A transformer loss reduction intelligent control method and system based on a power generation end

Abstract

The application discloses a transformer consumption reduction intelligent control method and system based on a power generation end, relates to the technical field of electric power control, and collects transformer operation data sets and historical load data sets in advance, constructs an ARIMA model based on the transformer operation data sets and the historical load data sets, generates residual data sets by using the ARIMA model, constructs and trains an LSTM model for predicting future load demand based on the residual data sets and the transformer operation data sets, collects real-time operation data in the actual use process of the transformer in real time, predicts the future load demand corresponding to the real-time operation data by using the combination of the ARIMA model and the LSTM model, constructs a consumption reduction optimization model based on the future load demand and the real-time operation data, and adjusts the transformer operation strategy by using an optimization algorithm based on the consumption reduction optimization model, so that the load change can be intelligently responded to and a regulation and control strategy can be generated in real time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power control, in particular to a transformer energy consumption reduction intelligent control method and system based on a power generation end. BACKGROUND

[0002] In modern power systems, transformers at the power generation end serve as a key link connecting power generation equipment and transmission networks, and their operating efficiency directly affects the energy transmission and distribution efficiency of the entire power system. With the increasing global energy consumption and the growing demand for sustainable energy utilization, reducing transformer energy consumption has become one of the important tasks of the power industry. In traditional transformer operation, fixed strategies are often used to cope with changing load demands. While this approach is simple, it is not very efficient, especially in cases of large load fluctuations, where fixed strategies can lead to energy waste and excessive wear and tear on equipment. In addition, current load forecasting often relies on simple historical averages or rule-based scheduling, lacking sufficient use of real-time data and dynamic adjustment capabilities.

[0003] The prior art has obvious shortcomings in dealing with complex load changes and improving transformer operating efficiency. Traditional load forecasting methods often ignore the dynamic characteristics and multi-dimensional interactions of real-time data, and the adjustment of transformer operating strategies relies on human experience and static rules, which cannot quickly respond to load changes. This not only leads to energy waste, but also increases the risk of equipment failure and shortens the service life.

[0004] Chinese patent CN117691594B discloses an energy-saving and consumption-reducing judgment method and system for a transformer, which collects and processes real-time operating data, historical data, and cost data of the transformer; analyzes the processed data; assigns a preset weight to each analysis result and scores it; calculates a comprehensive score based on the score of each analysis result and the corresponding weight; and determines whether energy saving and consumption reduction is needed based on whether the comprehensive score exceeds a preset comprehensive score threshold. However, this method does not solve the problem of intelligent energy consumption reduction control.

[0005] Therefore, the present application proposes a transformer energy consumption reduction intelligent control method and system based on a power generation end. SUMMARY

[0006] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application proposes a transformer energy consumption reduction intelligent control method and system based on a power generation end, which can intelligently respond to load changes in real time and generate control strategies.

[0007] To achieve the above-mentioned purpose, a transformer energy consumption reduction intelligent control method based on a power generation end is proposed, comprising the following steps:

[0008] Step one: pre-collecting transformer operation data set and historical load data set;

[0009] Step two: based on the transformer operation data set and the historical load data set, constructing an ARIMA model and using the ARIMA model to generate a residual data set;

[0010] Step three: based on the residual data set and the transformer operation data set, constructing and training an LSTM model for predicting future load demand;

[0011] Step four: collecting real-time operation data in the actual use process of the transformer in real time, and using the combination of the ARIMA model and the LSTM model to predict the future load demand corresponding to the real-time operation data;

[0012] Step five: based on the future load demand and the real-time operation data, constructing a consumption reduction optimization model;

[0013] Step six: based on the consumption reduction optimization model, using an optimization algorithm to adjust the transformer operation strategy;

[0014] The pre-collected transformer operation data set and historical load data set are as follows:

[0015] A sensor network is installed for the transformer, and the parameter values of each sensor are monitored and recorded in real time. By reading the parameter values of each sensor, a transformer operation data set is formed;

[0016] By reading the historical database in the transformer operation process and the smart meter, a historical load data set is collected;

[0017] The ARIMA model is constructed based on the historical load data set, and the residual data set is generated using the ARIMA model as follows:

[0018] For each load parameter in the historical load data set, the residual data set includes the following steps:

[0019] Step 21: arrange the parameter values of the load parameter in time order and arrange them into time series format to obtain the load parameter time series of the corresponding load parameter;

[0020] Step 22: analyze whether each load parameter time series has a stable mean and variance by using statistical methods. If it does not have a stable mean and variance, apply the difference method to make the load parameter time series stationary. The difference method eliminates trends or seasonality by calculating the difference between adjacent data points;

[0021] Step 23: draw autocorrelation function and partial autocorrelation function graphs for the load parameter time series to identify the internal patterns of the time series data;

[0022] Step 24: According to the autocorrelation function graph and the partial autocorrelation function graph, select the parameter combination (p, d, q) of the ARIMA model; wherein p is the order of the autoregressive part, indicating the number of lags in the model, d is the difference number, indicating the difference order required for stationarization, and q is the order of the moving average part, indicating the number of lagged prediction error terms;

[0023] Step 25: Use the selected (p, d, q) parameters to construct the ARIMA model and perform fitting; verify the accuracy of the model by checking the residuals of the model, and use statistical test methods to confirm the randomness of the residuals;

[0024] Step 26: In the ARIMA model fitting process, each unit of time in the parameter time series has a predicted value; the residual is the actual parameter value minus the predicted value given by the ARIMA model;

[0025] Step 27: Collect the residuals of each unit of time to form a residual data set;

[0026] The construction process of the LSTM model includes:

[0027] Standardize all parameters in the transformer operating data set and the residual data set to ensure that the parameters are on the same scale;

[0028] A fixed time window is preset to predict the load demand of the next time step;

[0029] The LSTM model takes the parameter time series composed of various transformer operating data parameters and the residual time series composed of residual values of various load parameters within a fixed time window as an input sample;

[0030] The LSTM model takes the load demand in the next preset time step as the output, and takes the actual load demand in the next preset time step in the historical load data set as the prediction target;

[0031] The structure of the LSTM model includes an input layer, an LSTM layer, a fully connected layer, and an output layer;

[0032] The input layer is used to receive multi-dimensional time series data, i.e., residuals and transformer operating data;

[0033] The LSTM layer includes one or more LSTM layers that can be stacked to capture long-term dependencies in time series; the fully connected layer connects the LSTM layer output to the final output node to generate a prediction value, and the output layer outputs a single prediction value, i.e., the load demand of the next time step;

[0034] The training process of the LSTM model includes:

[0035] All input samples are divided into a training set and a test set;

[0036] The mean square error is used as the loss function of the LSTM model to measure the difference between the predicted value and the actual value;

[0037] The Adam optimizer is used as the optimizer of the LSTM model to improve the efficiency and effect of training;

[0038] The LSTM model is trained on the training set, and the network weights are adjusted to minimize the loss function;

[0039] At the end of each iteration, the performance of the LSTM model is evaluated on the validation set, and the model hyperparameters are adjusted to prevent overfitting;

[0040] The way of using ARIMA model and LSTM model combination to predict the future load demand corresponding to the real-time running data is:

[0041] By using the ARIMA model to predict each unit of time for each load parameter, the predicted value of the corresponding load parameter is obtained;

[0042] Calculate the difference between the predicted value of each unit of time load parameter and the actual parameter value to form a residual set;

[0043] The residual set and the parameter values of each real-time running parameter form a new input sample, which is input into the LSTM model to obtain the predicted value of the load demand output by the LSTM model;

[0044] The way of constructing a consumption reduction optimization model based on future load demand and real-time running data is:

[0045] The predicted value of the load demand is marked as Pr, the voltage of the transformer is marked as V, the power factor of the transformer is marked as pf, and the equivalent resistance of the core is marked as Rc;

[0046] The winding resistance of the transformer is marked as R;

[0047] Based on the predicted value of the load demand, the voltage, the power factor, a consumption reduction objective function F is constructed;

[0048] According to the preset safety voltage range and safety current range of the transformer, a set of consumption reduction optimization conditions U is constructed;

[0049] A convex optimization model is constructed as a consumption reduction optimization model, with the minimization of the consumption reduction objective function F as the optimization objective and the set of consumption reduction optimization conditions U as the constraint condition set;

[0050] The transformer operation strategy is adjusted by using an optimization algorithm based on the loss reduction optimization model.

[0051] The loss reduction optimization model is solved by using a nonlinear programming solution algorithm or a heuristic algorithm to obtain a solved voltage value, and the transformer operation strategy is adjusting the voltage of the transformer to the solved voltage value.

[0052] A transformer loss reduction intelligent control system based on a power generation end is provided, comprising a sample data collection module, a residual data generation module, an LSTM model training module, a load prediction module, and a strategy adjustment module; wherein each module is connected through an electrical connection.

[0053] The sample data collection module pre-collects a transformer operation data set and a historical load data set, and sends the historical load data set to the residual data generation module and the transformer operation data set to the LSTM model training module.

[0054] The residual data generation module constructs an ARIMA model based on the transformer operation data set and the historical load data set, generates a residual data set using the ARIMA model, and sends the residual data set to the LSTM model training module and the trained ARIMA model to the load prediction module.

[0055] The LSTM model training module constructs and trains an LSTM model for predicting future load demand based on the residual data set and the transformer operation data set, and sends the trained LSTM model to the load prediction module.

[0056] The load prediction module collects real-time operation data in the actual use of the transformer in real time, uses an ARIMA model and an LSTM model combination to predict the future load demand corresponding to the real-time operation data, and sends the future load demand to the strategy adjustment module.

[0057] The strategy adjustment module constructs a loss reduction optimization model based on the future load demand and the real-time operation data, and adjusts the transformer operation strategy by using an optimization algorithm based on the loss reduction optimization model.

[0058] An electronic device is provided, comprising a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0059] The processor executes the above-mentioned transformer loss reduction intelligent control method based on the power generation end by calling the computer program stored in the memory.

[0060] A computer readable storage medium is provided, which stores an erasable computer program;

[0061] When the computer program runs on the computer device, the computer device is caused to execute the above-mentioned transformer loss reduction intelligent control method based on a power generation end.

[0062] Compared with the prior art, the beneficial effects of the present application are:

[0063] The present application pre-collects a transformer operation data set and a historical load data set, constructs an ARIMA model based on the transformer operation data set and the historical load data set, generates a residual data set using the ARIMA model, constructs and trains an LSTM model for predicting future load demand based on the residual data set and the transformer operation data set, collects real-time operation data in real time during actual use of the transformer, uses the ARIMA model and the LSTM model to predict future load demand corresponding to the real-time operation data, constructs a loss reduction optimization model based on the future load demand and the real-time operation data, and adjusts the transformer operation strategy using an optimization algorithm based on the loss reduction optimization model. First, by collecting the operation data and historical load data of the transformer, a refined load demand prediction model is established. The advanced ARIMA and LSTM combined model can effectively decompose and predict the linear and nonlinear parts of the load. Based on the prediction results, the tap switch position and cooling system settings of the transformer are dynamically adjusted to optimize the voltage level and reduce energy loss, which can respond to load changes and generate control strategies in real time. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 A flowchart of a transformer loss reduction intelligent control method based on a power generation end in embodiment 1 of the present application;

[0065] Figure 2 A module connection relationship diagram of a transformer loss reduction intelligent control system based on a power generation end in embodiment 2 of the present application. DETAILED DESCRIPTION

[0066] The technical solutions of the present application will be described below in conjunction with the embodiments, obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0067] Embodiment 1

[0068] As shown in Figure 1 A transformer loss reduction intelligent control method based on a power generation end, comprising the following steps:

[0069] Step 1: Pre-collect a transformer operation data set and a historical load data set;

[0070] Step two: based on the transformer operation data set and the historical load data set, an ARIMA model is constructed, and the ARIMA model is used to generate a residual data set;

[0071] Step three: based on the residual data set and the transformer operation data set, an LSTM model for predicting future load demand is constructed and trained;

[0072] Step four: real-time operation data during actual use of the transformer is collected in real time, and the ARIMA model and the LSTM model are combined to predict the future load demand corresponding to the real-time operation data;

[0073] Step five: based on the future load demand and the real-time operation data, a consumption reduction optimization model is constructed;

[0074] Step six: based on the consumption reduction optimization model, an optimization algorithm is used to adjust the transformer operation strategy;

[0075] Among them, the way to pre-collect the transformer operation data set and the historical load data set is:

[0076] A sensor network is installed for the transformer, and the parameter values of each sensor are monitored and recorded in real time, and the transformer operation data set is composed by reading the parameter values of each sensor;

[0077] The historical database during the operation of the transformer and the smart meter are read to collect the historical load data set;

[0078] Specifically, the sensor network is used to collect real-time operation parameters of the transformer in real time, and the real-time operation parameters include but are not limited to voltage, current, temperature, power coefficient, switch state and the like. Each sensor in the sensor network corresponds to a real-time operation parameter, for example, a voltage sensor collects real-time voltage, and a data collection device such as a data logger or a programmable logic controller is used to connect the sensor network, so as to realize real-time collection and storage of operation data. Further, the parameter data collected by the sensor can be transmitted to a centralized database or a cloud storage platform through a wired or wireless network, so as to be saved by using a SCADA system;

[0079] The historical database is an existing database of the power company or power station, which records the past transformer operation and load data;

[0080] The historical load data set includes load parameters of the transformer at each unit time;

[0081] Specifically, the load parameters include but are not limited to load demand, power consumption, seasonality and daily variation;

[0082] The load demand can be read by using the smart meter and the historical load data stored in the historical record database, the electricity consumption can be recorded by the meter or extracted from the historical electricity bill, and the seasonality and daily variation refer to the seasonal and daily variation patterns of the load demand, which can be formed by long-term data accumulation and are usually recorded by the power company; it should be noted that the collection frequency of each load parameter in the collected historical load data set should be consistent, for example, collected once an hour;

[0083] Further, the ARIMA model is constructed based on the historical load data set, and the residual data set is generated using the ARIMA model in the following manner:

[0084] For each load parameter in the historical load data set, generating the residual data set includes the following steps:

[0085] Step 21: arrange the parameter values of the load parameter in time order and arrange them into time series format to obtain the load parameter time series corresponding to the load parameter;

[0086] Step 22: analyze whether each load parameter time series has a stable mean and variance by using a statistical method, and if it does not have a stable mean and variance, apply a difference method to make the load parameter time series stationary, the difference method eliminates trends or seasonality by calculating the difference between adjacent data points; specifically, the determination of stationarity can be determined using the ADF test method;

[0087] Step 23: plot the autocorrelation function graph and the partial autocorrelation function graph for the load parameter time series to identify the internal pattern of the time series data;

[0088] Step 24: select the parameter combination (p, d, q) of the ARIMA model according to the autocorrelation function graph and the partial autocorrelation function graph; wherein p is the order of the autoregressive part, indicating the number of lags in the model, d is the difference number, indicating the difference order required for stationarization, and q is the order of the moving average part, indicating the number of lagged prediction error terms;

[0089] It should be noted that the autocorrelation function graph is the ACF graph, which shows the correlation between the time series and its lagged values, and is used to identify the order qq of the moving average (MA) component; in the ACF graph, significant truncation at lag (i.e., points significantly different from zero) indicates the order of MA;

[0090] The partial autocorrelation function graph is the PACF graph, which shows the correlation between the time series and its lagged values while controlling other lag effects, and is used to identify the order pp of the autoregressive (AR) component; in the PACF graph, significant truncation at lag indicates the order of AR;

[0091] Therefore, the difference order d can be generally set as 1, and p and q can be manually selected according to the truncated and trailing patterns by observing the ACF and PACF graphs, or the optimal parameters can be selected by using the automatic function of the statistical software;

[0092] For example, when the following features are observed:

[0093] ACF graph: significantly truncated at lag 2, indicating a possible MA(2);

[0094] PACF graph: significantly truncated at lag 1, indicating a possible AR(1);

[0095] Set the parameter combination (p, d, q) as (1, 2, 1);

[0096] Step 25: use the selected (p, d, q) parameters to construct an ARIMA model and perform fitting; verify the accuracy of the model by checking the residuals of the model, and use statistical test methods such as Ljung-Box test to confirm the randomness of the residuals;

[0097] Step 26: in the process of fitting the ARIMA model, each unit of time in the parameter time series has a predicted value; the residual is the actual parameter value minus the predicted value given by the ARIMA model;

[0098] Step 27: collect the residuals of each unit of time to form a residual data set;

[0099] It can be understood that the ARIMA model is usually constructed based on single variable time series data, that is, it mainly relies on the historical values of the load data itself for prediction, so it is good at handling linear trends and seasonal components, but it may not perform well in the face of complex nonlinear relationships. The residual of each unit of time is the linear part of the load data, and further analysis of the nonlinear changes of the transformer operating data and the nonlinear changes of the load data is needed;

[0100] Further, the construction process of the LSTM model includes:

[0101] Standardize all parameters in the transformer operating data set and the residual data set to ensure that the parameters are on the same scale;

[0102] A fixed time window is preset to predict the load demand of the next time step; specifically, the size of the time window can be set to the past 5 hours;

[0103] The LSTM model takes the parameter time series composed of parameters of each transformer operating data in a fixed time window and the residual time series composed of residual values of each load parameter as an input sample;

[0104] The LSTM model takes the load demand in the next preset time step of each input sample as the output, and takes the actual load demand in the next preset time step in the historical load data set as the prediction target;

[0105] The structure of the LSTM model includes an input layer, an LSTM layer, a fully connected layer, and an output layer;

[0106] The input layer is used to receive multi-dimensional time series data, i.e., residual and transformer operating data. The LSTM layer includes one or more LSTM layers that can be stacked to capture long-term dependencies in time series. The number of units in the LSTM layer is usually between 50 and 200, depending on the complexity of the data and the training resources. The fully connected layer connects the LSTM layer output to the final output node to generate the predicted value. The output layer outputs a single predicted value, i.e., the load demand in the next time step.

[0107] Further, the training process of the LSTM model includes:

[0108] All input samples are divided into a training set and a test set. Generally, the ratio of the training set and the test set is 70% training set and 30% test set.

[0109] The mean square error is used as the loss function of the LSTM model to measure the difference between the predicted value and the actual value.

[0110] The Adam optimizer is used as the optimizer of the LSTM model to improve the efficiency and effectiveness of the training.

[0111] The LSTM model is trained on the training set to adjust the network weights to minimize the loss function.

[0112] At the end of each iteration, the performance of the LSTM model is evaluated on the validation set to adjust the model hyperparameters to prevent overfitting.

[0113] It can be understood that the LSTM model can capture complex nonlinear dynamics in time series. Therefore, by first processing the linear part of the data with ARIMA and then processing the residual part (i.e., the nonlinear component) with LSTM, a more comprehensive model of the time series can be built, thereby improving the prediction accuracy. By combining the advantages of the two models, the change in power load can be better predicted, thereby optimizing the load management strategy of the transformer and ultimately achieving the goal of reducing consumption.

[0114] Further, the manner that the real-time running data of the transformer in actual use is collected in real time is:

[0115] Before the transformer is actually used, a sensor network is installed on the transformer, and parameter values of each real-time running parameter and parameter values of the actual load parameter of the transformer per unit time are collected through the sensor network to form the real-time running data;

[0116] Further, the manner that the real-time running data of the transformer in actual use is collected in real time is:

[0117] The parameter values of each load parameter per unit time are predicted by using the ARIMA model to obtain the predicted values of the corresponding load parameters;

[0118] The difference between the predicted values and the actual parameter values of the load parameter per unit time is calculated to form a residual set;

[0119] The residual set and the parameter values of each real-time running parameter are used to form a new input sample, which is input into the LSTM model to obtain the predicted value of the load demand output by the LSTM model;

[0120] Further, the manner that the real-time running data of the transformer in actual use is collected in real time is:

[0121] The predicted value of the load demand is marked as Pr, the voltage of the transformer is marked as V, the power factor of the transformer is marked as pf, and the equivalent resistance of the core is marked as Rc; It can be understood that for the transformer, the voltage value is a variable parameter, that is, the voltage is a controllable variable, and the power factor is a constant in the range of (0, 1), indicating the proportion of active power in apparent power;

[0122] The winding resistance of the transformer is marked as R;

[0123] Based on the predicted value of the load demand, the voltage, and the power factor, a loss reduction objective function F is constructed;

[0124] According to the preset safe voltage range and safe current range of the transformer, a loss reduction optimization condition set U is constructed;

[0125] A convex optimization model is constructed as a loss reduction optimization model by taking the minimization of the loss reduction objective function F as the optimization objective and taking the loss reduction optimization condition set U as the constraint condition set;

[0126] Since the energy consumption of the transformer under different loads and operating conditions is mainly reflected in copper loss and iron loss;

[0127] Therefore, preferably, the formula of the loss reduction objective function F can be:

[0128]

[0129] wherein, I is the current of the transformer, the calculation method of I is power divided by voltage, that is, the calculation formula of I is: I 2 x R is the copper loss;

[0130] wherein, represents the iron loss;

[0131] Further, the manner of constructing the loss reduction optimization condition set U is:

[0132] Constructing the voltage constraint: Vmin≤V≤Vmax, wherein Vmin and Vmax are the pre-set minimum voltage and maximum voltage, respectively;

[0133] Constructing the current constraint: Imin≤I≤Imax, wherein Imin and Imax are the pre-set minimum current and maximum current, respectively;

[0134] The voltage constraint and the current constraint together constitute the loss reduction optimization condition set U;

[0135] Further, the manner of adjusting the transformer operation strategy based on the loss reduction optimization model using an optimization algorithm is:

[0136] Using a nonlinear programming solution algorithm or a heuristic algorithm to solve the loss reduction optimization model, obtaining a solved voltage value, and the transformer operation strategy is to adjust the voltage of the transformer to the solved voltage value.

[0137] Embodiment 2

[0138] As shown in Figure 2 , a transformer loss reduction intelligent control system based on a power generation end includes a sample data collection module, a residual data generation module, an LSTM model training module, a load prediction module, and a strategy adjustment module; wherein each module is connected through electrical connection;

[0139] The sample data collection module pre-collects a transformer operation data set and a historical load data set, and sends the historical load data set to the residual data generation module and the transformer operation data set to the LSTM model training module;

[0140] The residual data generation module constructs an ARIMA model based on the transformer operation data set and the historical load data set, generates a residual data set using the ARIMA model, and sends the residual data set to the LSTM model training module and the trained ARIMA model to the load prediction module;

[0141] The LSTM model training module constructs and trains an LSTM model for predicting future load demand based on the residual data set and the transformer operation data set, and sends the trained LSTM model to the load prediction module.

[0142] The load prediction module collects real-time operation data in the actual use process of the transformer in real time, predicts future load demand corresponding to the real-time operation data using an ARIMA model and an LSTM model in combination, and sends the future load demand to the strategy adjustment module.

[0143] The strategy adjustment module constructs a consumption reduction optimization model based on the future load demand and the real-time operation data, and adjusts the transformer operation strategy using an optimization algorithm based on the consumption reduction optimization model.

[0144] Embodiment 3

[0145] According to another aspect of the present application, an electronic device is also provided. The electronic device can include one or more processors and one or more memories. The memory has computer readable code stored therein, which, when executed by the one or more processors, can perform the transformer consumption reduction intelligent control method based on a power generation end as described above.

[0146] The electronic device can include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output components, a hard disk, etc. The storage device in the electronic device, such as ROM or hard disk, can store the transformer consumption reduction intelligent control method based on a power generation end provided by the present application.

[0147] Further, the electronic device can also include a user interface. Of course, this architecture is only exemplary, and when implementing different devices, one or more components in the electronic device can be omitted according to actual needs.

[0148] Embodiment 4

[0149] According to an embodiment of the present application, a computer readable storage medium is provided. The computer readable storage medium has computer readable instructions stored thereon. When the computer readable instructions are executed by a processor, the transformer consumption reduction intelligent control method based on a power generation end according to the embodiment of the present application described with reference to the above figures can be executed. The computer readable storage medium includes but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and cache memory, etc. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0150] In addition, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions executable by a processor to perform instructions corresponding to the method steps provided by the present application, which perform the above-mentioned functions defined in the method of the present application when the computer program is executed by a central processing unit (CPU).

[0151] The method and device, apparatus of the present application can be implemented in many ways. For example, the method and device, apparatus of the present application can be implemented by software, hardware, firmware, or any combination of software, hardware, firmware. The above-mentioned order of steps for the method is only for illustration, and the steps of the method of the present application are not limited to the above specifically described order, unless otherwise specifically described. In addition, in some embodiments, the present application can also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present application. Thus, the present application also covers the recording medium storing the program for executing the method according to the present application.

[0152] In addition, the part of the above technical solutions provided in the embodiments of the present application that is consistent with the implementation principle of the corresponding technical solutions in the prior art is not described in detail to avoid excessive repetition.

[0153] The specific embodiments described above further illustrate the objects, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0154] The above-mentioned preset parameters or preset thresholds are set by a person skilled in the art according to actual conditions or obtained by a large amount of data simulation.

[0155] The above embodiments are only used to illustrate the technical method of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A power generation end-based transformer loss reduction intelligent control method, characterized in that, The method comprises the following steps: Step 1: collecting a transformer operation data set and a historical load data set in advance; Step 2: constructing an ARIMA model based on the historical load data set, and generating a residual data set using the ARIMA model; Step 3: constructing and training an LSTM model for predicting future load demand based on the residual data set and the transformer operation data set; Step 4: collecting real-time operation data in the actual use process of the transformer in real time, and predicting the future load demand corresponding to the real-time operation data using the combination of the ARIMA model and the LSTM model; Step 5: constructing a loss reduction optimization model based on the future load demand and the real-time operation data; marking the predicted value of the load demand as Pr, marking the voltage of the transformer as V, marking the power factor of the transformer as pf, and marking the equivalent resistance of the core as Rc; marking the winding resistance of the transformer as R; constructing a loss reduction objective function F based on the predicted value of the load demand, the voltage, and the power factor; constructing a loss reduction optimization condition set U according to a preset safe voltage range and a safe current range of the transformer; constructing a convex optimization model as the loss reduction optimization model by taking the minimization of the loss reduction objective function F as an optimization objective and taking the loss reduction optimization condition set U as a constraint condition set; the formula of the loss reduction objective function F is: ； Wherein, I is the current of the transformer, the calculation method of I is power divided by voltage, that is, the calculation formula of I is: , is the copper loss; wherein, represents the iron loss; the manner of constructing the loss reduction optimization condition set U is: constructing a voltage constraint: Vmin≤V≤Vmax, wherein Vmin and Vmax are respectively a preset minimum voltage and a preset maximum voltage; constructing a current constraint: Imin≤I≤Imax, wherein Imin and Imax are respectively a preset minimum current and a preset maximum current; the voltage constraint and the current constraint jointly constitute the loss reduction optimization condition set U; Step 6: solving the loss reduction optimization model using a nonlinear programming solving algorithm or a heuristic algorithm based on the loss reduction optimization model, obtaining a solved voltage value, and the transformer operation strategy is adjusting the voltage of the transformer to the solved voltage value, and adjusting the transformer operation strategy using an optimization algorithm.

2. The transformer loss reduction intelligent control method based on power generation end according to claim 1, characterized in that, The manner of collecting the transformer operation data set and the historical load data set in advance is: installing a sensor network for the transformer, and monitoring and recording parameter values of each sensor in real time, and composing the transformer operation data set by reading the parameter values of each sensor; collecting the historical load data set by reading a historical database in the operation process of the transformer and a smart meter.

3. The transformer loss reduction intelligent control method based on power generation end according to claim 2, characterized in that, The manner of constructing an ARIMA model based on the historical load data set, and generating a residual data set using the ARIMA model is: For each load parameter in the historical load data set, generating the residual data set comprises the following steps: Step 21: arranging the parameter values of the load parameter in time sequence, arranging into a time series format, and obtaining a load parameter time series corresponding to the load parameter; Step 22: Analyze whether each load parameter time series has stable mean and variance by using statistical methods. If not, apply difference method to make the load parameter time series stationary by calculating the difference between adjacent data points to eliminate trends or seasonality. Step 23: Plot the autocorrelation function and partial autocorrelation function for the load parameter time series to identify the internal patterns of the time series data. Step 24: Select the parameter combination (p, d, q) of the ARIMA model according to the autocorrelation function and partial autocorrelation function; where p is the order of the autoregressive part, indicating the number of lag periods in the model, d is the difference order, indicating the difference order required for stationarization, and q is the order of the moving average part, indicating the number of lagged prediction error terms. Step 25: Use the selected (p, d, q) parameters to construct the ARIMA model and perform fitting; verify the accuracy of the model by testing the residuals of the model, and use statistical testing methods to confirm the randomness of the residuals. Step 26: In the ARIMA model fitting process, each unit of time in the load parameter time series has a predicted value; the residual is the actual parameter value minus the predicted value given by the ARIMA model. Step 27: Collect the residuals of each unit of time to form a residual data set.

4. The transformer loss reduction intelligent control method based on the power generation end according to claim 3, characterized in that, The construction process of the LSTM model includes: Standardize all parameters in the transformer operating data set and the residual data set to ensure that each parameter is on the same scale; Pre-set a fixed time window to predict the load demand of the next time step; The LSTM model takes the parameter time series composed of each item of transformer operating data and the residual time series composed of the residual values of each load parameter within a fixed time window as an input sample; The LSTM model takes the load demand in the next preset time step as the output, and the actual load demand in the next preset time step in the historical load data set as the prediction target; The structure of the LSTM model includes an input layer, an LSTM layer, a fully connected layer, and an output layer; The input layer is used to receive multi-dimensional time series data, i.e., residuals and transformer operating data; The LSTM layer includes one or more LSTM layer stacks to capture long-term dependencies in time series; the fully connected layer connects the LSTM layer output to the final output node to produce predicted values; the output layer outputs a single predicted value, i.e., the load demand of the next time step; The training process of the LSTM model includes: Divide all input samples into training set and test set; Use mean squared error as the loss function of the LSTM model to measure the difference between predicted values and actual values; Use Adam optimizer as the optimizer of the LSTM model; Train the LSTM model on the training set and adjust the network weights to minimize the loss function; At the end of each iteration, evaluate the performance of the LSTM model on the validation set and adjust the model hyperparameters to prevent overfitting.

5. The power generation end-based transformer loss reduction intelligent control method according to claim 4, characterized in that, The method for predicting the future load demand corresponding to the real-time operation data by using the combination of the ARIMA model and the LSTM model is: The predicted value of each load parameter in each unit of time is obtained by using the ARIMA model to predict each load parameter in each unit of time; The difference between the predicted value of the load parameter in each unit of time and the actual parameter value is calculated to form a residual set; The residual set and the parameter values of each real-time operation parameter form a new input sample, which is input into the LSTM model to obtain the predicted value of the load demand output by the LSTM model.

6. A power generation terminal-based transformer loss reduction intelligent control system for implementing the power generation terminal-based transformer loss reduction intelligent control method of any one of claims 1-5, characterized in that, The method comprises a sample data collection module, a residual data generation module, an LSTM model training module, a load prediction module, and a strategy adjustment module, wherein the modules are connected in an electrical manner; The sample data collection module collects a transformer operation data set and a historical load data set in advance, and sends the historical load data set to the residual data generation module and the transformer operation data set to the LSTM model training module; The residual data generation module constructs an ARIMA model based on the transformer operation data set and the historical load data set, generates a residual data set using the ARIMA model, sends the residual data set to the LSTM model training module, and sends the trained ARIMA model to the load prediction module; The LSTM model training module constructs and trains an LSTM model for predicting future load demand based on the residual data set and the transformer operation data set, and sends the trained LSTM model to the load prediction module; The load prediction module collects real-time operation data in the actual use of the transformer in real time, predicts the future load demand corresponding to the real-time operation data by using the combination of the ARIMA model and the LSTM model, and sends the future load demand to the strategy adjustment module; The strategy adjustment module constructs a consumption reduction optimization model based on the future load demand and the real-time operation data, and adjusts the transformer operation strategy using an optimization algorithm based on the consumption reduction optimization model.

7. An electronic device, comprising: It comprises: a processor and a memory, wherein: the memory stores a computer program that can be called by the processor; the processor executes the transformer consumption reduction intelligent control method based on the power generation end in the background by calling the computer program stored in the memory.

8. A computer-readable storage medium, characterized in that, It has a writable computer program stored thereon; When the computer program runs on the computer device, the computer device executes the transformer consumption reduction intelligent control method based on the power generation end in the background.

Images