New energy combination prediction and optimization method in novel power system based on multiple neural networks

Abstract

The invention discloses a new energy combination prediction and optimization method in a novel power system based on multiple neural networks, and the method comprises the steps: S1, building a CNN-LSTM-GWO combination prediction model, comparing the CNN-LSTM-GWO combination prediction model with a single LSTM neural network and a CNN-LSTM combination prediction model, carrying out the training through inputting a same training set, and comparing the prediction results of a test set; and S2, applying prediction results of different models to a subsequent optimal scheduling project of the novel power system, applying a weight-updated intra-day rolling optimization method to an optimal scheduling process of the novel power system, and comparing the cost of the novel power system with that of a traditional rolling optimization method. The present invention explores the recent progress by integrating two particularly effective neural networks for predicting wind and solar energy. The combination improves the accuracy of model prediction. In addition, in combination with the grey wolf optimization algorithm, the training process is accelerated, and the practicability of the model to ultra-short-term prediction and optimization scheduling tasks is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of large-scale integration of renewable energy into new power systems, and in particular to a new energy combination prediction and optimization method in a new power system based on multiple neural networks. Background Art

[0002] As renewable energy sources are increasingly integrated into new power systems, their inherent randomness and volatility pose significant challenges to optimization and scheduling. Therefore, it is increasingly critical to accurately predict solar and wind power generation and effectively manage forecast data during the subsequent optimization of new power systems. In the field of wind and solar forecasting, advances in big data and artificial intelligence technologies have led to the emergence of many modern methods, mainly machine learning and deep learning techniques. These methods have strong data processing capabilities and the ability to model nonlinear relationships, and have gradually become mainstream forecasting methods. Machine learning-based load forecasting methods include linear regression, support vector machines, and artificial neural networks (ANN). With the increase in data complexity, some deep learning algorithm models have emerged as promising alternatives. Recurrent neural networks (RNN) are a deep learning model in which the output of RNN depends on the previous input, exploiting the temporal correlation in the sequence. In theory, RNN can model sequences of arbitrary length. Some researchers have demonstrated significant performance improvements over other modeling methods by training a large-scale recurrent neural language model using nearly 1 billion words as training data. However, its accuracy is still not ideal when facing more variable and complex forecasting scenarios.

[0003] First of all, for the wind and solar prediction model, the main disadvantage of the current machine learning method is that it cannot learn time features from the input data, has limited flexibility in identifying complex patterns, is difficult to capture intricate variable relationships, and is prone to overfitting problems; for neural networks, a single neural network is usually unable to perform comprehensive feature learning and information processing, and the generalization ability of the model is limited when dealing with complex tasks and data. The combined prediction model developed on the basis of a single neural network can better solve the above problems. The combined prediction model mainly achieves this goal by combining individual models with different characteristics, thereby utilizing the advantages of each model and alleviating the limitations of any single model. The present invention uses a combined model of convolutional neural network (CNN) and LSTM to perform wind and solar output, and the results show that the combined prediction model has the best prediction performance.

[0004] Secondly, for the refined scheduling within the day, the time of each scheduling window is very urgent, so the training process of the neural network used should not be too long, otherwise it cannot meet the refined scheduling needs within the day. Generally speaking, the training time of the neural network varies from tens of minutes to several hours with the increase of data and the complexity of the scene. Based on this, it is urgent to optimize the training process of the neural network to speed up the training speed and training accuracy. The present invention starts from the hyperparameter setting of the neural network and uses the Grey Wolf Optimizer (GWO) to quickly lock the optimal training hyperparameters and improve the training speed of the model.

[0005] Finally, as the forecast time range expands, any form of forecast model will show increasingly larger forecast errors. In order to solve this problem, the present invention proposes a new power system dispatch rolling optimization method that combines day-ahead and intraday weight updates. Compared with the current mainstream day-ahead dispatch method and the traditional rolling optimization method, this method improves the accuracy of forecast data and reduces the overall dispatch cost. Summary of the invention

[0006] The purpose of the present invention is to provide a new type of new energy combination prediction and optimization method in a power system based on multiple neural networks.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The present invention proposes a new type of power system new energy combination prediction and optimization method based on multi-neural network, comprising the following steps:

[0009] S1. Build a CNN-LSTM-GWO combined prediction model, compare it with a single LSTM neural network and a CNN-LSTM combined prediction model, train it by inputting the same training set, and compare the prediction effect of the test set;

[0010] S2. Use the prediction results of different models in the subsequent optimization and dispatching projects of new power systems. Use the weight-updated intraday rolling optimization method in the optimization and dispatching process of new power systems. Compare the cost of new power systems using this method and the traditional rolling optimization method.

[0011] Furthermore, the step S2 includes:

[0012] Establish a day-ahead dispatch optimization model: The purpose of day-ahead dispatch optimization is to minimize the operating cost of the new power system. This involves forecasting renewable energy power based on the previous day's forecast, taking into account the difference in peak and off-peak electricity prices, various traditional power generation costs, and energy storage charging and discharging costs. The optimization includes scheduling the discharge and charging behavior of the energy storage system and the power generation of each generator unit, and formulating corresponding buying and selling strategies based on electricity price fluctuations;

[0013] Establish an intraday rolling optimization model: Implement rolling optimization within one day, use the forecasting model to perform rolling forecasts with a resolution of 15 minutes across 96 nodes, each rolling optimization forecast window spans 4 time steps, execute the first step of the optimal forecast sequence, and then repeat the cyclic rolling optimization steps in sequence.

[0014] Furthermore, the day-ahead scheduling optimization model includes a day-ahead scheduling optimization objective function and system constraints. The goal of the day-ahead scheduling optimization objective function is to reduce the operating cost of the system during the entire scheduling cycle.

[0015] Furthermore, the day-ahead scheduling optimization objective function is as follows:

[0016]

[0017] In the formula The overall economic cost of buying and selling electricity, is the total operating cost of each generator set, It is the dispatching cost of the energy storage unit;

[0018]

[0019] In the formula, is the market electricity price during the time period, P t m The amount of electricity purchased and sold by the new power system participating in the market during period t;

[0020]

[0021] E(u n,t )=u n,t (1-u n,t-1 )S n (8)

[0022] Where N G is the number of dispatchable units, P g,n,t is the output of unit n during period t, a n ,b n ,c n is the fuel cost coefficient of unit n, S n is the startup cost of unit n, u n,tIndicates the operating status of unit n in period t. A value of 1 indicates that the unit is running, and a value of 0 indicates that the unit is shut down.

[0023]

[0024] In the formula, The unit cost of charging and discharging the energy storage battery; is the energy of battery discharge and charge; N s is the number of energy storage batteries.

[0025] Furthermore, the system constraints are as follows:

[0026]

[0027] In the formula, is the discharge power of energy storage unit s in time period t, is the charging power of energy storage unit s in time period t, P F,f,t is the power of distributed energy f in time period t, N f is the number of distributed energy sources. t L is the load power in the tth period;

[0028]

[0029] In the formula, and are the downward and upward climbing rates of the unit respectively;

[0030]

[0031] In the formula, E se,s,t represents the energy storage of storage unit s in time period t; the charging rate and discharging rate of the storage unit are respectively expressed as and The unit also has specified maximum and minimum power constraints, denoted as and The storage unit s is subject to predefined upper and lower energy limits, expressed as and

[0032] 0≤P F,f,t ≤P max F,f,t (17)

[0033] Where P max F,f,t is the output limit of distributed energy s.

[0034] Furthermore, the objective function of the intraday rolling optimization model is as follows:

[0035]

[0036] w t =4*F t / FA t (twenty two)

[0037] In the formula, r t represents the scheduling error between the day-ahead scheduling and the intraday scheduling of the same unit at the same time, in order to enhance the global optimality of intraday fine scheduling, They represent the output of the energy storage unit, the output of the dispatchable unit, and the day-ahead dispatching strategy for purchasing and selling electricity. They represent the output of the energy storage unit, the output of the dispatchable unit, and the intraday dispatch power purchase and sales strategy. u1, u2, and u3 represent the error coefficients. They represent the predicted output and actual output of the renewable energy unit at time t-4 respectively. Equations (21) and (22) define the updated weight values.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] In the context of forecasting models, this paper explores recent advances by integrating two particularly effective neural networks for forecasting wind and solar energy. This combination improves the accuracy of the model's forecasts. In addition, incorporating the Gray Wolf Optimization algorithm speeds up the training process and enhances the model's usefulness for ultra-short-term forecasting and optimization scheduling tasks.

[0040] After improving the accuracy of the prediction itself, this paper proposes a weight update method based on day-ahead collaborative rolling scheduling, which further solves the accuracy problem in subsequent optimization scheduling. This method not only retains the global optimality of traditional day-ahead scheduling, but also improves the accuracy of the prediction data at each rolling optimization time step, thereby reducing the scheduling cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Optimize the display graph for scrolling windows;

[0042] Figure 2 This is the structural diagram of the convolutional neural network;

[0043] Figure 3 This is the unit diagram of LSTM. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] The present invention provides a new energy combination prediction and optimization method in a new power system based on multiple neural networks, comprising the following steps:

[0046] S1. Build a CNN-LSTM-GWO combined prediction model, compare it with a single LSTM neural network and a CNN-LSTM combined prediction model, train it by inputting the same training set, and compare the prediction effect of the test set;

[0047] S2. Use the prediction results of different models in the subsequent optimization and dispatching projects of new power systems. Use the weight-updated intraday rolling optimization method in the optimization and dispatching process of new power systems. Compare the cost of new power systems using this method and the traditional rolling optimization method.

[0048] Step S2 includes:

[0049] Establish a day-ahead dispatch optimization model: The purpose of day-ahead dispatch optimization is to minimize the operating cost of the new power system. This involves forecasting renewable energy power based on the previous day's forecast, taking into account the difference in peak and off-peak electricity prices, various traditional power generation costs, and energy storage charging and discharging costs. The optimization includes scheduling the discharge and charging behavior of the energy storage system and the power generation of each generator unit, and formulating corresponding buying and selling strategies based on electricity price fluctuations;

[0050] Establish an intraday rolling optimization model: Implement rolling optimization within one day, use the forecasting model to perform rolling forecasts with a resolution of 15 minutes across 96 nodes, each rolling optimization forecast window spans 4 time steps, execute the first step of the optimal forecast sequence, and then repeat the cyclic rolling optimization steps in sequence.

[0051] The day-ahead scheduling optimization model includes a day-ahead scheduling optimization objective function and system constraints. The day-ahead scheduling optimization objective function aims to reduce the operating cost of the system during the entire scheduling cycle.

[0052] The details are as follows:

[0053] Day-ahead Optimal Scheduling Modeling

[0054] The purpose of day-ahead dispatch optimization is to minimize the operating costs of a new power system. This involves forecasting renewable energy power based on the previous day's forecast, taking into account peak and off-peak electricity price differences, various traditional generation costs, and energy storage charging and discharging costs. The optimization involves scheduling the discharging and charging behavior of the energy storage system and the generation of each generator unit, and formulating corresponding buying and selling strategies based on electricity price fluctuations. The goal of this method is to minimize the operating costs of the system.

[0055] a) Objective Function

[0056] The goal of the objective function is to reduce the operating cost of the system during the entire scheduling period.

[0057]

[0058] In the formula The overall economic cost of buying and selling electricity, is the total operating cost of each generator set, It is the dispatching cost of the energy storage unit.

[0059]

[0060] In the formula, is the market electricity price during the time period, P t m It is the amount of electricity purchased and sold by the new power system participating in the market during period t.

[0061]

[0062] E(u n,t )=u n,t (1-u n,t-1 )S n (8)

[0063] Where N G is the number of dispatchable units, P g,n,t is the output of unit n during period t, a n ,b n ,c n is the fuel cost coefficient of unit n, S n is the startup cost of unit n, u n,t Indicates the operating status of unit n during period t. A value of 1 indicates that the unit is running, and a value of 0 indicates that the unit is shut down.

[0064]

[0065] In the formula, The unit cost of charging and discharging energy storage batteries. It is the energy of battery discharge and charge. s is the number of energy storage batteries.

[0066] b) System constraints

[0067]

[0068] In the formula, is the discharge power of energy storage unit s in time period t, is the charging power of energy storage unit s in time period t, P F,f,t is the power of distributed energy f in time period t, N f is the number of distributed energy sources. t L is the load power in period t.

[0069]

[0070] In the formula, and They are the downward and upward climbing rates of the unit respectively.

[0071]

[0072]

[0073] In the formula, E se,s,t represents the energy storage of storage unit s in time period t. The charging rate and discharging rate of the storage unit are respectively expressed as and The unit also has specified maximum and minimum power constraints, denoted as and In addition, the storage unit s is subject to predefined upper and lower energy limits, expressed as and

[0074] 0≤P F,f,t ≤P max F,f,t (17)

[0075] Where P max F,f,t is the output limit of distributed energy s.

[0076] C. Intraday rolling optimization

[0077] Rolling optimization was implemented within one day, using the forecasting model to perform rolling forecasts across 96 nodes. Each rolling optimization forecast window spanned 4 time steps, executing the first step of the optimal forecast sequence, followed by repeated rolling optimization steps in sequence.

[0078] The results of advance scheduling show better global optimality but are associated with larger errors in solar and wind output data. In contrast, daytime scheduling benefits from the higher accuracy of solar and wind data but can usually only achieve local optimal solutions. Therefore, in the objective function of intraday scheduling, the day-ahead error and intraday scheduling error are used as penalty terms to enhance the global optimality of the intraday scheduling results.

[0079] In addition, it is necessary to define the weights of different time periods in the rolling time domain according to the error patterns of the neural network model in predicting the solar and wind data. Generally, the prediction accuracy of the neural network decreases as the time scale increases.

[0080] Taking the above two factors into consideration, an intraday scheduling optimization model is established. The main difference from day-ahead scheduling is the optimization adjustment of the objective function:

[0081]

[0082] w t =4*F t / FA t (twenty two)

[0083] It can be seen that in order to more accurately describe the objective function, two new parameters are defined here, namely r t and w t Among them, r t represents the scheduling error between the day-ahead scheduling and the intraday scheduling of the same unit at the same time, in order to enhance the global optimality of intraday fine scheduling, They represent the output of the energy storage unit, the output of the dispatchable unit, and the day-ahead dispatching strategy for purchasing and selling electricity. They represent the output of the energy storage unit, the output of the dispatchable unit, and the intraday dispatch power purchase and sales strategy. u1, u2, and u3 represent the error coefficients. They represent the predicted output and actual output of the renewable energy unit at time t-4 respectively. In addition, equations (21) and (22) define the updated weight values.

[0084] In addition, the model principles of CNN and LSTM.

[0085] 1. CNN

[0086] CNNs are based on the principle of neural perceptrons and are inspired by the animal vision system. They represent a special form of deep feedforward neural networks. They are widely used in fields such as image processing and natural language processing. Figure 2 As shown in Figure 1, the design of CNN consists of convolutional layers, pooling layers, and fully connected layers. The network first uses convolutional layers to extract features from the input data, then enhances these features through pooling layers, and finally integrates the results through fully connected layers.

[0087] 2. LSTM

[0088] In order to study and predict the mapping relationship between wind and solar energy output and daily time series, consider using an LSTM network from a deep neural network, whose network state update process is as follows Figure 3 shown.

[0089] In the above structure, the forget gate decides which information to discard. The forget gate takes the previous time step h t-1, and the current input information x t Read the user response and output a number between 0 and 1. The update formula of the forget gate is as follows:

[0090] f t =σ(w f ×[h t-1 ,x t ])+b f (1)

[0091] Where b f is the bias associated with the forget gate, σ represents the sigmoid function, and w f represents the weight of the forget gate, x t represents the input of the current time step, h t-1, Represents the output of the previous time step.

[0092] The input gate determines which values ​​will be used to update the state of the user response state recording unit. It first processes the input vector using the sigmoid activation function and generates a candidate value vector using the tanh function. The update formula of the input gate is as follows:

[0093] i t =σ(w i ×[h t-1 ,x t ]+b i ) (2)

[0094]

[0095] In the formula, w i and b i represents the weight and bias of the input gate, and w c and b c Represents the weight and bias of the candidate value vector, expressed as

[0096] The user response state recording unit maintains the state of the previous time step. After processing the current input data, it is necessary to update the state information in the cell. The data of the forget gate is used to select what to discard, while the data of the input gate is used to select what to update. The update formula of the user response state recording unit is as follows:

[0097]

[0098] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0099] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A new method for forecasting and optimizing new energy combinations in a new power system based on multiple neural networks, characterized by: The following steps are involved: S1. Build a CNN-LSTM-GWO combined prediction model, compare it with a single LSTM neural network and a CNN-LSTM combined prediction model, train it by inputting the same training set, and compare the prediction results of the test machine; S2. Use the prediction results of different models in the subsequent optimization and dispatching projects of new power systems. Use the weight-updated intraday rolling optimization method in the optimization and dispatching process of new power systems. Compare the cost of new power systems using this method and the traditional rolling optimization method.

2. The new energy combination prediction and optimization method in the new power system based on multi-neural network according to claim 1 is characterized by: The step S2 includes: Establish a day-ahead dispatch optimization model: The purpose of day-ahead dispatch optimization is to minimize the operating cost of the new power system. This involves forecasting renewable energy power based on the previous day's forecast, taking into account the difference in peak and off-peak electricity prices, various traditional power generation costs, and energy storage charging and discharging costs. The optimization includes scheduling the discharge and charging behavior of the energy storage system and the power generation of each generator unit, and formulating corresponding buying and selling strategies based on electricity price fluctuations; Establish an intraday rolling optimization model: Implement rolling optimization within one day, use the forecasting model to perform rolling forecasts with a resolution of 15 minutes across 96 nodes, each rolling optimization forecast window spans 4 time steps, execute the first step of the optimal forecast sequence, and then repeat the cyclic rolling optimization steps in sequence.

3. The new energy combination prediction and optimization method in the new power system based on multi-neural network according to claim 2 is characterized by: The day-ahead scheduling optimization model includes a day-ahead scheduling optimization objective function and system constraints. The day-ahead scheduling optimization objective function aims to reduce the operating cost of the system during the entire scheduling cycle.

4. The new energy combination prediction and optimization method in the new power system based on multi-neural network according to claim 3 is characterized by: The day-ahead scheduling optimization objective function is as follows: In the formula is the overall economic cost of purchasing and selling electricity, is the total operating cost of each generator set, It is the dispatching cost of the energy storage unit; In the formula, is the market electricity price during the time period, P t m The amount of electricity purchased and sold by the new power system participating in the market during period t; E(in n,t )=in n,t (1-in n,t-1 )WITH n (8) Where N G is the number of dispatchable units, P g,n,t is the output of unit n during period t, a n ,b n ,c n is the fuel cost coefficient of unit n, S n is the startup cost of unit n, u n,t Indicates the operating status of unit n in period t. A value of 1 indicates that the unit is running, and a value of 0 indicates that the unit is shut down. In the formula, The unit cost of charging and discharging the energy storage battery; is the energy of battery discharge and charge; N s is the number of energy storage batteries.

5. The new energy combination prediction and optimization method in the new power system based on multi-neural network according to claim 3 is characterized by: The system constraints are as follows: In the formula, is the discharge power of energy storage unit s in time period t, is the charging power of energy storage unit s in time period t, P F,f,t is the power of distributed energy f in time period t, N f is the number of distributed energy sources. t L is the load power in the tth period; In the formula, and are the downward and upward climbing rates of the unit respectively; In the formula, E se,s,t represents the energy storage of storage unit s in time period t; the charging rate and discharging rate of the storage unit are respectively expressed as and The unit also has specified maximum and minimum power constraints, denoted as and The storage unit s is subject to predefined upper and lower energy limits, expressed as and 0≤P F,f,t ≤P max F,f,t (17) Where P max F,f,t is the output limit of distributed energy s.

6. The new energy combination prediction and optimization method in a new power system based on multiple neural networks according to claim 2 is characterized by: The objective function of the intraday rolling optimization model is as follows: <h2 style=";text-align:left;direction:ltr">w<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> <4*F<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> / FA<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> (22) In the formula, r t represents the scheduling error between the day-ahead scheduling and the intraday scheduling of the same unit at the same time, in order to enhance the global optimality of intraday fine scheduling, They represent the output of the energy storage unit, the output of the dispatchable unit, and the day-ahead dispatching strategy for purchasing and selling electricity. They represent the output of the energy storage unit, the output of the dispatchable unit, and the intraday dispatch power purchase and sales strategy. u1, u2, and u3 represent the error coefficients. They represent the predicted output and actual output of the renewable energy unit at time t-4 respectively. Equations (21) and (22) define the updated weight values.

Images