and light data that affect the power generation of photovoltaic equipment and wind turbines, the weather and temperature data that affect load fluctuations, and the equipment performance data of the source-grid-load-storage. The matrix representation of each specific data is as follows:

[0012] D t =[P cell ,N cell ,L cell , B cell ,Pv,Wi,E cell ]; among them, D t represents the data at time t, P cell Indicates the output status of the power supply set, N cell Indicates the operation status of the transmission and distribution network, L cell Indicates the power consumption of the load, B cell represents the operating status of the energy storage device, Pv is the light intensity within the source-grid-load-storage system at that moment, Wi is the wind speed within the source-grid-load-storage system, and E cell It is the environmental weather factor within the scope of the source-grid-load-storage system.

[0013] Preferably, in S1, data preprocessing includes: text quantization, missing value check and data integrity check. First, the text in the data is segmented, banned words are removed and stem extraction is performed; then the text data is converted into numerical form using a text vectorization method, and data cleaning and normalization operations are performed at the same time to exclude abnormal values. If there are missing values, polynomial interpolation is used to supplement them.

[0014] Preferably, the construction of the long short-term memory network model comprises the following steps:

[0015] Sa, build the input layer;

[0016] Sb, build a multi-layer LSTM network layer, including determining the number of network layers, the number of units in each layer, and the input / output dimensions;

[0017] Sc, design the training process, including selecting activation functions and designing loss functions;

[0018] Sd, adjust the training parameters, including selecting the optimizer and configuring the callback function;

[0019] Se. Use the output layer of the last LSTM layer as the output of the LSTM network model.

[0020] Preferably, S2 includes: inputting wind, solar, and air temperature data and equipment operation boundaries into a pre-trained long short-term memory network model to obtain a load forecast sequence and a source-grid-storage forecast sequence.

[0021] Preferably, S3 includes:

[0022] S31, inputting the source-grid-load-storage prediction sequence into the multi-objective optimization model, and using the source-grid-load-storage prediction results as optimization boundary conditions;

[0023] S32, calculating a target planning scheme according to boundary conditions, the scheme including the power generation capacity of each power source, the power transmission and distribution network flow, and the energy storage configuration;

[0024] S33. Verify the feasibility of the above schemes, select feasible schemes, and apply them to the coordinated operation of source, grid, load and storage.

[0025] Preferably, in S33, the feasibility verification includes: verifying the power supply and demand balance constraint conditions, and excluding planning schemes that do not meet the power constraints; verifying the stability of the source-grid-load-storage system, and excluding planning schemes that do not meet the system stability; verifying the feasibility of emergency response, and excluding planning schemes that do not meet the feasibility of emergency response; verifying the cost feasibility, and excluding planning schemes that do not meet the cost feasibility; sorting the schemes that meet all constraints according to the verification condition weighted matrix, selecting the optimal scheme, and applying it to the operation of the power system source-grid-load-storage system.

[0026] Preferably, S4 comprises the following steps:

[0027] S41, comparing the source-grid-load-storage operation result with the load forecast sequence and the source-grid-storage forecast sequence, obtaining an error sequence, taking the actual sequence as a reference value, calculating the error fluctuation radius, and determining the error action range;

[0028] S42. Determine the error fluctuation radius. If the fluctuation radius is less than or equal to the operation stability threshold, the long short-term memory network parameters are not updated. If the fluctuation radius is greater than the operation stability threshold, the error sequence corresponding to the fluctuation is used as the training data of the error data set to update the long short-term memory network parameters.

[0029] Preferably, the calculation formula of the error sequence is:

[0030] C i =P i -R i ; Among them, the P sequence is the prediction result sequence including the load prediction sequence and the source network storage prediction sequence, and the R sequence is the real-time operation sequence including the load real-time sequence and the source network storage operation sequence;

[0031] The calculation formula of error fluctuation radius is:

[0032] Among them, the value l is the sequence characteristic number. First, the absolute value of the error sequence is summed up with the real-time operation sequence to find the percentage of the error capacity in the total operation capacity, and then multiply it by the maximum load power;

[0033] The calculation formula for the running stability threshold is:

[0034] Among them, μ is the source-grid-load-storage scale coefficient. According to the power supply capacity, it is divided into small source-grid-load-storage system, medium source-grid-load-storage system, large source-grid-load-storage system and super large source-grid-load-storage system, where μ is 1, 10 and 10 respectively. -1 , 10 -2 , 10 -3 .

[0035] A multi-objective planning system for source-grid-load-storage operation based on incremental update of intelligent prediction models, the system includes a data acquisition and processing module, which collects relevant operation data of target equipment during the operation of source-grid-load-storage equipment to form a source-grid-load-storage comprehensive data set, the data set including the operation data of the source-grid-load-storage equipment, wind-solar-light temperature data and equipment performance data in the past year; a model training module, the model training module uses the constructed pseudo source-grid-load-storage comprehensive data set to train a long-short-term memory model and obtain a pre-trained long-short-term memory model; a sequence prediction module: the sequence prediction module uses the wind-solar-light temperature and time series data of the day to predict the operation status of the source-grid-load-storage; an operation planning module: the operation planning module uses the predicted source-grid-load-storage sequence as a boundary condition, performs multi-objective optimization of the operation of the power grid, applies the optimal solution to the power operation control, reads the data after operation to calculate the error fluctuation radius, and compares it with the operation stability threshold to determine whether to transmit the data back to the model training module.

[0036] An electronic device includes a central processing unit, a storage medium, and a computer executable program that can be run on the central processing unit to implement the multi-objective planning method for source-grid-load-storage operation based on incremental update of an intelligent prediction model as described above.

[0037] A high-reading speed storage medium stores executable computer programs, test files and other related contents. The program and test files can be executed by a central processing unit to implement the above-mentioned multi-objective planning method for source-grid-load-storage operation based on incremental update of intelligent prediction model.

[0038] Beneficial effects of the present invention:

[0039] 1. Compared with the existing technology, this solution reduces the impact of load fluctuations on the stability of the power system, solves the operation lag problem of planning and scheduling lagging behind load fluctuations, ensures the high precision and high efficiency of the coordinated operation of source, grid, load and storage, effectively overcomes the shortcomings of control lag in related technologies, and can also update the intelligent prediction model according to real-time operation data, ensuring the sensitivity of the model to real-time data.

[0040] 2. This solution can use recent operating data to predict the operating demand and power generation capacity of source, grid, load and storage based on long short-term memory network, and perform multi-objective optimization of source, grid, load and storage equipment based on the predicted source, grid, load and storage sequence, and update the intelligent prediction model incrementally in real time according to the optimized operation results. At the same time, this method combines the deep neural network structure for training, increases the data-driven intensity of the planning model, and enables it to have a certain intelligent judgment ability.

[0041] 3. Compared with the prior art, this scheme evaluates the accuracy of the prediction model by calculating the difference between the prediction result sequence (P sequence) and the real-time operation sequence (R sequence). The error fluctuation radius is used as an indicator to quantify the impact of the prediction error on the network fluctuation. According to the different power supply, the source-grid-load-storage system is divided into small, medium, large and super-large, and different scale coefficients are specified for each scale. According to the scale of the source-grid-load-storage system, different scale coefficients (μ) are multiplied by the sum of the load values ​​to determine the operation stability threshold. By continuously analyzing the error sequence, evaluating the error fluctuation radius and judging the relationship with the size of the operation stability threshold, the prediction model is continuously updated iteratively based on this. Its beneficial effect is that the judgment process can quantify the parameter sensitivity of the prediction model, identify the deviation of the model in the prediction process, and provide a basis for the adjustment of the model parameters; provide a quantitative indicator to evaluate the impact of the prediction error on the actual operation, which helps to identify and deal with the situation of large prediction errors; can set a suitable fluctuation tolerance range according to systems of different scales, so that the update of the prediction model is more adapted to the actual operating conditions; make the setting of the operation stability threshold more refined, which helps to improve the applicability and accuracy of the prediction model in systems of different scales. This method improves the accuracy of the prediction model and the operational stability of the system through refined error analysis and system scale adaptability design. In particular, it provides an effective prediction verification and adjustment mechanism for the key link of operational stability judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0043] Figure 1 A flow chart of the method of this application;

[0044] Figure 2 A flowchart of the training and use of the method of the embodiment of the present application;

[0045] Figure 3 A structural flow chart of the system of an embodiment of the present application;

[0046] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application;

[0047] Figure 5 A schematic diagram of the structure provided for an embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0049] Example 1: Figure 1 As shown in FIG. 1 , the multi-objective planning method for source-grid-load-storage operation based on prediction incremental update includes: Figure 1 This is a flowchart of a multi-objective planning method for source-grid-load-storage operation based on incremental update of an intelligent prediction model according to an embodiment of the present application, which mainly includes the following steps:

[0050] Step S1: Construct a comprehensive data set of source, grid, load and storage, including collecting the operation data of source, grid, load and storage equipment, wind, solar, temperature data and equipment performance data in the past year, and preprocessing the data. Based on the above data, a comprehensive data set of source, grid, load and storage is constructed to form a source, grid, load and storage prediction data set suitable for long-term and short-term memory network model training.

[0051] It is understandable that data preprocessing will be performed when constructing a data set. The preprocessing operations mainly include text quantization, missing value verification and data integrity verification. First, the text in the data is segmented, banned words are removed and stem extraction operations are performed. The above steps can extract keywords and feature information. Then the text data is converted into a numerical form using a text vectorization method, and data cleaning and normalization operations are performed at the same time to exclude abnormal values. If there are missing values, polynomial interpolation is used to supplement them, and data normalization uses the Max-Min normalization method. The purpose of normalization is to normalize the numerical range of physical quantities of different scales to between 0-1, so as to speed up the convergence speed of the neural network, speed up the training speed of the network, and improve the fault prediction accuracy of the deep neural network. Specifically, the embodiment of the present application uses the Max-Min normalization method. The calculation formula is:

[0052] The largest value in the feature dimension is x max , the smallest value is x min , the standardized data is x^.

[0053] The specific structure of the data structure is shown in the formula:

[0054] D t =[P cell ,N cell ,L cell ,B cell ,Pv,Wi,E cell ].

[0055] Step S2: Build a long short-term memory network model, train it using the above dataset, and set the training boundary according to the device operating conditions.

[0056] Furthermore, the steps of constructing the long short-term memory network model include:

[0057] Build the input layer; build a multi-layer LSTM network layer, including determining the number of network layers, the number of units in each layer, the input / output dimensions, etc.; design the training process, including selecting activation functions, designing loss functions, etc.; adjust training parameters, including selecting optimizers, configuring callback functions, etc.; use the output layer of the last LSTM layer as the output of the LSTM network model.

[0058] Furthermore, in the step of constructing the input layer, the input layer specifies the format of the input data, including the batch size, training rounds, and feature dimensions. Generally, the batch size is set to 10, the training rounds are set to 50, and the feature dimension is set to n. Then a sample in the batch data is represented as a real number sequence matrix R 50×n , remember X ij For R 50×n The vector representation of the jth data in the i-th training round.

[0059] Furthermore, in the step of constructing a multi-layer LSTM network layer, each LSTM network layer contains a cell state and a hidden state. The cell state is used to store long-term information and pass information to the next layer to alleviate the gradient disappearance; the hidden state is the result of the LSTM network layer output at each time step, which mainly includes three gate processing, namely input gate, forget gate, and output gate. The forget gate can retain or abandon the memory. The input gate will add the stored information in the cell state to the current retained memory, and the output gate can output the information of the above operations. The hidden state is mainly used for the short-term memory operation of the network state at the current time step. It is calculated by the cell state and the input data, and will be passed to the next time step for prediction or generation; the output of the previous LSTM layer is used as the input of the next LSTM layer, and is passed down layer by layer.

[0060] Furthermore, in the design of the training process, the activation function selection includes Tanh, Sigmoid, ReLU, etc. In this example, Tanh is used as the activation function, and its function form is:

[0061] Its output range is between (-1,1);

[0062] The choices of loss functions include mean square error (MSE), cross entropy and sequence loss. In this example, sequence loss combined with cross entropy is used as the loss function, and its function form is:

[0063]

[0064] Where T′ is the length of the target sequence, Loss t is the cross entropy loss at time step t.

[0065] Furthermore, in the step of adjusting the training parameters, the optimizer options include SGD, NAG, Momentum, etc. In this example, SGD is used as the optimizer, and its update rule can be expressed as:

[0066]

[0067] where θ t is the parameter value at the tth iteration, η is the learning rate, a positive constant that determines the step size of each update, is the parameter θ t The gradient of the objective function J at .

[0068] The callback function selection includes data recording, parameter checking, model saving, etc. In this example, the model saving function and data recording function are used. The above dual prediction sequence results are input into the multi-objective optimization model, and the source network storage control plan is output to perform source network collaborative control operation.

[0069] Step S3: the pre-trained long short-term memory network model outputs a source grid load storage prediction sequence, including: inputting the wind, solar, and air temperature data of the day and the equipment operation boundary into the pre-trained long short-term memory network model; the long short-term memory network model outputs the prediction sequence results of the source grid load storage.

[0070] Step S4: Input the above-mentioned dual prediction sequence results into the multi-objective optimization model, output the source-grid-storage control plan, and perform source-grid-load-storage coordinated control operation, including: input the source-grid-load-storage prediction sequence into the multi-objective optimization model, and use the source-grid-load-storage prediction results as optimization boundary conditions; calculate the target planning plan according to the boundary conditions, the plan includes the power generation capacity of each power source, the transmission and distribution network flow, energy storage configuration, etc.; verify the feasibility of the above plans, select feasible plans, and apply them to the source-grid-load-storage coordinated operation.

[0071] Furthermore, feasibility verification is carried out to select feasible solutions, including: verifying the power supply and demand balance constraints and excluding planning solutions that do not meet the power constraints; verifying the stability of the source-grid-load-storage system and excluding planning solutions that do not meet the system stability; verifying the feasibility of emergency response and excluding planning solutions that do not meet the feasibility of emergency response; verifying the cost feasibility and excluding planning solutions that do not meet the cost feasibility; sorting the solutions that meet all constraints according to the weighted matrix of verification conditions, selecting the optimal solution, and applying it to the operation of the source-grid-load-storage system of the power system.

[0072] Step S5: Input the source-grid-load-storage operation error data into the long short-term memory model, and update the weight of the long short-term memory model with the error data set.

[0073] Furthermore, the source-grid-load-storage operation results are compared with the predicted operation sequence to obtain the error sequence. The error fluctuation radius is calculated with the actual sequence as a reference to determine the scope of the error. The error fluctuation radius is determined. If the fluctuation radius is greater than the operation stability threshold, the error sequence corresponding to the fluctuation is used as training data to update the long short-term memory network parameters.

[0074] Specifically, the calculation formula of the error sequence is:

[0075] C i =P i -R i ; Among them, the P sequence is the prediction result sequence, including the load prediction sequence and the source-grid-storage prediction sequence, and the R sequence is the real-time operation sequence, including the load real-time sequence and the source-grid-storage operation sequence.

[0076] Specifically, the error fluctuation radius calculation formula is:

[0077]

[0078] The l value is the sequence characteristic number. First, the absolute value of the error sequence is summed with the real-time operation sequence to find the percentage of the error capacity in the total operation capacity, and then multiply it by the maximum load power. The error fluctuation radius can maximize the network fluctuation range caused by the prediction sequence error, and whether the prediction is lower or higher, it deviates from the real-time operation.

[0079] Specifically, the calculation formula for the running stability threshold is:

[0080]

[0081] Among them, μ is the source-grid-load-storage scale coefficient. According to the power supply capacity, it is divided into small source-grid-load-storage system, medium source-grid-load-storage system, large source-grid-load-storage system and super large source-grid-load-storage system, where μ is 1, 10 and 10 respectively. -1 , 10 -2 , 10-3 .

[0082] Specifically, if Figure 2 As shown, the training method for incremental updating of the intelligent prediction model in the embodiment of the present application specifically includes:

[0083] Step S21, obtaining training data. The training data includes all operating data of power sources, transmission and distribution networks, loads, and energy storage equipment within one year, as well as wind and light data that affect photovoltaic equipment power generation and wind turbines, weather and temperature data that affect load fluctuations, and equipment performance data of source, grid, load, and storage.

[0084] Step S22, inputting the sample data into the long short-term memory network model to be trained to obtain a predicted source grid load-storage sequence.

[0085] It can be understood that after obtaining the numerical data in the training set, the sample data is calculated to obtain the cell state and gate state of the sample, and the hidden state value is obtained after adding them. After calculation by the multi-layer LSTM network, the final hidden state is obtained, that is, the predicted source grid load storage sequence.

[0086] Step S23, using the preset loss function, calculate the loss value based on the predicted sequence and the actual sequence

[0087] Step S24, update the LSTM network model parameters according to the loss value, and return to the step of obtaining training data until the loss value meets the preset training completion condition or the training round reaches the upper limit, and obtain the trained LSTM model

[0088] The preset training completion condition may be that the loss function corresponding to the loss value has converged, or that the number of iterations of the training data has reached the target number, etc. For example, the target number may be 1×105 to 2×105, etc., and the embodiment of the present invention does not specifically limit the target number.

[0089] Specifically, according to the loss value, the new network parameters of the long short-term memory network model to be trained are calculated, and then the long short-term memory network model to be trained is updated to obtain an updated long short-term memory network model. Then, the terminal re-inputs the training data into the updated long short-term memory network model, and re-executes the steps of the above embodiment method until the calculated loss value meets the preset training completion condition or the training round reaches the upper limit, thereby obtaining a trained long short-term memory network model.

[0090] Step S25, obtaining the error data set generated after the prediction sequence is used, and continuously updating the parameters of the long short-term memory model.

[0091] Secondly, refer to the accompanying drawings to describe a multi-objective planning system for source-grid-load-storage operation based on incremental updating of an intelligent prediction model proposed in an embodiment of the present application.

[0092] Figure 3 It is a schematic diagram of the use process of a multi-objective planning system for source-grid-load-storage operation based on incremental update of an intelligent prediction model according to an embodiment of the present application.

[0093] like Figure 3 As shown, the multi-objective planning system for source-grid-load-storage operation based on incremental update of intelligent prediction model includes: a data acquisition and processing module 100, a model training module 200, a sequence prediction module 300 and an operation planning module 400.

[0094] Among them, the data acquisition and processing module 100 can collect relevant operating data of the target equipment during the operation of the source grid load storage equipment to form a source grid load storage comprehensive data set. The model training module 200 uses the constructed pseudo source grid load storage comprehensive data set to train the long-term and short-term memory model and obtain a pre-trained long-term and short-term memory model, and can also receive a new error data set to update the parameters of the prediction model. The sequence prediction module 300 uses the wind, light, air temperature and time series data of the day to predict the operating status of the source grid load storage. The operation planning module 400 uses the predicted source grid load storage sequence as the boundary condition to perform multi-objective optimization of the power grid operation, and finally applies the optimal solution to the power operation control, and calculates the error sequence and error fluctuation radius to determine whether the prediction model needs to be updated.

[0095] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application, the electronic device may include:

[0096] Memory 401 , processor 402 , and a computer program stored in the memory 401 and executable on the processor 402 .

[0097] When the processor 402 executes the program, the X-ray image defect detection method based on regional feature extraction provided in the above embodiment is implemented.

[0098] Furthermore, the electronic device further comprises:

[0099] The communication interface 403 is used for communication between the memory 401 and the processor 402 .

[0100] The memory 401 is used to store computer programs that can be executed on the processor 402 .

[0101] The memory 401 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0102] If the memory 401, the processor 402 and the communication interface 403 are implemented independently, the communication interface 403, the memory 401 and the processor 402 can be connected to each other through a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0103] Optionally, in a specific implementation, if the memory 401, the processor 402 and the communication interface 403 are integrated on a chip, the memory 401, the processor 402 and the communication interface 403 can communicate with each other through an internal interface.

[0104] The processor 402 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0105] Figure 5 A schematic diagram of the structure of a multi-objective planning system for source-grid-load-storage operation based on incremental update of an intelligent prediction model provided in an embodiment of the present application is shown in FIG. Figure 5 As shown, the multi-objective planning system 500 for source-grid-load-storage operation based on incremental update of intelligent prediction model includes: data collection and processing equipment 501 and source-grid-load-storage prediction equipment 502, multi-objective operation optimization equipment 503, data storage management equipment 504, and operation status display equipment 505.

[0106] The data collection and processing device 501 uses sensors such as wind, light, temperature, current, and voltage to collect the operating data of the source grid load storage device, wind, light, temperature data, and equipment performance data and pre-process the data; with the data set constructed by the data collection and processing device 501 as input, the source grid load storage prediction device 502 uses GPU and CPU through program instructions to calculate and process the input data, and then outputs the source grid load storage prediction sequence, and can also accept error data sets to update the prediction model parameters; with the source grid load storage prediction sequence as input, the multi-objective operation optimization device 503 calls system resources through program instructions, calculates and processes the input data, and then outputs the source grid load storage operation plan. After the operation plan is executed, the error sequence and error fluctuation radius are output according to the actual operation results to determine whether the source grid load storage stable operation threshold is met. If not, the error data set is used to update the prediction model increment in real time; the source grid load storage operation planning results will be transmitted to the operation status display device 505 and the data storage management device 504. Among them, the operation status display device visualizes the source-grid-load-storage operation planning results, including power generation capacity, energy storage operation status, power grid flow, etc. The operation status display device 505 is controlled by the multi-objective operation optimization device 503 and uses an LED liquid crystal display to complete the display. The data storage management device 504 stores and manages the source-grid-load-storage prediction sequence and operation planning results, which can be read and queried at any time. The system is controlled by a computer program, and the storage medium is a mechanical hard disk.

[0107] The embodiment of the present application also provides a high-reading speed storage medium on which executable computer programs, test files and other related contents are stored. The program and test files can be executed by a central processing unit to implement a multi-objective planning method for source-grid-load-storage operation based on incremental update of an intelligent prediction model.

[0108] In summary, compared with the prior art, the embodiments of the present application have the following advantages:

[0109] 1) It is able to summarize the operation rules of the source-grid-load-storage system under the condition of learning past operation data and apply the rules to the control optimization of the source-grid-load-storage system, so as to sort out the internal relationship of the operation data. At the same time, this method can adjust and optimize the source-grid-storage equipment in advance, so that it can quickly adapt to large load fluctuations in the future.

[0110] 2) This method can significantly improve the operating characteristics of the system, reduce the impact of load fluctuations on the stability of the power system, solve the operational lag problem of planning and scheduling lagging behind load fluctuations, ensure the high precision and high efficiency of the coordinated operation of source, grid, load and storage, and effectively overcome the shortcomings of control lag in related technologies.

[0111] 3) Applicable to a variety of source-grid-load-storage structures. The embodiment of the present application does not limit the source-grid-load-storage equipment, and is theoretically applicable to the power generation and distribution operation control of all source-grid-load-storage equipment. It can be seamlessly combined with the traditional power system operation planning model, which increases the scope of application of the embodiment of the present application.

[0112] This method / system is a multi-objective optimization method / system for source-grid-load-storage, which can use recent operating data to predict the operating demand and power generation capacity of source-grid-load-storage based on long-short-term memory networks, and perform multi-objective optimization of source-grid-load-storage equipment based on the predicted source-grid-load-storage sequence, and can also incrementally update the prediction model in real time according to the actual operating conditions. At the same time, this method combines the deep neural network structure for training, increases the data-driven intensity of the planning model, and enables it to have a certain intelligent judgment ability.

[0113] The above specific implementations are preferred implementations of the present invention, and are not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the present specific implementations. All equivalent changes made in accordance with the shape, structure, and method of the present invention are within the protection scope of the present invention.

Claims

1. A multi-objective planning method for source-grid-load-storage operation based on prediction incremental update, characterized in that: The steps include: S1. Collect historical source-grid-load-storage equipment operation data, wind-solar-temperature data, and equipment performance data, pre-process the data, and construct a source-grid-load-storage prediction data set; S2. Build a long short-term memory network model, use the source-grid-load-storage prediction data set for training, and set the training boundary to output the load prediction sequence and the source-grid-storage prediction sequence according to the equipment operating conditions; S3, input the load forecast sequence and the source-grid-storage forecast sequence into a multi-objective optimization model, output a source-grid-storage control scheme, and perform source-grid-load-storage coordinated control operation; S4. Calculate the source-grid-load-storage operation error data and construct an error data set, input it into the long-short-term memory model for weight update, continuously iterate the long-short-term memory model, and output the latest source-grid-storage control plan.

2. The multi-objective planning method for source-grid-load-storage operation based on prediction increment update according to claim 1 is characterized in that: In S1, the operation data of the source-grid-load-storage equipment, the wind-solar-air temperature data, and the equipment performance data include: all historical operation data of the power source, the transmission and distribution network, the load, and the energy storage equipment, and also include the wind and light data that affect the power generation of photovoltaic equipment and wind turbines, the weather and temperature data that affect load fluctuations, and the equipment performance data of the source-grid-load-storage. The matrix of the specific data is as follows: D t =[P cell ,N cell ,L cell ,B cell ,Pv,Wi,E cell ]; Among them, D t represents the data at time t, P cell Indicates the output status of the power supply set, N cell Indicates the operation status of the transmission and distribution network, L cell Indicates the power consumption of the load, B cell represents the operating status of the energy storage device, Pv is the light intensity within the source-grid-load-storage system at that moment, Wi is the wind speed within the source-grid-load-storage system, and E cell It is the environmental weather factor within the scope of the source-grid-load-storage system.

3. The multi-objective planning method for source-grid-load-storage operation based on prediction increment update according to claim 2 is characterized in that: In S1, the data is preprocessed including: text quantization, missing value check and data integrity check. First, the text in the data is segmented, banned words are removed and stem extraction is performed; then the text data is converted into numerical form using a text vectorization method, and data cleaning and normalization operations are performed at the same time to exclude abnormal values. If there are missing values, polynomial interpolation is used to supplement them.

4. The multi-objective planning method for source-grid-load-storage operation based on prediction increment update according to claim 1 is characterized in that: The construction of the long short-term memory network model comprises the following steps: Sa, build the input layer; Sb, build a multi-layer LSTM network layer, including determining the number of network layers, the number of units in each layer, and the input / output dimensions; Sc, design the training process, including selecting activation functions and designing loss functions; Sd, adjust the training parameters, including selecting the optimizer and configuring the callback function; Se. Use the output layer of the last LSTM layer as the output of the LSTM network model.

5. The multi-objective planning method for source-grid-load-storage operation based on prediction increment update according to claim 1 is characterized in that: The S2 includes: inputting wind, solar, and air temperature data and equipment operation boundaries into a pre-trained long short-term memory network model to obtain a load forecast sequence and a source-grid-storage forecast sequence.

6. The multi-objective planning method for source-grid-load-storage operation based on prediction increment update according to claim 1 is characterized in that: The S3 includes: S31, inputting the source-grid-load-storage prediction sequence into the multi-objective optimization model, and using the source-grid-load-storage prediction results as optimization boundary conditions; S32, calculating a target planning scheme according to boundary conditions, the scheme including the power generation capacity of each power source, the power transmission and distribution network flow, and the energy storage configuration; S33. Verify the feasibility of the above schemes, select feasible schemes, and apply them to the coordinated operation of source, grid, load and storage.

7. The multi-objective planning method for source-grid-load-storage operation based on prediction increment update according to claim 6 is characterized in that: In S33, performing feasibility verification includes: Verify the power supply and demand balance constraints and eliminate planning schemes that do not meet the power constraints; Verify the stability of the source-grid-load-storage system and eliminate planning schemes that do not meet the system stability requirements; Verify the feasibility of emergency response and eliminate planning schemes that do not meet the feasibility requirements of emergency response; Verify cost feasibility and eliminate planning schemes that do not meet cost feasibility requirements; According to the verification condition weighted matrix, the solutions that meet all constraints are sorted, and the optimal solution is selected and applied to the operation of the power system source-grid-load-storage system.

8. The multi-objective planning method for source-grid-load-storage operation based on prediction increment update according to claim 1 is characterized in that: The S4 comprises the following steps: S41, comparing the source-grid-load-storage operation result with the load forecast sequence and the source-grid-storage forecast sequence, obtaining an error sequence, taking the actual sequence as a reference value, calculating the error fluctuation radius, and determining the error action range; S42, determining the error fluctuation radius, if the fluctuation radius is less than or equal to the operation stability threshold, then not updating the long short-term memory network parameters; If the fluctuation radius is greater than the operational stability threshold, the error sequence corresponding to the fluctuation is used as the training data of the error data set to update the long short-term memory network parameters.

9. The multi-objective planning method for source-grid-load-storage operation based on prediction increment update according to claim 8 is characterized in that: The calculation formula of the error sequence is: C i =P i -R i ; Among them, the P sequence is the prediction result sequence including the load prediction sequence and the source network storage prediction sequence, and the R sequence is the real-time operation sequence including the load real-time sequence and the source network storage operation sequence; The calculation formula of error fluctuation radius is: Among them, the value l is the sequence characteristic number. First, the absolute value of the error sequence is summed up with the real-time operation sequence to find the percentage of the error capacity in the total operation capacity, and then multiply it by the maximum load power; The calculation formula for the running stability threshold is: Among them, μ is the source-grid-load-storage scale coefficient. According to the power supply capacity, it is divided into small source-grid-load-storage system, medium source-grid-load-storage system, large source-grid-load-storage system and super large source-grid-load-storage system, where μ is 1, 10 and 10 respectively. -1 , 10 -2 , 10 -3 .

Images