2.2. Electric Vehicles

EVs play a significant role in the smart grid infrastructure and they also solve some of the environmental problems by facilitating green travel. However, EV users are facing challenge in charging, as the number of EVs are increasing while the number of available charging stations are decreased. Additionally, short communication ranges, the need for frequent communications and mobility of EVs add new security and privacy concerns ^[25][84]. On the other hand, the high penetration of EVs and less-coordinated charging schedules might lead to overloading in the grid. Hence, there in an open challenge to integrate all these EVs in the grid, schedule the charging operations effectively and develop a transparent and secure charging system. Many researchers adopt decentralized BC technology to solve these issues and enhance the use of EVs and its charging management ^{[18][26][27][28][29][30][31][32]}[62,85,86,87,88,89,90,91].

2.3. Demand Response Management

Several research works explore different techniques to develop an innovative DRM system with the adoption of BC technology. The main idea of most of the works was to develop a secure trading platform for distributed prosumers and EVs. Smart contracts which can execute upon reaching desired conditions gained great acceptance in the DR application ^{[33][34][35][36][37][38][39][40]}[92,93,94,95,96,97,98,99]. Some of the works which aims to develop better DR prototypes are listed in Table 17.

Combining BC and ML for various applications is increasing. The learning capabilities of machine learning can be applied with blockchain to make the chain smarter ^[47][105]. Additionally, the computational power for ML models can minimize the time taken for the nonce calculation. On the other hand, using the decentralized data architecture of blockchain, better ML models can be built. The main benefits of combining BC and ML are (a) security, (b) improving energy consumption, (c) improving the smart contracts, (d) data trading and more. BC is a continuously growing data ledger, whereas ML requires large volumes of data. Both techniques complement each other and have high potential when combined for different applications. With the decentralized and immutable structure, BC builds a trust environment without third parties for any transactions occurring. ML-based solutions mainly help in demand management using prediction techniques and help in handling large amount of data from smart meters ^[48][33]. Additionally, optimization techniques can be used for scheduling the energy usage of users. Applying BC into ML systems helps such systems to use information collected by BCs to clear problems quickly, and applying ML techniques to BC adds powerful intelligence to BC systems for data processing and other intensive applications ^[11][24]. The storing of large-scale data in a secure and tamper-proof manner by BC-based systems ensures the confidentiality and auditability of the collaborative data training process and trained model using cryptographic techniques. Additionally, it enables decentralization that ensure access control with central authorities. Smart contracts and Dapps allow to model interactions between entities in the decentralized ML application. In the same way, adding ML helps the BC system to achieve energy efficiency, scalability, security and intelligent smart contracts. As ML can predict and speedily calculate data, it provides a feasible way for miners in selecting important transactions. Using predictive analysis can ensure energy and resource requirements to be met and improve overall efficiency. Hence, considering the future SG scenario, both BC and ML can together solve the issues regarding energy trading, EV charging and scheduling, load prediction, DRM, security and privacy concerns. Table 29 summarizes the existing works that combine BC, ML and SGs.

2.4. Security and Privacy

Smart meters placed at each home in a smart grid provide real-time information about electricity consumption and pricing. This information may contain aspects about users which are private and thus require mechanisms to preserve privacy. If a malicious attacker gains access to this data, they can track the energy usage behavior of users based on their consumption profile. The major concern for every user or entity in the SG environment is privacy, and the associated security of such information exchange ^[22][65]. Most of the works discussed in previous sections (Section 3.1, Section 3.2 and Section 3.3) aim to provide security and privacy in smart grid operations using BC technology. Similarly, security of power information system is also an important area of research. SG incorporates several communications for proper functioning and power delivery, and this information has to be stored securely. If the system is hacked, it may impact the grid and create unpredictable effects in the grid, leading grid operators toward making wrong energy decisions, eventually resulting in interruptions or even large-scale blackouts. Providing security for all parts of the SG is critical. The use of BC alone does not directly guarantee security and privacy; however, adding advanced cryptographic mechanisms such as multi-signatures, elliptic curve cryptography, etc., are also needed.

3.1. Attack Detection and Security

Ref. ^[42][20] addresses the problem of false data injection in smart grid and ML-based techniques to detect such attacks. Information exchanges in SG are mainly unidirectional, for example, information to control centers based on measurements. These data should be clearly monitored for studying malicious behaviors in the SG. Their work mainly uses two ML techniques to detect attacks in SG (supervised and unsupervised techniques) and classify normal data and tampered data. Ref. ^[43][19] uses clustering methods to find representative periods for optimizing energy systems.

3.2. Data Analysis

In ^[44][22], a comprehensive study on the application of big data and ML and the associated security concerns in the electricity grid is presented. The massive data generated from the grid, termed ‘Big Data’, require proper optimization from generation to distribution side and have to be processed/analyzed by implementing effective security solutions. Two major concerns regarding any data generated in SGs are processing the gathered data by satisfying the time limits and overcoming the involved security concerns ^[44][22]. ML techniques can be applied to predict the usage patterns and preferences of users in SG. Solar power plants and other RESs are affected greatly by seasonal changes. Thus, a system that can predict, in advance, solar energy information is required to minimize the operational costs caused.

3.3. Demand Response Management

Ref. ^[45][7] studies demand response algorithm implementation for SG residential buildings using ML models. They combine the optimization technique and ML models to find the optimal strategy for DRM and use metered data to train and test the algorithm. Two demand response algorithms—rule-based and predictive-based (ML-based)—were deployed and evaluated using a common demand price scheme. Ref. ^[46][21] compares two ANN-based load forecasting techniques: echo state network (ESN) and feed forward neural networks (FFNN) for demand response programs in SGs. The novelty of their method is that they compare two ANN techniques over a complete and heterogeneous data set which are pre-processed with adaptive clustering techniques for achieving optimized forecasting performance. Their work can help power system planners for forecasting user-based profiles and implement DR programs.