Photovoltaic power generation electricity stealing detection method and verification method based on hybrid deep learning model

Abstract

The invention discloses a photovoltaic power generation electricity larceny detection method and verification method based on a hybrid deep learning model, relates to the technical field of electricity larceny detection, and solves the technical problems that the photovoltaic power generation electricity larceny detection method is difficult to capture complex time dependence and is insufficient in multi-source data integration. The electricity larceny detection method comprises the following steps: using a sequence group # imgabs0 # to represent multi-source data of a producer and a consumer on the # imgabs1 # day within a time window T = 24; the method comprises the following steps: processing input data through a multi-scale convolutional neural network (CNN) to obtain a group of CNN features # imgabs2; modeling the CNN feature # imgabs3 # through a long short-term memory (LSTM) network, and outputting a hidden state # imgabs4 #; the hidden state # imgabs5 # is processed through a Transform encoder, and a Transform calculation result # imgabs6 # is obtained; temperature data embedding is carried out through a temperature embedding function # imgabs7 #; and the photovoltaic power generation electricity stealing detection probability is obtained. According to the invention, complex time dependence is captured, multi-source data including temperature statistical data are integrated, and the reliability and precision of electricity stealing detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity theft detection, and in particular to a photovoltaic power generation electricity theft detection method and verification method based on a hybrid deep learning model. Background Art

[0002] With the widespread deployment of smart grids, modern power systems are increasingly vulnerable to evolving electricity theft. As renewable energy continues to grow, renewable energy users, as prosumers, may manipulate power generation data and report higher power generation to power companies. This behavior threatens the stability and security of the power system, leading to an imbalance in power supply, local blackouts, and even large-scale power grid failures.

[0003] Electricity Theft Detection (ETD) has made significant progress in the consumer sector. However, as electricity theft increasingly shifts from the consumer sector to the power generation sector, especially among distributed photovoltaic users, more advanced electricity theft detection methods are needed. Typically, data related to normal electricity users dominate, while data related to fake electricity users is rare. This imbalance usually causes the electricity theft detection model to be biased towards the category where the majority of data is located during training, resulting in a decrease in the ability to accurately identify theft behavior. Therefore, in the prior art, distributed photovoltaic electricity theft sample data is enhanced by adversarial networks, and a data-driven system based on machine learning is developed to capture the intrinsic characteristics of benign data, and then detect deviations from this baseline for unsupervised anomaly detection. At the same time, a deep learning-based network attack detection model using a single data source is also proposed. However, information from a single data source may not be sufficient to fully capture the complexity and diversity of electricity theft. In order to improve the performance of the detection system, the existing technology also uses the integration of multiple data sources to develop more robust detection models. For example, a set of network attack functions is proposed to simulate the way fake consumers steal electricity by manipulating renewable energy smart meters, and a deep learning-based electricity theft detection system is developed using various data sources to detect electricity theft network attacks on solar panels.

[0004] As the amount of data increases and theft becomes more complex, traditional photovoltaic power theft detection methods have the following problems: (1) Limited time dependency capture. Traditional machine learning and deep learning models have difficulty capturing complex time dependencies, resulting in inaccurate detection of complex, time-varying power theft activities, thereby ignoring the temporal patterns necessary to identify energy theft. (2) Insufficient integration of multi-source data. Traditional methods often fail to fully integrate multi-source data and external factors, which limits their ability to fully utilize environmental influences (such as temperature changes), resulting in reduced detection reliability.

[0005] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art: The photovoltaic power generation electricity theft detection method is difficult to capture the complex time dependency, and the multi-source data integration is insufficient, resulting in low reliability of electricity theft detection and difficulty in accurately identifying electricity theft in photovoltaic power generation. Summary of the invention

[0006] The purpose of the present invention is to provide a photovoltaic power generation theft detection method and verification method based on a hybrid deep learning model to solve the technical problems that the photovoltaic power generation theft detection method in the prior art is difficult to capture complex time dependencies, and the multi-source data integration is insufficient, resulting in low reliability of theft detection and difficulty in accurately identifying theft in photovoltaic power generation. The many technical effects that can be produced by the preferred technical solution among the many technical solutions provided by the present invention are described in detail below.

[0007] To achieve the above object, the present invention provides the following technical solutions: The present invention provides a photovoltaic power generation theft detection method based on a hybrid deep learning model, comprising the following steps: S100: using a sequence group Indicates that the prosumer Days in time window The multi-source data in are used as input variables of the hybrid deep learning model, where represents the average electricity load of N prosumers, represents the actual load of the prosumer, represents the photovoltaic electricity generation reported by prosumers, represents the average power generation pattern of N prosumers; S200: the input data is processed by a multi-scale convolutional neural network (CNN) to obtain the short-term time dependency of the univariate sequence and the multivariate sequence, and a set of CNN features is obtained. ; S300: CNN features through long short-term memory network LSTM network Modeling and outputting hidden states ; S400: hidden state Processed through the Transformer encoder to obtain the Transformer calculation result ; S500: Using fully connected neural network FCNN, through temperature embedding function Embed temperature data; S600: Based on Transformer calculation results Embedded function with temperature , and the detection probability of photovoltaic power theft is obtained.

[0008] Preferably, in step S200, the sequence group All four sequences in Kernel, The convolution filter of the kernel processes the CNN features The expression is: , , Respectively represent the kernel size , The convolution operation.

[0009] Preferably, the The kernel is used to capture local short-term dependencies and patterns. Kernels are used to detect broader trends and correlations between consecutive data points.

[0010] Preferably, in step S300, the LSTM network models the time dependency in the time series data, and the hidden state The expression is: in, Represents the parameters of the LSTM network.

[0011] Preferably, in step S400, Transformer calculates the result The expression is: ,in, Represents the parameters of the Transformer network.

[0012] Preferably, in step S500, embed temperature data, the temperature embedding function of the temperature data The expression is: in, For the 5 statistical temperature values ​​for the day, are the parameters of the fully connected neural network FCNN used to embed temperature data.

[0013] Preferably, in step S600, the temperature is embedded in the function Computation results with Transformer Concatenate to form a combined feature vector , the combined feature vector Input the fully connected neural network FCNN and output the probability of photovoltaic power theft detection ,in, represents the detection model, Indicates that the detection model detects electricity theft. Parameters of the fully connected neural network FCNN that outputs the final result.

[0014] Preferably, a cross entropy loss function is used to train the hybrid deep learning model. The cross entropy loss function The expression is: ,in = , The ground truth label indicating the presence or absence of PV power theft.

[0015] A verification method for a photovoltaic power generation electricity theft detection method based on a hybrid deep learning model, the verification method is used to verify any of the above photovoltaic power generation electricity theft detection methods based on a hybrid deep learning model, and comprises the following steps: S10: using a synthetic data set of normal residential electricity producers and consumers as a normal data set; S20: applying a network attack function to part of the normal data set to simulate malicious sample data of electricity theft, and forming a verification data set together with the remaining normal data; S30: based on the verification data set, using accuracy, F1 score, and area under the curve verification indicators to verify and compare the hybrid deep learning model with a machine learning model and a deep learning model; S40: repeating the verification in step S30 5 times, calculating the average results of the verification indicators corresponding to the hybrid deep learning model, the machine learning model, and the deep learning model, and obtaining the verification result of the hybrid deep learning model.

[0016] Preferably, the network attack function includes a first network attack function and a second network attack function, wherein the first network attack function and the second network attack function are two different types of network attack functions, wherein the first network attack function increases the daily reported power generation by a fixed percentage, and the second network attack function randomly increases the power generation reported each time.

[0017] Implementing one of the above technical solutions of the present invention has the following advantages or beneficial effects: The present invention integrates CNN, LSTM and Transformer to effectively detect electricity theft in photovoltaic power generation systems. The method captures complex time dependencies and integrates multi-source data including temperature statistics, thereby improving the reliability and accuracy of electricity theft detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. It is obvious that the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 It is a schematic diagram of a photovoltaic power generation theft detection method based on a hybrid deep learning model according to Embodiment 1 of the present invention; Figure 2 This is a flow chart of a photovoltaic power generation theft detection method based on a hybrid deep learning model in Embodiment 1 of the present invention; Figure 3 It is a flow chart of a verification method of a photovoltaic power generation theft detection method based on a hybrid deep learning model according to a second embodiment of the present invention; Figure 4 This is a confusion matrix based on the comparison between the hybrid deep learning model and the existing model in embodiment 2 of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the present invention clearer, the various exemplary embodiments to be described below will refer to the corresponding drawings, which constitute a part of the exemplary embodiments, wherein various exemplary embodiments that may be used to implement the present invention are described. Unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation methods described in the following exemplary embodiments do not represent all implementation methods consistent with the present disclosure. It should be understood that they are only examples of processes, methods, devices, etc. that are consistent with some aspects of the present disclosure as detailed in the attached claims, and other embodiments may also be used, or the embodiments listed herein may be modified in structure and function without departing from the scope and essence of the present invention.

[0020] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", etc. indicate the orientation or positional relationship based on the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the elements referred to must have a specific orientation, be constructed and operated in a specific orientation. The terms "first", "second", etc. are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. The term "multiple" means two or more. The terms "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, an integral connection, a mechanical connection, an electrical connection, a communication connection, a direct connection, an indirect connection through an intermediate medium, and can be the internal connection of two elements or the interaction relationship between two elements. The term "and / or" includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0021] In order to illustrate the technical solution of the present invention, a specific embodiment is used below for description, and only the parts related to the embodiment of the present invention are shown.

[0022] Embodiment 1: like Figure 1 , Figure 2 As shown, the present invention provides a photovoltaic power generation theft detection method based on a hybrid deep learning model, comprising the following steps: S100: using a sequence group Indicates that the prosumer day at time t in the time window The multi-source data in are used as input variables of the hybrid deep learning model, where represents the average electricity load of N prosumers, represents the actual load of the prosumer, represents the photovoltaic electricity generation reported by prosumers, Represents the average power generation pattern of N prosumers. S200: Process the input data through multi-scale convolutional neural networks (CNNs) to obtain the short-term time dependencies of univariate and multivariate sequences, and obtain a set of CNN features S300: CNN features are trained using the LSTM (Long Short-Term Memory) network. Modeling, Long Short-Term Memory Network LSTM network models the time dependency in time series data and outputs hidden state ; S400: In order to enhance the model's ability to model the relationship between multiple data sources within 24 hours, the hidden state of LSTM Further processing through the Transformer encoder to obtain the Transformer calculation result S500: Using the fully connected neural network FCNN (Fully Connected Neural Network), through the temperature embedding function Embed temperature data. The temperature data may include the daily maximum, minimum, median, standard deviation value and corresponding season. S600: Transformer-based calculation results Embedded function with temperature , and obtain the probability of photovoltaic power theft detection. Tiandi The actual load and power generation of the prosumer at the moment are recorded as and , prosumers report their electricity generation to the power company as In general, the electricity generated by prosumers should be consistent with the actual electricity generated, i.e. However, in the case of power theft, prosumers may report inflated power generation values ​​to illegally gain additional profits. This behavior can be represented as: ,in Indicates the inflated electricity generation reported by prosumers. , indicating that the power generation reported by the prosumer is higher than the actual value in order to obtain excess profits. This false report leads to an overestimation of the total power generation received by the power company. , accurately identifying instances of false generation reporting by prosumers, especially detecting In order to prevent the situation caused by the original A threshold is set to tolerate small differences due to prediction errors or other physical factors that cause inaccurate readings and misjudgments. The specific size of the threshold is set as needed. , the model output is defined as: ,in Indicates that power theft has been detected. Indicates that no electricity theft is detected. Assuming that the power company has A prosumer, Indicates If there are prosumers, the smart meter records The cumulative power generation report mode at time t of the day is: , while the actual power load and power generation mode are: , The present invention integrates CNN, LSTM and Transformer to effectively detect electricity theft in photovoltaic power generation systems. This method captures complex time dependencies and seamlessly integrates multi-source data including temperature statistics, thereby improving the reliability and accuracy of electricity theft detection.

[0023] As an optional implementation, in step S200, the sequence group All four sequences in Kernel, The kernel's convolution filter is used to process the CNN feature The expression is: , , Respectively represent the kernel size , The convolution operation. The kernel is used to capture local short-term dependencies and patterns, such as sudden changes in photovoltaic power generation, which may indicate abnormal behavior. This kernel effectively looks at a 4-hour segment at a time, identifying fine-grained temporal patterns. Kernels are used to detect broader trends and correlations between consecutive data points, which is critical for capturing anomalies that are gradual and systematic in the data, such as small changes that appear continuous and regular over multiple points in time that can gradually accumulate and evolve into significant anomalies.

[0024] As an optional implementation, in step S300, the LSTM network models the time dependency in the time series data, and the hidden state The expression is: in, Represents the parameters of the LSTM network.

[0025] As an optional implementation, in step S400, the Transformer provides a global context that enables the hybrid deep model to detect subtle and complex patterns that indicate photovoltaic power theft. Transformer calculation results The expression is: ,in, Represents the parameters of the Transformer network.

[0026] As an optional implementation, in step S500, embed Temperature data, temperature embedding function of temperature data The expression is: in, For the 5 statistical temperature values ​​for the day, are the parameters of the fully connected neural network FCNN used to embed temperature data.

[0027] As an optional implementation, in step S600, the temperature is embedded in the function Computation results with Transformer Concatenate to form a combined feature vector , the combined feature vector Input the fully connected neural network FCNN and output the probability of photovoltaic power theft detection ,in, represents the detection model, Indicates that the detection model detects electricity theft. Parameters of the fully connected neural network FCNN that outputs the final result.

[0028] As an optional implementation, a cross entropy loss function is used to train the hybrid deep learning model. The cross entropy loss function The expression is: ,in = , represents the detection model, The ground truth label indicates the presence or absence of photovoltaic power theft. During the training process, the hybrid deep learning model minimizes the cross entropy loss function , optimizing the parameters of CNN, LSTM, Transformer, and temperature embedding functions, this comprehensive training process enables the hybrid deep learning model to effectively detect PV power theft by capturing the temporal patterns in multi-source time series data and the impact of daily temperature variations.

[0029] The embodiment is only a specific example and does not represent only one implementation mode of the present invention.

[0030] Embodiment 2: A verification method for photovoltaic power generation theft detection method based on a hybrid deep learning model, the verification method is used to verify the hybrid deep learning model in Example 1 to verify whether the model can better detect photovoltaic power theft behavior than the existing machine learning model and deep learning model, such as Figure 3As shown, the following steps are included. S10: Use a synthetic dataset of normal residential electricity prosumers as a normal dataset. For example, the synthetic dataset of Danish residential electricity prosumers (data source: https: / / doi.org / 10.1038 / s41597-023-02271-3, R. Yuan, SAPourmousavi, WL Soong, AJ Black, JA Liisberg, and J. Lemos-Vinasco, "A synthetic dataset of danish residential electricity prosumers," ScientificData, vol. 10, no. 1, p. 371, 2023.) is used. This dataset is a comprehensive dataset containing electricity data from residential prosumers. Specifically, the dataset includes photovoltaic power generation data, which details the hourly power output of photovoltaic systems installed by individual residential prosumers, which is then fed into the grid. These data are annotated with date type, season, and daily temperature, and include a total of 50,000 days of solar power generation records. Therefore, all samples in this dataset are normal datasets without electricity theft. S20: Since the normal data set lacks public theft data, a network attack function can be used to simulate malicious sample data of electricity theft. The network attack function is applied to part of the normal data set to simulate malicious sample data of electricity theft, and together with the remaining normal data, a verification data set is formed, so that the verification data set contains both normal data and electricity theft data. S30: Based on the verification data set, the accuracy (Accuracy), F1 score (F1-Score), and area under the curve (Area Under the Curve, AUC) verification indicators are used to verify and compare the hybrid deep learning model with the machine learning model and the deep learning model. The three verification indicators can be used to comprehensively measure the performance of the hybrid model of this embodiment. For the machine learning model, classic algorithms such as Support Vector Classifier (SVC), Random Forest (RF) and Shape Dynamic Time Warping (Shape DTW) are used; at the same time, the deep learning model includes CNN, MACNN (Multi-scale AttentionCNN), LSTM, Transformer. By comparing the performance of these two types of models with the hybrid model proposed in the present invention, the effectiveness of various models can be comprehensively evaluated.S40: Repeat the verification in step S30 for 5 times (of course, other times can be selected as needed), calculate the average results of the verification indicators corresponding to the hybrid deep learning model, the machine learning model, and the deep learning model, and obtain the verification results of the hybrid deep learning model. The detection performance in Table 1 provides a comprehensive comparison of different methods. The model proposed in this patent outperforms all other models in most evaluation indicators, achieving the highest accuracy and F1 score, as well as excellent AUC performance. This shows that the method not only correctly identifies a high percentage of instances, but also effectively balances precision and recall.

[0031] Table 1 Comparison of verification results between the model of this patent and the existing model Among the deep learning models, MACNN has the highest AUC of all models, which is 95.77%, indicating that MACNN is particularly effective in distinguishing positive and negative classes at different decision thresholds, although it lags slightly behind the method proposed in the present invention in terms of accuracy and F1 score. CNN, LSTM, and Transformer also showed strong performance, among which LSTM achieved 93.06% accuracy and 92.93% F1-score. On the other hand, compared with deep learning models, traditional machine learning models (such as SVC and RF) showed lower performance, indicating potential weaknesses in processing complex data patterns. Another traditional method, Shape DTW, performed generally, better than RF and SVC, but still lagged behind deep learning models. Therefore, the model of this patent has obvious advantages over traditional machine learning models and deep learning-based baseline models, especially in terms of accuracy and F1 score. Figure 4 As shown, the method proposed in this patent is significantly superior to the methods of other models, especially in terms of the ability to accurately detect true positives and true negatives, with the lowest incidence of false positives and false negatives. The method of the present invention shows strong performance in detecting benign samples, but its accuracy is reduced when detecting malicious samples. In particular, the random forest model has difficulty predicting malicious samples, which may be an inherent limitation of traditional machine learning techniques. These limitations may include the inability of the model to capture complex patterns and relationships in the data, which are crucial for accurately identifying malicious samples. In summary, the hybrid deep learning model proposed in the present invention integrates multi-scale CNN, LSTM and Transformer to effectively detect energy theft in photovoltaic power generation systems. The method excels in capturing complex time dependencies and seamlessly integrates multi-source data, including temperature statistics, to improve detection accuracy. The model of the present invention was verified by extensive simulation using real-world data, demonstrating its robustness and effectiveness in maintaining the stability and fairness of smart city energy systems.

[0032] As an optional implementation, the network attack function includes a first network attack function and a second network attack function. The first network attack function and the second network attack function are two different types of network attack functions. The first network attack function increases the daily reported power generation by a fixed percentage, and the second network attack function randomly increases the power generation reported each time. In the verification data set, it is preferred that normal samples without power theft account for 50% of the total data, malicious sample data generated by the first network attack function accounts for 25% of the total data, and malicious sample data generated by the first network attack function also accounts for 25% of the total data. Assume that the daily power generation vector for ,in express Therefore, the electricity theft corresponding to the first network attack function and the second network attack function can be expressed as: , , in, Represents a randomly generated fixed ratio value in the range , for all hourly power generation in a day Increase the same proportion, A stable and consistent electricity stealing behavior was simulated. Represents a randomly generated ratio value every hour, ranging from , Indicates the amount of electricity generated per hour Increase different proportions separately , simulating a random and irregular electricity theft behavior.

[0033] The above description is only the preferred embodiment of the present invention. It is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application belong to the protection scope of the present invention.

Claims

1. A photovoltaic power generation theft detection method based on a hybrid deep learning model, characterized in that: The following steps are included: S100: using sequence group Indicates that the prosumer Days in time window The multi-source data in are used as input variables of the hybrid deep learning model, where represents the average electricity load of N prosumers, represents the actual load of the prosumer, represents the photovoltaic electricity generation reported by prosumers, represents the average power generation pattern of N prosumers; S200: the input data is processed by a multi-scale convolutional neural network (CNN) to obtain the short-term time dependency of the univariate sequence and the multivariate sequence, and a set of CNN features is obtained. ; S300: CNN features through long short-term memory network LSTM network Modeling and outputting hidden states ; S400: hidden state Processed through the Transformer encoder to obtain the Transformer calculation result ; S500: Using fully connected neural network FCNN, through temperature embedding function Embed temperature data; S600: Based on Transformer calculation results Embedded function with temperature , and the detection probability of photovoltaic power theft is obtained.

2. According to claim 1, a photovoltaic power generation theft detection method based on a hybrid deep learning model is characterized in that: In the step S200, the sequence group All four sequences in Kernel, The convolution filter of the kernel processes the CNN features The expression is: , , They represent the kernel size respectively. , The convolution operation.

3. The photovoltaic power generation theft detection method based on a hybrid deep learning model according to claim 2 is characterized in that: Said The kernel is used to capture local short-term dependencies and patterns. Kernels are used to detect broader trends and correlations between consecutive data points.

4. The photovoltaic power generation theft detection method based on a hybrid deep learning model according to claim 1 is characterized in that: In the step S300, the LSTM network models the time dependency in the time series data, and the hidden state The expression is: in, Represents the parameters of the LSTM network.

5. The photovoltaic power theft detection method based on a hybrid deep learning model according to claim 1 is characterized in that: In step S400, Transformer calculates the result The expression is: ,in, Represents the parameters of the Transformer network.

6. The photovoltaic power generation theft detection method based on a hybrid deep learning model according to claim 1 is characterized in that: In the S500 step, every day temperature data, the temperature embedding function of the temperature data The expression is: ,in, For the 5 statistical temperature values ​​for the day, are the parameters of the fully connected neural network FCNN used to embed temperature data.

7. The photovoltaic power generation theft detection method based on a hybrid deep learning model according to claim 1 is characterized in that: In step S600, the temperature is embedded in the function Computation results with Transformer Concatenate to form a combined feature vector , the combined feature vector Input the fully connected neural network FCNN and output the probability of photovoltaic power theft detection ,in, represents the detection model, Indicates that the detection model detects electricity theft. Parameters of the fully connected neural network FCNN that outputs the final result.

8. The photovoltaic power generation theft detection method based on a hybrid deep learning model according to claim 1 is characterized in that: The cross entropy loss function is used to train the hybrid deep learning model. The expression is: ,in = , The ground truth label indicating the presence or absence of PV power theft.

9. A verification method for photovoltaic power generation theft detection method based on a hybrid deep learning model, characterized in that: The verification method is used to verify a photovoltaic power generation theft detection method based on a hybrid deep learning model according to any one of claims 1 to 8, comprising the following steps: S10: Use a synthetic dataset of normal residential electricity producers and consumers as a normal dataset; S20: Apply a network attack function to part of the normal dataset to simulate malicious sample data of electricity theft, and form a verification dataset together with the remaining normal data; S30: Based on the verification dataset, use accuracy, F1 score, and area under the curve verification indicators to verify and compare the hybrid deep learning model with the machine learning model and the deep learning model; S40: Repeat the verification in step S30 5 times, calculate the average results of the verification indicators corresponding to the hybrid deep learning model, the machine learning model, and the deep learning model, and obtain the verification result of the hybrid deep learning model.

10. The verification method of photovoltaic power generation theft detection method based on hybrid deep learning model according to claim 9 is characterized in that: The network attack function includes a first network attack function and a second network attack function. The first network attack function and the second network attack function are two different types of network attack functions. The first network attack function increases the daily reported power generation by a fixed percentage, and the second network attack function randomly increases the power generation reported each time.

Images