A load prediction method and system based on an incremental echo state network

Abstract

The application discloses a load prediction method and system based on an incremental echo state network, relates to the technical field of power load prediction, and comprises the following steps: defining a similarity measurement standard, and acquiring a prediction sample set based on similar time series retrieval; calculating and determining the number of neurons of the incremental echo state network according to an empirical formula, and setting network parameters according to a task; training the incremental echo state network by using the prediction sample set, and performing prediction by using the trained network. The method disclosed by the application can quickly locate similar historical data to a target load mode by quantitatively analyzing and comparing the characteristics of time series and by adopting the similar time series retrieval technology, thereby providing a more accurate sample set for a load prediction model; and the number of hidden layer neurons is adaptively adjusted according to the characteristics of load data and the requirements of a prediction task by using an empirical formula, the dynamic optimization of the network structure is realized, and the flexibility and accuracy of load prediction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power load prediction, in particular to a load prediction method and system based on incremental echo state network. BACKGROUND

[0002] Load prediction of power systems is a key technology to ensure grid stability and optimize energy distribution. With the development of social economy and the diversification of energy consumption patterns, load data presents increasingly strong nonlinearity and dynamic change characteristics. Traditional load prediction methods, such as linear regression based on statistics and time series analysis, are difficult to accurately capture the complexity and variability of load data due to their inherent linear assumptions.

[0003] To address this challenge, deep learning techniques, especially recurrent neural networks (RNN) and their variants such as long short-term memory networks (LSTM) and echo state networks (ESN), have received widespread attention for their strong ability to handle time series data. ESN, as a special type of RNN, excels in handling large-scale time series data due to its random and fixed internal nodes and ease of training.

[0004] A major problem with ESN in existing technology when dealing with incremental or real-time data is that once the network structure is determined, the number of neurons is fixed. This limits the network's ability to adapt to new data, as a fixed structure network may not effectively capture new patterns and trends when the data distribution changes. In addition, retraining the entire network to adapt to new data is not only time-consuming but also may cause excessive forgetting of historical data.

[0005] To address this problem, the present application proposes a load prediction method based on incremental echo state network. The incremental echo state network can dynamically adjust the number of neurons to adapt to changes in data. Compared with traditional ESN, the incremental echo state network can dynamically increase or decrease the number of neurons according to the characteristics of the data and the requirements of the prediction task when receiving new data. This dynamic adjustment mechanism enables the incremental echo state network to more flexibly adapt to changes in load data while retaining memory of historical data. SUMMARY

[0006] In view of the above problems, the present application is proposed.

[0007] Therefore, the technical problem solved by the present application is that the existing load prediction method has a fixed number of neurons once the network structure is determined, which makes it difficult for the network to effectively capture new patterns and trends, and how to dynamically adjust the number of neurons to more flexibly adapt to changes in load data.

[0008] To solve the above technical problems, the application provides the following technical scheme: a load prediction method based on an incremental echo state network, comprising defining a similarity measurement standard, obtaining a prediction sample set based on a similar time sequence retrieval, calculating and determining the number of neurons of the incremental echo state network according to an empirical formula, and setting network parameters according to a task, training the incremental echo state network by using the prediction sample set, and performing prediction by using the trained network.

[0009] As a preferred scheme of the load prediction method based on the incremental echo state network, the obtaining of the prediction sample set comprises setting the length before a point to be predicted as N, and representing the load sequence as:

[0010] S=(s[i], i=0, 1,..., N-1)

[0011] The set of retrieved space historical sequences is set as T h , the set of similar sequences is set as T s ; T h contains historical load sequences of the same period in the previous L days of the day to be predicted, and T s is the prediction sample set, the upper limit of the capacity is Q, and the given threshold is ε; for the load sequence X ∈ T, if D(S, X) < ε is satisfied, X ∈ T s ; the determination is repeated until T s reaches the upper limit of the capacity or the elements in T h have been traversed.

[0012] As a preferred scheme of the load prediction method based on the incremental echo state network, the determination of the number of hidden layer neurons of the incremental echo state network comprises determining the number of neurons according to an empirical formula; the empirical formula is represented as:

[0013]

[0014] wherein N h represents the determined number of neurons, N i represents the number of input layer neurons, N o represents the number of output layer neurons, N s represents the number of samples of the training set, and α represents an arbitrary variable, and the range is usually 2-10.

[0015] As a preferred scheme of the load prediction method based on the incremental echo state network, the network parameters comprise a spectral radius of a reserve pool internal connection weight matrix, an input unit scale, and a reserve pool sparsity degree.

[0016] As a preferred scheme of the load prediction method based on the incremental echo state network, the spectral radius of the internal connection weight matrix of the reservoir includes a reservoir internal connection matrix W initialization process, which is to randomly generate a connection matrix W0 with a size of N*N, and the connection weight value is selected from [-1, 1] with a uniform probability distribution; the spectral radius λ of W0 is calculated max , which is expressed as:

[0017]

[0018] Normalization processing is performed, and W1 is multiplied by a coefficient α', 0 < α' < 1, which is expressed as:

[0019] W = α'W1.

[0020] As a preferred scheme of the load prediction method based on the incremental echo state network, the training of the incremental echo state network includes initializing the input weight matrix W in , W and the feedback weight W back according to the parameters set according to the network size; at the initialization time, the reservoir state vector X(0) at time 0 = 0; the network state is collected, assuming that the training sample set has M pairs of (u(i), y(i)) (i = 1, 2, 3,..., M), then the reservoir state is updated, which is expressed as:

[0021] x(n+1) = f(W in u(n+1) + Wx(n) + W back y(n))

[0022] Wherein, u(n+1) represents the input of the n+1 step, x(n) represents the state vector of the reservoir, and y(n) represents the output vector; assuming that the state length discarded is m, then the network state matrix S formed by the training sample set is (M-m) × (K+N+L) dimension; wherein, the (i+1) row of the network state matrix is composed of the input, the reservoir state and the training sample output, which is expressed as:

[0023] S(i+1,:) = [u(n+1)', X(n+1)', y(n)]

[0024] Collecting signal state, assuming that the output function f out is a tanh function, then the target output of the training sample is taken as the inverse function, that is, tanh -1 (y(i)), thereby forming the output state matrix D.

[0025] As a preferred scheme of the load prediction method based on the incremental echo state network, wherein: the training of the incremental echo state network further comprises that the input weight, the internal connection weight and the feedback weight of the echo state network are randomly determined, and the output weight is represented as

[0026] W out = (S T S) -1 S T D

[0027] For (S T S) -1 , the pseudo-inverse is usually replaced, that is (S T S) +1 .

[0028] Another object of the present application is to provide a load prediction system based on an incremental echo state network, which can make the incremental echo state network more flexible to adapt to changes in load data through a dynamic adjustment mechanism, solving the problem that the current technology is difficult to dynamically adjust the number of neurons.

[0029] As a preferred scheme of the load prediction system based on the incremental echo state network, wherein: it comprises a sample retrieval module, a network construction module and a network training prediction module; the sample retrieval module is used to define a similarity measurement standard, and retrieve a prediction sample set based on similar time series; the network construction module is used to calculate and determine the number of neurons of the incremental echo state network according to an empirical formula, and set network parameters according to tasks; the network training prediction module is used to train the incremental echo state network by using the prediction sample set, and make prediction by using the trained network.

[0030] A computer device comprising a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the steps of the load prediction method based on the incremental echo state network.

[0031] A computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of the load prediction method based on the incremental echo state network.

[0032] The beneficial effects of this invention are as follows: The load forecasting method based on incremental echo state networks provided by this invention employs similar time series retrieval technology to quickly locate historical data similar to the target load pattern by quantitatively analyzing and comparing the characteristics of time series, thereby providing a more accurate sample set for the load forecasting model; by proposing an empirical formula to adaptively adjust the number of hidden layer neurons according to the characteristics of load data and the requirements of the forecasting task, dynamic optimization of the network structure is achieved, improving the flexibility and accuracy of load forecasting; the load forecasting method adopted integrates multiple technical advantages of ESN, including random and fixed internal node connections, dynamic time series processing capabilities, and significantly improves the performance of load forecasting through innovative similarity retrieval and neuron number adjustment mechanisms; by controlling the sparsity of the reserve pool, the amount of computation during network training is reduced, thus lowering the computational complexity. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 The first embodiment of the present invention provides an overall flowchart of a load prediction method based on incremental echo state network.

[0035] Figure 2 The topology diagram of an ESN based on an incremental echo state network for load forecasting provided in the first embodiment of the present invention is shown.

[0036] Figure 3 The following is an overall flowchart of a load forecasting system based on an incremental echo state network, provided for the third embodiment of the present invention. Detailed Implementation

[0037] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0038] Example 1, referring to Figures 1-2 As one embodiment of the present invention, a load forecasting method based on incremental echo state networks is provided, comprising:

[0039] S1: Define similarity metrics and obtain a predicted sample set based on similar time series retrieval.

[0040] Furthermore, obtaining the prediction sample set includes setting the length before the point to be predicted to be N, and representing the load sequence as follows:

[0041] S = (s[i], i = 0, 1, ..., N-1)

[0042] Let T be the set of historical sequences in the retrieval space. h Let T be the set of similar sequences. s ;T h Includes historical load sequences for the same period L days before the forecast date, T s This is the prediction sample set, with a maximum capacity of Q and a given threshold of ε; for a load sequence X∈T, if D(S,X)<ε, then X∈T s Repeat the judgment until T. s Reaching capacity limit or already having T h Traversing elements in the array.

[0043] It should be noted that determining the number of hidden layer neurons in an incremental echo state network includes determining the number of neurons based on an empirical formula; the empirical formula is expressed as:

[0044]

[0045] Where, N h N represents a specific number of neurons. i N represents the number of neurons in the input layer. o N represents the number of neurons in the output layer. s This represents the number of samples in the training set, and α represents any value that can be chosen, typically ranging from 2 to 10.

[0046] It should also be noted that, such as Figure 2 The diagram shows the topology of the Incremental Echo State Network (ESN). From left to right, the layers are the input layer, hidden layer, and output layer. During training, the input connection weight matrix is ​​randomly initialized, while the output connection weight matrix is ​​changed.

[0047] It should also be noted that the number of neurons is determined as follows: Assume the network input nodes are K, the number of neurons in the pool is N, the number of output nodes is L, and the relationships between the layers of the network are as follows:

[0048] The input layer with K nodes and the hidden layer with N nodes are connected by input weights. This matrix has dimensions N×K and is denoted as W. inThe N nodes in the reserve pool are connected by an internal connection weight matrix, which has an N×N dimension and is denoted as W. The connection from the output layer to the reserve pool is called the feedback weight, which has an N×L dimension and is denoted as W. back The output weights are formed by the connections between the input layer, the pool layer, and the output layer itself, and are matrixed as (K+N+L)×L, denoted as W. out The input weights, reserve pool weights, and feedback weights are generated randomly, and the output weights are obtained through linear regression.

[0049] Suppose the input to the network at step n is u(n), the state vector of the reservoir is x(n), and the output vector is y(n), expressed as:

[0050] u(n) = [u1(n), u2(n), ..., u k (n)] T

[0051] x(n) = [x1(n), x2(n), ..., x N (n)] T

[0052] y(n)=[y1(n),y2(n),...,y L (n)] T

[0053] The status of each unit in the reserve pool is updated, as shown below:

[0054] x(n+1)=f(W in u(n+1)+Wx(n)+W back y(n))

[0055] The network output is represented as:

[0056] y(n+1)=f out (W out (u(n+1),x(n+1),y(n)))

[0057] Where f = (f1, f2, ..., f N () represents the activation function of each node in the reservoir, typically taken as the sigmoid function or the hyperbolic tangent function (tanh). The output function of the network output node is represented by (u(n+1),x(n+1),y(n)), and the network state vector is composed of the input, internal unit function, and previous output.

[0058] S2: Calculate and determine the number of neurons in the incremental echo state network based on empirical formulas, and set the network parameters according to the task.

[0059] Furthermore, the network parameters include the spectral radius of the internal connection weight matrix of the reservoir, the scale of the input cells, and the sparsity of the reservoir.

[0060] It should be noted that the spectral radius of the internal connection weight matrix of the reservoir includes the initialization process of the internal connection matrix W, which involves randomly generating an N×N connection matrix W0, with the connection weights selected from the uniform probability distribution of [-1,1]; and calculating the spectral radius λ of W0. max , is represented as:

[0061]

[0062] After normalization, W1 is multiplied by the coefficient α', where 0 < α' < 1, and can be expressed as:

[0063] W = α'W1.

[0064] It should also be noted that the radius of the internal linkage weight spectrum of the reservoir refers to the maximum absolute value among the eigenvalues ​​of the internal linkage weight matrix W of the reservoir, denoted as λ. max .

[0065] The spectral radius is an important parameter of the reservoir and has a significant impact on the echo-state properties of the echo-state network. When λ max When the spectral radius is less than 1, the echo-state network has the echo-state property, which can ensure that the influence of the historical reservoir state and input on the network will disappear after a sufficiently long time. The spectral radius is less than 1, which is a sufficient but not necessary condition for the echo-state network to have the echo-state property.

[0066] It should also be noted that the input unit scale includes measures to make W in Uniform distribution, by assigning the input scale a as a variation interval [-a, a], W in The value of W is selected from the range of variation; for linear tasks... in When W is small, the tanh() activation function is close to a linear state. in At larger scales, neurons are prone to saturation, approaching 1 and -1, and exhibiting more nonlinear changes.

[0067] S3: Train the incremental echo state network using the prediction sample set, and then use the trained network for prediction.

[0068] Furthermore, training the incremental echo state network includes initializing the input weight matrix W based on parameters set according to the network size. in W and feedback weight W back During initialization, the reservoir state vector X(0) = 0 at time 0; network state collection, assuming there are M pairs (u(i), y(i)) (i = 1, 2, 3, ..., M) in the training sample set, then the reservoir state update is expressed as:

[0069] x(n+1)=f(W in u(n+1)+Wx(n)+W back y(n))

[0070] Where u(n+1) represents the input at step (n+1), x(n) represents the state vector of the reservoir, and y(n) represents the output vector; assuming the length of the discarded state is m, the network state matrix S formed by the training sample set is (Mm)×(K+N+L) dimensional; where the (i+1)th row of the network state matrix consists of the input, the reservoir state, and the training sample output, expressed as:

[0071] S(i+1,:)=[u(n+1)',X(n+1)',y(n)']

[0072] Collect signal states, assuming the output function f out If the function is tanh, then the inverse function of the target output over the training samples is tanh. -1 (y(i)), thus forming the output state matrix D.

[0073] It should be noted that training the incremental echo state network also includes randomly determining the input weights, internal connection weights, and feedback weights of the echo state network. Therefore, the output weights are expressed as:

[0074] W out =(S T S) -1 S T D

[0075] For (S) T S) -1 It is usually replaced by a pseudo-inverse, i.e. (S) T S) +1 .

[0076] Example 2 is an embodiment of the present invention, which provides a load forecasting method based on incremental echo state network. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiment.

[0077] First, this embodiment aims to verify a power load forecasting method based on incremental echo state networks, which improves the accuracy and adaptability of load forecasting by combining similar time series retrieval and dynamic neuron adjustment mechanisms.

[0078] Historical load data for a region's power system from January 2022 to June 2023 was collected, including hourly load values ​​and relevant meteorological data such as temperature, humidity, and wind speed. The data was categorized by date type into weekdays and weekends, and further subdivided into morning, afternoon, and evening time periods. Euclidean distance was used as the similarity metric to calculate the difference between two load sequences. Considering the influence of meteorological factors, temperature, humidity, and wind speed were also included in the similarity calculation. Taking the load forecast for 8:00 AM on June 1, 2023 as an example, similar weekday and weekend load sequences at 8:00 AM on June 1, along with their corresponding meteorological data, were retrieved from historical data. A similarity threshold ε = 0.1 was set, and the top 10 sequences most similar to the current load pattern were selected as the prediction sample set.

[0079] The number of hidden layer neurons was calculated using an empirical formula. The input layer neurons were set to M=6, the output layer neurons to L=1, and the training set samples to K=10, resulting in N=10 hidden layer neurons. The spectral radius of the internal connection weight matrix of the reservoir was set to 0.9, the input unit scale to 1, and the reservoir sparsity to 0.5. The incremental echo state network was trained using the predicted sample set to learn patterns and trends in the load data. The trained network was then used to predict the load at 8:00 AM on June 1, 2023, and compared with the actual load value.

[0080] Table 1. Incremental Echo State Network Load Forecast Data

[0081] Test subject Date Period Predicted load MW Actual load MW Error MW Prediction 1 2023-06-01 8 a.m. 1300 1297 3 Prediction 2 2023-06-02 8 a.m. 1285 1280 5 Prediction 3 2023-06-03 8 a.m. 1270 1271 1 Prediction 4 2023-06-04 8 a.m. 1254 1255 1 Prediction 5 2023-06-05 8 a.m. 1240 1238 2 Prediction 6 2023-06-06 8 a.m. 1225 1228 3

[0082] As shown in Table 1, the prediction errors of this invention are all less than 5MW, indicating that the prediction accuracy of this method is high. By comparing the prediction results with the actual load values, it can be seen that the method of this invention has the following beneficial effects: In terms of high prediction accuracy, the method of this invention can effectively capture complex patterns and trends in load data, and the prediction results have a high degree of agreement with the actual load values ​​with small errors; in terms of strong adaptability, the incremental echo state network can dynamically adjust the number of neurons according to the changes in load data, making it more flexible to adapt to load prediction tasks in different scenarios; in terms of efficiency, the method of this invention can quickly obtain the prediction sample set through similar time series retrieval and dynamically adjust the network structure using empirical formulas, thereby improving prediction efficiency.

[0083] Example 3, referring to Figure 3 As an embodiment of the present invention, a load prediction system based on incremental echo state network is provided, including a sample retrieval module, a network construction module, and a network training and prediction module.

[0084] The sample retrieval module is used to define similarity metrics and obtain a prediction sample set based on similar time series retrieval; the network construction module is used to calculate and determine the number of neurons in the incremental echo state network according to empirical formulas and set network parameters according to the task; the network training and prediction module is used to train the incremental echo state network using the prediction sample set and use the trained network for prediction.

[0085] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0086] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0087] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0088] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0089] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A load forecasting method based on incremental echo state networks, characterized in that, include: Define similarity metrics and obtain a prediction sample set based on similar time series retrieval; The number of neurons in the incremental echo state network is calculated and determined based on empirical formulas, and the network parameters are set according to the task. The incremental echo state network is trained using a prediction sample set, and the trained network is used for prediction. The process of obtaining the prediction sample set includes setting the length before the point to be predicted to be... The load sequence is represented as: ； Let the set of retrieval spatial historical sequences be denoted as Let the set of similar sequences be denoted as ; Including the date to be predicted Historical load sequences for the same period of the day This is the prediction sample set, with a maximum capacity of [value missing]. Given a threshold ; For load sequence If satisfied Then there is ; Repeat the judgment until... Reaching capacity limit or already Traversing elements in the array; Determining the number of hidden layer neurons in the incremental echo state network includes determining the number of neurons based on empirical formulas; The empirical formula is expressed as: ； in, This indicates a specific number of neurons. This indicates the number of neurons in the input layer. Indicates the number of neurons in the output layer. This represents the number of samples in the training set. This represents a variable that can take any value, ranging from 2 to 10. The network parameters include the spectral radius of the internal connection weight matrix of the reservoir, the scale of the input unit, and the sparsity of the reservoir. The spectral radius of the internal connectivity weight matrix of the storage pool includes the internal connectivity matrix of the storage pool. The initialization process involves randomly generating a size of Connection matrix The uniform probability distribution of connection weights from Selected from; right Normalization is represented as: ； Wherein, the radius of the internal connectivity weight spectrum of the reservoir refers to the internal connectivity weight matrix of the reservoir. The maximum absolute value among the eigenvalues ​​is denoted as . ; Will Multiply by a coefficient , , is represented as: 。 2. The load forecasting method based on incremental echo state networks as described in claim 1, characterized in that: Training the incremental echo state network includes initializing the input weight matrix according to parameters set based on the network size. , and feedback weights ; The state vector of the reservoir at time 0 during initialization. ; Network state collection, assuming the training sample set has a total of right Then, the reserve pool status is updated, as shown below: ； in, Indicates the first +1 step input, This represents the state vector of the reserve pool. Indicates the output vector; Assume the length of the discarded state is The network state matrix formed by the training sample set for dimension; Among them, the network state matrix is ​​numbered... A row consists of the input, the reservoir state, and the training sample output, represented as: ； Collect signal states, and assume the output function. for The function is the inverse function of the target output of the training samples, which is... This forms the output state matrix. .

3. The load forecasting method based on incremental echo state networks as described in claim 2, characterized in that: The training of the incremental echo state network further includes randomly determining the input weights, internal connection weights, and feedback weights of the echo state network. The output weights are then expressed as: ； for Replace with pseudo-inverse, that is .

4. A system employing the load forecasting method based on incremental echo state networks as described in any one of claims 1 to 3, characterized in that: It includes a sample retrieval module, a network construction module, and a network training and prediction module; The sample retrieval module is used to define similarity measurement standards and obtain a predicted sample set based on similar time series retrieval. The network construction module is used to calculate and determine the number of neurons in the incremental echo state network based on empirical formulas, and to set network parameters according to the task. The network training and prediction module is used to train the incremental echo state network using the prediction sample set, and then use the trained network for prediction.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the load forecasting method based on incremental echo state network as described in any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the load forecasting method based on incremental echo state network as described in any one of claims 1 to 3.

Images