Power system load prediction method, system, equipment and medium

Abstract

The invention relates to the technical field of load prediction, and discloses a power system load prediction method, system and device and a medium, and the method comprises the steps: comprehensively considering multiple factors such as historical load data, meteorological information and economic data, and avoiding the limitation of a conventional method. By establishing the first load characteristic analysis model and the second load prediction model, the load characteristics of the power grid are deeply analyzed, and the change rule of the load can be more accurately captured, so that the accuracy of load prediction is effectively improved. In the aspect of model updating, actual load data are fed back to the model for parameter iteration updating, it is guaranteed that the model can adapt to changes of power grid load characteristics in time, and the reliability and adaptability of the model are improved. In terms of display and application of the prediction result, different strategy rules are preset, the prediction display content and mode are flexibly adjusted according to the power grid state and the environmental factors, and comprehensive and accurate support is provided for decision making of the power system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of load prediction, in particular to a power system load prediction method, system, device and medium. BACKGROUND

[0002] In the operation and management of power systems, accurate load prediction is crucial. However, existing load prediction methods have some limitations. Traditional load prediction models often only consider a single factor, such as predicting only based on historical load data, without fully combining meteorological information and economic data to affect the load, resulting in a large deviation between the predicted results and the actual load. Moreover, some models lack in-depth analysis of load characteristics during the establishment process, making it difficult to accurately capture the changing patterns of the load, and making it difficult to meet the needs of fine management of power systems in terms of prediction accuracy.

[0003] In addition, some load prediction methods are not timely and effective in model updating. With the development of the power grid, changes in weather conditions and fluctuations in the economic situation, the load characteristics will also change. However, some existing methods cannot quickly update the model parameters based on actual load data feedback, resulting in the model gradually losing its adaptability to load changes, reducing the accuracy and reliability of the prediction.

[0004] In addition, in terms of the display and application of the prediction results, the existing methods are relatively simple, mostly limited to simple chart display, lacking in-depth analysis and mining of the prediction results, and unable to provide comprehensive and accurate support for decision-making in the power system. For example, it is difficult to flexibly adjust the prediction display content and method according to different power grid states and environmental factors, making it difficult to meet the actual needs in different scenarios. SUMMARY

[0005] In view of the above existing problems, the present application is proposed.

[0006] Therefore, the present application provides a power system load prediction method, system, device and medium, which can solve the problem of large deviation in the prediction results caused by only considering a single factor and lacking in-depth analysis of load characteristics in existing load prediction methods.

[0007] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a power system load prediction method, comprising: obtaining historical load data, meteorological information and economic data of a target regional power grid, and preprocessing the historical load data, meteorological information and economic data of the target regional power grid; establishing a first load characteristic analysis model according to the preprocessed data; establishing a second load prediction model according to the first load characteristic analysis model and historical load data; generating an optimal load prediction result based on the first load characteristic analysis model and the second load prediction model; calling the optimal load prediction result by a preset upper-layer strategy function and triggering a corresponding lower-layer execution function to implement a prediction display operation.

[0008] As a preferred scheme of the power system load prediction method, the method further comprises: obtaining actual load data of the optimal load prediction result; feeding back the actual load data to the first load characteristic analysis model and the second load prediction model for parameter iterative updating; until an iteration condition is met.

[0009] As a preferred scheme of the power system load prediction method, the method further comprises: presetting a load characteristic index set, the load characteristic index set comprising a plurality of load characteristic indexes for evaluating a target regional power grid; selecting load characteristic indexes based on the preprocessed data; characterizing the selected load characteristic indexes by a scoring formula, and calculating a load characteristic score of the target regional power grid according to the scoring formula; grading the load characteristics of the power grid according to the load characteristic score to form the first load characteristic analysis model.

[0010] The preferred scheme has the advantage of accurately grasping the load characteristics of the target regional power grid. The pre-set load characteristic index set comprehensively covers various factors that may affect the load characteristics of the power grid, making the subsequent index selection more targeted and comprehensive. The selection of the load characteristic indexes based on the preprocessed data ensures that the selected indexes can truly reflect the actual operation of the power grid, avoiding analysis deviation caused by inaccurate or incomplete data. The selected load characteristic indexes are characterized by a scoring formula and the load characteristic score is calculated, providing a quantitative evaluation standard for the load characteristics of the power grid. The quantitative score can intuitively show the advantages and disadvantages of the load characteristics of the power grid, facilitating power system managers to quickly understand the operation status of the power grid.

[0011] As a preferred scheme of the power system load prediction method, the method further comprises: presetting a load type set, the load type set comprising a plurality of different load types and corresponding features; Feature extraction is performed on the output of the first load characteristic analysis model, and feature information related to the features in the preset load type set is extracted; The extracted features are matched and compared with different features in the preset load type set, and the matching degree of each load type is calculated; According to the matching degree and the historical load data, the probability of each load type occurring under the current power grid state is determined, and a second load prediction model is constructed.

[0012] As a preferred scheme of the power system load prediction method provided by the application, the generation of the optimal load prediction result based on the first load characteristic analysis model and the second load prediction model comprises: Determine the load prediction period and time resolution of the target regional power grid; Establish an optimization configuration model characterized by maximizing load prediction accuracy as the target; Combine the load characteristic analysis result and the load prediction model to generate the optimal load prediction result.

[0013] As a preferred scheme of the power system load prediction method provided by the application, the generation of the optimal load prediction result based on the first load characteristic analysis model and the second load prediction model comprises: Preset strategy rules under different combinations of power grid state and environmental factors; The upper strategy function selects the corresponding rule from the preset strategy rules according to the current obtained power grid state and environmental factors; The selected rule is passed to the optimization configuration model as an input condition; Comprehensive evaluation is performed on different prediction behaviors to find a scheme that can achieve the optimal load prediction accuracy under the current state.

[0014] As a preferred scheme of the power system load prediction method provided by the application, the triggering of the corresponding lower execution function to implement the prediction display operation comprises: When the optimal load prediction result is determined, the upper strategy function triggers the corresponding lower execution function; The lower execution function generates specific display instructions according to the optimal load prediction result; The instructions include but are not limited to drawing load curve graphs, column charts, displaying predicted load and actual load deviation, and providing correlation analysis reports; The specific display instructions are sent to the corresponding display devices or personnel to implement the specific display operation.

[0015] In a second aspect, the application provides a power system load prediction system, comprising: The data acquisition and processing module is configured to acquire historical load data, meteorological information and economic data of the target regional power grid, and preprocess the historical load data, meteorological information and economic data of the target regional power grid. The first model establishing module is configured to establish a first load characteristic analysis model according to the preprocessed data. The second model establishing module is configured to establish a second load prediction model according to the first load characteristic analysis model and the historical load data. The prediction module is configured to generate an optimal load prediction result based on the first load characteristic analysis model and the second load prediction model. The implementation module is configured to call the optimal load prediction result through a preset upper-layer strategy function, and trigger a corresponding lower-layer execution function to implement a prediction display operation.

[0016] In a third aspect, the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method described above when executing the computer program.

[0017] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the method described above.

[0018] Compared with the prior art, the present application has the following beneficial effects: the present application proposes a power system load prediction method, which comprehensively considers historical load data, meteorological information and economic data and other factors, avoiding the limitation of traditional methods which only consider a single factor. By establishing a first load characteristic analysis model and a second load prediction model, the load characteristics of the power grid are analyzed in depth, which can more accurately capture the change rule of the load, thereby effectively improving the accuracy of load prediction. In terms of model updating, the present application updates the parameters by feeding back the actual load data to the model, which ensures that the model can timely adapt to the change of the load characteristics of the power grid, and improves the reliability and adaptability of the model. In terms of display and application of the prediction result, the present application adjusts the prediction display content and mode flexibly according to the state of the power grid and environmental factors by presetting different strategy rules, which provides comprehensive and accurate support for the decision-making of the power system. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0020] Figure 1A method flow chart of a power system load prediction method is provided for an embodiment of the present application.

[0021] Figure 2 An internal structure diagram of an electronic device of a power system load prediction method is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0023] Embodiment 1, reference Figure 1 For the first embodiment of the present application, the embodiment provides a power system load prediction method, comprising: In the prior art, there are some problems, such as the load prediction method only considers a single factor, resulting in inaccurate prediction results. Some methods only rely on historical load data, but ignore the influence of meteorological information and economic data on load. At the same time, part of the model lacks in-depth analysis of the characteristics of the load, and it is difficult to capture the load change rule. In addition, the model is not updated in time, and cannot be updated according to the actual load data feedback iteration, so that the model gradually loses adaptability. Moreover, the display method of the prediction result is single, mostly simple charts, lacking in-depth analysis and mining, and cannot provide comprehensive support for power system decision-making.

[0024] The present application provides a method that can effectively solve the above-mentioned problems. Next, how to realize the power system load prediction method will be described in detail in combination with multiple embodiments; Figure 1 A method flow chart of a power system load prediction method is shown, comprising: S101, obtaining historical load data, meteorological information and economic data of a target regional power grid, and preprocessing the historical load data, meteorological information and economic data of the target regional power grid, wherein: It should be noted that in order to realize the load prediction of the power system, the related data of the target regional power grid need to be obtained, and the calculation logic for load prediction is obtained from these related data; In some specific embodiments, the relevant data can include weather data such as temperature, humidity, wind speed, etc. in different time periods within the target area, which are closely related to power load. For example, in hot weather, residents and businesses increase the use of air conditioning and other cooling equipment, which will cause a significant increase in power load; while in cold weather, the use of heating equipment will also increase the load. At the same time, economic development index data of the region should also be included, such as GDP growth rate, industrial structure proportion, etc. Different industrial structures have different demands for electricity, and industrialized areas usually have higher power load than service-oriented areas, and the speed of economic growth will also directly affect the consumption of electricity.

[0025] In some specific embodiments, population data is also one of the important relevant data, including population size, population density, and population distribution, etc. The power demand in densely populated areas is relatively large, and the flow and distribution of population will also affect the power load. There are also special event data such as large-scale activities, holidays, etc., which will cause abnormal fluctuations in power load. By comprehensively considering these relevant data, a more comprehensive and accurate information basis can be provided for subsequent load prediction, thereby improving the accuracy and reliability of load prediction.

[0026] However, in the embodiments of the present application, the historical load data of the target area power grid, weather information and economic data are obtained, the historical load data of the target area power grid reflects the power consumption in the past period of time, the weather information such as temperature, humidity, wind speed, etc. will significantly affect the power demand, and the economic data is related to the industrial development and resident consumption of the region, which will all affect the power grid load.

[0027] It should be noted that after obtaining the historical load data of the target area power grid, weather information and economic data, considering the quality and availability of the data, the data needs to be preprocessed.

[0028] In some specific embodiments, the preprocessing process includes data cleaning, missing value processing, outlier processing, and data normalization operations. Data cleaning is to remove noise and error information in the data to ensure the accuracy and consistency of the data. For missing values, interpolation method, mean method, etc. can be used for filling to ensure the integrity of the data. The processing of outliers is to identify and correct data points that deviate significantly from the normal range to avoid their adverse effects on subsequent analysis and modeling. Data normalization is to unify data of different ranges and scales to the same interval, which facilitates the model to process and analyze the data, and improves the stability and accuracy of the model.

[0029] Specifically, three types of key information are extracted from the historical database of the target area power grid: historical load data : represents the hourly electricity load value in the past period (e.g., the past 3 years), in units of megawatts (MW).

[0030] weather information : includes daily / hourly temperature , humidity , wind speed , whether it is raining (0 means no rain, 1 means rain), etc.

[0031] economic data : such as quarterly GDP growth rate , industrial added value , holiday identifier (0 means weekday, 1 means holiday), etc.

[0032] It should be noted that these data come from the power dispatching system, the weather bureau interface and the macroeconomic database, respectively.

[0033] Furthermore, since the sampling frequencies of data from different sources may be different (e.g., load is hourly, GDP is quarterly), time alignment and interpolation expansion are required. For low-frequency data (e.g. ), the forward filling method is used:

[0034] wherein represents the GDP growth rate at time point , and represents the value published in the latest quarter, represents the start time of the quarter, represents the end time of the quarter; Furthermore, all data are finally unified to the same time granularity (e.g., one record per hour) to form the original data set , wherein the subscript represents the time stamp.

[0035] Furthermore, in a real system, some sensor data may be missing. If there is no record of a certain feature at time t, linear interpolation is used to repair it:

[0036] wherein represents the feature value at the time point t to be filled, , represents its adjacent valid observation value sequence. If the first and last are missing, the nearest neighbor value is filled.

[0037] Further, consider that load surges or sensor failures can cause anomalies. The 3σ rule can be used to identify outliers: if , then determine as an anomaly, denotes the mean of the feature over the entire sequence, denotes the standard deviation of the feature. For anomaly points, replace with the median of its sliding window (e.g., 5 points) before and after.

[0038] Further, to eliminate dimensional differences, scale all features to the [0, 1] interval:

[0039] where, denotes the normalized feature value, denotes the original feature value, denotes the minimum and maximum values of the feature over the entire training set; Further, after processing, output the cleaned and standardized dataset as input for the next stage.

[0040] It should be noted that the output in this stage is the basis for all subsequent modeling. It is not only used to establish the load characteristic model, but will also be repeatedly called in feature extraction, type matching, and prediction training, and is the "data foundation" of the entire system.

[0041] S102, a first load characteristic analysis model is established according to the preprocessed data, wherein: It should be noted that after obtaining the preprocessed data, the first load characteristic analysis model can be constructed using these data. The model aims to deeply mine the internal relationship and variation law between the target regional power grid load and meteorological information and economic data.

[0042] In some specific embodiments, the first load characteristic analysis model can be established using a machine learning algorithm, such as a decision tree algorithm to establish the first load characteristic analysis model. The decision tree is a model that makes decisions based on tree structure, which can classify and predict load characteristics according to input features (such as temperature and humidity in meteorological information, GDP growth rate in economic data, etc.). By training the preprocessed data, the decision tree can learn the load variation trend under different feature combinations.

[0043] In some specific implementations, the first load characteristic analysis model can also be established using neural network algorithms. Neural networks have powerful nonlinear mapping capabilities and can handle complex input-output relationships. A multilayer perceptron neural network can be constructed, using preprocessed data as the input layer, performing feature extraction and nonlinear transformation through neurons in the hidden layers, and finally obtaining the predicted load characteristics at the output layer. Alternatively, the support vector machine algorithm is also an effective choice, as it can perform classification and regression analysis of load characteristics by finding the optimal hyperplane.

[0044] In this embodiment of the invention, establishing a first load characteristic analysis model based on preprocessed data includes: A preset set of load characteristic indicators is provided, which includes several load characteristic indicators of the power grid in the target area for evaluation. Select load characteristic indicators based on preprocessed data; The selected load characteristic indicators are represented by a scoring formula, and the load characteristic score of the target area power grid is calculated based on the scoring formula. The load characteristics of the power grid are classified according to the load characteristic scores to form the first load characteristic analysis model.

[0045] Specifically, establish the first load characteristic analysis model. To quantify the operating characteristics of the power grid, a set of predefined load characteristic indicators can be established. .

[0046] For example, design Peak-to-valley difference rate:

[0047] in, Indicates the daily maximum load. Indicates the daily minimum load. This represents the daily average load, while the peak-to-valley difference rate reflects the degree of load fluctuation. design Load factor:

[0048] Among them, the closer the load factor is to 1, the more stable the load is; design Daily volatility:

[0049] in, This represents the daily load standard deviation, while the daily volatility is used to measure load stability. design Temperature sensitivity:

[0050] in, The Pearson correlation coefficient is represented by the temperature sensitivity, which represents the intensity of the load's response to temperature changes. design Impact index of holidays:

[0051] Among them, a holiday impact index of less than 1 indicates that electricity consumption is reduced during holidays; Furthermore, correlation analysis is used to screen for indicators strongly correlated with future loads. Calculate each... Load for the next hour Correlation coefficient:

[0052] in, Indicates the first The correlation between these indicators and future load. Indicates time Calculated Individual indicator values, Indicates the respective mean; By way of example, the present invention reserves the right to satisfy (like The indicators form an effective subset. This step enables a focus from the "complete set" to the "key features," avoiding redundant interference.

[0053] Furthermore, to facilitate comparison and classification, each effective indicator... Convert to standardized score The scoring function is designed as follows:

[0054] in, Indicates the first Scoring of each indicator This represents the minimum / maximum value of the indicator in historical data, and then a weighted sum is taken to obtain the total characteristic score:

[0055] in, This represents the overall score of load characteristics (the higher the score, the more significant the characteristics). Indicates the first The weights of each indicator satisfy... The weights can be determined through expert scoring or the entropy weight method. This reflects the importance of different indicators in judging load behavior.

[0056] Further, the load characteristic classification is based on the score The current grid state is divided into different categories: If , it is determined to be stable type: small load change, easy to predict; If , it is determined to be general type: certain fluctuations; If , it is determined to be complex type: affected by many factors, high prediction difficulty; It should be noted that the model , the output is two parts of characteristic grade Feature vector , The output is the key basis for building a prediction model. Feature vector Will be used as input to extract the behavior pattern of the current load; and characteristic grade Will be used for subsequent strategy selection and model fusion weight allocation to achieve "class-based strategy".

[0057] S103, a second load prediction model is established according to the first load characteristic analysis model and historical load data, wherein: It should be noted that when the output of the first load characteristic analysis model is obtained, a second load prediction model can be constructed in combination with historical load data. The purpose of this model is to more accurately predict the future load of the target area power grid.

[0058] In some embodiments, the second load prediction model can use time series analysis method. Time series analysis can predict future load based on the change rule of historical load data. For example, autoregressive integrated moving average model (ARIMA), which can predict future load based on past load values, eliminate data non-stationarity through difference processing of historical load data, and then use autoregressive and moving average method to fit the change rule of data.

[0059] In some embodiments, a deep learning model such as long short-term memory network (LSTM) can also be used to construct the second load prediction model. LSTM can handle data with long-term dependencies, which is very suitable for load prediction problems that need to consider the long-term impact of historical data. It can learn complex patterns and trends in historical load data, and can adjust the prediction strategy and weight according to the current load characteristic feature vector and characteristic grade, thereby improving the accuracy of prediction.

[0060] In the embodiments of the present application, the second load prediction model is established according to the first load characteristic analysis model and historical load data, comprising: A preset load type set, the load type set including several different load types and corresponding features; Feature extraction is performed on the output of the first load characteristic analysis model, and feature information related to the features in the preset load type set is extracted; The extracted features are matched and compared with different features in the preset load type set, and the matching degree of each load type is calculated; According to the matching degree and the historical load data, the probability of each load type occurring under the current power grid state is determined, and a second load prediction model is constructed.

[0061] Specifically, the second load prediction model is established , several typical load behavior patterns are defined in advance through the preset load type set, forming a type set . Each type has a corresponding feature template ; Exemplarily, set to represent a weekday smooth type ; to represent a high-temperature air conditioning type ; to represent a holiday low type , these templates are summarized from historical data analysis. The high and low of the data can be obtained through a preset threshold or historical experience.

[0062] Further, from the output , a feature subset related to type identification is extracted .

[0063] Exemplarily, if pay attention to temperature sensitivity, extract ; if pay attention to holiday effect, extract , that is:

[0064] Among them, represents a feature vector for type matching, represents the key index relied on by each type; Further, the similarity between the current feature and each template is calculated, using the cosine similarity:

[0065] Among them, represents the matching degree of the current state and the jth type, ranging from 0 to 1; denotes vector dot product; |·| denotes vector length

[0066] Further, the historical frequency , i.e. the proportion in the past year , the real-time matching degree is fused to obtain the occurrence probability of each type in the current state:

[0067] wherein, denotes the probability of the jth load type in the current occurrence, and a e (0, 1) denotes a dynamic fusion coefficient (such as 0.6) controlling the weight of "current performance" and "historical habit" Further, normalization is performed:

[0068] It should be noted that the model , the core of which is the probability distribution P=[ , , …, ], is used to guide the selection and fusion of subsequent prediction models. The output P of the model is the core weight for generating the optimal prediction. It tells the system: "which load mode is the most likely at present", so as to decide which sub-model should be focused on for prediction, and realize "on-demand prediction".

[0069] S104, generating an optimal load prediction result based on the first load characteristic analysis model and the second load prediction model, wherein: It should be noted that after the output of the first load characteristic analysis model and the output of the second load prediction model are obtained, the two can be combined to generate an optimal load prediction result.

[0070] In some specific embodiments, the optimal load prediction result can be generated by using a model fusion method according to the characteristic level and the feature vector output by the first load characteristic analysis model, and the load type probability distribution output by the second load prediction model. For example, for the case of stable type of characteristic level, the prediction sub-model constructed based on time series analysis method can be given higher weight, because the change rule of stable type of load is relatively simple, and the time series analysis method can better capture its rule; and for the case of complex type of characteristic level, the weight of the prediction sub-model constructed by the deep learning model can be increased to cope with the complex and changeable load situation.

[0071] In some specific embodiments, the optimal load prediction result can also be generated by using a weighted average method according to the feature vector output by the first load characteristic analysis model and the load type probability distribution output by the second load prediction model. Specifically, weights corresponding to the load type probability distribution are assigned to different prediction sub-models, and the prediction results of the sub-models are weighted and summed according to the weights to obtain the final load prediction value. For example, if the second load prediction model output shows that a certain load type has a high probability of occurrence, the corresponding prediction sub-model will be given a greater weight in the weighted summation.

[0072] In the embodiments of the present application, generating the optimal load prediction result based on the first load characteristic analysis model and the second load prediction model comprises: determining the load prediction period and the time resolution of the target regional power grid; establishing an optimization configuration model for maximizing the load prediction accuracy; combining the load characteristic analysis result and the load prediction model to generate the optimal load prediction result.

[0073] Specifically, the optimal load prediction result can be further generated by setting the prediction task parameters, specifying the prediction target, setting the prediction period to H hours in the future (e.g., H=24), and setting the time resolution to one step every Δt hours, which can be set to Δt=1. Further, the loss function is defined as the mean absolute percentage error (MAPE):

[0074] wherein, represents the predicted load at the hth step, represents the true load; Further, a regularization term is introduced to prevent overfitting:

[0075] wherein, represents the adjustable parameters of the prediction model (such as neural network weights, regression coefficients), and λ represents the regularization coefficient, which can be set to 0.01; Further, the optimization objective is:

[0076] Further, the outputs of and are used to weight and combine the outputs of multiple special-purpose predictors:

[0077] wherein, represents the final prediction result, representing the occurrence probability of the type (from ), representing the output of the dedicated prediction model trained for the type, which can be LSTM, XGBoost, etc. For example, if (high temperature type) is very high, the system automatically increases the weight of the temperature-related model.

[0078] It should be noted that the optimal load prediction sequence The prediction result is not only an output, but also a "signal source" that triggers subsequent intelligent display and strategy response. Its accuracy directly affects the quality of decision-making.

[0079] S105, the optimal load prediction result is called by the preset upper strategy function, and the corresponding lower execution function is triggered to implement the prediction display operation, wherein: It should be noted that the preset upper strategy function can make judgments and decisions according to the optimal load prediction result. For example, when the optimal load prediction result shows that the load will rise sharply in the future, the upper strategy function will judge according to the preset rules that corresponding measures need to be taken.

[0080] In some specific embodiments, if the load rise exceeds a certain preset safety threshold, the upper strategy function will call the corresponding lower execution function. These lower execution functions can be functions for notifying the power dispatching department to allocate power, or functions for starting standby power generation equipment.

[0081] In some specific embodiments, when performing the prediction display operation, the lower execution function can display the optimal load prediction result in an intuitive way. For example, by displaying the load prediction values of different time periods in the form of a chart, relevant personnel can clearly see the trend of load changes. It can also list various prediction data in the form of a report, including the occurrence probability of different load types, the output of each dedicated prediction model, and other information, to facilitate in-depth analysis and decision-making.

[0082] In some specific embodiments, the lower execution function can also provide personalized display content according to different user roles and needs. For power dispatch personnel, they may be more concerned about the overall trend of load changes and peak prediction in different time periods, in order to reasonably arrange power resources; for management personnel, they may be more concerned about the impact of prediction results on costs and benefits.

[0083] In some specific embodiments, in the process of displaying the prediction result, real-time data can also be combined for dynamic updating. When new real-time data is transmitted into the system, the prediction result is adjusted in time and re-displayed, ensuring the timeliness and accuracy of the displayed content. In this way, relevant personnel can make scientific and reasonable decisions in a timely manner based on the latest prediction information, improving the operation efficiency and stability of the power system.

[0084] In the embodiments of the present application, the optimal load prediction result is obtained by calling the preset upper strategy function, which includes: presetting strategy rules under different combinations of power grid states and environmental factors; the upper strategy function selects the corresponding rule from the preset strategy rules according to the current obtained power grid state and environmental factor; the selected rule is passed to the optimization configuration model as an input condition; different prediction behaviors are comprehensively evaluated to find a scheme that can achieve the optimal load prediction accuracy under the current state.

[0085] In the embodiments of the present application, triggering the corresponding lower execution function to implement the prediction display operation includes: after determining the optimal load prediction result, the upper strategy function triggers the corresponding lower execution function; the lower execution function generates specific display instructions according to the optimal load prediction result; the instructions include but are not limited to drawing load curve graphs, column charts, displaying the deviation between predicted load and actual load, and providing correlation analysis reports; the specific display instructions are sent to the corresponding display devices or personnel to implement specific display operations.

[0086] In the embodiments of the present application, the actual load data for obtaining the optimal load prediction result is obtained; the actual load data is fed back to the first load characteristic analysis model and the second load prediction model for parameter iteration update; until the iteration condition is met.

[0087] In summary, this invention proposes a power system load forecasting method that comprehensively considers multiple factors, including historical load data, meteorological information, and economic data, avoiding the limitations of traditional methods that only consider a single factor. By establishing a first load characteristic analysis model and a second load forecasting model, the load characteristics of the power grid are analyzed in depth, enabling more accurate capture of load variation patterns and thus effectively improving the accuracy of load forecasting. Regarding model updates, this invention feeds actual load data back to the model for iterative parameter updates, ensuring that the model can adapt to changes in power grid load characteristics in a timely manner, improving the model's reliability and adaptability. In terms of the display and application of forecast results, this invention uses preset different strategy rules to flexibly adjust the content and method of forecast display according to power grid conditions and environmental factors, providing comprehensive and accurate support for power system decision-making.

[0088] Example 2, in a preferred embodiment, the specific operation of calling the optimal load prediction result through a preset upper-level strategy function and triggering the corresponding lower-level execution function to perform the prediction display operation can be as follows: Preset a set of policy rule bases Each rule corresponds to a specific combination of power grid state and environment. (Policy function) Based on the current state Choose the optimal strategy:

[0089] in, Indicates from The load characteristic level, Indicates from The type probability, Indicates the current temperature. Indicate whether it is a holiday; For example, if Complex and If so, a "high temperature and high load warning" will be triggered; if If this occurs, the "Holiday Power Supply Guarantee Plan" will be triggered. Furthermore, once a strategy is selected... The corresponding execution function is triggered immediately. Generate specific operation instructions: To plot a load curve, specific operations may include setting the horizontal axis to time and the vertical axis to... With subsequent actual values ; Displaying deviations, specific operations may include setting up the calculation of error at each step. ; Generate a report; specific operations may include setting up correlation analysis, ranking of key influencing factors, confidence intervals, etc. The instructions are sent to the dispatch large screen, mobile terminal or manager mailbox through the API.

[0090] Further, when the real load arrives, the error is calculated immediately and the model parameters are updated by back propagation:

[0091] wherein, denotes the learning rate, which can be set to 0.001, denotes the gradient of the loss function with respect to the parameters, and the parameters are updated simultaneously The index statistical range in the above formula is, for example, , The historical frequency in the above formula is, for example, The iteration continues until the convergence condition is met:

[0092] wherein, denotes the convergence threshold, which is designed to be 0.005 in the present application, and the entire system thus realizes the self-evolution ability of "the more you use, the more accurate you are".

[0093] In Embodiment 3, with reference to Figure 2 , the present embodiment also provides a power system load prediction system, comprising: a data acquisition and processing module, configured to acquire historical load data, meteorological information and economic data of a target regional power grid, and to preprocess the historical load data, meteorological information and economic data of the target regional power grid; a first model establishing module, configured to establish a first load characteristic analysis model according to the preprocessed data; a second model establishing module, configured to establish a second load prediction model according to the first load characteristic analysis model and the historical load data; a prediction module, configured to generate an optimal load prediction result based on the first load characteristic analysis model and the second load prediction model; an implementation module, configured to call the optimal load prediction result through a preset upper-layer strategy function, and trigger a corresponding lower-layer execution function to implement a prediction display operation.

[0094] The above various unit modules can be embedded in or independent of the processor in the electronic device in hardware form, or can be stored in the memory in the electronic device in software form, so as to be called and executed by the processor to perform the operations corresponding to the above various modules.

[0095] The present embodiment also provides an electronic device, which can be a terminal, and the internal structure diagram thereof can be as shown in Figure 2As shown in the figure. The electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capability. The memory of the electronic device includes non-volatile storage medium, internal memory. The non-volatile storage medium stores the operating system and the computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the electronic device is used for wired or wireless communication with external terminals. Wireless mode can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement a power system load prediction method. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen. The input device of the electronic device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the electronic device, or an external keyboard, touchpad or mouse, etc.

[0096] The embodiment also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the following steps: Obtain historical load data, weather information and economic data of a target regional power grid, and preprocess the historical load data, weather information and economic data of the target regional power grid; Establish a first load characteristic analysis model according to the preprocessed data; Establish a second load prediction model according to the first load characteristic analysis model and the historical load data; Generate an optimal load prediction result based on the first load characteristic analysis model and the second load prediction model; Call the optimal load prediction result through a preset upper-layer strategy function, and trigger a corresponding lower-layer execution function to implement a prediction display operation.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all modifications and replacements should be covered in the scope of the claims of the present application.

[0098] Although the preferred embodiments of the present application have been described, those skilled in the art can make further changes and modifications to the embodiments once they understand the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all modifications and replacements falling within the scope of the present application.

[0099] Obviously, many modifications and variations of the present application are possible in light of the above teachings. It is, therefore, to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A method of load forecasting for an electric power system, characterized by, The method comprises the following steps: obtaining historical load data, meteorological information and economic data of a target regional power grid, and preprocessing the historical load data, meteorological information and economic data of the target regional power grid; establishing a first load characteristic analysis model according to the preprocessed data; establishing a second load prediction model according to the first load characteristic analysis model and the historical load data; generating an optimal load prediction result based on the first load characteristic analysis model and the second load prediction model; calling the optimal load prediction result through a preset upper-layer strategy function, and triggering a corresponding lower-layer execution function to implement a prediction display operation.

2. A power system load forecasting method as claimed in claim 1, characterized by, Further comprising: obtaining actual load data of the optimal load prediction result; feeding back the actual load data to the first load characteristic analysis model and the second load prediction model for parameter iteration update; until the iteration condition is met.

3. A power system load forecasting method as claimed in claim 2, wherein, The step of establishing the first load characteristic analysis model according to the preprocessed data comprises: presetting a load characteristic index set, wherein the load characteristic index set comprises a plurality of load characteristic indexes for evaluating the target regional power grid; selecting load characteristic indexes based on the preprocessed data; characterizing the selected load characteristic indexes by a scoring formula, and calculating a load characteristic score of the target regional power grid according to the scoring formula; grading the load characteristics of the power grid according to the load characteristic score to form the first load characteristic analysis model.

4. A power system load forecasting method as claimed in claim 3, wherein, The step of establishing the second load prediction model according to the first load characteristic analysis model and the historical load data comprises: presetting a load type set, wherein the load type set comprises a plurality of different load types and corresponding features; extracting features from the output of the first load characteristic analysis model, and extracting feature information related to the features in the preset load type set; matching and comparing the extracted features with different features in the preset load type set, and calculating a matching degree of each load type; determining a probability of each load type occurring under the current power grid state according to the matching degree and the historical load data, and constructing the second load prediction model.

5. A power system load forecasting method as claimed in claim 4, characterized by, The step of generating the optimal load prediction result based on the first load characteristic analysis model and the second load prediction model comprises: determining a load prediction period and a time resolution of the target regional power grid; establishing an optimization configuration model for maximizing the load prediction accuracy; combining the load characteristic analysis result and the load prediction model to generate the optimal load prediction result.

6. A power system load forecasting method as claimed in claim 5, characterized by, The step of calling the optimal load prediction result through the preset upper-layer strategy function comprises: presetting strategy rules under different combinations of power grid states and environmental factors; selecting corresponding rules from the preset strategy rules according to the current acquired power grid state and environmental factors by the upper-layer strategy function; passing the selected rules as input conditions to the optimization configuration model; comprehensively evaluating different prediction behaviors to find a scheme that can achieve the optimal load prediction accuracy under the current state.

7. A method of power system load forecasting as defined in claim 6, wherein, The step of triggering the corresponding lower-layer execution function to implement the prediction display operation comprises: triggering the corresponding lower-layer execution function by the upper-layer strategy function after determining the optimal load prediction result; generating specific display instructions by the lower-layer execution function according to the optimal load prediction result; The instructions include, but are not limited to, drawing load curve charts, column charts, displaying predicted load and actual load deviation, and providing correlation analysis reports; Specific display instructions are sent to corresponding display devices or personnel to implement specific display operations.

8. A power system load forecasting system applying the method according to any one of claims 1 to 7, characterized in that, Comprise: A data acquisition and processing module is configured to acquire historical load data, meteorological information and economic data of a target regional power grid, and preprocess the historical load data, meteorological information and economic data of the target regional power grid; A first model establishing module is configured to establish a first load characteristic analysis model according to the preprocessed data; A second model establishing module is configured to establish a second load prediction model according to the first load characteristic analysis model and historical load data; A prediction module is configured to generate an optimal load prediction result based on the first load characteristic analysis model and the second load prediction model; An implementation module is configured to call the optimal load prediction result through a preset upper-layer strategy function, and trigger a corresponding lower-layer execution function to implement a prediction display operation. 9.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to implement the steps of the power system load prediction method in any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the power system load prediction method in any one of claims 1-7.

Images