Small-scale reactor power grid dispatching EMS control method based on machine-network coordination control
By adopting the EMS method of machine-network coordination control in a small stack grid, combining linear and nonlinear prediction models, the problem of large load prediction deviations of the equipment layer and network layer is solved, the accuracy and security of the system are improved, and the more Accurate load prediction and adjustment.
Patent Information
- Application Number
- CN202410963224.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-12
AI Technical Summary
The existing small stack grid scheduling EMS control system has shortcomings in achieving accurate coordinated operation between the equipment layer and the network layer, resulting in large deviations in load prediction, affecting the accuracy and safety of the system.
Using the EMS control method based on machine-network coordination control, the load historical data prediction and deviation calculation of the equipment layer and network layer is gradually adjusted, and the power output or output prompt information is combined with the fusion of linear and nonlinear prediction models to improve the accuracy of load prediction. sex.
The accuracy and safety of machine-network coordination control are improved, and the stability and reliability of the system are enhanced through accurate load prediction and adjustment.
Smart Images

Figure CN118944056A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power grid dispatching, and in particular to an EMS control method for small-scale reactor power grid dispatching based on machine-grid coordinated control. Background Art
[0002] The small reactor power grid is an integrated energy system that coordinates the operation of small reactor power sources, wind power, photovoltaics, energy storage devices, and loads through a control system to form a single controllable unit, which directly serves as the power supply and power consumption system at the field end. According to the hierarchical control concept, the operation control of the small reactor power grid can be divided into device-level local control that relies on local measurement and status information, regional-level control that relies on weak communication and can achieve coordinated operation of multiple devices, and system-level control with green economic operation as the main goal.
[0003] How to achieve accurate machine (i.e., device layer) and network (i.e., regional layer or system layer) coordination is the core content of EMS control for small reactor power grid dispatching, and there is still a need for further improvement. Summary of the invention
[0004] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides an EMS control method for small-scale reactor power grid dispatching based on machine-grid coordinated control.
[0005] A small-scale reactor power grid dispatching EMS control method based on machine-grid coordinated control, the method comprising: The device layer predicts the first load forecast data of the target period based on the first user load history data, and the network layer predicts the second load forecast data of the device layer in the target period based on the second user load history data; wherein the first user load history data corresponds to the device layer and includes a number of historical load forecast data sent by the network layer, and the second user load history data includes the first user load history data; Calculating a deviation value between the first load forecast data and the second load forecast data, wherein the deviation value corresponds to each target time period; When the deviation value is greater than the deviation threshold, the device layer gradually predicts and obtains a plurality of third load prediction data based on the third user load history data, and predicts and obtains a fourth load prediction data corresponding to the target time period according to each of the third load prediction data; If the deviation value between the fourth load forecast data and the second load forecast data is less than the deviation threshold, the device layer adjusts the power output according to the second load forecast data; otherwise, a prompt message is output to the network side.
[0006] Optionally, the third user load history data includes part of the user load history data close to the current moment in the first user load history data, and current user load real-time data.
[0007] Optionally, the device layer gradually predicts and obtains a number of third load prediction data based on the third user load history data, including: Calculating the difference between the deviation value and the deviation threshold, and determining the prediction step length according to the difference; Taking the prediction step as a reference, the device layer gradually predicts and obtains a number of the third load prediction data based on the third user load historical data.
[0008] Optionally, the device layer predicts and obtains first load forecast data for a target period based on the first user load history data, including: Collecting and preprocessing the historical load data of each first user in the small-scale reactor power grid to obtain a user load data set of the small-scale reactor power grid; Performing curve fitting processing on each data element in the user load data set according to date and time to obtain a load fluctuation curve, and determining the load fluctuation data corresponding to the time period during which the small-scale reactor power grid needs to perform ultra-short-term and short-term load forecasting according to the load fluctuation curve; The linear prediction model and the nonlinear prediction model respectively make ultra-short-term and short-term load forecasts for the small reactor power grid based on the user load data set; The optimal weights of the linear prediction model and the nonlinear prediction model are determined according to the load fluctuation data, and the prediction results of the linear prediction model and the nonlinear prediction model are fused according to the optimal weights to obtain a target prediction result.
[0009] Optionally, determining the optimal weights of the linear prediction model and the nonlinear prediction model according to the load fluctuation data includes: Determine a first initial weight of the linear prediction model and a second initial weight of the nonlinear prediction model according to the load fluctuation data; Obtain the first prediction result of the linear prediction model for the specified day and the second prediction result of the nonlinear prediction model for the specified day, combine the actual load data of the specified day, use the least squares method to fit the load curve of the specified day, obtain the optimal weight combination λ0, λ1, λ2, establish the least squares combination model, and predict the load data of the next day of the specified day. The final prediction model is: Y=λ0+λ1P1+λ2P2 In the formula, Y is the predicted value, λ0, λ1, λ2 are weight coefficients, P1 is the first prediction result, and P2 is the second prediction result; The average value of the first initial weight and the weight coefficient λ1 is calculated as the optimal weight of the linear prediction model, and the average value of the second initial weight and the weight coefficient λ2 is calculated as the optimal weight of the linear prediction model.
[0010] Optionally, determining the first initial weight of the linear prediction model and the second initial weight of the nonlinear prediction model according to the load fluctuation data includes: If the load fluctuation data is smaller than the fluctuation threshold, the first initial weight is set to be greater than the second initial weight; otherwise, the first initial weight is set to be smaller than the second initial weight.
[0011] The present invention realizes a small-scale reactor power grid dispatching EMS control strategy based on machine-network coordinated control. The control strategy takes into account the processing of load data prediction deviations at the equipment layer and the network layer, making the machine-network coordinated control more accurate and safer. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.
[0013] Figure 1 The present invention discloses a flowchart of an EMS control method for small-scale reactor power grid dispatching based on machine-grid coordinated control. DETAILED DESCRIPTION
[0014] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0015] Figure 1 The present invention discloses a small-scale reactor power grid dispatching EMS control method based on machine-grid coordinated control, and the method includes: The device layer predicts the first load forecast data of the target period based on the first user load history data, and the network layer predicts the second load forecast data of the device layer in the target period based on the second user load history data; wherein the first user load history data corresponds to the device layer and includes a number of historical load forecast data sent by the network layer, and the second user load history data includes the first user load history data; Calculating a deviation value between the first load forecast data and the second load forecast data, wherein the deviation value corresponds to each target time period; When the deviation value is greater than the deviation threshold, the device layer gradually predicts and obtains a plurality of third load prediction data based on the third user load history data, and predicts and obtains a fourth load prediction data corresponding to the target time period according to each of the third load prediction data; If the deviation value between the fourth load forecast data and the second load forecast data is less than the deviation threshold, the device layer adjusts the power output according to the second load forecast data; otherwise, a prompt message is output to the network side.
[0016] The present invention improves the traditional EMS control system. The network layer (regional layer or system layer) generates the second load forecast data of each equipment layer in the target time period based on the operating parameters of each power-consuming equipment and power-generating equipment in the region, and uses conventional EMS control strategies to send it to the corresponding equipment layers. The equipment layer also predicts the first load forecast data in the target time period based on the first user load history data related to its own operation and the recorded historical load forecast data sent by the network layer (i.e., the second load forecast data sent in the earlier scheduling cycle), and compares the first matching forecast data with the second load forecast data predicted by the network layer. If the deviation is too large, it indicates that there may be a prediction error. At this time, the equipment layer gradually predicts towards the target time period based on part of the historical load data and the current real-time load data, obtains multiple third load forecast data, and then predicts the fourth load forecast data in the target time period based on each third load forecast data. If the deviation between the fourth load prediction data and the second load prediction data of the network layer is not greater than the deviation threshold, it means that there are more abnormal data in the first user load history data based on the device layer, resulting in distortion of the first load prediction data predicted by it. At this time, the second load prediction data sent by the network layer is accepted, and its own power output is adjusted accordingly; but if the deviation is still greater than the deviation threshold, it means that there are errors in the prediction data of the network layer. This error may be caused by a deviation in the prediction algorithm of the network layer, or it may be caused by the presence of more abnormal data in the second user load history data based on the network layer. These abnormal data, for example, come from other device layers. At this time, prompt information is output to the network layer.
[0017] Optionally, the third user load history data includes part of the user load history data close to the current moment in the first user load history data, and current user load real-time data.
[0018] In the embodiment of the present invention, the third user load historical data used for local re-prediction by the device layer includes a part of the historical data, i.e., the part of the user load historical data close to the current time in the first user load historical data, and the current user load real-time data. That is, the historical data and real-time data are continuously used to perform multi-step prediction in the direction of the target time period.
[0019] Optionally, the device layer gradually predicts and obtains a number of third load prediction data based on the third user load history data, including: Calculating the difference between the deviation value and the deviation threshold, and determining the prediction step length according to the difference; Taking the prediction step as a reference, the device layer gradually predicts and obtains a number of the third load prediction data based on the third user load historical data.
[0020] In an embodiment of the present invention, when re-predicting to obtain a number of third load prediction data, the prediction step length needs to be considered. The appropriateness of the prediction step length selection will directly affect the prediction accuracy of the subsequent fourth load prediction data. Generally speaking, the more third load prediction data there are, the higher the accuracy of the predicted fourth load prediction data. In this regard, the present invention selects the corresponding prediction step length according to the difference between the aforementioned deviation value and the deviation threshold. When the prediction deviation of the network layer and the device layer is larger, a shorter prediction step length is adopted to obtain more third load prediction data, thereby improving the accuracy of the fourth load prediction data.
[0021] Optionally, the device layer predicts and obtains first load forecast data for a target period based on the first user load history data, including: Collecting and preprocessing the historical load data of each first user in the small-scale reactor power grid to obtain a user load data set of the small-scale reactor power grid; Performing curve fitting processing on each data element in the user load data set according to date and time to obtain a load fluctuation curve, and determining the load fluctuation data corresponding to the time period during which the small-scale reactor power grid needs to perform ultra-short-term and short-term load forecasting according to the load fluctuation curve; The linear prediction model and the nonlinear prediction model respectively make ultra-short-term and short-term load forecasts for the small reactor power grid based on the user load data set; The optimal weights of the linear prediction model and the nonlinear prediction model are determined according to the load fluctuation data, and the prediction results of the linear prediction model and the nonlinear prediction model are fused according to the optimal weights to obtain a target prediction result.
[0022] In the embodiment of the present invention, the present invention adopts a combined result prediction method to perform load prediction on a small-scale reactor power grid, mainly involving load prediction of ultra-short-term and short-term loads. The load prediction methods adopted by the device layer and the network layer are the same, and the only difference is that the first user load data and the second user load data based on them cover different ranges. Specifically, the first user load data is limited to the data of the area associated with each device layer, while the second user load data is the data of the area associated with the network layer, and the second user load data does not contain the historical load prediction data sent by the network layer.
[0023] Among them, the combined result prediction method includes linear models and nonlinear models. The linear model requires less historical data, less computational overhead, and has a good prediction effect on loads with smaller fluctuations. The nonlinear model can better describe complex nonlinear mapping relationships, has adaptive functions for a large amount of irregular data, and can autonomously learn, reason about knowledge, and optimize calculations. Therefore, the present invention is based on the above-mentioned respective characteristics of the linear model and the nonlinear model for fusion use, which is conducive to making the prediction results more accurate. Specifically, the load data of each user in the small-scale reactor power grid is first collected, and the user load data set is obtained after processing. Based on the user load data set, the load fluctuation curve corresponding to the small-scale reactor power station can be fitted, and the load fluctuation curve includes the load fluctuation data corresponding to the period when the small-scale reactor power grid needs to perform ultra-short-term and short-term load prediction. The load fluctuation data is used to optimize the weights of the linear prediction model and the nonlinear prediction model. Finally, the prediction results of the linear prediction model and the nonlinear prediction model are fused and calculated using the optimal weight to obtain the final target prediction result.
[0024] Preprocessing can be processing operations such as supplementing, replacing, deleting and normalizing missing data and abnormal data in load data.
[0025] Optionally, determining the optimal weights of the linear prediction model and the nonlinear prediction model according to the load fluctuation data includes: Determine a first initial weight of the linear prediction model and a second initial weight of the nonlinear prediction model according to the load fluctuation data; Obtain the first prediction result of the linear prediction model for the specified day and the second prediction result of the nonlinear prediction model for the specified day, combine the actual load data of the specified day, use the least squares method to fit the load curve of the specified day, obtain the optimal weight combination λ0, λ1, λ2, establish the least squares combination model, and predict the load data of the next day of the specified day. The final prediction model is: Y=λ0+λ1P1+λ2P2 In the formula, Y is the predicted value, λ0, λ1, λ2 are weight coefficients, P1 is the first prediction result, and P2 is the second prediction result; The average value of the first initial weight and the weight coefficient λ1 is calculated as the optimal weight of the linear prediction model, and the average value of the second initial weight and the weight coefficient λ2 is calculated as the optimal weight of the linear prediction model.
[0026] In an embodiment of the present invention, the present invention first determines the load fluctuation size of the target period for which the small-scale reactor power grid needs to perform ultra-short-term and short-term load forecasting based on the load fluctuation data, and determines the initial weights of the linear model and the nonlinear model. Then, the historical forecast data of the linear prediction model and the nonlinear prediction model are obtained, and the optimal weight combination is calculated using the least squares method. The average of the optimal weight and the aforementioned initial weight is used as the final weight of the present invention.
[0027] After multiple calculations, when the difference between the predicted value Y and the actual load data of the next day of the specified day is within a predetermined range, the corresponding weight coefficients λ0, λ1, and λ2 are determined to be the optimal weight combination.
[0028] Optionally, determining the first initial weight of the linear prediction model and the second initial weight of the nonlinear prediction model according to the load fluctuation data includes: If the load fluctuation data is smaller than the fluctuation threshold, the first initial weight is set to be greater than the second initial weight; otherwise, the first initial weight is set to be smaller than the second initial weight.
[0029] In the embodiment of the present invention, when the load fluctuation in the period during which the small-scale reactor power grid needs to perform ultra-short-term and short-term load forecasting is small, the confidence of the prediction result using the linear prediction model is improved, and vice versa, the confidence of the prediction result using the nonlinear prediction model is improved. Therefore, regardless of whether the load fluctuation is large or small, the combined result prediction method of the present invention can obtain a relatively more accurate load forecast result.
[0030] Among them, the linear prediction model is preferably a time series model, and the nonlinear prediction model is preferably an artificial neural network model.
[0031] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A small-scale reactor power grid dispatching EMS control method based on machine-grid coordinated control, characterized in that: The method comprises: The device layer predicts the first load forecast data of the target period based on the first user load history data, and the network layer predicts the second load forecast data of the device layer in the target period based on the second user load history data; wherein the first user load history data corresponds to the device layer and includes a number of historical load forecast data sent by the network layer, and the second user load history data includes the first user load history data; Calculating a deviation value between the first load forecast data and the second load forecast data, wherein the deviation value corresponds to each target time period; When the deviation value is greater than the deviation threshold, the device layer gradually predicts and obtains a plurality of third load prediction data based on the third user load history data, and predicts and obtains a fourth load prediction data corresponding to the target time period according to each of the third load prediction data; If the deviation value between the fourth load forecast data and the second load forecast data is less than the deviation threshold, the device layer adjusts the power output according to the second load forecast data; otherwise, a prompt message is output to the network side.
2. The EMS control method for small-scale reactor power grid dispatching based on machine-grid coordinated control according to claim 1 is characterized by: The third user load history data includes a portion of user load history data close to the current time in the first user load history data, and current user load real-time data.
3. The EMS control method for small-scale reactor power grid dispatching based on machine-grid coordinated control according to claim 2 is characterized by: The device layer gradually predicts and obtains a number of third load prediction data based on the third user load history data, including: Calculating the difference between the deviation value and the deviation threshold, and determining the prediction step length according to the difference; Taking the prediction step as a reference, the device layer gradually predicts and obtains a number of the third load prediction data based on the third user load historical data.
4. The EMS control method for small-scale reactor power grid dispatching based on machine-grid coordinated control according to claim 3 is characterized by: The device layer predicts and obtains first load forecast data for a target period based on the first user load history data, including: Collecting and preprocessing the historical load data of each first user in the small-scale reactor power grid to obtain a user load data set of the small-scale reactor power grid; Performing curve fitting processing on each data element in the user load data set according to date and time to obtain a load fluctuation curve, and determining the load fluctuation data corresponding to the time period during which the small-scale reactor power grid needs to perform ultra-short-term and short-term load forecasting according to the load fluctuation curve; The linear prediction model and the nonlinear prediction model respectively make ultra-short-term and short-term load forecasts for the small reactor power grid based on the user load data set; The optimal weights of the linear prediction model and the nonlinear prediction model are determined according to the load fluctuation data, and the prediction results of the linear prediction model and the nonlinear prediction model are fused according to the optimal weights to obtain a target prediction result.
5. The EMS control method for small-scale reactor power grid dispatching based on machine-grid coordinated control according to claim 4 is characterized by: The step of determining the optimal weights of the linear prediction model and the nonlinear prediction model according to the load fluctuation data includes: Determine a first initial weight of the linear prediction model and a second initial weight of the nonlinear prediction model according to the load fluctuation data; Obtain the first prediction result of the linear prediction model for the specified day and the second prediction result of the nonlinear prediction model for the specified day, combine the actual load data of the specified day, use the least squares method to fit the load curve of the specified day, obtain the optimal weight combination λ0, λ1, λ2, establish the least squares combination model, and predict the load data of the next day of the specified day. The final prediction model is: Y=λ0+λ1P1+λ2P2 In the formula, Y is the predicted value, λ0, λ1, λ2 are weight coefficients, P1 is the first prediction result, and P2 is the second prediction result; The average value of the first initial weight and the weight coefficient λ1 is calculated as the optimal weight of the linear prediction model, and the average value of the second initial weight and the weight coefficient λ2 is calculated as the optimal weight of the linear prediction model.
6. The EMS control method for small-scale reactor power grid dispatching based on machine-grid coordinated control according to claim 5 is characterized by: The step of determining the first initial weight of the linear prediction model and the second initial weight of the nonlinear prediction model according to the load fluctuation data includes: If the load fluctuation data is smaller than the fluctuation threshold, the first initial weight is set to be greater than the second initial weight; otherwise, the first initial weight is set to be smaller than the second initial weight.
