Multi-modal power load response capability optimization method and system

Abstract

The invention discloses a multi-mode power load response capability optimization method and system, and relates to the technical field of power load response, and the method comprises the steps: carrying out the data preprocessing; carrying out load decomposition; establishing a peak valley line prediction model; and various load response capabilities are quantified and adjusted. According to the multi-mode power load response capability optimization method provided by the invention, the load of the whole province is decomposed into various typical user loads through a load decomposition method, and the component of each typical user load is subjected to proportion evaluation. Peak and valley point detection is carried out on load data, the maximum value and the minimum value of the load in an alternating current period are determined according to the peak and the valley, the self-adaption purpose is achieved by dynamically adjusting system parameters, then the load demand interval is determined, and response capability quantification is achieved. And calculating a total sequence index of the whole province by adopting a volatility index, and comparing the total sequence index with the volatility of a real load value of the whole province.

Description

Technical Field

[0001] The present invention relates to the technical field of power load response, and in particular to a method and system for optimizing multi-modal power load response capability. Background Art

[0002] As the proportion of fluctuating power sources such as wind power and photovoltaic power continues to rise, the strong randomness and volatility of wind and light resources pose severe challenges to the balance of power supply and demand in the system. The diversification of energy structure also brings increasingly complex challenges to the power system. How to optimize the power load response capability according to the characteristics and supply and demand conditions of various energy sources has become an important issue. Traditional energy usually relies on experience or fixed rules, and it is difficult to achieve a dynamic balance between multi-modal energy sources. Affected by the complexity of real-time supply and demand of the power grid, in the process of advancing the power load demand side response process, it is necessary to conduct capacity assessments under different modes to determine the load classification status in different seasons and different time periods. By optimizing the power load response capability, multi-modal energy can be used more effectively, the operating efficiency and reliability of the power system can be improved, energy waste and environmental pollution can be reduced, and contributions can be made to achieving sustainable development and energy conservation and emission reduction goals.

[0003] At present, there are many methods for optimizing the response capacity of multimodal power loads, but there are certain limitations. Among them, time series decomposition is mainly based on the different changing trends of the power system load in different cycles, and detailed descriptions are made on a monthly and weekly basis to build an evaluation model for the load time series in different time periods. The periodicity of the data is analyzed, and strategies such as load superposition and separate detection of peak and valley values ​​are not considered; the active response capability based on cascade hydropower and high-energy load power changes is to optimize the load response capability through the mathematical characterization of the regulation capability between source data and load power, and lacks a certain regulation capability evaluation. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: by classifying the province's load into large industrial electricity, commercial electricity, residential electricity, and others as demand response entities, analyzing the response characteristics, weight ratios, and periodic trends of each type of demand response entity, and forming a response capacity prediction model.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for optimizing the response capability of multi-modal power loads, comprising:

[0007] Perform data preprocessing;

[0008] Perform load decomposition;

[0009] Establish a peak-valley line prediction model;

[0010] Quantify and adjust the response capabilities of various load types.

[0011] As a preferred solution of the multi-modal power load response capability optimization method described in the present invention, wherein: the data preprocessing includes data normalization and data discretization;

[0012] The data normalization is to unify all features into the same numerical range;

[0013] The data discretization is to discretize continuous data by using equal width interval method and maximum entropy method, divide the value range or value interval of continuous attribute into small intervals, and assign a discrete symbol to these intervals;

[0014] The load decomposition includes decomposing the load of the whole province through combinatorial optimization or factor hidden Markov, dividing it into large industrial electricity, commercial electricity, residential electricity and others, analyzing the electricity consumption trend of typical load types, the proportion relationship between the loads in the whole province, and determining the flexible regulation of the load;

[0015] The establishing of the peak-valley line prediction model includes determining the maximum and minimum values ​​of the load in the AC cycle by detecting the peaks and valleys.

[0016] As a preferred solution of the multi-modal power load response capacity optimization method described in the present invention, the quantification of the response capacity of various loads and the adjustment thereof include preliminary processing of each reconstructed sequence after load decomposition by a cubic exponential smoothing algorithm, and then predicting the conditional variance sequence of each type of load by an ARIMA-GARCH-based method, and converting it into various adjustment factors, accumulating the weights on each decomposed load sequence, and summarizing it into the total sequence P of the whole province. all , the volatility index is used to calculate the total sequence index of the province and compared with the volatility of the actual value of the load in the province.

[0017] As a preferred solution of the multi-modal power load response capacity optimization method of the present invention, the preliminary processing includes using a cubic exponential smoothing algorithm to perform initial prediction on each subsequence after load decomposition and reconstruction. The cubic exponential smoothing algorithm predicts a time series containing both trend and seasonality, which is expressed as:

[0018] s i =αx i / p i-k +(1-α)(s i-1 +t i-1 )

[0019] t i =β(s i -s i-1 )+(1-β)t i-1

[0020] p i =γx i / s i +(1-γ)p i-k

[0021] Among them, k is a quarterly cycle, α, β, γ are parameters, and their values ​​are all between [0,1]. s is the smoothed observation value, t is the trend factor value, p is the seasonal factor value, and x is the observation value, that is, each subsequence of the load. The prediction formula of the cumulative triple exponential smoothing is:

[0022] x i+h =(s i +ht i ) i-k

[0023] Among them, h is the number of periods. The prediction results of each subsequence obtained after three exponential smoothing predictions are:

[0024] [other,T1,T2,T3]

[0025] Among them, T1 is large industrial electricity, T2 is commercial electricity, T3 is residential electricity, and other is others.

[0026] As a preferred solution of the multi-modal power load response capability optimization method of the present invention, the conversion into various adjustment factors includes converting the non-stationary time series y into a stationary time series {yt},

[0027] Establish an ARMA(p,q) model for {yt}, expressed as,

[0028]

[0029] Among them, c a is a constant, p and q are both non-negative integers, representing the order of the autoregressive term and the moving average term in the model, respectively. is the coefficient of the i-th autoregressive term, θ j is the coefficient of the jth moving average term ε t is the residual term, which is an independent and identically distributed random variable and obeys a normal distribution with a mean of 0;

[0030] The residual term is expressed as,

[0031]

[0032] e t ~1N(0,1)

[0033] Where t is the time index, h tis the conditional variance at time t, a one-step forecast based on information from time t-1 and before, u is the autoregressive order of the GARCH process, u>0, v is the moving average order of the GARCH process v>0, α0 is a positive constant coefficient, α i,i-1,...,v is the lagged sample coefficient ε t-i 2 The non-negative coefficient of i,i-1,...,u is the non-negative coefficient of the lagged conditional variance, e t are independent normal variables with zero mean and unit variance.

[0034] As a preferred solution of the multi-modal power load response capability optimization method described in the present invention, wherein: the total sequence of the whole province is expressed as,

[0035] P all =α1*β1*P 工 +α2*β2*P 商 +α3*β3*P 居 +α4*β4*P 其他

[0036] Among them, P all is the accumulated provincial load, P 工 It is a large industrial power load, P 商 is the commercial electricity load, P 居 is the residential electricity load, P 其他 is other load, α1, α2, α3, α4 are the converted thresholds of the peak and valley line detection model results, representing large industrial electricity, commercial electricity, residential electricity and other four categories respectively; β1, β2, β3, β4 are the conditional variance sequence results of ARIMA-GARCH, representing large industrial electricity, commercial electricity, residential electricity and other four categories respectively.

[0037] As a preferred solution of the multi-modal power load response capability optimization method of the present invention, wherein: the volatility index is expressed as,

[0038]

[0039] Here, x is a sequence.

[0040] Another object of the present invention is to provide a multimodal power load response capability optimization system, which can more effectively utilize multimodal energy by constructing a multimodal power load response capability optimization system and optimizing the power load response capability, thereby improving the operating efficiency and reliability of the power system, reducing energy waste and environmental pollution, and contributing to the realization of sustainable development and energy conservation and emission reduction goals.

[0041] To solve the above technical problems, the present invention provides the following technical solutions: a multimodal power load response capability optimization system, comprising: an acquisition module, a decomposition module, a prediction module and an adjustment module; the acquisition module is used for data preprocessing; the decomposition module is used for load decomposition; the prediction module is used to establish a peak-valley line prediction model; the adjustment module is used to quantify the response capabilities of various types of loads and adjust them.

[0042] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned multi-modal power load response capability optimization method when executing the computer program.

[0043] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the multi-modal power load response capability optimization method as described above.

[0044] Beneficial effects of the present invention: The multi-modal power load response capability optimization method provided by the present invention decomposes the load of the entire province into various types of typical user loads through a load decomposition method, and evaluates the proportion of the components of various types of typical user loads.

[0045] A peak-valley line prediction model is proposed to detect the peak and valley points of load data, determine the maximum and minimum values ​​of the load in the AC cycle by peaks and valleys, dynamically adjust system parameters to achieve the purpose of self-adaptation, and then determine the load demand range to achieve quantification of response capacity.

[0046] The variance sequence of each reconstructed sequence after load decomposition is predicted as a factor adjustment factor to obtain the conditional variance sequence of each time period of the predicted day, which is converted into various adjustment factors and weighted added to each decomposed load sequence to form a total sequence for the province. The volatility index is used to calculate the total sequence index for the province and compared with the volatility of the actual value of the provincial load. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0048] Figure 1 An overall flow chart of a multi-modal power load response capability optimization method provided by an embodiment of the present invention.

[0049] Figure 2 A peak-valley line detection diagram of a multi-modal power load response capability optimization method provided by an embodiment of the present invention.

[0050] Figure 3 A standardized process diagram of a multi-modal power load response capability optimization method provided by an embodiment of the present invention.

[0051] Figure 4 An ARIMA-GARCH result calculation diagram of a multi-modal power load response capability optimization method provided by an embodiment of the present invention.

[0052] Figure 5 A volatility result diagram of a multi-modal power load response capability optimization method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0055] Example 1

[0056] Reference Figure 1 , as an embodiment of the present invention, provides a multi-modal power load response capability optimization method, comprising:

[0057] Data preprocessing, data preprocessing technology can improve the quality of data, thus helping to improve the accuracy and performance of the mining process. Because the data determines the upper limit of the result, the model and optimization only gradually approach this upper limit. Through data preprocessing, the data sources related to the task are extracted and discovered, and the legality of the data is checked according to the binding rules in the task context.

[0058] Load decomposition: Through combinatorial optimization or factor hidden Markov model, the province's load is decomposed into large industrial electricity, commercial electricity, residential electricity and others, and the electricity consumption trend of typical load types and the proportion relationship between the loads in the province are analyzed to determine the flexible regulation of the load.

[0059] A peak-valley line prediction model is established, which is mainly divided into two parts: the provincial load peak-valley line prediction model and the typical type load peak-valley line prediction model, to carry out load change peak-valley line detection.

[0060] The response capacity of various loads is quantified and adjusted. The reconstructed sequences after load decomposition are preliminarily processed by using the cubic exponential smoothing algorithm. Then, the conditional variance sequence of each type of load is predicted by the ARIMA-GARCH method, and converted into various adjustment factors. The weights are accumulated on each decomposed load sequence and summarized into the total sequence P of the province. all , the volatility index is used to calculate the total sequence index of the province, and compared with the volatility of the actual value of the load in the province, so as to obtain the response capacity of the load data. And the response capacity quantitative curves of the typical load and the load of the whole province are obtained respectively, so as to realize the two-way adjustment of the model.

[0061] The data collected are the total load data of a province with a collection frequency of 15 minutes. These provincial loads are the total added value of loads of various electricity users including large industrial electricity, commercial electricity and residential electricity at the same time. In order to analyze the relationship between various types of electricity consumption and the total load, the load category decomposition was carried out in the second step.

[0062] Step 1: Data preprocessing:

[0063] Extract and discover task-related data sources through data preprocessing. Common data preprocessing methods include:

[0064] Data Normalization:

[0065] Normalizing the features of numerical types can unify all features into a roughly similar numerical range. When the model learning speed is the same, the normalized data converges faster than the unnormalized data.

[0066] Data discretization:

[0067] When performing data preprocessing, some feature data need to be discretized. The usual practice is to discretize continuous data using methods such as equal-width interval method and maximum entropy, divide the value range or value interval of the continuous attribute into several small intervals, and assign a discrete symbol to these intervals.

[0068] Step 2: Load decomposition:

[0069] Through combinatorial optimization or factor hidden Markov model, the province's load is decomposed into large industrial electricity, commercial electricity, residential electricity and others, and the electricity consumption trend of typical load types and the proportion relationship between the province's loads are analyzed to determine the flexible regulation of the load.

[0070] The factor hidden Markov scheme is better for decomposition.

[0071] Before the decomposition, there is a sequence of length n, which refers to the load of the entire province; after the decomposition, it becomes 4 sequences of length n, which refer to large industrial electricity consumption, commercial electricity consumption, residential electricity consumption and other four types of electricity users.

[0072] Through the API of the NILMTK official website, we know that the algorithms of the load decomposition package include combinatorial optimization and factor hidden Markov. Combinatorial optimization is also called discrete optimization. It uses mathematical methods to find the optimal arrangement, grouping, order or screening of discrete events. Factor hidden Markov (FHMM) was proposed by Ghahramani in 1997. It is a multi-chain hidden Markov model suitable for modeling dynamic process time series and has a strong classification ability of time series models. It is particularly suitable for the analysis of non-stationary and poorly reproducible sequences.

[0073] Step 3: Establish a peak and valley line detection model:

[0074] Most signals are in a time-varying state, and the signal is in a fluctuating state similar to a sine wave in the time domain. By detecting the peaks and troughs, the maximum and minimum values ​​of the load in the AC cycle can be determined, so as to dynamically adjust the system parameters to achieve the purpose of self-adaptation. Common peak and trough extraction algorithms include comparison discrimination method, first-order difference combined discrimination method, second-order difference discrimination method, etc.

[0075] In a curve, the maximum value is the peak value and the minimum value is the valley value. Peak-valley line detection is to obtain the extreme value in the curve and distinguish between the maximum value and the minimum value.

[0076] The province's load before decomposition and the four user loads after decomposition were tested with the peak-valley model to obtain five curves.

[0077] The selection method is to find the maximum (minimum) value of a certain sequence.

[0078] On the premise of decomposition, the data was tested for peak and valley lines, which can be understood as extracting the peak and valley line characteristics of the load data for subsequent use.

[0079] Step 4: Quantify and adjust the response capabilities of various loads:

[0080] Firstly, the reconstructed sequences after load decomposition are processed by using the cubic exponential smoothing algorithm, and then the conditional variance sequence of each type of load is predicted by the ARIMA-GARCH method, and converted into various adjustment factors, and the weights are accumulated to each decomposed load sequence to form the total sequence P of the province. all , the volatility index is used to calculate the total sequence index of the province and compared with the volatility of the actual value of the load in the province.

[0081] Each reconstructed sequence is an operation of the second step: 5 sequences of the province's load, large industrial electricity, commercial electricity, residential electricity and other types are obtained, which characterize the fluctuation of each type of load. The peak and valley line detection and the conditional variance sequence here are both used to formulate various adjustment factors.

[0082] Step 4.1: Preliminary forecast of load data characteristics based on triple exponential smoothing:

[0083] For each subsequence after load decomposition and reconstruction, the triple exponential smoothing algorithm is used for initial prediction. The triple exponential smoothing algorithm (Holtwinters) can predict time series containing both trends and seasonality. There are two methods of triple exponential smoothing: cumulative and multiplicative. This paper uses the multiplicative triple exponential smoothing. The following formula is the multiplicative triple exponential smoothing:

[0084] s i =αx i / p i-k +(1-α)(s i-1 +t i-1 )

[0085] t i =β(s i -s i-1 )+(1-β)t i-1

[0086] p i =γx i / s i +(1-γ)p i-k

[0087] Among them, k is a quarterly cycle, α, β, γ are parameters, and their values ​​are all between [0,1]. s is the smoothed observation value, t is the trend factor value, p is the seasonal factor value, and x is the observation value, that is, each subsequence of the load. The prediction formula of cumulative triple exponential smoothing is:

[0088] x i+h =(s i +ht i ) i-k

[0089] Among them, h is the number of periods. The prediction results of each subsequence obtained after three exponential smoothing predictions are:

[0090] [other,T1,T2,T3]

[0091] Among them, T1 is large industrial electricity, T2 is commercial electricity, T3 is residential electricity, and other is others.

[0092] Step 4.2: Adjustment factor construction:

[0093] This paper adopts the ARIMA-GARCH-based method to predict the variance fluctuation of each sequence, and uses this result as the adjustment factor of each factor to adjust the prediction result.

[0094] The ARIMA model is a typical method for describing the autocorrelation of non-stationary time series in time series analysis. The ARIMA model uses the d-order difference method to transform the non-stationary time series y into a stationary time series {yt}, and then establishes an ARMA(p,q) model for {yt}:

[0095]

[0096] In the formula, c a is a constant, p and q are both non-negative integers, representing the order of the autoregressive term and the moving average term in the model, respectively. is the coefficient of the i-th autoregressive term, θ j is the coefficient of the jth moving average term ε t is the residual term, which is an independent and identically distributed random variable and obeys a normal distribution with a mean of 0.

[0097] Since the residual sequence has heteroscedasticity, GARCH is a common method for modeling the second-order conditional moments of heteroscedastic sequences. The GARCH model assumes that the residual sequence ε t has a non-constant variance. In time series modeling, it is assumed that the residual series ε t Following the GARCH (u, v) model, the definition is as follows:

[0098]

[0099]

[0100] e t ~1N(0,1)

[0101] Where t is the time index, h t is the conditional variance at time t, a one-step forecast based on information from time t-1 and before, u is the autoregressive order of the GARCH process, u>0, v is the moving average order of the GARCH process v>0, α0 is a positive constant coefficient, α i,i-1,...,v is the lagged sample coefficient ε t-i 2 The non-negative coefficient of i,i-1,...,u is the non-negative coefficient of the lagged conditional variance, e tIt is an independent normal variable with zero mean and unit variance, such as 1N(0,1). Based on the ARIMA forecast residual results of each sequence, this paper uses the GARCH(1,1) model to model each residual sequence and predict the conditional variance sequence of each time period of the predicted day. The result can be regarded as the degree of influence of different influencing factors on load changes.

[0102] Step 4.3: Model Evaluation:

[0103] After the above steps, the conditional variance sequence of each time period of the predicted day is obtained and converted into various adjustment factors. The weights are accumulated to each decomposed load sequence and summarized into the total sequence P for the whole province. all The volatility index is used to calculate the total sequence index of the province and compared with the volatility of the actual value of the load in the province. The smaller the volatility, the more stable the load.

[0104] P all =α1*β1*P 工 +α2*β2*P 商 +α3*β3*P 居 +α4*β4*P 其他

[0105] Among them, P all is the accumulated provincial load, P 工 It is a large industrial power load, P 商 is the commercial electricity load, P 居 is the residential electricity load, P 其他 is other loads, α1, α2, α3, α4 are the converted thresholds of the peak-valley line detection model results, representing large industrial electricity, commercial electricity, residential electricity and other four categories respectively; β1, β2, β3, β4 are the conditional variance sequence results of ARIMA-GARCH, representing large industrial electricity, commercial electricity, residential electricity and other four categories respectively;

[0106] P is the adjustment factor determined by the conditional variance sequence of each type of load, which is the result of ARIMA-GARCH.

[0107] It can be understood that after each of the four load lines is given a weight, they are added up to a total provincial load Pall (predicted value), and then the volatility index is used to calculate the volatility of Pall. The provincial load data can calculate a volatility index, and the optimized Pall volatility index is smaller than the provincial load data.

[0108] α i First, the peak-valley line detection model results (extreme values ​​of the load sequence) of step 2 are standardized, and the data range is controlled between [0,1]. The positions where there are no peak-valley values ​​are filled with 0, which is expressed as,

[0109]

[0110] Among them, X min is the minimum value of the sequence, X max is the maximum value of the sequence.

[0111] α i The purpose of standardization is to control the data range between [0,1] to prevent individual values ​​from being too large or too small, causing large variance fluctuations and affecting the final calculation results. The data is then added to the Pall formula as a threshold, and the subsequent steps are to participate in the Pall formula calculation.

[0112] Volatility Indicators:

[0113]

[0114] Among them, x is a sequence, and one sequence calculates a volatility result.

[0115] X is a sequence, which refers to the calculated Pall, Pprovincial, Pindustrial, Pcommercial, Presidential, and Pother load sequences. A volatility result can be calculated for each sequence.

[0116] Example 2

[0117] According to an embodiment of the present invention, a multi-modal power load response capability optimization system is provided, comprising:

[0118] Acquisition module, decomposition module, prediction module and adjustment module;

[0119] The acquisition module is used for data preprocessing;

[0120] The decomposition module is used to perform load decomposition;

[0121] The prediction module is used to establish a peak-valley line prediction model;

[0122] The regulation module is used to quantify the response capabilities of various loads and perform regulation.

[0123] Example 3

[0124] An embodiment of the present invention is different from the first two embodiments in that:

[0125] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0126] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0127] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0128] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0129] Example 4

[0130] Reference Figure 2-Figure 5 , which is an embodiment of the present invention, provides a method for optimizing the response capability of a multi-modal power load. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0131] 1. Load decomposition

[0132] The province's load is broken down into four categories: industrial, commercial, residential and other.

[0133] like Figure 2 As shown, peak and valley line detection (get α i , taking the whole province's load as an example), peak_indexes represents the index position of the peak line, peak_y represents the value of the peak line, valley_indexes represents the index position of the valley line, and valley_indexes represents the value of the valley line.

[0134] like Figure 3 As shown, the values ​​are standardized (except for the index position of the peak and valley line, the values ​​of other indexes are set to 0), and α is obtained. i (Partial data).

[0135] like Figure 4 As shown, calculate the ARIMA-GARCH results (get β i , taking the provincial load as an example), calculate the Pall sequence, then calculate the volatility, and determine that the Pall volatility value is 0.0268, and the P provincial load volatility is 0.2086.

[0136] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for optimizing multi-modal power load response capability, characterized in that: include: Perform data preprocessing; Perform load decomposition; Establish a peak-valley line prediction model; Quantify and adjust the response capabilities of various load types.

2. The multi-modal power load response capability optimization method according to claim 1, characterized in that: The data preprocessing includes data normalization and data discretization; The data normalization is to unify all features into the same numerical range; The data discretization is to discretize continuous data by using equal width interval method and maximum entropy method, divide the value range or value interval of continuous attribute into small intervals, and assign a discrete symbol to these intervals; The load decomposition includes decomposing the load of the whole province through combinatorial optimization or factor hidden Markov, dividing it into large industrial electricity, commercial electricity, residential electricity and others, analyzing the electricity consumption trend of typical load types, the proportion relationship between the loads in the whole province, and determining the flexible regulation of the load; The establishing of the peak-valley line prediction model includes determining the maximum and minimum values ​​of the load in the AC cycle by detecting the peaks and valleys.

3. The multi-modal power load response capability optimization method according to claim 2, characterized in that: The quantification and adjustment of the response capacity of various loads includes preliminary processing of each reconstructed sequence after load decomposition using a cubic exponential smoothing algorithm, and then predicting the conditional variance sequence of each type of load using an ARIMA-GARCH-based method, and converting it into various adjustment factors, accumulating the weights on each decomposed load sequence, and summarizing it into the total sequence P of the province. all , the volatility index is used to calculate the total sequence index of the province and compared with the volatility of the actual value of the load in the province.

4. The multi-modal power load response capability optimization method according to claim 3, characterized in that: The preliminary processing includes using the triple exponential smoothing algorithm to make initial predictions for each subsequence after load decomposition and reconstruction. The triple exponential smoothing algorithm predicts time series containing both trends and seasonality, which is expressed as: s i =αx i / p i-k +(1-α)(s i-1 +t i-1 ) t i =β(s i -s i-1 )+(1-β)t i-1 p i =γx i / s i +(1-γ)p i-k Among them, k is a quarterly cycle, α, β, γ are parameters, and their values ​​are all between [0,1]. s is the smoothed observation value, t is the trend factor value, p is the seasonal factor value, and x is the observation value, that is, each subsequence of the load. The prediction formula of the cumulative triple exponential smoothing is: x i+h =(s i +ht i )p i-k Among them, h is the number of periods. The prediction results of each subsequence obtained after three exponential smoothing predictions are: [other,T1,T2,T3] Among them, T1 is large industrial electricity, T2 is commercial electricity, T3 is residential electricity, and other is others.

5. The multi-modal power load response capability optimization method according to claim 4, characterized in that: The conversion into various adjustment factors includes converting the non-stationary time series y into a stationary time series {yt}, and establishing an ARMA (p, q) model for {yt}, which is expressed as: Among them, c a is a constant, p and q are both non-negative integers, representing the order of the autoregressive term and the moving average term in the model, respectively. is the coefficient of the i-th autoregressive term, θ j is the coefficient of the jth moving average term ε t is the residual term, which is an independent and identically distributed random variable and obeys a normal distribution with a mean of 0; The residual term is expressed as, and t ~1N(0.1) Where t is the time index, h t is the conditional variance at time t, a one-step forecast based on information from time t-1 and before, u is the autoregressive order of the GARCH process, u>0, v is the moving average order of the GARCH process v>0, α0 is a positive constant coefficient, α i,i-1,…,v is the lagged sample coefficient ε t-i 2 The non-negative coefficient of i,i-1,…,u is the non-negative coefficient of the lagged conditional variance, e t are independent normal variables with zero mean and unit variance.

6. The multi-modal power load response capability optimization method according to claim 5, characterized in that: The total sequence of the province is expressed as, P all =α1*β1*P 工 +α2*β2*P 商 +α3*β3*P 居 +α4*β4*P 其他 Among them, P all is the accumulated provincial load, P 工 It is a large industrial power load, P 商 is the commercial electricity load, P 居 is the residential electricity load, P 其他 is other load, α1, α2, α3, α4 are the converted thresholds of the peak and valley line detection model results, representing large industrial electricity, commercial electricity, residential electricity and other four categories respectively; β1, β2, β3, β4 are the conditional variance sequence results of ARIMA-GARCH, representing large industrial electricity, commercial electricity, residential electricity and other four categories respectively.

7. The multi-modal power load response capability optimization method according to claim 6, characterized in that: The volatility index is expressed as, Here, x is a sequence.

8. A system using the multi-modal power load response capability optimization method according to any one of claims 1 to 7, characterized in that: include: Acquisition module, decomposition module, prediction module and adjustment module; The acquisition module is used for data preprocessing; The decomposition module is used to perform load decomposition; The prediction module is used to establish a peak-valley line prediction model; The regulation module is used to quantify the response capabilities of various loads and perform regulation.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the multi-modal power load responsiveness optimization method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multi-modal power load response capability optimization method described in any one of claims 1 to 7 are implemented.

Images