Regional power grid new energy consumption prediction and decision support system

Abstract

The invention discloses a regional power grid new energy consumption prediction and decision support system, which relates to the technical field of smart power grids and new energy, and comprises a data acquisition unit, a characteristic analysis unit, a prediction model construction unit, a decision generation unit and an interactive display unit, multi-source data such as new energy power generation, meteorological information and power grid operation are collected in real time through the data collection unit, key features are extracted through the feature analysis unit, a solid foundation is provided for prediction model construction, the prediction model constructed through the quantum fusion algorithm can accurately predict the new energy power generation capacity and the power grid load, and the prediction efficiency is improved. Meanwhile, the decision generation unit can quickly generate an optimal new energy consumption decision scheme based on a dynamic fuzzy neural network technology, effectively cope with power grid fluctuation and ensure stable operation of a power grid, and the functions jointly improve the utilization rate of new energy and the overall efficiency of the power grid.

Description

Technical Field

[0001] The present invention relates to the field of smart grid and new energy technology, and specifically to a regional power grid new energy consumption prediction and decision support system. Background Art

[0002] With the rapid development of renewable energy power generation technology, the proportion of renewable energy in regional power grids has continued to increase, bringing new challenges to the stable operation of the power grid and the effective consumption of renewable energy. However, the current renewable energy consumption prediction and decision support system has some obvious shortcomings and is difficult to fully meet actual needs.

[0003] On the one hand, traditional new energy consumption prediction systems often rely on simple mathematical models and statistical methods, with limited prediction accuracy, and it is difficult to accurately reflect the complex and changing characteristics of new energy generation and grid load. This leads to serious waste of new energy generation and affects the stability of the grid. Due to inaccurate prediction results, it is difficult for the decision-making unit to generate the optimal new energy consumption decision plan, thus limiting the effective use of new energy.

[0004] On the other hand, the intelligence and automation level of the existing system needs to be improved. The traditional system lacks self-learning and optimization functions and cannot dynamically adjust the prediction model and decision-making strategy according to real-time data and grid status. In addition, the sensor fault detection and self-repair functions of the data acquisition unit are not perfect, which easily leads to data anomalies and error accumulation, further affecting the accuracy of prediction and decision-making.

[0005] Therefore, there is an urgent need for a more advanced and intelligent new energy consumption prediction and decision support system to improve the efficiency of new energy consumption, enhance the stability of the power grid, and improve the intelligence and automation level of the system. Summary of the invention

[0006] The purpose of the present invention is to make up for the shortcomings of the prior art and provide a regional power grid new energy consumption prediction and decision support system, which can collect multi-source data such as new energy power generation, meteorological information and power grid operation in real time through the data acquisition unit, and extract key features through the characteristic analysis unit, which provides a solid foundation for the construction of the prediction model. The prediction model constructed using the quantum fusion algorithm can accurately predict the new energy power generation and power grid load, providing a reliable basis for the decision generation unit. At the same time, the decision generation unit is based on dynamic fuzzy neural network technology, which can quickly generate the optimal new energy consumption decision plan, effectively respond to power grid fluctuations, and ensure the stable operation of the power grid. These functions jointly improve the utilization rate of new energy and the overall efficiency of the power grid.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: a regional power grid new energy consumption prediction and decision support system, the system includes a data acquisition unit, a characteristic analysis unit, a prediction model construction unit, a decision generation unit and an interactive display unit;

[0008] The data acquisition unit collects the output data of the new energy power generation equipment in the area, meteorological information and multi-source data of the power grid operation parameters in real time through a variety of sensors and data interfaces, and performs preliminary data cleaning and sorting. The sensor adopts self-calibration technology, and the self-calibration formula is M conrected =M meusured +k×(TT reference ), where M corrected Represents the measured value after calibration, M measured represents the original measurement value, k is the calibration coefficient, and the determination process is as follows: a large number of tests are conducted on different types of sensors under various typical environmental conditions, and measurement data under different temperature, humidity and air pressure environmental factors are collected. The relationship model between the calibration coefficients of different sensors and environmental factors is determined through data analysis, so as to accurately calculate the calibration coefficient k according to the actual environmental parameters. T is the current ambient temperature, which is obtained by real-time measurement with a high-precision temperature sensor, and T reference The reference temperature is usually set to the temperature value in a standard laboratory environment;

[0009] The characteristic analysis unit conducts in-depth analysis on the collected new energy power generation characteristics and grid load characteristics, and extracts key characteristic parameters. The key characteristic parameter extraction method is as follows: First, the new energy power generation data and grid load data are analyzed in the time domain and frequency domain to obtain the characteristic parameters of power spectrum density and fluctuation coefficient. Then, the principal component analysis (PCA) method is used to reduce the dimension of the characteristic parameters and extract the main features. The calculation formula of PCA is: X new =X×W, where X is the original feature matrix, W is the eigenvector matrix, and W is determined by performing covariance matrix calculation and eigenvalue decomposition on a large number of sample data. The specific steps are: collect a large amount of renewable energy power generation and grid load data in different time periods and weather conditions, calculate its covariance matrix, perform eigenvalue decomposition on the covariance matrix, and select eigenvectors with larger eigenvalues ​​to form the eigenvector matrix W, thereby achieving dimensionality reduction processing of high-dimensional data, X new is the feature matrix after dimensionality reduction;

[0010] The prediction model building unit uses a quantum fusion algorithm to combine the parallelism of quantum computing with a traditional machine learning algorithm to build a prediction model for renewable energy power generation and grid load. The quantum fusion algorithm introduces the superposition state and entanglement state of quantum bits to represent data features, and performs feature fusion and optimization through quantum gate operations. The structure of the prediction model includes an input layer, a quantum fusion layer, a deep learning layer, and an output layer. The input layer receives feature parameters processed by the feature analysis unit. The quantum fusion layer uses a quantum rotation gate and a controlled NOT gate to encode and fuse data. The quantum rotation gate operation formula is: |ψ> new =R(θ)|ψ>, where |ψ> is the input quantum state, R(θ) is the quantum rotating gate matrix, and its parameter θ is dynamically adjusted according to the prediction error. The adjustment formula is: θ t+1 =θ t +α×e t , where θ t is the parameter of the current iteration, α is the adjustment step size, and its value is determined by experimentally comparing the convergence speed and prediction accuracy of the model under different step sizes. t is the current prediction error, which is calculated by the difference between the actual prediction value and the true value. The deep learning layer uses a multi-layer neural network structure to further extract features and predict, and the output layer outputs the predicted renewable energy power generation and grid load;

[0011] The decision generation unit is based on the dynamic fuzzy neural network (DFNN) technology. According to the prediction results and the real-time power grid status, the optimal decision-making plan for new energy consumption is generated. The DFNN consists of a fuzzy inference system and a neural network. Through fuzzification processing of input data, fuzzy rule reasoning and defuzzification operations, the decision control signal is output. The weight of the fuzzy rule is determined by an adaptive learning algorithm. The weight update formula is: in is the weight of the jth input variable in the ith rule at the tth iteration, β is the learning rate, and its value is determined by testing on different types of power grid data and comprehensively considering the convergence speed and decision accuracy of the model. i is the output of the i-th rule, y i is the expected output, determined by analysis of historical data and expert experience, x j is the jth input variable, and the fuzzy set and membership function of the input variable are determined as follows: the input variables include the predicted renewable energy power generation, grid load and grid transmission capacity, and the membership function adopts a Gaussian function, and its formula is: Among them, μ(x) represents the membership degree of the input variable x to the fuzzy set, c is the center value, and the center position of the data distribution is determined as the center value c by statistical analysis of a large amount of historical data. σ is the standard deviation, and its value is determined according to the degree of dispersion of the data. The specific calculation method is to calculate the variance of the data and then take the square root to get the standard deviation σ. The defuzzification method uses the centroid method, and its formula is: Among them, y is the output value after defuzzification, μ i (x) is the membership degree of the i-th fuzzy set, y i is the output value corresponding to the i-th fuzzy set, and n is the number of fuzzy sets;

[0012] The interactive display unit provides users with an intuitive visual interface to display prediction results, decision-making plans and various data analysis charts. Users can set parameters and intervene through the interface. Virtual reality (VR) and augmented reality (AR) technology are used. The calculation formula for VR / AR display is: display =f(D data ,P user ), where D display Indicates the displayed content, D data The data provided by the data acquisition unit and the prediction model building unit, P user are the user's position and viewing angle parameters, which are obtained in real time through sensors in VR / AR devices. f is the display function, which dynamically generates a visualization interface suitable for the user to watch based on the user's position and viewing angle parameters and data content.

[0013] Furthermore, the system also includes a security protection unit, which uses encryption technology and access control mechanism to encrypt and protect data transmission and storage to prevent data leakage and tampering. The encryption algorithm uses quantum encryption technology, the principle of which is to use the characteristics of quantum mechanics, including but not limited to the quantum non-cloning theorem and the uncertainty principle, to ensure the security of communication. The encryption formula is: C = E K (M), where C represents the ciphertext, E KRepresents encryption function, M represents plain text, K is encryption key, key generation uses quantum random number generator to generate high-strength random key, the working principle of quantum random number generator is based on the randomness of quantum physical phenomenon, that is, the uncertainty of quantum state measurement result, and obtains random number sequence by measuring quantum system, and generates encryption key K after processing, and the access control mechanism adopts role-based access control (RBAC), which assigns different access levels according to user roles and permissions. The specific implementation method is to establish a user role database, define different permission sets for different roles, and when a user logs into the system, his role is determined according to his identity authentication information, and then the corresponding permissions are assigned according to the role to ensure that only authorized users can access specific data and functions.

[0014] Furthermore, the system has the function of self-learning and optimization. The self-learning adopts the reinforcement learning algorithm, and its formula is: Q(s,a)=Q(s,a)+α×(r+γ×max a′ Q(s′,a′)-Q(s,a)), where Q(s,a) represents the value function of taking action a in state s, α is the learning rate, and its value is determined based on experience and experiments. The specific determination method is to adjust the value of the learning rate in different learning scenarios, observe the learning speed and convergence of the model, and select a learning rate value that can achieve good convergence effect within a reasonable time. r is the current reward, and the reward is set according to the prediction accuracy and decision effect. The reward is calculated by comprehensively determining the reward value based on the degree of proximity between the prediction result and the actual value and the effect of improving the consumption of new energy after the implementation of the decision plan. γ is the discount factor, and its value is determined according to the importance of future rewards. By evaluating the long-term decision-making effect of the model under different discount factor values, the appropriate discount factor value is selected to balance the importance of current and future rewards. s′ is the next state, and a′ is the optimal action in the next state. Through continuous learning and optimization, the system automatically adjusts the parameters of the prediction model and the decision generation unit to improve the accuracy of prediction and decision.

[0015] Furthermore, the system is integrated and works in collaboration with other energy management systems and smart grid equipment. Through standardized interfaces and protocols, the system can exchange data and conduct collaborative control with other systems to achieve comprehensive management and optimized scheduling of regional energy. The interfaces and protocols use distributed ledger technology to ensure the security, reliability and traceability of data. The calculation formula of the distributed ledger is H=hash(B), where H represents the hash value and hash is a hash function. A secure hash algorithm, including but not limited to SHA-256, is used to perform hash calculation on the data block B in the ledger to obtain a unique hash value H, which is used to verify the integrity and authenticity of the data. The data exchange and collaborative control use a message queue mechanism to achieve efficient data transmission and real-time response. The message queue uses an advanced distributed message queue system, namely Kafka, to handle a large number of concurrent data transmission requests to ensure fast and reliable data transmission between different systems.

[0016] Furthermore, the sensor in the data acquisition unit has fault detection and self-repair functions. The fault detection of the sensor is judged by monitoring the stability and rationality of the sensor output data. The specific method is to set the data stability threshold and rationality range. When the fluctuation of the sensor output data exceeds the stability threshold or exceeds the rationality range, it is determined that the sensor may be faulty. When a fault is detected, the self-repair mechanism is started. The self-repair formula is: S repuired =S faulty +d×(FS faulty ), where S repaired Represents the sensor output after repair, S fasllty represents the sensor output under fault conditions, d is the repair coefficient, and its value is determined through experiments. The experimental process is to adjust the value of the repair coefficient under simulated fault conditions, observe the degree of proximity between the repaired data and the true value, and select the repair coefficient value that can achieve a better repair effect. F is the reasonable output value predicted based on historical data and other normal sensor data. By analyzing and modeling historical data, the machine learning algorithm is used to predict the reasonable output value F of the sensor under current environmental conditions.

[0017] Furthermore, the time domain and frequency domain analysis in the characteristic analysis unit adopts a time-frequency joint analysis method, which combines short-time Fourier transform (STFT) and wavelet transform, analyzes data in the time and frequency domains at the same time, and extracts richer feature information. The calculation formula of the time-frequency joint analysis is: TF(t,f)=W(t,f)×STFT(t,f), wherein TF(t,f) represents the result of the time-frequency joint analysis, t is time, f is frequency, and W(t,f) is the wavelet transform coefficient. By selecting a suitable wavelet basis function, the data is subjected to wavelet transform to obtain the wavelet transform coefficient W(t,f), and STFT(t,f) is the short-time Fourier transform result. The data is subjected to short-time Fourier transform in a sliding window manner to obtain frequency distribution information under different time windows.

[0018] Furthermore, the deep learning layer in the prediction model construction unit adopts a deep residual network (ResNet) structure. ResNet effectively solves the gradient vanishing problem in deep neural networks by introducing residual blocks, thereby improving the training effect and prediction accuracy of the model. The calculation formula of the residual block is: y=F(x)+x, where y is the output of the residual block and F(x) is the result of a series of convolution and activation function operations. By selecting the appropriate convolution kernel size and activation function, feature extraction and nonlinear transformation are performed on the input data, and x is the input.

[0019] Furthermore, the fuzzy rule reasoning process in the decision generation unit is optimized by using fuzzy Petri net (FPN). FPN combines fuzzy reasoning with the graphical representation and dynamic analysis capabilities of Petri net, thereby improving the efficiency and accuracy of fuzzy rule reasoning. The state transition formula of FPN is: new =M current +A×I, where M new Represents the new state identification vector, M current Represents the current state identification vector, A is the state transfer matrix, and the element value of the state transfer matrix A is determined by analyzing and modeling the fuzzy rules. I is the input identification vector, and the value of the input identification vector I is determined according to the fuzzification result of the input data.

[0020] Furthermore, the VR / AR technology in the interactive display unit supports gesture recognition and voice interaction functions. Gesture recognition is implemented through a depth camera and a machine learning algorithm, and can recognize the user's gestures and perform corresponding operations. The specific implementation process is to use a depth camera to obtain the user's hand image data, extract and classify the hand image, and recognize different gestures. Voice interaction uses voice recognition and natural language processing technology. The user can interact with the system through voice commands. The voice recognition module converts the user's voice signal into text, and the natural language processing module analyzes and understands the text, determines the user's intention, and generates a corresponding system response based on the intention. The formula for gesture recognition and voice interaction is: A actiom =G(V)+S(T), where A actiom represents the response action of the system, G(V) represents the action determined according to the gesture recognition result, V is the gesture image data, S(T) represents the action determined according to the speech recognition result, and T is the speech data.

[0021] Compared with the existing technology, the regional power grid new energy consumption forecasting and decision support system has the following beneficial effects:

[0022] 1. The present invention uses a data acquisition unit to collect output data, meteorological information and power grid operation parameters of renewable energy power generation equipment in the region in real time, uses a characteristic analysis unit to conduct in-depth analysis of renewable energy power generation characteristics and power grid load characteristics, and extracts key characteristic parameters. The prediction model construction unit uses a quantum fusion algorithm to construct a prediction model of renewable energy power generation and power grid load to achieve high-precision prediction. The decision generation unit is based on dynamic fuzzy neural network (DFNN) technology to generate the optimal renewable energy consumption decision plan based on the prediction results and real-time power grid status. The optimization of this series of processes significantly improves the efficiency of renewable energy consumption, reduces the waste of renewable energy power generation, and enhances the stability and reliability of the power grid.

[0023] 2. The present invention continuously adjusts and optimizes the parameters of the prediction model and decision-making unit through the reinforcement learning algorithm to improve the accuracy of prediction and decision-making. In addition, the sensor in the data acquisition unit has fault detection and self-repair functions, and can monitor the output data of the sensor in real time. Once a fault is detected, the self-repair mechanism is activated to ensure the accuracy and reliability of the data. These functions enhance the intelligence and automation level of the system, reduce the need for manual intervention, and improve the operating efficiency and reliability of the system.

[0024] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 The figure is a flow chart of a regional power grid new energy consumption forecasting and decision support system. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] Embodiment 1:

[0029] In a specific regional power grid environment, the region has multiple distributed wind farms and solar power stations. The regional power grid management department decides to adopt the regional power grid new energy consumption prediction and decision support system of the present invention to improve the new energy consumption efficiency and power grid stability.

[0030] Data acquisition unit:

[0031] A variety of high-precision sensors were deployed, including wind speed sensors, light intensity sensors, and smart meters. For wind speed sensors, a large number of tests were conducted under different wind speed, temperature, and humidity conditions to establish a relationship model between the calibration coefficient and environmental factors. For example, tests were conducted under a series of specific environmental combinations. After analyzing the data, it was determined that when the current ambient temperature was 22°C, the humidity was 45%, and the average wind speed was 8m / s, the calibration coefficient k was 0.45, and its reference temperature was set to 20°C. A similar method was used for the light intensity sensor to determine the calibration coefficient based on different light intensity ranges and ambient temperatures. The smart meter collected the load data of the power grid in real time, and performed preliminary data sorting and outlier removal.

[0032] The data interface is connected to the monitoring system of the new energy power plant to ensure that the operating status data of the power generation equipment can be obtained in real time.

[0033] Characteristic analysis unit:

[0034] The output power data of wind farms and solar power stations and the grid load data are analyzed in the time domain and frequency domain, and the principal component analysis (PCA) method is used to reduce the dimension of the characteristic parameters. For example, a week of renewable energy power generation and grid load data is collected, the covariance matrix is ​​calculated and the eigenvalue decomposition is performed, and the first three eigenvectors with larger eigenvalues ​​are selected to form the eigenvector matrix W. It is assumed that these three eigenvectors correspond to the volatility of renewable energy power generation, the periodicity of grid load, and the impact of environmental factors on power generation. In this way, the high-dimensional data is reduced to a dimension that is easier to handle.

[0035] Prediction model building unit:

[0036] The prediction model is constructed using the quantum fusion algorithm. In the quantum fusion layer, the quantum rotating gate operation formula is |ψ> new =R(θ)|ψ>, where the initial value of parameter θ is set to 0.15. During the model training process, it is dynamically adjusted according to the prediction error. Assuming that the current prediction error is 0.25, the adjustment step α is determined to be 0.06 through previous experiments, then θ t+1 =θ t +α×e t =0.15+0.06×0.25=0.165. The deep learning layer adopts the deep residual network (ResNet) structure to extract deep features of the data through a series of convolution and activation function operations. Finally, the output layer outputs the predicted values ​​of new energy power generation and grid load in the future.

[0037] Decision making unit:

[0038] The decision-making scheme is generated based on the dynamic fuzzy neural network (DFNN) technology. The weight of the fuzzy rule is determined by the adaptive learning algorithm. The weight update formula is: Assume that the learning rate is 0.04 after testing different data sets, and the output f of the current i-th rule is i is 75, the expected output is y i is 12, and the jth input variable is x j , if the initial weight The updated weight is

[0039] The membership function of the input variable adopts a Gaussian function, and its formula is: For the input variable of renewable energy power generation, statistical analysis is performed on the data of the past month to determine the center position of the data distribution as the center value c, which is assumed to be 480MW. The standard deviation σ is calculated based on the degree of dispersion of the data. It is assumed that the standard deviation obtained through variance calculation is 45. Through fuzzy rule reasoning and defuzzification operations, decision-making control signals are output, such as adjusting the output power of renewable energy power plants and optimizing grid scheduling.

[0040] Interactive display unit:

[0041] Virtual reality (VR) and augmented reality (AR) technologies are used to provide users with an intuitive visual interface. For example, users can use VR devices to immersively view the real-time operating status of the regional power grid, the consumption of new energy sources, and the forecast results, and display the function D display =f(D data ,P user ) data The data provided by the data acquisition unit and the prediction model building unit, P user are the user's position and viewing angle parameters, which are obtained in real time through the sensors in the VR device. f is the display function, which dynamically generates a visualization interface suitable for the user to watch based on the user's position and viewing angle parameters and data content.

[0042] Embodiment 2:

[0043] In an industrial area covered by a large regional power grid, which includes various types of renewable energy power generation facilities, such as wind power, solar power and biomass power stations, the regional power grid operator hopes to improve the utilization efficiency of renewable energy and the reliability of the power grid with the help of the system of the present invention.

[0044] Data acquisition unit:

[0045] Wind direction sensors, light intensity sensors, fuel supply sensors for biomass power plants, etc. are installed. The calibration coefficients of wind direction sensors are determined by experiments under different wind direction, temperature and air pressure conditions. For example, tests are conducted under a series of specific environmental combinations. After analyzing the data, it is determined that when the current ambient temperature is 28°C and the air pressure is 101kP a , when the wind direction is northeast, the calibration coefficient k is 0.38, and its reference temperature is set to 25°C. The fuel supply sensor is used to monitor the fuel supply of the biomass power station and provide data for power generation forecasting. The smart meter collects the load data of the power grid in real time and performs data cleaning and sorting;

[0046] The data interface is connected to the control system of each power station to obtain the operating status and performance parameter data of the power generation equipment in real time.

[0047] Characteristic analysis unit:

[0048] Perform time domain and frequency domain analysis on different types of renewable energy power generation data and grid load data, and adopt an innovative time-frequency joint analysis method, combining short-time Fourier transform (STFT) and wavelet transform. For example, select a suitable wavelet basis function, perform wavelet transform on the output power data of the biomass power station to obtain the wavelet transform coefficient W(t,f), and then perform short-time Fourier transform on the data through a sliding window to obtain STFT(t,f), and finally obtain the time-frequency joint analysis result TF(t,f)=W(t,f)×STFT(t,f). In this way, more abundant feature information can be extracted for subsequent analysis;

[0049] The PCA method was used to reduce the dimension of the characteristic parameters. One month's renewable energy power generation and grid load data were collected, and the characteristic vector matrix was obtained after calculation. It was assumed that this matrix reflects the main characteristics of different types of renewable energy power generation and the changing trend of grid load.

[0050] Prediction model building unit:

[0051] The prediction model is constructed using the quantum fusion algorithm. In the quantum fusion layer, the initial value of the parameter θ of the quantum rotating gate is set to 0.18. During the model training process, it is dynamically adjusted according to the prediction error. Assuming that the current prediction error is 0.32, the adjustment step α is determined to be 0.055 through preliminary experiments, then θ t+1 =θ t +α×e t =0.18+0.055×0.32=0.1976. The deep learning layer uses the ResNet structure to perform deep feature extraction on the data. Finally, the output layer outputs the predicted values ​​of new energy power generation and grid load in the future.

[0052] Decision making unit:

[0053] Fuzzy Petri Net (FPN) is used to optimize the fuzzy rule reasoning process, determine the element value of the state transfer matrix A, and determine the value of the input identification vector I according to the fuzzification result of the input data. For example, for input variables such as renewable energy power generation, grid load, and grid transmission capacity, their membership is determined through fuzzification processing, and then the input identification vector I is determined. Assuming that the current state identification vector M current =(0.22, 0.31, 0.47), and the new state identification vector M is obtained after calculation new =M current +A×I;

[0054] The weight of the fuzzy rule is determined by an adaptive learning algorithm, and the weight update formula is: Assume that the learning rate β is determined to be 0.035 after testing different data sets, and the output f of the current i-th rule isi is 90, the expected output is y i is 85, the jth input variable x j is 14, if the initial weight The updated weight is Through fuzzy rule reasoning and defuzzification operations, decision control signals are output, such as adjusting the output ratio of different types of renewable energy power generation, optimizing the operation of grid energy storage equipment, etc.

[0055] Interactive display unit:

[0056] Through AR technology, users can view virtual data analysis and decision support information through mobile devices at the actual power grid equipment site. For example, at the solar power station site, users can see real-time light intensity information, solar power generation forecast results and corresponding decision suggestions through the AR application on their mobile phones. Gesture recognition and voice interaction functions facilitate user operation. Suppose the user makes a specific gesture, after recognition by the depth camera and machine learning algorithm, the system performs the corresponding operation, such as switching to display the consumption of different types of renewable energy power generation. The voice interaction function allows users to query specific information through voice commands, such as "What is the current operating efficiency of the biomass power station?" The system uses voice recognition and natural language processing technology to answer users' questions and display function D. display =f(D data ,P user ) data The data provided by the data acquisition unit and the prediction model building unit, P user are the user's position and viewing angle parameters, which are obtained in real time through the sensors of the mobile device. f is the display function, which dynamically generates a visualization interface suitable for the user to watch based on the user's position and viewing angle parameters and data content.

[0057] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A regional power grid new energy consumption forecasting and decision support system, characterized in that: The system includes a data acquisition unit, a characteristic analysis unit, a prediction model building unit, a decision generation unit, and an interactive display unit; The data acquisition unit collects the output data of the new energy power generation equipment in the area, meteorological information and multi-source data of the power grid operation parameters in real time through a variety of sensors and data interfaces, and performs preliminary data cleaning and sorting. The sensor adopts self-calibration technology, and the self-calibration formula is M conrected =M meusured +k×TT reference ), where M corrected Represents the measured value after calibration, M measured represents the original measurement value, k is the calibration coefficient, and the determination process is as follows: a large number of tests are conducted on different types of sensors under various typical environmental conditions, and measurement data under different temperature, humidity and air pressure environmental factors are collected. The relationship model between the calibration coefficients of different sensors and environmental factors is determined through data analysis, so as to accurately calculate the calibration coefficient k according to the actual environmental parameters. T is the current ambient temperature, which is obtained by real-time measurement with a high-precision temperature sensor, and T reference The reference temperature is usually set to the temperature value in a standard laboratory environment; The characteristic analysis unit conducts in-depth analysis on the collected new energy power generation characteristics and grid load characteristics, and extracts key characteristic parameters. The key characteristic parameter extraction method is as follows: First, the new energy power generation data and grid load data are analyzed in the time domain and frequency domain to obtain the characteristic parameters of power spectrum density and fluctuation coefficient. Then, the principal component analysis (PCA) method is used to reduce the dimension of the characteristic parameters and extract the main features. The calculation formula of PCA is: X new =X×W, where X is the original feature matrix, W is the eigenvector matrix, and W is determined by performing covariance matrix calculation and eigenvalue decomposition on a large number of sample data. The specific steps are: collect a large amount of renewable energy power generation and grid load data in different time periods and weather conditions, calculate its covariance matrix, perform eigenvalue decomposition on the covariance matrix, and select eigenvectors with larger eigenvalues ​​to form the eigenvector matrix W, thereby achieving dimensionality reduction processing of high-dimensional data, X new is the feature matrix after dimensionality reduction; The prediction model building unit uses a quantum fusion algorithm to combine the parallelism of quantum computing with a traditional machine learning algorithm to build a prediction model for renewable energy power generation and grid load. The quantum fusion algorithm introduces the superposition state and entanglement state of quantum bits to represent data features, and performs feature fusion and optimization through quantum gate operations. The structure of the prediction model includes an input layer, a quantum fusion layer, a deep learning layer, and an output layer. The input layer receives feature parameters processed by the feature analysis unit. The quantum fusion layer uses a quantum rotation gate and a controlled NOT gate to encode and fuse data. The quantum rotation gate operation formula is: |ψ> new =R(θ)|ψ>, where |ψ> is the input quantum state, R(θ) is the quantum rotating gate matrix, and its parameter θ is dynamically adjusted according to the prediction error. The adjustment formula is: θ t+1 =θ t +α×e t , where θ t is the parameter of the current iteration, α is the adjustment step size, and its value is determined by experimentally comparing the convergence speed and prediction accuracy of the model under different step sizes. t is the current prediction error, which is calculated by the difference between the actual prediction value and the true value. The deep learning layer uses a multi-layer neural network structure to further extract features and predict, and the output layer outputs the predicted renewable energy power generation and grid load; The decision generation unit is based on the dynamic fuzzy neural network (DFNN) technology. According to the prediction results and the real-time power grid status, the optimal decision-making plan for new energy consumption is generated. The DFNN consists of a fuzzy inference system and a neural network. Through fuzzification processing of input data, fuzzy rule reasoning and defuzzification operations, the decision control signal is output. The weight of the fuzzy rule is determined by an adaptive learning algorithm. The weight update formula is: in is the weight of the jth input variable in the ith rule at the tth iteration, β is the learning rate, and its value is determined by testing on different types of power grid data and comprehensively considering the convergence speed and decision accuracy of the model. i is the output of the i-th rule, y i is the expected output, determined by analysis of historical data and expert experience, x j is the jth input variable, and the fuzzy set and membership function of the input variable are determined as follows: the input variables include the predicted renewable energy power generation, grid load and grid transmission capacity, and the membership function adopts a Gaussian function, and its formula is: Among them, μ(x) represents the membership degree of the input variable x to the fuzzy set, c is the center value, and the center position of the data distribution is determined as the center value c by statistical analysis of a large amount of historical data. σ is the standard deviation, and its value is determined according to the degree of dispersion of the data. The specific calculation method is to calculate the variance of the data and then take the square root to get the standard deviation σ. The defuzzification method uses the centroid method, and its formula is: Among them, y is the output value after defuzzification, μ i (x) is the membership degree of the i-th fuzzy set, y i is the output value corresponding to the i-th fuzzy set, and n is the number of fuzzy sets; The interactive display unit provides users with an intuitive visual interface to display prediction results, decision-making plans and various data analysis charts. Users can set parameters and intervene through the interface. Virtual reality (VR) and augmented reality (AR) technology are used. The calculation formula for VR / AR display is: display =f(D data ,P user ), where D display Indicates the displayed content, D data The data provided by the data acquisition unit and the prediction model building unit, P user are the user's position and viewing angle parameters, which are obtained in real time through sensors in VR / AR devices. f is the display function, which dynamically generates a visualization interface suitable for the user to watch based on the user's position and viewing angle parameters and data content.

2. The regional power grid new energy consumption prediction and decision support system according to claim 1 is characterized in that: The system also includes a security protection unit, which uses encryption technology and access control mechanism to encrypt and protect data transmission and storage to prevent data leakage and tampering. The encryption algorithm uses quantum encryption technology, the principle of which is to use the characteristics of quantum mechanics, including but not limited to the quantum non-cloning theorem and the uncertainty principle, to ensure the security of communication. The encryption formula is: C = E K (M), where C represents the ciphertext, E K Represents encryption function, M represents plain text, K is encryption key, key generation uses quantum random number generator to generate high-strength random key, the working principle of quantum random number generator is based on the randomness of quantum physical phenomenon, that is, the uncertainty of quantum state measurement result, and obtains random number sequence by measuring quantum system, and generates encryption key K after processing, and the access control mechanism adopts role-based access control (RBAC), which assigns different access levels according to user roles and permissions. The specific implementation method is to establish a user role database, define different permission sets for different roles, and when a user logs into the system, his role is determined according to his identity authentication information, and then the corresponding permissions are assigned according to the role to ensure that only authorized users can access specific data and functions.

3. The regional power grid new energy consumption prediction and decision support system according to claim 1 is characterized in that: The system has the function of self-learning and optimization. The self-learning adopts the reinforcement learning algorithm, and its formula is: Q(s,a)=Q(s,a)+α×(r+γ×max a′ Q(s′,a′)-Q(s,a)), where Q(s,a) represents the value function of taking action a in state s, α is the learning rate, and its value is determined based on experience and experiments. The specific determination method is to adjust the value of the learning rate in different learning scenarios, observe the learning speed and convergence of the model, and select a learning rate value that can achieve good convergence effect within a reasonable time. r is the current reward, and the reward is set according to the prediction accuracy and decision effect. The reward is calculated by comprehensively determining the reward value based on the degree of proximity between the prediction result and the actual value and the effect of improving the consumption of new energy after the implementation of the decision plan. γ is the discount factor, and its value is determined according to the importance of future rewards. By evaluating the long-term decision-making effect of the model under different discount factor values, the appropriate discount factor value is selected to balance the importance of current and future rewards. s′ is the next state, and a′ is the optimal action in the next state. Through continuous learning and optimization, the system automatically adjusts the parameters of the prediction model and the decision generation unit to improve the accuracy of prediction and decision.

4. The regional power grid new energy consumption prediction and decision support system according to claim 1 is characterized in that: The system is integrated and works in collaboration with other energy management systems and smart grid equipment. Through standardized interfaces and protocols, the system can exchange data and conduct collaborative control with other systems to achieve comprehensive management and optimized scheduling of regional energy. The interfaces and protocols use distributed ledger technology to ensure data security, reliability and traceability. The calculation formula of the distributed ledger is H=hash(B), where H represents the hash value and hash is a hash function. A secure hash algorithm, including but not limited to SHA-256, is used to perform hash calculation on the data block B in the ledger to obtain a unique hash value H, which is used to verify the integrity and authenticity of the data. The message queue mechanism is used for data exchange and collaborative control to achieve efficient data transmission and real-time response. The message queue uses an advanced distributed message queue system, namely Kafka, to handle a large number of concurrent data transmission requests to ensure fast and reliable data transmission between different systems.

5. The regional power grid new energy consumption prediction and decision support system according to claim 1 is characterized in that: The sensor in the data acquisition unit has fault detection and self-repair functions. The fault detection of the sensor is judged by monitoring the stability and rationality of the sensor output data. The specific method is to set the data stability threshold and rationality range. When the fluctuation of the sensor output data exceeds the stability threshold or exceeds the rationality range, it is determined that the sensor may be faulty. When a fault is detected, the self-repair mechanism is started. The self-repair formula is: S repuired =S faulty +d×(FS faulty ), where S repaired Represents the sensor output after repair, S fasllty represents the sensor output under fault conditions, d is the repair coefficient, and its value is determined through experiments. The experimental process is to adjust the value of the repair coefficient under simulated fault conditions, observe the degree of proximity between the repaired data and the true value, and select the repair coefficient value that can achieve a better repair effect. F is the reasonable output value predicted based on historical data and other normal sensor data. By analyzing and modeling historical data, the machine learning algorithm is used to predict the reasonable output value F of the sensor under current environmental conditions.

6. The regional power grid new energy consumption prediction and decision support system according to claim 1 is characterized in that: The time domain and frequency domain analysis in the characteristic analysis unit adopts a time-frequency joint analysis method, which combines short-time Fourier transform (STFT) and wavelet transform, analyzes data in the time and frequency domains at the same time, and extracts richer feature information. The calculation formula of the time-frequency joint analysis is: TF(t,f)=W(t,f)×STFT(t,f), wherein TF(t,f) represents the result of the time-frequency joint analysis, t is time, f is frequency, and W(t,f) is the wavelet transform coefficient. By selecting a suitable wavelet basis function, the data is subjected to wavelet transform to obtain the wavelet transform coefficient W(t,f), and STFT(t,f) is the short-time Fourier transform result. The data is subjected to short-time Fourier transform in a sliding window manner to obtain frequency distribution information under different time windows.

7. The regional power grid new energy consumption prediction and decision support system according to claim 1 is characterized in that: The deep learning layer in the prediction model construction unit adopts a deep residual network (ResNet) structure. ResNet effectively solves the gradient vanishing problem in deep neural networks by introducing residual blocks, improves the training effect and prediction accuracy of the model, and the calculation formula of the residual block is: y=F(x)+x, where y is the output of the residual block and F(x) is the result of a series of convolution and activation function operations. By selecting the appropriate convolution kernel size and activation function, feature extraction and nonlinear transformation are performed on the input data, and x is the input.

8. The regional power grid new energy consumption forecasting and decision support system according to claim 1 is characterized in that: The fuzzy rule reasoning process in the decision generation unit is optimized by using fuzzy Petri net (FPN). FPN combines fuzzy reasoning with the graphical representation and dynamic analysis capabilities of Petri net, which improves the efficiency and accuracy of fuzzy rule reasoning. The state transfer formula of FPN is: new =M current +A×I, where M new Represents the new state identification vector, M current Represents the current state identification vector, A is the state transfer matrix, and the element value of the state transfer matrix A is determined by analyzing and modeling the fuzzy rules. I is the input identification vector, and the value of the input identification vector I is determined according to the fuzzification result of the input data.

9. The regional power grid new energy consumption prediction and decision support system according to claim 1, characterized in that: The VR / AR technology in the interactive display unit supports gesture recognition and voice interaction functions. Gesture recognition is implemented through a depth camera and a machine learning algorithm, and can recognize the user's gestures and perform corresponding operations. The specific implementation process is to use a depth camera to obtain the user's hand image data, extract and classify the hand image features, and recognize different gestures. Voice interaction uses voice recognition and natural language processing technology. The user can interact with the system through voice commands. The voice recognition module converts the user's voice signal into text, and the natural language processing module analyzes and understands the text, determines the user's intention, and generates a corresponding system response based on the intention. The formula for gesture recognition and voice interaction is: A actiom =G(V)+S(T), where A actiom represents the response action of the system, G(V) represents the action determined according to the gesture recognition result, V is the gesture image data, S(T) represents the action determined according to the speech recognition result, and T is the speech data.

Images